/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 624 by ph10, Tue Jul 19 10:43:28 2011 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) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
# Line 76  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       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -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       0,      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,      0, -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 106  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
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 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       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,     0, -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,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 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 164  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 171  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 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  };    "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      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  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 262  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 298  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 373  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
549      int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553  /*************************************************  /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579    /*************************************************
580  *            Handle escapes                      *  *            Handle escapes                      *
581  *************************************************/  *************************************************/
582    
# Line 399  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
# Line 417  ptr--; /* Set Line 615  ptr--; /* Set
615    
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* 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
619  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.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 442  else Line 640  else
640      /* 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
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650      /* \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:
     is an absolute backreference. If negative, it is a relative backreference.  
     This is a Perl 5.10 feature. */  
651    
652      case 'g':      (1) A number, either plain or braced. If positive, it is an absolute
653      if (ptr[1] == '{')      backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668        {        {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676          {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685        braced = TRUE;        braced = TRUE;
686        ptr++;        ptr++;
687        }        }
688      else braced = FALSE;      else braced = FALSE;
689    
690      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
691        {        {
692        negated = TRUE;        negated = TRUE;
693        ptr++;        ptr++;
# Line 471  else Line 696  else
696    
697      c = 0;      c = 0;
698      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
699        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        return 0;        break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717        }        }
718    
719      if (negated)      if (negated)
# Line 484  else Line 721  else
721        if (c > bracount)        if (c > bracount)
722          {          {
723          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
724          return 0;          break;
725          }          }
726        c = bracount - (c - 1);        c = bracount - (c - 1);
727        }        }
# Line 504  else Line 741  else
741      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
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      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:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 767  else
767      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.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 538  else Line 780  else
780      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
781      than 3 octal digits. */      than 3 octal digits. */
782    
783      case '0':      case CHAR_0:
784      c -= '0';      c -= CHAR_0;
785      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
787      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
# Line 549  else Line 791  else
791      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
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 559  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 587  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      if (c >= 'a' && c <= 'z') c -= 32;        {
859          *errorcodeptr = ERR68;
860          break;
861          }
862        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      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
873      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
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 677  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \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
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 713  top = _pcre_utt_size; Line 973  top = _pcre_utt_size;
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top) >> 1;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0)    if (c == 0)
978      {      {
979      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 756  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 797  int max = -1; Line 1057  int max = -1;
1057  /* 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
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* 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.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1101  *************************************************/  *************************************************/
1102    
1103  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106  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
1107  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
1108  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1109  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  be terminated by '>' because that is checked in the first pass.  
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1126    count        current count of capturing parens so far encountered    cd           compile background data
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1134  */  */
1135    
1136  static int  static int
1137  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,
1138    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  const uschar *thisname;  uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1232      {      {
1233      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1234      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1235        {        {
1236        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1238        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1239        }        }
1240      continue;      continue;
1241      }      }
1242    
1243    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250      {      {
1251      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1252        for (;;)
1253          {
1254          if (ptr[1] == CHAR_BACKSLASH)
1255            {
1256            if (ptr[2] == CHAR_E)
1257              ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280        {        {
1281        if (*ptr == '\\')        if (*ptr == 0) return -1;
1282          if (*ptr == CHAR_BACKSLASH)
1283          {          {
1284          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1285          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1286            {            {
1287            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1289            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1290            }            }
1291          continue;          continue;
1292          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1296  for (; *ptr != 0; ptr++)
1296    
1297    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1298    
1299    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1302      if (*ptr == 0) return -1;      while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
1313    
1314    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1315    
1316    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1317      {      {
1318      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1320      continue;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
1322    
1323    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1337         *ptr != '\'')  *ptrptr = ptr;
1338      continue;  return -1;
1339    }
1340    
1341    
1342    
1343    
1344    /*************************************************
1345    *       Find forward referenced subpattern       *
1346    *************************************************/
1347    
1348    /* This function scans along a pattern's text looking for capturing
1349    subpatterns, and counting them. If it finds a named pattern that matches the
1350    name it is given, it returns its number. Alternatively, if the name is NULL, it
1351    returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    count++;  static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    if (name == NULL && count == lorn) return count;  for (;;)
1383    term = *ptr++;    {
1384    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385    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;  
1386    }    }
1387    
1388  return -1;  return rc;
1389  }  }
1390    
1391    
1392    
1393    
1394  /*************************************************  /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1400  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1401  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1406    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1407    
1408  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1409  */  */
1410    
1411  static const uschar*  static const uschar*
1412  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1413  {  {
1414  for (;;)  for (;;)
1415    {    {
1416    switch ((int)*code)    switch ((int)*code)
1417      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1418      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1419      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1420      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 996  for (;;) Line 1430  for (;;)
1430    
1431      case OP_CALLOUT:      case OP_CALLOUT:
1432      case OP_CREF:      case OP_CREF:
1433        case OP_NCREF:
1434      case OP_RREF:      case OP_RREF:
1435        case OP_NRREF:
1436      case OP_DEF:      case OP_DEF:
1437      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1438      break;      break;
# Line 1012  for (;;) Line 1448  for (;;)
1448    
1449    
1450  /*************************************************  /*************************************************
1451  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1452  *************************************************/  *************************************************/
1453    
1454  /* 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,
1455  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.
1456  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
1457    temporarily terminated with OP_END when this function is called.
1458    
1459    This function is called when a backward assertion is encountered, so that if it
1460    fails, the error message can point to the correct place in the pattern.
1461    However, we cannot do this when the assertion contains subroutine calls,
1462    because they can be forward references. We solve this by remembering this case
1463    and doing the check at the end; a flag specifies which mode we are running in.
1464    
1465  Arguments:  Arguments:
1466    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1467    options  the compiling options    utf8     TRUE in UTF-8 mode
1468      atend    TRUE if called when the pattern is complete
1469      cd       the "compile data" structure
1470    
1471  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1472                 or -1 if there is no fixed length,
1473               or -2 if \C was encountered               or -2 if \C was encountered
1474                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1475  */  */
1476    
1477  static int  static int
1478  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1479  {  {
1480  int length = -1;  int length = -1;
1481    
# Line 1041  branch, check the length against that of Line 1488  branch, check the length against that of
1488  for (;;)  for (;;)
1489    {    {
1490    int d;    int d;
1491      uschar *ce, *cs;
1492    register int op = *cc;    register int op = *cc;
   
1493    switch (op)    switch (op)
1494      {      {
1495        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1496        OP_BRA (normal non-capturing bracket) because the other variants of these
1497        opcodes are all concerned with unlimited repeated groups, which of course
1498        are not of fixed length. They will cause a -1 response from the default
1499        case of this switch. */
1500    
1501      case OP_CBRA:      case OP_CBRA:
1502      case OP_BRA:      case OP_BRA:
1503      case OP_ONCE:      case OP_ONCE:
1504      case OP_COND:      case OP_COND:
1505      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1506      if (d < 0) return d;      if (d < 0) return d;
1507      branchlength += d;      branchlength += d;
1508      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1058  for (;;) Line 1511  for (;;)
1511    
1512      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1513      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1514      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1515        Note that we must not include the OP_KETRxxx opcodes here, because they
1516        all imply an unlimited repeat. */
1517    
1518      case OP_ALT:      case OP_ALT:
1519      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1520      case OP_END:      case OP_END:
1521      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1522        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1072  for (;;) Line 1525  for (;;)
1525      branchlength = 0;      branchlength = 0;
1526      break;      break;
1527    
1528        /* A true recursion implies not fixed length, but a subroutine call may
1529        be OK. If the subroutine is a forward reference, we can't deal with
1530        it until the end of the pattern, so return -3. */
1531    
1532        case OP_RECURSE:
1533        if (!atend) return -3;
1534        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1535        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1536        if (cc > cs && cc < ce) return -1;                /* Recursion */
1537        d = find_fixedlength(cs + 2, utf8, atend, cd);
1538        if (d < 0) return d;
1539        branchlength += d;
1540        cc += 1 + LINK_SIZE;
1541        break;
1542    
1543      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1544    
1545      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1553  for (;;)
1553    
1554      case OP_REVERSE:      case OP_REVERSE:
1555      case OP_CREF:      case OP_CREF:
1556        case OP_NCREF:
1557      case OP_RREF:      case OP_RREF:
1558        case OP_NRREF:
1559      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1560      case OP_CALLOUT:      case OP_CALLOUT:
1561      case OP_SOD:      case OP_SOD:
1562      case OP_SOM:      case OP_SOM:
1563        case OP_SET_SOM:
1564      case OP_EOD:      case OP_EOD:
1565      case OP_EODN:      case OP_EODN:
1566      case OP_CIRC:      case OP_CIRC:
1567        case OP_CIRCM:
1568      case OP_DOLL:      case OP_DOLL:
1569        case OP_DOLLM:
1570      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1571      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1572      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1103  for (;;) Line 1575  for (;;)
1575      /* Handle literal characters */      /* Handle literal characters */
1576    
1577      case OP_CHAR:      case OP_CHAR:
1578      case OP_CHARNC:      case OP_CHARI:
1579      case OP_NOT:      case OP_NOT:
1580        case OP_NOTI:
1581      branchlength++;      branchlength++;
1582      cc += 2;      cc += 2;
1583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1584      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1585  #endif  #endif
1586      break;      break;
1587    
# Line 1122  for (;;) Line 1592  for (;;)
1592      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1593      cc += 4;      cc += 4;
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1596  #endif  #endif
1597      break;      break;
1598    
1599      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1600      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1601        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1602      cc += 4;      cc += 4;
1603      break;      break;
1604    
# Line 1148  for (;;) Line 1616  for (;;)
1616      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1617      case OP_WORDCHAR:      case OP_WORDCHAR:
1618      case OP_ANY:      case OP_ANY:
1619        case OP_ALLANY:
1620      branchlength++;      branchlength++;
1621      cc++;      cc++;
1622      break;      break;
# Line 1202  for (;;) Line 1671  for (;;)
1671    
1672    
1673  /*************************************************  /*************************************************
1674  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1675  *************************************************/  *************************************************/
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1679    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1680    so that it can be called from pcre_study() when finding the minimum matching
1681    length.
1682    
1683  Arguments:  Arguments:
1684    code        points to start of expression    code        points to start of expression
1685    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1686    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1687    
1688  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
1689  */  */
1690    
1691  static const uschar *  const uschar *
1692  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1693  {  {
1694  for (;;)  for (;;)
1695    {    {
1696    register int c = *code;    register int c = *code;
1697    
1698    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1699    
1700    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1230  for (;;) Line 1703  for (;;)
1703    
1704    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1705    
1706      /* Handle recursion */
1707    
1708      else if (c == OP_REVERSE)
1709        {
1710        if (number < 0) return (uschar *)code;
1711        code += _pcre_OP_lengths[c];
1712        }
1713    
1714    /* Handle capturing bracket */    /* Handle capturing bracket */
1715    
1716    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1717               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1718      {      {
1719      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1720      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1721      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1722      }      }
1723    
1724    /* 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
1725    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
1726    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1727      must add in its length. */
1728    
1729    else    else
1730      {      {
1731        switch(c)
1732          {
1733          case OP_TYPESTAR:
1734          case OP_TYPEMINSTAR:
1735          case OP_TYPEPLUS:
1736          case OP_TYPEMINPLUS:
1737          case OP_TYPEQUERY:
1738          case OP_TYPEMINQUERY:
1739          case OP_TYPEPOSSTAR:
1740          case OP_TYPEPOSPLUS:
1741          case OP_TYPEPOSQUERY:
1742          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1743          break;
1744    
1745          case OP_TYPEUPTO:
1746          case OP_TYPEMINUPTO:
1747          case OP_TYPEEXACT:
1748          case OP_TYPEPOSUPTO:
1749          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1750          break;
1751    
1752          case OP_MARK:
1753          case OP_PRUNE_ARG:
1754          case OP_SKIP_ARG:
1755          code += code[1];
1756          break;
1757    
1758          case OP_THEN_ARG:
1759          code += code[1+LINK_SIZE];
1760          break;
1761          }
1762    
1763        /* Add in the fixed length from the table */
1764    
1765      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1766    
1767      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1768      a multi-byte character. The length in the table is a minimum, so we have to
1769      arrange to skip the extra bytes. */
1770    
1771  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1772      if (utf8) switch(c)      if (utf8) switch(c)
1773        {        {
1774        case OP_CHAR:        case OP_CHAR:
1775        case OP_CHARNC:        case OP_CHARI:
1776        case OP_EXACT:        case OP_EXACT:
1777          case OP_EXACTI:
1778        case OP_UPTO:        case OP_UPTO:
1779          case OP_UPTOI:
1780        case OP_MINUPTO:        case OP_MINUPTO:
1781          case OP_MINUPTOI:
1782        case OP_POSUPTO:        case OP_POSUPTO:
1783          case OP_POSUPTOI:
1784        case OP_STAR:        case OP_STAR:
1785          case OP_STARI:
1786        case OP_MINSTAR:        case OP_MINSTAR:
1787          case OP_MINSTARI:
1788        case OP_POSSTAR:        case OP_POSSTAR:
1789          case OP_POSSTARI:
1790        case OP_PLUS:        case OP_PLUS:
1791          case OP_PLUSI:
1792        case OP_MINPLUS:        case OP_MINPLUS:
1793          case OP_MINPLUSI:
1794        case OP_POSPLUS:        case OP_POSPLUS:
1795          case OP_POSPLUSI:
1796        case OP_QUERY:        case OP_QUERY:
1797          case OP_QUERYI:
1798        case OP_MINQUERY:        case OP_MINQUERY:
1799          case OP_MINQUERYI:
1800        case OP_POSQUERY:        case OP_POSQUERY:
1801          case OP_POSQUERYI:
1802        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1803        break;        break;
1804        }        }
1805    #else
1806        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1807  #endif  #endif
1808      }      }
1809    }    }
# Line 1303  for (;;) Line 1840  for (;;)
1840    
1841    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1842    
1843    /* 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
1844    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
1845    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
1846    bytes. */    must add in its length. */
1847    
1848    else    else
1849      {      {
1850        switch(c)
1851          {
1852          case OP_TYPESTAR:
1853          case OP_TYPEMINSTAR:
1854          case OP_TYPEPLUS:
1855          case OP_TYPEMINPLUS:
1856          case OP_TYPEQUERY:
1857          case OP_TYPEMINQUERY:
1858          case OP_TYPEPOSSTAR:
1859          case OP_TYPEPOSPLUS:
1860          case OP_TYPEPOSQUERY:
1861          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1862          break;
1863    
1864          case OP_TYPEPOSUPTO:
1865          case OP_TYPEUPTO:
1866          case OP_TYPEMINUPTO:
1867          case OP_TYPEEXACT:
1868          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1869          break;
1870    
1871          case OP_MARK:
1872          case OP_PRUNE_ARG:
1873          case OP_SKIP_ARG:
1874          code += code[1];
1875          break;
1876    
1877          case OP_THEN_ARG:
1878          code += code[1+LINK_SIZE];
1879          break;
1880          }
1881    
1882        /* Add in the fixed length from the table */
1883    
1884      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1885    
1886        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1887        by a multi-byte character. The length in the table is a minimum, so we have
1888        to arrange to skip the extra bytes. */
1889    
1890  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1891      if (utf8) switch(c)      if (utf8) switch(c)
1892        {        {
1893        case OP_CHAR:        case OP_CHAR:
1894        case OP_CHARNC:        case OP_CHARI:
1895        case OP_EXACT:        case OP_EXACT:
1896          case OP_EXACTI:
1897        case OP_UPTO:        case OP_UPTO:
1898          case OP_UPTOI:
1899        case OP_MINUPTO:        case OP_MINUPTO:
1900          case OP_MINUPTOI:
1901        case OP_POSUPTO:        case OP_POSUPTO:
1902          case OP_POSUPTOI:
1903        case OP_STAR:        case OP_STAR:
1904          case OP_STARI:
1905        case OP_MINSTAR:        case OP_MINSTAR:
1906          case OP_MINSTARI:
1907        case OP_POSSTAR:        case OP_POSSTAR:
1908          case OP_POSSTARI:
1909        case OP_PLUS:        case OP_PLUS:
1910          case OP_PLUSI:
1911        case OP_MINPLUS:        case OP_MINPLUS:
1912          case OP_MINPLUSI:
1913        case OP_POSPLUS:        case OP_POSPLUS:
1914          case OP_POSPLUSI:
1915        case OP_QUERY:        case OP_QUERY:
1916          case OP_QUERYI:
1917        case OP_MINQUERY:        case OP_MINQUERY:
1918          case OP_MINQUERYI:
1919        case OP_POSQUERY:        case OP_POSQUERY:
1920          case OP_POSQUERYI:
1921        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1922        break;        break;
1923        }        }
1924    #else
1925        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1926  #endif  #endif
1927      }      }
1928    }    }
# Line 1347  for (;;) Line 1938  for (;;)
1938  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()
1939  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
1940  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1941  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
1942  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
1943    bracket whose current branch will already have been scanned.
1944    
1945  Arguments:  Arguments:
1946    code        points to start of search    code        points to start of search
1947    endcode     points to where to stop    endcode     points to where to stop
1948    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1949      cd          contains pointers to tables etc.
1950    
1951  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1952  */  */
1953    
1954  static BOOL  static BOOL
1955  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1956      compile_data *cd)
1957  {  {
1958  register int c;  register int c;
1959  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1960       code < endcode;       code < endcode;
1961       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1962    {    {
1963    const uschar *ccode;    const uschar *ccode;
1964    
1965    c = *code;    c = *code;
1966    
1967    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1968      first_significant_code() with a TRUE final argument. */
1969    
1970      if (c == OP_ASSERT)
1971      {      {
1972      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1973      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1974        continue;
1975        }
1976    
1977      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1978      implies a backward reference subroutine call, we can scan it. If it's a
1979      forward reference subroutine call, we can't. To detect forward reference
1980      we have to scan up the list that is kept in the workspace. This function is
1981      called only when doing the real compile, not during the pre-compile that
1982      measures the size of the compiled pattern. */
1983    
1984      if (c == OP_RECURSE)
1985        {
1986        const uschar *scode;
1987        BOOL empty_branch;
1988    
1989        /* Test for forward reference */
1990    
1991        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1992          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1993    
1994        /* Not a forward reference, test for completed backward reference */
1995    
1996      empty_branch = FALSE;      empty_branch = FALSE;
1997        scode = cd->start_code + GET(code, 1);
1998        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1999    
2000        /* Completed backwards reference */
2001    
2002      do      do
2003        {        {
2004        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2005            {
2006          empty_branch = TRUE;          empty_branch = TRUE;
2007        code += GET(code, 1);          break;
2008            }
2009          scode += GET(scode, 1);
2010        }        }
2011      while (*code == OP_ALT);      while (*scode == OP_ALT);
2012      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2013        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2014        continue;
2015        }
2016    
2017      /* Move past the KET and fudge things so that the increment in the "for"    /* Groups with zero repeats can of course be empty; skip them. */
     above has no effect. */  
2018    
2019      c = OP_END;    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2020      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];        c == OP_BRAPOSZERO)
2021        {
2022        code += _pcre_OP_lengths[c];
2023        do code += GET(code, 1); while (*code == OP_ALT);
2024        c = *code;
2025        continue;
2026        }
2027    
2028      /* A nested group that is already marked as "could be empty" can just be
2029      skipped. */
2030    
2031      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2032          c == OP_SCBRA || c == OP_SCBRAPOS)
2033        {
2034        do code += GET(code, 1); while (*code == OP_ALT);
2035        c = *code;
2036        continue;
2037        }
2038    
2039      /* For other groups, scan the branches. */
2040    
2041      if (c == OP_BRA  || c == OP_BRAPOS ||
2042          c == OP_CBRA || c == OP_CBRAPOS ||
2043          c == OP_ONCE || c == OP_COND)
2044        {
2045        BOOL empty_branch;
2046        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2047    
2048        /* If a conditional group has only one branch, there is a second, implied,
2049        empty branch, so just skip over the conditional, because it could be empty.
2050        Otherwise, scan the individual branches of the group. */
2051    
2052        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2053          code += GET(code, 1);
2054        else
2055          {
2056          empty_branch = FALSE;
2057          do
2058            {
2059            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2060              empty_branch = TRUE;
2061            code += GET(code, 1);
2062            }
2063          while (*code == OP_ALT);
2064          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2065          }
2066    
2067        c = *code;
2068      continue;      continue;
2069      }      }
2070    
# Line 1399  for (code = first_significant_code(code Line 2072  for (code = first_significant_code(code
2072    
2073    switch (c)    switch (c)
2074      {      {
2075      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2076        cannot be represented just by a bit map. This includes negated single
2077        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2078        actual length is stored in the compiled code, so we must update "code"
2079        here. */
2080    
2081  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2082      case OP_XCLASS:      case OP_XCLASS:
2083      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2084      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2085  #endif  #endif
2086    
# Line 1447  for (code = first_significant_code(code Line 2124  for (code = first_significant_code(code
2124      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2125      case OP_WORDCHAR:      case OP_WORDCHAR:
2126      case OP_ANY:      case OP_ANY:
2127        case OP_ALLANY:
2128      case OP_ANYBYTE:      case OP_ANYBYTE:
2129      case OP_CHAR:      case OP_CHAR:
2130      case OP_CHARNC:      case OP_CHARI:
2131      case OP_NOT:      case OP_NOT:
2132        case OP_NOTI:
2133      case OP_PLUS:      case OP_PLUS:
2134      case OP_MINPLUS:      case OP_MINPLUS:
2135      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1465  for (code = first_significant_code(code Line 2144  for (code = first_significant_code(code
2144      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2145      return FALSE;      return FALSE;
2146    
2147        /* These are going to continue, as they may be empty, but we have to
2148        fudge the length for the \p and \P cases. */
2149    
2150        case OP_TYPESTAR:
2151        case OP_TYPEMINSTAR:
2152        case OP_TYPEPOSSTAR:
2153        case OP_TYPEQUERY:
2154        case OP_TYPEMINQUERY:
2155        case OP_TYPEPOSQUERY:
2156        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2157        break;
2158    
2159        /* Same for these */
2160    
2161        case OP_TYPEUPTO:
2162        case OP_TYPEMINUPTO:
2163        case OP_TYPEPOSUPTO:
2164        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2165        break;
2166    
2167      /* End of branch */      /* End of branch */
2168    
2169      case OP_KET:      case OP_KET:
2170      case OP_KETRMAX:      case OP_KETRMAX:
2171      case OP_KETRMIN:      case OP_KETRMIN:
2172        case OP_KETRPOS:
2173      case OP_ALT:      case OP_ALT:
2174      return TRUE;      return TRUE;
2175    
# Line 1478  for (code = first_significant_code(code Line 2178  for (code = first_significant_code(code
2178    
2179  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2180      case OP_STAR:      case OP_STAR:
2181        case OP_STARI:
2182      case OP_MINSTAR:      case OP_MINSTAR:
2183        case OP_MINSTARI:
2184      case OP_POSSTAR:      case OP_POSSTAR:
2185        case OP_POSSTARI:
2186      case OP_QUERY:      case OP_QUERY:
2187        case OP_QUERYI:
2188      case OP_MINQUERY:      case OP_MINQUERY:
2189        case OP_MINQUERYI:
2190      case OP_POSQUERY:      case OP_POSQUERY:
2191        case OP_POSQUERYI:
2192        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2193        break;
2194    
2195      case OP_UPTO:      case OP_UPTO:
2196        case OP_UPTOI:
2197      case OP_MINUPTO:      case OP_MINUPTO:
2198        case OP_MINUPTOI:
2199      case OP_POSUPTO:      case OP_POSUPTO:
2200      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2201        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2202      break;      break;
2203  #endif  #endif
2204    
2205        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2206        string. */
2207    
2208        case OP_MARK:
2209        case OP_PRUNE_ARG:
2210        case OP_SKIP_ARG:
2211        code += code[1];
2212        break;
2213    
2214        case OP_THEN_ARG:
2215        code += code[1+LINK_SIZE];
2216        break;
2217    
2218        /* None of the remaining opcodes are required to match a character. */
2219    
2220        default:
2221        break;
2222      }      }
2223    }    }
2224    
# Line 1505  return TRUE; Line 2235  return TRUE;
2235  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2236  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2237  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2238    This function is called only during the real compile, not during the
2239    pre-compile.
2240    
2241  Arguments:  Arguments:
2242    code        points to start of the recursion    code        points to start of the recursion
2243    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2244    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2245    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2246      cd          pointers to tables etc
2247    
2248  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2249  */  */
2250    
2251  static BOOL  static BOOL
2252  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2253    BOOL utf8)    BOOL utf8, compile_data *cd)
2254  {  {
2255  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2256    {    {
2257    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2258        return FALSE;
2259    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2260    }    }
2261  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2268  return TRUE;
2268  *************************************************/  *************************************************/
2269    
2270  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2271  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
2272  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2273  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2274    
2275    Originally, this function only recognized a sequence of letters between the
2276    terminators, but it seems that Perl recognizes any sequence of characters,
2277    though of course unknown POSIX names are subsequently rejected. Perl gives an
2278    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2279    didn't consider this to be a POSIX class. Likewise for [:1234:].
2280    
2281    The problem in trying to be exactly like Perl is in the handling of escapes. We
2282    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2283    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2284    below handles the special case of \], but does not try to do any other escape
2285    processing. This makes it different from Perl for cases such as [:l\ower:]
2286    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2287    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2288    I think.
2289    
2290  Argument:  Arguments:
2291    ptr      pointer to the initial [    ptr      pointer to the initial [
2292    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2293    
2294  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2295  */  */
2296    
2297  static BOOL  static BOOL
2298  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2299  {  {
2300  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2301  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2302  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2303    {    {
2304    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2305    return TRUE;      {
2306        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2307        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2308          {
2309          *endptr = ptr;
2310          return TRUE;
2311          }
2312        }
2313    }    }
2314  return FALSE;  return FALSE;
2315  }  }
# Line 1581  Returns: a value representing the na Line 2334  Returns: a value representing the na
2334  static int  static int
2335  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2336  {  {
2337    const char *pn = posix_names;
2338  register int yield = 0;  register int yield = 0;
2339  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2340    {    {
2341    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2342      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2343      pn += posix_name_lengths[yield] + 1;
2344    yield++;    yield++;
2345    }    }
2346  return -1;  return -1;
# Line 1600  return -1; Line 2355  return -1;
2355  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2356  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2357  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
2358  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
2359  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
2360  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
2361  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
2362  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2363    OP_END.
2364    
2365  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2366  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1627  adjust_recurse(uschar *group, int adjust Line 2383  adjust_recurse(uschar *group, int adjust
2383    uschar *save_hwm)    uschar *save_hwm)
2384  {  {
2385  uschar *ptr = group;  uschar *ptr = group;
2386    
2387  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2388    {    {
2389    int offset;    int offset;
# Line 1680  auto_callout(uschar *code, const uschar Line 2437  auto_callout(uschar *code, const uschar
2437  {  {
2438  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2439  *code++ = 255;  *code++ = 255;
2440  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2441  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2442  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2443  }  }
2444    
# Line 1706  Returns: nothing Line 2463  Returns: nothing
2463  static void  static void
2464  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2465  {  {
2466  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2467  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2468  }  }
2469    
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2495  get_othercase_range(unsigned int *cptr,
2495  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2496    
2497  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2498    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2499    
2500  if (c > d) return FALSE;  if (c > d) return FALSE;
2501    
# Line 1747  next = othercase + 1; Line 2504  next = othercase + 1;
2504    
2505  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2506    {    {
2507    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2508    next++;    next++;
2509    }    }
2510    
# Line 1756  for (++c; c <= d; c++) Line 2513  for (++c; c <= d; c++)
2513    
2514  return TRUE;  return TRUE;
2515  }  }
2516    
2517    
2518    
2519    /*************************************************
2520    *        Check a character and a property        *
2521    *************************************************/
2522    
2523    /* This function is called by check_auto_possessive() when a property item
2524    is adjacent to a fixed character.
2525    
2526    Arguments:
2527      c            the character
2528      ptype        the property type
2529      pdata        the data for the type
2530      negated      TRUE if it's a negated property (\P or \p{^)
2531    
2532    Returns:       TRUE if auto-possessifying is OK
2533    */
2534    
2535    static BOOL
2536    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2537    {
2538    const ucd_record *prop = GET_UCD(c);
2539    switch(ptype)
2540      {
2541      case PT_LAMP:
2542      return (prop->chartype == ucp_Lu ||
2543              prop->chartype == ucp_Ll ||
2544              prop->chartype == ucp_Lt) == negated;
2545    
2546      case PT_GC:
2547      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2548    
2549      case PT_PC:
2550      return (pdata == prop->chartype) == negated;
2551    
2552      case PT_SC:
2553      return (pdata == prop->script) == negated;
2554    
2555      /* These are specials */
2556    
2557      case PT_ALNUM:
2558      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2559              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2560    
2561      case PT_SPACE:    /* Perl space */
2562      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2563              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2564              == negated;
2565    
2566      case PT_PXSPACE:  /* POSIX space */
2567      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2568              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2569              c == CHAR_FF || c == CHAR_CR)
2570              == negated;
2571    
2572      case PT_WORD:
2573      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2574              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2575              c == CHAR_UNDERSCORE) == negated;
2576      }
2577    return FALSE;
2578    }
2579  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2580    
2581    
# Line 1769  whether the next thing could possibly ma Line 2589  whether the next thing could possibly ma
2589  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2590    
2591  Arguments:  Arguments:
2592    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2593    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2594    ptr           next character in pattern    ptr           next character in pattern
2595    options       options bits    options       options bits
2596    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns: TRUE if possessifying is Line 2599  Returns: TRUE if possessifying is
2599  */  */
2600    
2601  static BOOL  static BOOL
2602  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2603    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2604  {  {
2605  int next;  int c, next;
2606    int op_code = *previous++;
2607    
2608  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2609    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2612  if ((options & PCRE_EXTENDED) != 0)
2612    for (;;)    for (;;)
2613      {      {
2614      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2615      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2616        {        {
2617        while (*(++ptr) != 0)        ptr++;
2618          while (*ptr != 0)
2619            {
2620          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2621            ptr++;
2622    #ifdef SUPPORT_UTF8
2623            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2624    #endif
2625            }
2626        }        }
2627      else break;      else break;
2628      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2631  if ((options & PCRE_EXTENDED) != 0)
2631  /* 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
2632  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2633    
2634  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2635    {    {
2636    int temperrorcode = 0;    int temperrorcode = 0;
2637    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2656  if ((options & PCRE_EXTENDED) != 0)
2656    for (;;)    for (;;)
2657      {      {
2658      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2659      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2660        {        {
2661        while (*(++ptr) != 0)        ptr++;
2662          while (*ptr != 0)
2663            {
2664          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2665            ptr++;
2666    #ifdef SUPPORT_UTF8
2667            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2668    #endif
2669            }
2670        }        }
2671      else break;      else break;
2672      }      }
2673    }    }
   
 /* If the next thing is itself optional, we have to give up. */  
   
 if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  
   return FALSE;  
2674    
2675  /* Now compare the next item with the previous opcode. If the previous is a  /* If the next thing is itself optional, we have to give up. */
 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. */  
2676    
2677    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2678      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2679        return FALSE;
2680    
2681  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2682    the next item is a character. */
2683    
2684  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2685    {    {
2686    case OP_CHAR:    case OP_CHAR:
2687  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2688    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2689    #else
2690      c = *previous;
2691  #endif  #endif
2692    return item != next;    return c != next;
2693    
2694    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2695    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
2696    high-valued characters. */    high-valued characters. */
2697    
2698    case OP_CHARNC:    case OP_CHARI:
2699  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2700    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2701    #else
2702      c = *previous;
2703  #endif  #endif
2704    if (item == next) return FALSE;    if (c == next) return FALSE;
2705  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2706    if (utf8)    if (utf8)
2707      {      {
2708      unsigned int othercase;      unsigned int othercase;
2709      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2710  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2711      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2712  #else  #else
2713      othercase = NOTACHAR;      othercase = NOTACHAR;
2714  #endif  #endif
2715      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2716      }      }
2717    else    else
2718  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2719    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2720    
2721    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2722      opcodes are not used for multi-byte characters, because they are coded using
2723      an XCLASS instead. */
2724    
2725    case OP_NOT:    case OP_NOT:
2726    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2727    if (item == next) return TRUE;  
2728    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2729      if ((c = *previous) == next) return TRUE;
2730  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2731    if (utf8)    if (utf8)
2732      {      {
2733      unsigned int othercase;      unsigned int othercase;
2734      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2735  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2736      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2737  #else  #else
2738      othercase = NOTACHAR;      othercase = NOTACHAR;
2739  #endif  #endif
2740      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2741      }      }
2742    else    else
2743  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2744    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2745    
2746      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2747      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2748    
2749    case OP_DIGIT:    case OP_DIGIT:
2750    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1925  if (next >= 0) switch(op_code) Line 2764  if (next >= 0) switch(op_code)
2764    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2765    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2766    
2767      case OP_HSPACE:
2768      case OP_NOT_HSPACE:
2769      switch(next)
2770        {
2771        case 0x09:
2772        case 0x20:
2773        case 0xa0:
2774        case 0x1680:
2775        case 0x180e:
2776        case 0x2000:
2777        case 0x2001:
2778        case 0x2002:
2779        case 0x2003:
2780        case 0x2004:
2781        case 0x2005:
2782        case 0x2006:
2783        case 0x2007:
2784        case 0x2008:
2785        case 0x2009:
2786        case 0x200A:
2787        case 0x202f:
2788        case 0x205f:
2789        case 0x3000:
2790        return op_code == OP_NOT_HSPACE;
2791        default:
2792        return op_code != OP_NOT_HSPACE;
2793        }
2794    
2795      case OP_ANYNL:
2796      case OP_VSPACE:
2797      case OP_NOT_VSPACE:
2798      switch(next)
2799        {
2800        case 0x0a:
2801        case 0x0b:
2802        case 0x0c:
2803        case 0x0d:
2804        case 0x85:
2805        case 0x2028:
2806        case 0x2029:
2807        return op_code == OP_NOT_VSPACE;
2808        default:
2809        return op_code != OP_NOT_VSPACE;
2810        }
2811    
2812    #ifdef SUPPORT_UCP
2813      case OP_PROP:
2814      return check_char_prop(next, previous[0], previous[1], FALSE);
2815    
2816      case OP_NOTPROP:
2817      return check_char_prop(next, previous[0], previous[1], TRUE);
2818    #endif
2819    
2820    default:    default:
2821    return FALSE;    return FALSE;
2822    }    }
2823    
2824    
2825  /* 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
2826    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2827    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2828    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2829    replaced by OP_PROP codes when PCRE_UCP is set. */
2830    
2831  switch(op_code)  switch(op_code)
2832    {    {
2833    case OP_CHAR:    case OP_CHAR:
2834    case OP_CHARNC:    case OP_CHARI:
2835  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2836    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2837    #else
2838      c = *previous;
2839  #endif  #endif
2840    switch(-next)    switch(-next)
2841      {      {
2842      case ESC_d:      case ESC_d:
2843      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2844    
2845      case ESC_D:      case ESC_D:
2846      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2847    
2848      case ESC_s:      case ESC_s:
2849      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2850    
2851      case ESC_S:      case ESC_S:
2852      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2853    
2854      case ESC_w:      case ESC_w:
2855      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2856    
2857      case ESC_W:      case ESC_W:
2858      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2859    
2860        case ESC_h:
2861        case ESC_H:
2862        switch(c)
2863          {
2864          case 0x09:
2865          case 0x20:
2866          case 0xa0:
2867          case 0x1680:
2868          case 0x180e:
2869          case 0x2000:
2870          case 0x2001:
2871          case 0x2002:
2872          case 0x2003:
2873          case 0x2004:
2874          case 0x2005:
2875          case 0x2006:
2876          case 0x2007:
2877          case 0x2008:
2878          case 0x2009:
2879          case 0x200A:
2880          case 0x202f:
2881          case 0x205f:
2882          case 0x3000:
2883          return -next != ESC_h;
2884          default:
2885          return -next == ESC_h;
2886          }
2887    
2888        case ESC_v:
2889        case ESC_V:
2890        switch(c)
2891          {
2892          case 0x0a:
2893          case 0x0b:
2894          case 0x0c:
2895          case 0x0d:
2896          case 0x85:
2897          case 0x2028:
2898          case 0x2029:
2899          return -next != ESC_v;
2900          default:
2901          return -next == ESC_v;
2902          }
2903    
2904        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2905        their substitutions and process them. The result will always be either
2906        -ESC_p or -ESC_P. Then fall through to process those values. */
2907    
2908    #ifdef SUPPORT_UCP
2909        case ESC_du:
2910        case ESC_DU:
2911        case ESC_wu:
2912        case ESC_WU:
2913        case ESC_su:
2914        case ESC_SU:
2915          {
2916          int temperrorcode = 0;
2917          ptr = substitutes[-next - ESC_DU];
2918          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2919          if (temperrorcode != 0) return FALSE;
2920          ptr++;    /* For compatibility */
2921          }
2922        /* Fall through */
2923    
2924        case ESC_p:
2925        case ESC_P:
2926          {
2927          int ptype, pdata, errorcodeptr;
2928          BOOL negated;
2929    
2930          ptr--;      /* Make ptr point at the p or P */
2931          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2932          if (ptype < 0) return FALSE;
2933          ptr++;      /* Point past the final curly ket */
2934    
2935          /* If the property item is optional, we have to give up. (When generated
2936          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2937          to the original \d etc. At this point, ptr will point to a zero byte. */
2938    
2939          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2940            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2941              return FALSE;
2942    
2943          /* Do the property check. */
2944    
2945          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2946          }
2947    #endif
2948    
2949      default:      default:
2950      return FALSE;      return FALSE;
2951      }      }
2952    
2953      /* In principle, support for Unicode properties should be integrated here as
2954      well. It means re-organizing the above code so as to get hold of the property
2955      values before switching on the op-code. However, I wonder how many patterns
2956      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2957      these op-codes are never generated.) */
2958    
2959    case OP_DIGIT:    case OP_DIGIT:
2960    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2961             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2962    
2963    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2964    return next == -ESC_d;    return next == -ESC_d;
2965    
2966    case OP_WHITESPACE:    case OP_WHITESPACE:
2967    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2968    
2969    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2970    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2971    
2972      case OP_HSPACE:
2973      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2974             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2975    
2976      case OP_NOT_HSPACE:
2977      return next == -ESC_h;
2978    
2979      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2980      case OP_ANYNL:
2981      case OP_VSPACE:
2982      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2983    
2984      case OP_NOT_VSPACE:
2985      return next == -ESC_v || next == -ESC_R;
2986    
2987    case OP_WORDCHAR:    case OP_WORDCHAR:
2988    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2989             next == -ESC_v || next == -ESC_R;
2990    
2991    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2992    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2040  BOOL inescq = FALSE; Line 3050  BOOL inescq = FALSE;
3050  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3051  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3052  const uschar *tempptr;  const uschar *tempptr;
3053    const uschar *nestptr = NULL;
3054  uschar *previous = NULL;  uschar *previous = NULL;
3055  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3056  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2049  uschar classbits[32]; Line 3060  uschar classbits[32];
3060  BOOL class_utf8;  BOOL class_utf8;
3061  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3062  uschar *class_utf8data;  uschar *class_utf8data;
3063    uschar *class_utf8data_base;
3064  uschar utf8_char[6];  uschar utf8_char[6];
3065  #else  #else
3066  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
3067  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3068  #endif  #endif
3069    
3070  #ifdef DEBUG  #ifdef PCRE_DEBUG
3071  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3072  #endif  #endif
3073    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 3100  req_caseopt = ((options & PCRE_CASELESS)
3100  for (;; ptr++)  for (;; ptr++)
3101    {    {
3102    BOOL negate_class;    BOOL negate_class;
3103      BOOL should_flip_negation;
3104    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3105    BOOL is_quantifier;    BOOL is_quantifier;
3106    BOOL is_recurse;    BOOL is_recurse;
3107      BOOL reset_bracount;
3108    int class_charcount;    int class_charcount;
3109    int class_lastchar;    int class_lastchar;
3110    int newoptions;    int newoptions;
3111    int recno;    int recno;
3112      int refsign;
3113    int skipbytes;    int skipbytes;
3114    int subreqbyte;    int subreqbyte;
3115    int subfirstbyte;    int subfirstbyte;
# Line 2106  for (;; ptr++) Line 3121  for (;; ptr++)
3121    
3122    c = *ptr;    c = *ptr;
3123    
3124      /* If we are at the end of a nested substitution, revert to the outer level
3125      string. Nesting only happens one level deep. */
3126    
3127      if (c == 0 && nestptr != NULL)
3128        {
3129        ptr = nestptr;
3130        nestptr = NULL;
3131        c = *ptr;
3132        }
3133    
3134    /* 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
3135    previous cycle of this loop. */    previous cycle of this loop. */
3136    
3137    if (lengthptr != NULL)    if (lengthptr != NULL)
3138      {      {
3139  #ifdef DEBUG  #ifdef PCRE_DEBUG
3140      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3141  #endif  #endif
3142      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3143        {        {
3144        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3145        goto FAILED;        goto FAILED;
# Line 2127  for (;; ptr++) Line 3152  for (;; ptr++)
3152      */      */
3153    
3154      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3155      *lengthptr += code - last_code;  
3156        /* Paranoid check for integer overflow */
3157    
3158        if (OFLOW_MAX - *lengthptr < code - last_code)
3159          {
3160          *errorcodeptr = ERR20;
3161          goto FAILED;
3162          }
3163    
3164        *lengthptr += (int)(code - last_code);
3165      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));
3166    
3167      /* 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 2154  for (;; ptr++) Line 3188  for (;; ptr++)
3188    /* 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
3189    reference list. */    reference list. */
3190    
3191    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3192      {      {
3193      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3194      goto FAILED;      goto FAILED;
# Line 2164  for (;; ptr++) Line 3198  for (;; ptr++)
3198    
3199    if (inescq && c != 0)    if (inescq && c != 0)
3200      {      {
3201      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3202        {        {
3203        inescq = FALSE;        inescq = FALSE;
3204        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 3224  for (;; ptr++)
3224    /* 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
3225    a quantifier. */    a quantifier. */
3226    
3227    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3228      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3229        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3230    
3231    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3232         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 3241  for (;; ptr++)
3241    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3242      {      {
3243      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3244      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3245        {        {
3246        while (*(++ptr) != 0)        ptr++;
3247          while (*ptr != 0)
3248          {          {
3249          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3250            ptr++;
3251    #ifdef SUPPORT_UTF8
3252            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3253    #endif
3254          }          }
3255        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3256    
# Line 2231  for (;; ptr++) Line 3271  for (;; ptr++)
3271      {      {
3272      /* ===================================================================*/      /* ===================================================================*/
3273      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3274      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3275      case ')':      case CHAR_RIGHT_PARENTHESIS:
3276      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3277      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3278      *codeptr = code;      *codeptr = code;
3279      *ptrptr = ptr;      *ptrptr = ptr;
3280      if (lengthptr != NULL)      if (lengthptr != NULL)
3281        {        {
3282        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3283            {
3284            *errorcodeptr = ERR20;
3285            goto FAILED;
3286            }
3287          *lengthptr += (int)(code - last_code);   /* To include callout length */
3288        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3289        }        }
3290      return TRUE;      return TRUE;
# Line 2249  for (;; ptr++) Line 3294  for (;; ptr++)
3294      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3295      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3296    
3297      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3298        previous = NULL;
3299      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3300        {        {
3301        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3302          *code++ = OP_CIRCM;
3303        }        }
3304      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3305      break;      break;
3306    
3307      case '$':      case CHAR_DOLLAR_SIGN:
3308      previous = NULL;      previous = NULL;
3309      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3310      break;      break;
3311    
3312      /* 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
3313      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3314    
3315      case '.':      case CHAR_DOT:
3316      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3317      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3318      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3319      previous = code;      previous = code;
3320      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3321      break;      break;
3322    
3323    
# Line 2286  for (;; ptr++) Line 3332  for (;; ptr++)
3332      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,
3333      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3334      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.
     */  
3335    
3336      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3337        default (Perl) mode, it is treated as a data character. */
3338    
3339        case CHAR_RIGHT_SQUARE_BRACKET:
3340        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3341          {
3342          *errorcodeptr = ERR64;
3343          goto FAILED;
3344          }
3345        goto NORMAL_CHAR;
3346    
3347        case CHAR_LEFT_SQUARE_BRACKET:
3348      previous = code;      previous = code;
3349    
3350      /* 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
3351      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. */
3352    
3353      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3354          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3355            check_posix_syntax(ptr, &tempptr))
3356        {        {
3357        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3358        goto FAILED;        goto FAILED;
3359        }        }
3360    
3361      /* 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,
3362        if the first few characters (either before or after ^) are \Q\E or \E we
3363        skip them too. This makes for compatibility with Perl. */
3364    
3365      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3366        for (;;)
3367        {        {
       negate_class = TRUE;  
3368        c = *(++ptr);        c = *(++ptr);
3369          if (c == CHAR_BACKSLASH)
3370            {
3371            if (ptr[1] == CHAR_E)
3372              ptr++;
3373            else if (strncmp((const char *)ptr+1,
3374                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3375              ptr += 3;
3376            else
3377              break;
3378            }
3379          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3380            negate_class = TRUE;
3381          else break;
3382        }        }
3383      else  
3384        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3385        an initial ']' is taken as a data character -- the code below handles
3386        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3387        [^] must match any character, so generate OP_ALLANY. */
3388    
3389        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3390            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3391        {        {
3392        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3393          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3394          zerofirstbyte = firstbyte;
3395          break;
3396        }        }
3397    
3398        /* If a class contains a negative special such as \S, we need to flip the
3399        negation flag at the end, so that support for characters > 255 works
3400        correctly (they are all included in the class). */
3401    
3402        should_flip_negation = FALSE;
3403    
3404      /* 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
3405      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
3406      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 3418  for (;; ptr++)
3418  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3419      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3420      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3421        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3422  #endif  #endif
3423    
3424      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2345  for (;; ptr++) Line 3434  for (;; ptr++)
3434          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3435          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3436          }          }
3437    
3438          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3439          data and reset the pointer. This is so that very large classes that
3440          contain a zillion UTF-8 characters no longer overwrite the work space
3441          (which is on the stack). */
3442    
3443          if (lengthptr != NULL)
3444            {
3445            *lengthptr += class_utf8data - class_utf8data_base;
3446            class_utf8data = class_utf8data_base;
3447            }
3448    
3449  #endif  #endif
3450    
3451        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3452    
3453        if (inescq)        if (inescq)
3454          {          {
3455          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3456            {            {
3457            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3458            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3467  for (;; ptr++)
3467        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3468        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3469    
3470        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3471            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3472            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3473          {          {
3474          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3475          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3476          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3477          uschar pbits[32];          uschar pbits[32];
3478    
3479          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3480            {            {
3481            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3482            goto FAILED;            goto FAILED;
3483            }            }
3484    
3485          ptr += 2;          ptr += 2;
3486          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3487            {            {
3488            local_negate = TRUE;            local_negate = TRUE;
3489              should_flip_negation = TRUE;  /* Note negative special */
3490            ptr++;            ptr++;
3491            }            }
3492    
3493          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3494          if (posix_class < 0)          if (posix_class < 0)
3495            {            {
3496            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2402  for (;; ptr++) Line 3504  for (;; ptr++)
3504          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3505            posix_class = 0;            posix_class = 0;
3506    
3507          /* 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
3508          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3509          subtract bits that may be in the main map already. At the end we or the  
3510          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3511            if ((options & PCRE_UCP) != 0)
3512              {
3513              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3514              if (posix_substitutes[pc] != NULL)
3515                {
3516                nestptr = tempptr + 1;
3517                ptr = posix_substitutes[pc] - 1;
3518                continue;
3519                }
3520              }
3521    #endif
3522            /* In the non-UCP case, we build the bit map for the POSIX class in a
3523            chunk of local store because we may be adding and subtracting from it,
3524            and we don't want to subtract bits that may be in the main map already.
3525            At the end we or the result into the bit map that is being built. */
3526    
3527          posix_class *= 3;          posix_class *= 3;
3528    
# Line 2449  for (;; ptr++) Line 3566  for (;; ptr++)
3566    
3567        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3568        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
3569        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
3570        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
3571        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
3572        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3573          PCRE_EXTRA is set. */
3574    
3575        if (c == '\\')        if (c == CHAR_BACKSLASH)
3576          {          {
3577          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3578          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3579    
3580          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 */  
3581          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3582            {            {
3583            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3584              {              {
3585              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3586              }              }
3587            else inescq = TRUE;            else inescq = TRUE;
3588            continue;            continue;
3589            }            }
3590            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3591    
3592          if (c < 0)          if (c < 0)
3593            {            {
3594            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3595            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3596    
3597            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3598              {              {
3599    #ifdef SUPPORT_UCP
3600                case ESC_du:     /* These are the values given for \d etc */
3601                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3602                case ESC_wu:     /* escape sequence with an appropriate \p */
3603                case ESC_WU:     /* or \P to test Unicode properties instead */
3604                case ESC_su:     /* of the default ASCII testing. */
3605                case ESC_SU:
3606                nestptr = ptr;
3607                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3608                class_charcount -= 2;                /* Undo! */
3609                continue;
3610    #endif
3611              case ESC_d:              case ESC_d:
3612              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3613              continue;              continue;
3614    
3615              case ESC_D:              case ESC_D:
3616                should_flip_negation = TRUE;
3617              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3618              continue;              continue;
3619    
# Line 2494  for (;; ptr++) Line 3622  for (;; ptr++)
3622              continue;              continue;
3623    
3624              case ESC_W:              case ESC_W:
3625                should_flip_negation = TRUE;
3626              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3627              continue;              continue;
3628    
3629                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3630                if it was previously set by something earlier in the character
3631                class. */
3632    
3633              case ESC_s:              case ESC_s:
3634              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3635              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3636                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3637              continue;              continue;
3638    
3639              case ESC_S:              case ESC_S:
3640                should_flip_negation = TRUE;
3641              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3642              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3643              continue;              continue;
3644    
3645              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3646                SETBIT(classbits, 0x09); /* VT */
3647                SETBIT(classbits, 0x20); /* SPACE */
3648                SETBIT(classbits, 0xa0); /* NSBP */
3649    #ifdef SUPPORT_UTF8
3650                if (utf8)
3651                  {
3652                  class_utf8 = TRUE;
3653                  *class_utf8data++ = XCL_SINGLE;
3654                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3655                  *class_utf8data++ = XCL_SINGLE;
3656                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3657                  *class_utf8data++ = XCL_RANGE;
3658                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3659                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3660                  *class_utf8data++ = XCL_SINGLE;
3661                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3662                  *class_utf8data++ = XCL_SINGLE;
3663                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3664                  *class_utf8data++ = XCL_SINGLE;
3665                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3666                  }
3667    #endif
3668              continue;              continue;
3669    
3670              default:    /* Not recognized; fall through */              case ESC_H:
3671              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3672              }                {
3673                  int x = 0xff;
3674                  switch (c)
3675                    {
3676                    case 0x09/8: x ^= 1 << (0x09%8); break;
3677                    case 0x20/8: x ^= 1 << (0x20%8); break;
3678                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3679                    default: break;
3680                    }
3681                  classbits[c] |= x;
3682                  }
3683    
3684            /* In the pre-compile phase, just do the recognition. */  #ifdef SUPPORT_UTF8
3685                if (utf8)
3686                  {
3687                  class_utf8 = TRUE;
3688                  *class_utf8data++ = XCL_RANGE;
3689                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3690                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3691                  *class_utf8data++ = XCL_RANGE;
3692                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3693                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3694                  *class_utf8data++ = XCL_RANGE;
3695                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3696                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3697                  *class_utf8data++ = XCL_RANGE;
3698                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3699                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3700                  *class_utf8data++ = XCL_RANGE;
3701                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3702                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3703                  *class_utf8data++ = XCL_RANGE;
3704                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3705                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3706                  *class_utf8data++ = XCL_RANGE;
3707                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3708                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3709                  }
3710    #endif
3711                continue;
3712    
3713            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_v:
3714                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              SETBIT(classbits, 0x0a); /* LF */
3715                SETBIT(classbits, 0x0b); /* VT */
3716                SETBIT(classbits, 0x0c); /* FF */
3717                SETBIT(classbits, 0x0d); /* CR */
3718                SETBIT(classbits, 0x85); /* NEL */
3719    #ifdef SUPPORT_UTF8
3720                if (utf8)
3721                  {
3722                  class_utf8 = TRUE;
3723                  *class_utf8data++ = XCL_RANGE;
3724                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3725                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3726                  }
3727    #endif
3728                continue;
3729    
3730            /* We need to deal with \P and \p in both phases. */              case ESC_V:
3731                for (c = 0; c < 32; c++)
3732                  {
3733                  int x = 0xff;
3734                  switch (c)
3735                    {
3736                    case 0x0a/8: x ^= 1 << (0x0a%8);
3737                                 x ^= 1 << (0x0b%8);
3738                                 x ^= 1 << (0x0c%8);
3739                                 x ^= 1 << (0x0d%8);
3740                                 break;
3741                    case 0x85/8: x ^= 1 << (0x85%8); break;
3742                    default: break;
3743                    }
3744                  classbits[c] |= x;
3745                  }
3746    
3747  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UTF8
3748            if (-c == ESC_p || -c == ESC_P)              if (utf8)
3749              {                {
3750              BOOL negated;                class_utf8 = TRUE;
3751              int pdata;                *class_utf8data++ = XCL_RANGE;
3752              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3753              if (ptype < 0) goto FAILED;                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3754              class_utf8 = TRUE;                *class_utf8data++ = XCL_RANGE;
3755              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3756                XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3757              *class_utf8data++ = ptype;                }
3758              *class_utf8data++ = pdata;  #endif
             class_charcount -= 2;   /* Not a < 256 character */  
3759              continue;              continue;
3760              }  
3761    #ifdef SUPPORT_UCP
3762                case ESC_p:
3763                case ESC_P:
3764                  {
3765                  BOOL negated;
3766                  int pdata;
3767                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3768                  if (ptype < 0) goto FAILED;
3769                  class_utf8 = TRUE;
3770                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3771                    XCL_PROP : XCL_NOTPROP;
3772                  *class_utf8data++ = ptype;
3773                  *class_utf8data++ = pdata;
3774                  class_charcount -= 2;   /* Not a < 256 character */
3775                  continue;
3776                  }
3777  #endif  #endif
3778            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3779            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3780            treated as literals. */              treated as literals. */
3781    
3782            if ((options & PCRE_EXTRA) != 0)              default:
3783              {              if ((options & PCRE_EXTRA) != 0)
3784              *errorcodeptr = ERR7;                {
3785              goto FAILED;                *errorcodeptr = ERR7;
3786                  goto FAILED;
3787                  }
3788                class_charcount -= 2;  /* Undo the default count from above */
3789                c = *ptr;              /* Get the final character and fall through */
3790                break;
3791              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3792            }            }
3793    
3794          /* 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 2562  for (;; ptr++) Line 3802  for (;; ptr++)
3802        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3803    
3804        CHECK_RANGE:        CHECK_RANGE:
3805        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3806          {          {
3807          inescq = FALSE;          inescq = FALSE;
3808          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3810  for (;; ptr++)
3810    
3811        oldptr = ptr;        oldptr = ptr;
3812    
3813        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3814    
3815          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3816    
3817          /* Check for range */
3818    
3819          if (!inescq && ptr[1] == CHAR_MINUS)
3820          {          {
3821          int d;          int d;
3822          ptr += 2;          ptr += 2;
3823          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3824    
3825          /* 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
3826          mode. */          mode. */
3827    
3828          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3829            {            {
3830            ptr += 2;            ptr += 2;
3831            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3832                { ptr += 2; continue; }
3833            inescq = TRUE;            inescq = TRUE;
3834            break;            break;
3835            }            }
3836    
3837          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3838            {            {
3839            ptr = oldptr;            ptr = oldptr;
3840            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3853  for (;; ptr++)
3853          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
3854          in such circumstances. */          in such circumstances. */
3855    
3856          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3857            {            {
3858            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3859            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3860    
3861            /* \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 */  
3862    
3863            if (d < 0)            if (d < 0)
3864              {              {
3865              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  
3866                {                {
3867                ptr = oldptr;                ptr = oldptr;
3868                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3881  for (;; ptr++)
3881    
3882          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3883    
3884            /* Remember \r or \n */
3885    
3886            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3887    
3888          /* 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
3889          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3890          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3907  for (;; ptr++)
3907              unsigned int origd = d;              unsigned int origd = d;
3908              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3909                {                {
3910                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3911                      ocd <= (unsigned int)d)
3912                    continue;                          /* Skip embedded ranges */
3913    
3914                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3915                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3916                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3917                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3918                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3919                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3920                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3921                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3922                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3923                  d = ocd;                  d = ocd;
3924                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 4003  for (;; ptr++)
4003          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4004            {            {
4005            unsigned int othercase;            unsigned int othercase;
4006            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4007              {              {
4008              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4009              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2776  for (;; ptr++) Line 4028  for (;; ptr++)
4028          }          }
4029        }        }
4030    
4031      /* 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.
4032        If we are at the end of an internal nested string, revert to the outer
4033        string. */
4034    
4035        while (((c = *(++ptr)) != 0 ||
4036               (nestptr != NULL &&
4037                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4038               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4039    
4040      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4041    
4042      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4043        {        {
4044        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4045        goto FAILED;        goto FAILED;
4046        }        }
4047    
4048      /* 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
4049      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
4050      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
4051      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4052      single-bytes only. This is an historical hangover. Maybe one day we can  
4053      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4054        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4055        operate on single-bytes characters only. This is an historical hangover.
4056        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4057    
4058      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
4059      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4060      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4061      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4062      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4063      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4064    
4065  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4066      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4067            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4068  #else  #else
4069      if (class_charcount == 1)      if (class_charcount == 1)
4070  #endif  #endif
4071        {        {
4072        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4073    
4074        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4075    
4076        if (negate_class)        if (negate_class)
4077          {          {
4078          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4079          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4080          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4081          *code++ = class_lastchar;          *code++ = class_lastchar;
4082          break;          break;
4083          }          }
# Line 2847  for (;; ptr++) Line 4107  for (;; ptr++)
4107      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4108    
4109      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4110      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4111      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
4112        characters > 255 are in the class, so any that were explicitly given as
4113        well can be ignored. If (when there are explicit characters > 255 that must
4114        be listed) there are no characters < 256, we can omit the bitmap in the
4115        actual compiled code. */
4116    
4117  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4118      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4119        {        {
4120        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4121        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 4141  for (;; ptr++)
4141        }        }
4142  #endif  #endif
4143    
4144      /* 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
4145      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
4146      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
4147      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4148        negating it if necessary. */
4149    
4150        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4151      if (negate_class)      if (negate_class)
4152        {        {
       *code++ = OP_NCLASS;  
4153        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4154          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4155        }        }
4156      else      else
4157        {        {
       *code++ = OP_CLASS;  
4158        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4159        }        }
4160      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 4165  for (;; ptr++)
4165      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4166      has been tested above. */      has been tested above. */
4167    
4168      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4169      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4170      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4171      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4172      goto REPEAT;      goto REPEAT;
4173    
4174      case '*':      case CHAR_ASTERISK:
4175      repeat_min = 0;      repeat_min = 0;
4176      repeat_max = -1;      repeat_max = -1;
4177      goto REPEAT;      goto REPEAT;
4178    
4179      case '+':      case CHAR_PLUS:
4180      repeat_min = 1;      repeat_min = 1;
4181      repeat_max = -1;      repeat_max = -1;
4182      goto REPEAT;      goto REPEAT;
4183    
4184      case '?':      case CHAR_QUESTION_MARK:
4185      repeat_min = 0;      repeat_min = 0;
4186      repeat_max = 1;      repeat_max = 1;
4187    
# Line 2952  for (;; ptr++) Line 4216  for (;; ptr++)
4216      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
4217      repeat type to the non-default. */      repeat type to the non-default. */
4218    
4219      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4220        {        {
4221        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4222        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4223        ptr++;        ptr++;
4224        }        }
4225      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4226        {        {
4227        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4228        ptr++;        ptr++;
4229        }        }
4230      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4231    
4232        /* If previous was a recursion call, wrap it in atomic brackets so that
4233        previous becomes the atomic group. All recursions were so wrapped in the
4234        past, but it no longer happens for non-repeated recursions. In fact, the
4235        repeated ones could be re-implemented independently so as not to need this,
4236        but for the moment we rely on the code for repeating groups. */
4237    
4238        if (*previous == OP_RECURSE)
4239          {
4240          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4241          *previous = OP_ONCE;
4242          PUT(previous, 1, 2 + 2*LINK_SIZE);
4243          previous[2 + 2*LINK_SIZE] = OP_KET;
4244          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4245          code += 2 + 2 * LINK_SIZE;
4246          length_prevgroup = 3 + 3*LINK_SIZE;
4247    
4248          /* When actually compiling, we need to check whether this was a forward
4249          reference, and if so, adjust the offset. */
4250    
4251          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4252            {
4253            int offset = GET(cd->hwm, -LINK_SIZE);
4254            if (offset == previous + 1 - cd->start_code)
4255              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4256            }
4257          }
4258    
4259        /* Now handle repetition for the different types of item. */
4260    
4261      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4262      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 2971  for (;; ptr++) Line 4264  for (;; ptr++)
4264      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4265      instead.  */      instead.  */
4266    
4267      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4268        {        {
4269          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4270    
4271        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4272        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4273        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3005  for (;; ptr++) Line 4300  for (;; ptr++)
4300    
4301        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4302            repeat_max < 0 &&            repeat_max < 0 &&
4303            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4304          {          {
4305          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4306          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3018  for (;; ptr++) Line 4312  for (;; ptr++)
4312      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4313      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4314      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4315      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4316      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4317    
4318      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4319        {        {
4320        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4321        c = previous[1];        c = previous[1];
4322        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4323            repeat_max < 0 &&            repeat_max < 0 &&
4324            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4325          {          {
4326          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4327          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3051  for (;; ptr++) Line 4345  for (;; ptr++)
4345    
4346        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4347            repeat_max < 0 &&            repeat_max < 0 &&
4348            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4349          {          {
4350          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4351          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3073  for (;; ptr++) Line 4367  for (;; ptr++)
4367    
4368        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4369    
4370          /*--------------------------------------------------------------------*/
4371          /* This code is obsolete from release 8.00; the restriction was finally
4372          removed: */
4373    
4374        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4375        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4376    
4377        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4378          /*--------------------------------------------------------------------*/
4379    
4380        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4381    
# Line 3215  for (;; ptr++) Line 4514  for (;; ptr++)
4514  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4515               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4516  #endif  #endif
4517               *previous == OP_REF)               *previous == OP_REF ||
4518                 *previous == OP_REFI)
4519        {        {
4520        if (repeat_max == 0)        if (repeat_max == 0)
4521          {          {
# Line 3223  for (;; ptr++) Line 4523  for (;; ptr++)
4523          goto END_REPEAT;          goto END_REPEAT;
4524          }          }
4525    
4526          /*--------------------------------------------------------------------*/
4527          /* This code is obsolete from release 8.00; the restriction was finally
4528          removed: */
4529    
4530        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4531        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4532    
4533        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4534          /*--------------------------------------------------------------------*/
4535    
4536        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4537          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3244  for (;; ptr++) Line 4549  for (;; ptr++)
4549        }        }
4550    
4551      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4552      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4553        KET opcodes, as this is the place where they get converted into the more
4554        special varieties. */
4555    
4556      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4557               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4558        {        {
4559        register int i;        register int i;
4560        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4561        uschar *bralink = NULL;        uschar *bralink = NULL;
4562          uschar *brazeroptr = NULL;
4563    
4564        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4565    
4566        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 3262  for (;; ptr++) Line 4569  for (;; ptr++)
4569          goto FAILED;          goto FAILED;
4570          }          }
4571    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
         }  
   
4572        /* 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
4573        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4574        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 3292  for (;; ptr++) Line 4578  for (;; ptr++)
4578    
4579        if (repeat_min == 0)        if (repeat_min == 0)
4580          {          {
4581          /* 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
4582          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4583    
4584          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4585          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4586          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4587          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4588          the start of the whole regex. Temporarily terminate the pattern while          **   }
4589          doing this. */  
4590            However, that fails when a group is referenced as a subroutine from
4591            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4592            so that it is skipped on execution. As we don't have a list of which
4593            groups are referenced, we cannot do this selectively.
4594    
4595            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4596            and do no more at this point. However, we do need to adjust any
4597            OP_RECURSE calls inside the group that refer to the group itself or any
4598            internal or forward referenced group, because the offset is from the
4599            start of the whole regex. Temporarily terminate the pattern while doing
4600            this. */
4601    
4602          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4603            {            {
4604            *code = OP_END;            *code = OP_END;
4605            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4606            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4607            code++;            code++;
4608              if (repeat_max == 0)
4609                {
4610                *previous++ = OP_SKIPZERO;
4611                goto END_REPEAT;
4612                }
4613              brazeroptr = previous;    /* Save for possessive optimizing */
4614            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4615            }            }
4616    
# Line 3338  for (;; ptr++) Line 4635  for (;; ptr++)
4635            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4636            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4637    
4638            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4639            bralink = previous;            bralink = previous;
4640            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4641            }            }
# Line 3358  for (;; ptr++) Line 4655  for (;; ptr++)
4655          if (repeat_min > 1)          if (repeat_min > 1)
4656            {            {
4657            /* 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
4658            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4659              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4660              integer type when available, otherwise double. */
4661    
4662            if (lengthptr != NULL)            if (lengthptr != NULL)
4663              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4664                int delta = (repeat_min - 1)*length_prevgroup;
4665                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4666                      (INT64_OR_DOUBLE)length_prevgroup >
4667                        (INT64_OR_DOUBLE)INT_MAX ||
4668                    OFLOW_MAX - *lengthptr < delta)
4669                  {
4670                  *errorcodeptr = ERR20;
4671                  goto FAILED;
4672                  }
4673                *lengthptr += delta;
4674                }
4675    
4676            /* This is compiling for real */            /* This is compiling for real */
4677    
# Line 3399  for (;; ptr++) Line 4709  for (;; ptr++)
4709          /* 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
4710          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
4711          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
4712          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4713            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4714            a 64-bit integer type when available, otherwise double. */
4715    
4716   &nbs