/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 150 by ph10, Tue Apr 17 08:22:40 2007 UTC revision 708 by ph10, Fri Sep 23 11:03:03 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"  /** DEAD **/
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      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  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 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    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 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    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 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
553    
554    /*************************************************
555    *            Find an error text                  *
556    *************************************************/
557    
558    /* The error texts are now all in one long string, to save on relocations. As
559    some of the text is of unknown length, we can't use a table of offsets.
560    Instead, just count through the strings. This is not a performance issue
561    because it happens only when there has been a compilation error.
562    
563    Argument:   the error number
564    Returns:    pointer to the error string
565    */
566    
567    static const char *
568    find_error_text(int n)
569    {
570    const char *s = error_texts;
571    for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576    return s;
577    }
578    
579    
580    /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
# Line 399  Arguments: Line 631  Arguments:
631    
632  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
633                   negative => a special escape sequence                   negative => a special escape sequence
634                   on error, errorptr is set                   on error, errorcodeptr is set
635  */  */
636    
637  static int  static int
# Line 417  ptr--; /* Set Line 649  ptr--; /* Set
649    
650  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
651    
652  /* 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
653  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.
654  Otherwise further processing may be required. */  Otherwise further processing may be required. */
655    
656  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
657  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
658  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
659    
660  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
661  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
662  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
663  #endif  #endif
664    
# Line 442  else Line 674  else
674      /* 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
675      error. */      error. */
676    
677      case 'l':      case CHAR_l:
678      case 'L':      case CHAR_L:
679      case 'N':      case CHAR_u:
680      case 'u':      case CHAR_U:
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by a number, either plain or braced. If positive, it      /* In a character class, \g is just a literal "g". Outside a character
685      is an absolute backreference. If negative, it is a relative backreference.      class, \g must be followed by one of a number of specific things:
     This is a Perl 5.10 feature. */  
686    
687      case 'g':      (1) A number, either plain or braced. If positive, it is an absolute
688      if (ptr[1] == '{')      backreference. If negative, it is a relative backreference. This is a Perl
689        5.10 feature.
690    
691        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
692        is part of Perl's movement towards a unified syntax for back references. As
693        this is synonymous with \k{name}, we fudge it up by pretending it really
694        was \k.
695    
696        (3) For Oniguruma compatibility we also support \g followed by a name or a
697        number either in angle brackets or in single quotes. However, these are
698        (possibly recursive) subroutine calls, _not_ backreferences. Just return
699        the -ESC_g code (cf \k). */
700    
701        case CHAR_g:
702        if (isclass) break;
703        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705          c = -ESC_g;
706          break;
707          }
708    
709        /* Handle the Perl-compatible cases */
710    
711        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
712          {
713          const uschar *p;
714          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
715            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
716          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
717            {
718            c = -ESC_k;
719            break;
720            }
721        braced = TRUE;        braced = TRUE;
722        ptr++;        ptr++;
723        }        }
724      else braced = FALSE;      else braced = FALSE;
725    
726      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
727        {        {
728        negated = TRUE;        negated = TRUE;
729        ptr++;        ptr++;
# Line 471  else Line 732  else
732    
733      c = 0;      c = 0;
734      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
735        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
736    
737      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
738          {
739          *errorcodeptr = ERR61;
740          break;
741          }
742    
743        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
744        {        {
745        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
746        return 0;        break;
747          }
748    
749        if (c == 0)
750          {
751          *errorcodeptr = ERR58;
752          break;
753        }        }
754    
755      if (negated)      if (negated)
# Line 484  else Line 757  else
757        if (c > bracount)        if (c > bracount)
758          {          {
759          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
760          return 0;          break;
761          }          }
762        c = bracount - (c - 1);        c = bracount - (c - 1);
763        }        }
# Line 504  else Line 777  else
777      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
778      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
779    
780      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:
781      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
782    
783      if (!isclass)      if (!isclass)
784        {        {
785        oldptr = ptr;        oldptr = ptr;
786        c -= '0';        c -= CHAR_0;
787        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
788          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
789          if (c < 0)    /* Integer overflow */
790            {
791            *errorcodeptr = ERR61;
792            break;
793            }
794        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
795          {          {
796          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 803  else
803      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.
804      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
805    
806      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
807        {        {
808        ptr--;        ptr--;
809        c = 0;        c = 0;
# Line 538  else Line 816  else
816      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
817      than 3 octal digits. */      than 3 octal digits. */
818    
819      case '0':      case CHAR_0:
820      c -= '0';      c -= CHAR_0;
821      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
822          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
823      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
824      break;      break;
825    
# Line 549  else Line 827  else
827      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
828      treated as a data character. */      treated as a data character. */
829    
830      case 'x':      case CHAR_x:
831      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
832        {        {
833        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
834        int count = 0;        int count = 0;
# Line 559  else Line 837  else
837        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
838          {          {
839          register int cc = *pt++;          register int cc = *pt++;
840          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
841          count++;          count++;
842    
843  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
844          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
845          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
846  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
847          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
848          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
849  #endif  #endif
850          }          }
851    
852        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
853          {          {
854          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
855          ptr = pt;          ptr = pt;
# Line 587  else Line 865  else
865      c = 0;      c = 0;
866      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
867        {        {
868        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
869        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
870  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
872        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
874        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
875        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876  #endif  #endif
877        }        }
878      break;      break;
879    
880      /* 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.
881      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
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case 'c':      case CHAR_c:
886      c = *(++ptr);      c = *(++ptr);
887      if (c == 0)      if (c == 0)
888        {        {
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        return 0;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894      if (c >= 'a' && c <= 'z') c -= 32;        {
895          *errorcodeptr = ERR68;
896          break;
897          }
898        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
904      break;      break;
905    
906      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
907      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
908      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
909      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
910      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
911    
912      default:      default:
913      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 920  else
920      }      }
921    }    }
922    
923    /* Perl supports \N{name} for character names, as well as plain \N for "not
924    newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929      *errorcodeptr = ERR37;
930    
931    /* If PCRE_UCP is set, we change the values for \d etc. */
932    
933    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
934      c -= (ESC_DU - ESC_D);
935    
936    /* Set the pointer to the final character before returning. */
937    
938  *ptrptr = ptr;  *ptrptr = ptr;
939  return c;  return c;
940  }  }
# Line 677  if (c == 0) goto ERROR_RETURN; Line 975  if (c == 0) goto ERROR_RETURN;
975  /* \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
976  negation. */  negation. */
977    
978  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
979    {    {
980    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
981      {      {
982      *negptr = TRUE;      *negptr = TRUE;
983      ptr++;      ptr++;
984      }      }
985    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
986      {      {
987      c = *(++ptr);      c = *(++ptr);
988      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
989      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
990      name[i] = c;      name[i] = c;
991      }      }
992    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
993    name[i] = 0;    name[i] = 0;
994    }    }
995    
# Line 713  top = _pcre_utt_size; Line 1011  top = _pcre_utt_size;
1011  while (bot < top)  while (bot < top)
1012    {    {
1013    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1014    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1015    if (c == 0)    if (c == 0)
1016      {      {
1017      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 737  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 797  int max = -1; Line 1062  int max = -1;
1062  /* 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
1063  an integer overflow. */  an integer overflow. */
1064    
1065  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1066  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1067    {    {
1068    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1072  if (min < 0 || min > 65535)
1072  /* 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.
1073  Also, max must not be less than min. */  Also, max must not be less than min. */
1074    
1075  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1076    {    {
1077    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1078      {      {
1079      max = 0;      max = 0;
1080      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1081      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1082        {        {
1083        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1102  return p;
1102    
1103    
1104  /*************************************************  /*************************************************
1105  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1106  *************************************************/  *************************************************/
1107    
1108  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1109    top-level call starts at the beginning of the pattern. All other calls must
1110    start at a parenthesis. It scans along a pattern's text looking for capturing
1111  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
1112  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
1113  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
1114  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  be terminated by '>' because that is checked in the first pass.  
1116    This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1131    count        current count of capturing parens so far encountered    cd           compile background data
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136      count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1139  */  */
1140    
1141  static int  static int
1142  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,
1143    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  const uschar *thisname;  uschar *ptr = *ptrptr;
1146    int start_count = *count;
1147    int hwm_count = start_count;
1148    BOOL dup_parens = FALSE;
1149    
1150    /* If the first character is a parenthesis, check on the type of group we are
1151    dealing with. The very first call may not start with a parenthesis. */
1152    
1153  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156    
1157      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159      /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161      else if (ptr[1] != CHAR_QUESTION_MARK)
1162        {
1163        *count += 1;
1164        if (name == NULL && *count == lorn) return *count;
1165        ptr++;
1166        }
1167    
1168      /* All cases now have (? at the start. Remember when we are in a group
1169      where the parenthesis numbers are duplicated. */
1170    
1171      else if (ptr[2] == CHAR_VERTICAL_LINE)
1172        {
1173        ptr += 3;
1174        dup_parens = TRUE;
1175        }
1176    
1177      /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179      else if (ptr[2] == CHAR_NUMBER_SIGN)
1180        {
1181        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182        goto FAIL_EXIT;
1183        }
1184    
1185      /* Handle a condition. If it is an assertion, just carry on so that it
1186      is processed as normal. If not, skip to the closing parenthesis of the
1187      condition (there can't be any nested parens). */
1188    
1189      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1190        {
1191        ptr += 2;
1192        if (ptr[1] != CHAR_QUESTION_MARK)
1193          {
1194          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1195          if (*ptr != 0) ptr++;
1196          }
1197        }
1198    
1199      /* Start with (? but not a condition. */
1200    
1201      else
1202        {
1203        ptr += 2;
1204        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1205    
1206        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1207    
1208        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1209            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1210          {
1211          int term;
1212          const uschar *thisname;
1213          *count += 1;
1214          if (name == NULL && *count == lorn) return *count;
1215          term = *ptr++;
1216          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1217          thisname = ptr;
1218          while (*ptr != term) ptr++;
1219          if (name != NULL && lorn == ptr - thisname &&
1220              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1221            return *count;
1222          term++;
1223          }
1224        }
1225      }
1226    
1227    /* Past any initial parenthesis handling, scan for parentheses or vertical
1228    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232    for (; ptr < cd->end_pattern; ptr++)
1233      {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
1236    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1237      {      {
1238      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1239      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1240        {        {
1241        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1242        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1243        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1244        }        }
1245      continue;      continue;
1246      }      }
1247    
1248    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1249      are handled for real. If the first character is '^', skip it. Also, if the
1250      first few characters (either before or after ^) are \Q\E or \E we skip them
1251      too. This makes for compatibility with Perl. Note the use of STR macros to
1252      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1253    
1254    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1255      {      {
1256      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1257        for (;;)
1258        {        {
1259        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1260          {          {
1261          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1262          if (*ptr == 'Q') for (;;)            ptr+= 2;
1263            else if (strncmp((const char *)ptr+2,
1264                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1265              ptr += 4;
1266            else
1267              break;
1268            }
1269          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1270            {
1271            negate_class = TRUE;
1272            ptr++;
1273            }
1274          else break;
1275          }
1276    
1277        /* If the next character is ']', it is a data character that must be
1278        skipped, except in JavaScript compatibility mode. */
1279    
1280        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1281            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1282          ptr++;
1283    
1284        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1285          {
1286          if (*ptr == 0) return -1;
1287          if (*ptr == CHAR_BACKSLASH)
1288            {
1289            if (*(++ptr) == 0) goto FAIL_EXIT;
1290            if (*ptr == CHAR_Q) for (;;)
1291            {            {
1292            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1293            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1294            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1295            }            }
1296          continue;          continue;
1297          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1301  for (; *ptr != 0; ptr++)
1301    
1302    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1303    
1304    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1307      if (*ptr == 0) return -1;      while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315        if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
1318    
1319    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1320    
1321    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1322      {      {
1323      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1325      continue;      if (*ptr == 0) goto FAIL_EXIT;
1326        }
1327    
1328      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329        {
1330        if (dup_parens && *count < hwm_count) *count = hwm_count;
1331        goto FAIL_EXIT;
1332      }      }
1333    
1334    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1335    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1336        if (*count > hwm_count) hwm_count = *count;
1337        *count = start_count;
1338        }
1339      }
1340    
1341    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1342    *ptrptr = ptr;
1343    return -1;
1344    }
1345    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1346    
   count++;  
1347    
1348    if (name == NULL && count == lorn) return count;  
1349    term = *ptr++;  /*************************************************
1350    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1351    thisname = ptr;  *************************************************/
1352    while (*ptr != term) ptr++;  
1353    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1354        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1355      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1356    returns when it reaches a given numbered subpattern. This is used for forward
1357    references to subpatterns. We used to be able to start this scan from the
1358    current compiling point, using the current count value from cd->bracount, and
1359    do it all in a single loop, but the addition of the possibility of duplicate
1360    subpattern numbers means that we have to scan from the very start, in order to
1361    take account of such duplicates, and to use a recursive function to keep track
1362    of the different types of group.
1363    
1364    Arguments:
1365      cd           compile background data
1366      name         name to seek, or NULL if seeking a numbered subpattern
1367      lorn         name length, or subpattern number if name is NULL
1368      xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371    Returns:       the number of the found subpattern, or -1 if not found
1372    */
1373    
1374    static int
1375    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376      BOOL utf8)
1377    {
1378    uschar *ptr = (uschar *)cd->start_pattern;
1379    int count = 0;
1380    int rc;
1381    
1382    /* If the pattern does not start with an opening parenthesis, the first call
1383    to find_parens_sub() will scan right to the end (if necessary). However, if it
1384    does start with a parenthesis, find_parens_sub() will return when it hits the
1385    matching closing parens. That is why we have to have a loop. */
1386    
1387    for (;;)
1388      {
1389      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390      if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
1393  return -1;  return rc;
1394  }  }
1395    
1396    
1397    
1398    
1399  /*************************************************  /*************************************************
1400  *      Find first significant op code            *  *      Find first significant op code            *
1401  *************************************************/  *************************************************/
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  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
1405  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
1406  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
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    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  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 996  for (;;) Line 1435  for (;;)
1435    
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1443      break;      break;
# Line 1012  for (;;) Line 1453  for (;;)
1453    
1454    
1455  /*************************************************  /*************************************************
1456  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1457  *************************************************/  *************************************************/
1458    
1459  /* 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,
1460  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.
1461  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
1462    temporarily terminated with OP_END when this function is called.
1463    
1464    This function is called when a backward assertion is encountered, so that if it
1465    fails, the error message can point to the correct place in the pattern.
1466    However, we cannot do this when the assertion contains subroutine calls,
1467    because they can be forward references. We solve this by remembering this case
1468    and doing the check at the end; a flag specifies which mode we are running in.
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473      atend    TRUE if called when the pattern is complete
1474      cd       the "compile data" structure
1475    
1476  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1477                 or -1 if there is no fixed length,
1478               or -2 if \C was encountered               or -2 if \C was encountered
1479                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1041  branch, check the length against that of Line 1493  branch, check the length against that of
1493  for (;;)  for (;;)
1494    {    {
1495    int d;    int d;
1496      uschar *ce, *cs;
1497    register int op = *cc;    register int op = *cc;
   
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509      case OP_COND:      case OP_COND:
1510      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1511      if (d < 0) return d;      if (d < 0) return d;
1512      branchlength += d;      branchlength += d;
1513      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1058  for (;;) Line 1516  for (;;)
1516    
1517      /* 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
1518      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
1519      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.
1520        Note that we must not include the OP_KETRxxx opcodes here, because they
1521        all imply an unlimited repeat. */
1522    
1523      case OP_ALT:      case OP_ALT:
1524      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1525      case OP_END:      case OP_END:
1526      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1527        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1072  for (;;) Line 1530  for (;;)
1530      branchlength = 0;      branchlength = 0;
1531      break;      break;
1532    
1533        /* A true recursion implies not fixed length, but a subroutine call may
1534        be OK. If the subroutine is a forward reference, we can't deal with
1535        it until the end of the pattern, so return -3. */
1536    
1537        case OP_RECURSE:
1538        if (!atend) return -3;
1539        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1540        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1541        if (cc > cs && cc < ce) return -1;                /* Recursion */
1542        d = find_fixedlength(cs + 2, utf8, atend, cd);
1543        if (d < 0) return d;
1544        branchlength += d;
1545        cc += 1 + LINK_SIZE;
1546        break;
1547    
1548      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1549    
1550      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1558  for (;;)
1558    
1559      case OP_REVERSE:      case OP_REVERSE:
1560      case OP_CREF:      case OP_CREF:
1561        case OP_NCREF:
1562      case OP_RREF:      case OP_RREF:
1563        case OP_NRREF:
1564      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
1572        case OP_CIRCM:
1573      case OP_DOLL:      case OP_DOLL:
1574        case OP_DOLLM:
1575      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1576      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1577      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1103  for (;;) Line 1580  for (;;)
1580      /* Handle literal characters */      /* Handle literal characters */
1581    
1582      case OP_CHAR:      case OP_CHAR:
1583      case OP_CHARNC:      case OP_CHARI:
1584      case OP_NOT:      case OP_NOT:
1585        case OP_NOTI:
1586      branchlength++;      branchlength++;
1587      cc += 2;      cc += 2;
1588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1589      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1590  #endif  #endif
1591      break;      break;
1592    
# Line 1122  for (;;) Line 1597  for (;;)
1597      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1598      cc += 4;      cc += 4;
1599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1600      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1601  #endif  #endif
1602      break;      break;
1603    
1604      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1605      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1606        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1607      cc += 4;      cc += 4;
1608      break;      break;
1609    
# Line 1148  for (;;) Line 1621  for (;;)
1621      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1622      case OP_WORDCHAR:      case OP_WORDCHAR:
1623      case OP_ANY:      case OP_ANY:
1624        case OP_ALLANY:
1625      branchlength++;      branchlength++;
1626      cc++;      cc++;
1627      break;      break;
# Line 1202  for (;;) Line 1676  for (;;)
1676    
1677    
1678  /*************************************************  /*************************************************
1679  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1680  *************************************************/  *************************************************/
1681    
1682  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1683  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1684    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1685    so that it can be called from pcre_study() when finding the minimum matching
1686    length.
1687    
1688  Arguments:  Arguments:
1689    code        points to start of expression    code        points to start of expression
1690    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1691    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1692    
1693  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
1694  */  */
1695    
1696  static const uschar *  const uschar *
1697  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1698  {  {
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* 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 1708  for (;;)
1708    
1709    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1710    
1711      /* Handle recursion */
1712    
1713      else if (c == OP_REVERSE)
1714        {
1715        if (number < 0) return (uschar *)code;
1716        code += _pcre_OP_lengths[c];
1717        }
1718    
1719    /* Handle capturing bracket */    /* Handle capturing bracket */
1720    
1721    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1722               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1723      {      {
1724      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1725      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1726      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1727      }      }
1728    
1729    /* 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
1730    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
1731    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1732      must add in its length. */
1733    
1734    else    else
1735      {      {
1736        switch(c)
1737          {
1738          case OP_TYPESTAR:
1739          case OP_TYPEMINSTAR:
1740          case OP_TYPEPLUS:
1741          case OP_TYPEMINPLUS:
1742          case OP_TYPEQUERY:
1743          case OP_TYPEMINQUERY:
1744          case OP_TYPEPOSSTAR:
1745          case OP_TYPEPOSPLUS:
1746          case OP_TYPEPOSQUERY:
1747          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1748          break;
1749    
1750          case OP_TYPEUPTO:
1751          case OP_TYPEMINUPTO:
1752          case OP_TYPEEXACT:
1753          case OP_TYPEPOSUPTO:
1754          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1755          break;
1756    
1757          case OP_MARK:
1758          case OP_PRUNE_ARG:
1759          case OP_SKIP_ARG:
1760          code += code[1];
1761          break;
1762    
1763          case OP_THEN_ARG:
1764          code += code[1+LINK_SIZE];
1765          break;
1766          }
1767    
1768        /* Add in the fixed length from the table */
1769    
1770      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1771    
1772      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1773      a multi-byte character. The length in the table is a minimum, so we have to
1774      arrange to skip the extra bytes. */
1775    
1776  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1777      if (utf8) switch(c)      if (utf8) switch(c)
1778        {        {
1779        case OP_CHAR:        case OP_CHAR:
1780        case OP_CHARNC:        case OP_CHARI:
1781        case OP_EXACT:        case OP_EXACT:
1782          case OP_EXACTI:
1783        case OP_UPTO:        case OP_UPTO:
1784          case OP_UPTOI:
1785        case OP_MINUPTO:        case OP_MINUPTO:
1786          case OP_MINUPTOI:
1787        case OP_POSUPTO:        case OP_POSUPTO:
1788          case OP_POSUPTOI:
1789        case OP_STAR:        case OP_STAR:
1790          case OP_STARI:
1791        case OP_MINSTAR:        case OP_MINSTAR:
1792          case OP_MINSTARI:
1793        case OP_POSSTAR:        case OP_POSSTAR:
1794          case OP_POSSTARI:
1795        case OP_PLUS:        case OP_PLUS:
1796          case OP_PLUSI:
1797        case OP_MINPLUS:        case OP_MINPLUS:
1798          case OP_MINPLUSI:
1799        case OP_POSPLUS:        case OP_POSPLUS:
1800          case OP_POSPLUSI:
1801        case OP_QUERY:        case OP_QUERY:
1802          case OP_QUERYI:
1803        case OP_MINQUERY:        case OP_MINQUERY:
1804          case OP_MINQUERYI:
1805        case OP_POSQUERY:        case OP_POSQUERY:
1806          case OP_POSQUERYI:
1807        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1808        break;        break;
1809        }        }
1810    #else
1811        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1812  #endif  #endif
1813      }      }
1814    }    }
# Line 1303  for (;;) Line 1845  for (;;)
1845    
1846    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1847    
1848    /* 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
1849    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
1850    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
1851    bytes. */    must add in its length. */
1852    
1853    else    else
1854      {      {
1855        switch(c)
1856          {
1857          case OP_TYPESTAR:
1858          case OP_TYPEMINSTAR:
1859          case OP_TYPEPLUS:
1860          case OP_TYPEMINPLUS:
1861          case OP_TYPEQUERY:
1862          case OP_TYPEMINQUERY:
1863          case OP_TYPEPOSSTAR:
1864          case OP_TYPEPOSPLUS:
1865          case OP_TYPEPOSQUERY:
1866          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1867          break;
1868    
1869          case OP_TYPEPOSUPTO:
1870          case OP_TYPEUPTO:
1871          case OP_TYPEMINUPTO:
1872          case OP_TYPEEXACT:
1873          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1874          break;
1875    
1876          case OP_MARK:
1877          case OP_PRUNE_ARG:
1878          case OP_SKIP_ARG:
1879          code += code[1];
1880          break;
1881    
1882          case OP_THEN_ARG:
1883          code += code[1+LINK_SIZE];
1884          break;
1885          }
1886    
1887        /* Add in the fixed length from the table */
1888    
1889      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1890    
1891        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1892        by a multi-byte character. The length in the table is a minimum, so we have
1893        to arrange to skip the extra bytes. */
1894    
1895  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1896      if (utf8) switch(c)      if (utf8) switch(c)
1897        {        {
1898        case OP_CHAR:        case OP_CHAR:
1899        case OP_CHARNC:        case OP_CHARI:
1900        case OP_EXACT:        case OP_EXACT:
1901          case OP_EXACTI:
1902        case OP_UPTO:        case OP_UPTO:
1903          case OP_UPTOI:
1904        case OP_MINUPTO:        case OP_MINUPTO:
1905          case OP_MINUPTOI:
1906        case OP_POSUPTO:        case OP_POSUPTO:
1907          case OP_POSUPTOI:
1908        case OP_STAR:        case OP_STAR:
1909          case OP_STARI:
1910        case OP_MINSTAR:        case OP_MINSTAR:
1911          case OP_MINSTARI:
1912        case OP_POSSTAR:        case OP_POSSTAR:
1913          case OP_POSSTARI:
1914        case OP_PLUS:        case OP_PLUS:
1915          case OP_PLUSI:
1916        case OP_MINPLUS:        case OP_MINPLUS:
1917          case OP_MINPLUSI:
1918        case OP_POSPLUS:        case OP_POSPLUS:
1919          case OP_POSPLUSI:
1920        case OP_QUERY:        case OP_QUERY:
1921          case OP_QUERYI:
1922        case OP_MINQUERY:        case OP_MINQUERY:
1923          case OP_MINQUERYI:
1924        case OP_POSQUERY:        case OP_POSQUERY:
1925          case OP_POSQUERYI:
1926        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1927        break;        break;
1928        }        }
1929    #else
1930        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1931  #endif  #endif
1932      }      }
1933    }    }
# Line 1347  for (;;) Line 1943  for (;;)
1943  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()
1944  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
1945  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1946  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
1947  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
1948    bracket whose current branch will already have been scanned.
1949    
1950  Arguments:  Arguments:
1951    code        points to start of search    code        points to start of search
1952    endcode     points to where to stop    endcode     points to where to stop
1953    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1954      cd          contains pointers to tables etc.
1955    
1956  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1957  */  */
1958    
1959  static BOOL  static BOOL
1960  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1961      compile_data *cd)
1962  {  {
1963  register int c;  register int c;
1964  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1965       code < endcode;       code < endcode;
1966       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1967    {    {
1968    const uschar *ccode;    const uschar *ccode;
1969    
1970    c = *code;    c = *code;
1971    
1972    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1973      first_significant_code() with a TRUE final argument. */
1974    
1975      if (c == OP_ASSERT)
1976      {      {
1977        do code += GET(code, 1); while (*code == OP_ALT);
1978        c = *code;
1979        continue;
1980        }
1981    
1982      /* For a recursion/subroutine call, if its end has been reached, which
1983      implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989      if (c == OP_RECURSE)
1990        {
1991        const uschar *scode;
1992      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
1993    
1994      /* Scan a closed bracket */      /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001      empty_branch = FALSE;      empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007      do      do
2008        {        {
2009        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2010            {
2011          empty_branch = TRUE;          empty_branch = TRUE;
2012        code += GET(code, 1);          break;
2013            }
2014          scode += GET(scode, 1);
2015        }        }
2016      while (*code == OP_ALT);      while (*scode == OP_ALT);
2017      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2018        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019        continue;
2020        }
2021    
2022      /* Groups with zero repeats can of course be empty; skip them. */
2023    
2024      /* Move past the KET and fudge things so that the increment in the "for"    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2025      above has no effect. */        c == OP_BRAPOSZERO)
2026        {
2027        code += _pcre_OP_lengths[c];
2028        do code += GET(code, 1); while (*code == OP_ALT);
2029        c = *code;
2030        continue;
2031        }
2032    
2033      /* A nested group that is already marked as "could be empty" can just be
2034      skipped. */
2035    
2036      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2037          c == OP_SCBRA || c == OP_SCBRAPOS)
2038        {
2039        do code += GET(code, 1); while (*code == OP_ALT);
2040        c = *code;
2041        continue;
2042        }
2043    
2044      /* For other groups, scan the branches. */
2045    
2046      if (c == OP_BRA  || c == OP_BRAPOS ||
2047          c == OP_CBRA || c == OP_CBRAPOS ||
2048          c == OP_ONCE || c == OP_COND)
2049        {
2050        BOOL empty_branch;
2051        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2052    
2053        /* If a conditional group has only one branch, there is a second, implied,
2054        empty branch, so just skip over the conditional, because it could be empty.
2055        Otherwise, scan the individual branches of the group. */
2056    
2057      c = OP_END;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2058      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];        code += GET(code, 1);
2059        else
2060          {
2061          empty_branch = FALSE;
2062          do
2063            {
2064            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2065              empty_branch = TRUE;
2066            code += GET(code, 1);
2067            }
2068          while (*code == OP_ALT);
2069          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2070          }
2071    
2072        c = *code;
2073      continue;      continue;
2074      }      }
2075    
# Line 1399  for (code = first_significant_code(code Line 2077  for (code = first_significant_code(code
2077    
2078    switch (c)    switch (c)
2079      {      {
2080      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2081        cannot be represented just by a bit map. This includes negated single
2082        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2083        actual length is stored in the compiled code, so we must update "code"
2084        here. */
2085    
2086  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2087      case OP_XCLASS:      case OP_XCLASS:
2088      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2089      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2090  #endif  #endif
2091    
# Line 1447  for (code = first_significant_code(code Line 2129  for (code = first_significant_code(code
2129      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2130      case OP_WORDCHAR:      case OP_WORDCHAR:
2131      case OP_ANY:      case OP_ANY:
2132        case OP_ALLANY:
2133      case OP_ANYBYTE:      case OP_ANYBYTE:
2134      case OP_CHAR:      case OP_CHAR:
2135      case OP_CHARNC:      case OP_CHARI:
2136      case OP_NOT:      case OP_NOT:
2137        case OP_NOTI:
2138      case OP_PLUS:      case OP_PLUS:
2139      case OP_MINPLUS:      case OP_MINPLUS:
2140      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1465  for (code = first_significant_code(code Line 2149  for (code = first_significant_code(code
2149      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2150      return FALSE;      return FALSE;
2151    
2152        /* These are going to continue, as they may be empty, but we have to
2153        fudge the length for the \p and \P cases. */
2154    
2155        case OP_TYPESTAR:
2156        case OP_TYPEMINSTAR:
2157        case OP_TYPEPOSSTAR:
2158        case OP_TYPEQUERY:
2159        case OP_TYPEMINQUERY:
2160        case OP_TYPEPOSQUERY:
2161        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2162        break;
2163    
2164        /* Same for these */
2165    
2166        case OP_TYPEUPTO:
2167        case OP_TYPEMINUPTO:
2168        case OP_TYPEPOSUPTO:
2169        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2170        break;
2171    
2172      /* End of branch */      /* End of branch */
2173    
2174      case OP_KET:      case OP_KET:
2175      case OP_KETRMAX:      case OP_KETRMAX:
2176      case OP_KETRMIN:      case OP_KETRMIN:
2177        case OP_KETRPOS:
2178      case OP_ALT:      case OP_ALT:
2179      return TRUE;      return TRUE;
2180    
# Line 1478  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183    
2184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2185      case OP_STAR:      case OP_STAR:
2186        case OP_STARI:
2187      case OP_MINSTAR:      case OP_MINSTAR:
2188        case OP_MINSTARI:
2189      case OP_POSSTAR:      case OP_POSSTAR:
2190        case OP_POSSTARI:
2191      case OP_QUERY:      case OP_QUERY:
2192        case OP_QUERYI:
2193      case OP_MINQUERY:      case OP_MINQUERY:
2194        case OP_MINQUERYI:
2195      case OP_POSQUERY:      case OP_POSQUERY:
2196        case OP_POSQUERYI:
2197        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2198        break;
2199    
2200      case OP_UPTO:      case OP_UPTO:
2201        case OP_UPTOI:
2202      case OP_MINUPTO:      case OP_MINUPTO:
2203        case OP_MINUPTOI:
2204      case OP_POSUPTO:      case OP_POSUPTO:
2205      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2206        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2207      break;      break;
2208  #endif  #endif
2209    
2210        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2211        string. */
2212    
2213        case OP_MARK:
2214        case OP_PRUNE_ARG:
2215        case OP_SKIP_ARG:
2216        code += code[1];
2217        break;
2218    
2219        case OP_THEN_ARG:
2220        code += code[1+LINK_SIZE];
2221        break;
2222    
2223        /* None of the remaining opcodes are required to match a character. */
2224    
2225        default:
2226        break;
2227      }      }
2228    }    }
2229    
# Line 1505  return TRUE; Line 2240  return TRUE;
2240  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
2241  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,
2242  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.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
2248    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2249    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2250    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2251      cd          pointers to tables etc
2252    
2253  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2254  */  */
2255    
2256  static BOOL  static BOOL
2257  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2258    BOOL utf8)    BOOL utf8, compile_data *cd)
2259  {  {
2260  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2261    {    {
2262    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2263        return FALSE;
2264    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2265    }    }
2266  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2273  return TRUE;
2273  *************************************************/  *************************************************/
2274    
2275  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2276  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
2277  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2278  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2279    
2280    Originally, this function only recognized a sequence of letters between the
2281    terminators, but it seems that Perl recognizes any sequence of characters,
2282    though of course unknown POSIX names are subsequently rejected. Perl gives an
2283    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2284    didn't consider this to be a POSIX class. Likewise for [:1234:].
2285    
2286    The problem in trying to be exactly like Perl is in the handling of escapes. We
2287    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2288    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2289    below handles the special case of \], but does not try to do any other escape
2290    processing. This makes it different from Perl for cases such as [:l\ower:]
2291    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2292    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2293    I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit.
2299    
2300    In Perl, unescaped square brackets may also appear as part of class names. For
2301    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2302    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2303    seem right at all. PCRE does not allow closing square brackets in POSIX class
2304    names.
2305    
2306  Argument:  Arguments:
2307    ptr      pointer to the initial [    ptr      pointer to the initial [
2308    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2309    
2310  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2311  */  */
2312    
2313  static BOOL  static BOOL
2314  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2315  {  {
2316  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2317  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2318  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2319    {    {
2320    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2321    return TRUE;      ptr++;
2322      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2323      else
2324        {
2325        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2326          {
2327          *endptr = ptr;
2328          return TRUE;
2329          }
2330        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2331             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2332              ptr[1] == CHAR_EQUALS_SIGN) &&
2333            check_posix_syntax(ptr, endptr))
2334          return FALSE;
2335        }
2336    }    }
2337  return FALSE;  return FALSE;
2338  }  }
# Line 1581  Returns: a value representing the na Line 2357  Returns: a value representing the na
2357  static int  static int
2358  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2359  {  {
2360    const char *pn = posix_names;
2361  register int yield = 0;  register int yield = 0;
2362  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2363    {    {
2364    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2365      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2366      pn += posix_name_lengths[yield] + 1;
2367    yield++;    yield++;
2368    }    }
2369  return -1;  return -1;
# Line 1600  return -1; Line 2378  return -1;
2378  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2379  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2380  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
2381  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
2382  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
2383  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
2384  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
2385  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2386    OP_END.
2387    
2388  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2389  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 2406  adjust_recurse(uschar *group, int adjust
2406    uschar *save_hwm)    uschar *save_hwm)
2407  {  {
2408  uschar *ptr = group;  uschar *ptr = group;
2409    
2410  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2411    {    {
2412    int offset;    int offset;
# Line 1680  auto_callout(uschar *code, const uschar Line 2460  auto_callout(uschar *code, const uschar
2460  {  {
2461  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2462  *code++ = 255;  *code++ = 255;
2463  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2464  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2465  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2466  }  }
2467    
# Line 1706  Returns: nothing Line 2486  Returns: nothing
2486  static void  static void
2487  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2488  {  {
2489  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2490  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2491  }  }
2492    
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2518  get_othercase_range(unsigned int *cptr,
2518  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2519    
2520  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2521    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2522    
2523  if (c > d) return FALSE;  if (c > d) return FALSE;
2524    
# Line 1747  next = othercase + 1; Line 2527  next = othercase + 1;
2527    
2528  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2529    {    {
2530    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2531    next++;    next++;
2532    }    }
2533    
# Line 1756  for (++c; c <= d; c++) Line 2536  for (++c; c <= d; c++)
2536    
2537  return TRUE;  return TRUE;
2538  }  }
2539    
2540    
2541    
2542    /*************************************************
2543    *        Check a character and a property        *
2544    *************************************************/
2545    
2546    /* This function is called by check_auto_possessive() when a property item
2547    is adjacent to a fixed character.
2548    
2549    Arguments:
2550      c            the character
2551      ptype        the property type
2552      pdata        the data for the type
2553      negated      TRUE if it's a negated property (\P or \p{^)
2554    
2555    Returns:       TRUE if auto-possessifying is OK
2556    */
2557    
2558    static BOOL
2559    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2560    {
2561    const ucd_record *prop = GET_UCD(c);
2562    switch(ptype)
2563      {
2564      case PT_LAMP:
2565      return (prop->chartype == ucp_Lu ||
2566              prop->chartype == ucp_Ll ||
2567              prop->chartype == ucp_Lt) == negated;
2568    
2569      case PT_GC:
2570      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2571    
2572      case PT_PC:
2573      return (pdata == prop->chartype) == negated;
2574    
2575      case PT_SC:
2576      return (pdata == prop->script) == negated;
2577    
2578      /* These are specials */
2579    
2580      case PT_ALNUM:
2581      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2582              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2583    
2584      case PT_SPACE:    /* Perl space */
2585      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2586              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2587              == negated;
2588    
2589      case PT_PXSPACE:  /* POSIX space */
2590      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2591              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2592              c == CHAR_FF || c == CHAR_CR)
2593              == negated;
2594    
2595      case PT_WORD:
2596      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2597              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2598              c == CHAR_UNDERSCORE) == negated;
2599      }
2600    return FALSE;
2601    }
2602  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2603    
2604    
# Line 1769  whether the next thing could possibly ma Line 2612  whether the next thing could possibly ma
2612  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2613    
2614  Arguments:  Arguments:
2615    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2616    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2617    ptr           next character in pattern    ptr           next character in pattern
2618    options       options bits    options       options bits
2619    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns: TRUE if possessifying is Line 2622  Returns: TRUE if possessifying is
2622  */  */
2623    
2624  static BOOL  static BOOL
2625  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2626    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2627  {  {
2628  int next;  int c, next;
2629    int op_code = *previous++;
2630    
2631  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2632    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2635  if ((options & PCRE_EXTENDED) != 0)
2635    for (;;)    for (;;)
2636      {      {
2637      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2638      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2639        {        {
2640        while (*(++ptr) != 0)        ptr++;
2641          while (*ptr != 0)
2642            {
2643          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2644            ptr++;
2645    #ifdef SUPPORT_UTF8
2646            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2647    #endif
2648            }
2649        }        }
2650      else break;      else break;
2651      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2654  if ((options & PCRE_EXTENDED) != 0)
2654  /* 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
2655  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2656    
2657  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2658    {    {
2659    int temperrorcode = 0;    int temperrorcode = 0;
2660    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 2679  if ((options & PCRE_EXTENDED) != 0)
2679    for (;;)    for (;;)
2680      {      {
2681      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2682      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2683        {        {
2684        while (*(++ptr) != 0)        ptr++;
2685          while (*ptr != 0)
2686            {
2687          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2688            ptr++;
2689    #ifdef SUPPORT_UTF8
2690            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2691    #endif
2692            }
2693        }        }
2694      else break;      else break;
2695      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2697  if ((options & PCRE_EXTENDED) != 0)
2697    
2698  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2699    
2700  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2701    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2702        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2703    
2704  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2705    the next item is a character. */
2706    
2707  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2708    {    {
2709    case OP_CHAR:    case OP_CHAR:
2710  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2711    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2712    #else
2713      c = *previous;
2714  #endif  #endif
2715    return item != next;    return c != next;
2716    
2717    /* 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
2718    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
2719    high-valued characters. */    high-valued characters. */
2720    
2721    case OP_CHARNC:    case OP_CHARI:
2722  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2723    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2724    #else
2725      c = *previous;
2726  #endif  #endif
2727    if (item == next) return FALSE;    if (c == next) return FALSE;
2728  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2729    if (utf8)    if (utf8)
2730      {      {
2731      unsigned int othercase;      unsigned int othercase;
2732      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2733  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2734      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2735  #else  #else
2736      othercase = NOTACHAR;      othercase = NOTACHAR;
2737  #endif  #endif
2738      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2739      }      }
2740    else    else
2741  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2742    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2743    
2744    /* 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
2745      opcodes are not used for multi-byte characters, because they are coded using
2746      an XCLASS instead. */
2747    
2748    case OP_NOT:    case OP_NOT:
2749    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2750    if (item == next) return TRUE;  
2751    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2752      if ((c = *previous) == next) return TRUE;
2753  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2754    if (utf8)    if (utf8)
2755      {      {
2756      unsigned int othercase;      unsigned int othercase;
2757      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2758  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2759      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2760  #else  #else
2761      othercase = NOTACHAR;      othercase = NOTACHAR;
2762  #endif  #endif
2763      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2764      }      }
2765    else    else
2766  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2767    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2768    
2769      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2770      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2771    
2772    case OP_DIGIT:    case OP_DIGIT:
2773    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1919  if (next >= 0) switch(op_code) Line 2781  if (next >= 0) switch(op_code)
2781    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2782    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2783    
2784    case OP_WORDCHAR:    case OP_WORDCHAR:
2785    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2786    
2787      case OP_NOT_WORDCHAR:
2788      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2789    
2790      case OP_HSPACE:
2791      case OP_NOT_HSPACE:
2792      switch(next)
2793        {
2794        case 0x09:
2795        case 0x20:
2796        case 0xa0:
2797        case 0x1680:
2798        case 0x180e:
2799        case 0x2000:
2800        case 0x2001:
2801        case 0x2002:
2802        case 0x2003:
2803        case 0x2004:
2804        case 0x2005:
2805        case 0x2006:
2806        case 0x2007:
2807        case 0x2008:
2808        case 0x2009:
2809        case 0x200A:
2810        case 0x202f:
2811        case 0x205f:
2812        case 0x3000:
2813        return op_code == OP_NOT_HSPACE;
2814        default:
2815        return op_code != OP_NOT_HSPACE;
2816        }
2817    
2818      case OP_ANYNL:
2819      case OP_VSPACE:
2820      case OP_NOT_VSPACE:
2821      switch(next)
2822        {
2823        case 0x0a:
2824        case 0x0b:
2825        case 0x0c:
2826        case 0x0d:
2827        case 0x85:
2828        case 0x2028:
2829        case 0x2029:
2830        return op_code == OP_NOT_VSPACE;
2831        default:
2832        return op_code != OP_NOT_VSPACE;
2833        }
2834    
2835    #ifdef SUPPORT_UCP
2836      case OP_PROP:
2837      return check_char_prop(next, previous[0], previous[1], FALSE);
2838    
2839    case OP_NOT_WORDCHAR:    case OP_NOTPROP:
2840    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return check_char_prop(next, previous[0], previous[1], TRUE);
2841    #endif
2842    
2843    default:    default:
2844    return FALSE;    return FALSE;
2845    }    }
2846    
2847    
2848  /* 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
2849    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2850    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2851    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2852    replaced by OP_PROP codes when PCRE_UCP is set. */
2853    
2854  switch(op_code)  switch(op_code)
2855    {    {
2856    case OP_CHAR:    case OP_CHAR:
2857    case OP_CHARNC:    case OP_CHARI:
2858  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2859    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2860    #else
2861      c = *previous;
2862  #endif  #endif
2863    switch(-next)    switch(-next)
2864      {      {
2865      case ESC_d:      case ESC_d:
2866      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2867    
2868      case ESC_D:      case ESC_D:
2869      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2870    
2871      case ESC_s:      case ESC_s:
2872      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2873    
2874      case ESC_S:      case ESC_S:
2875      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2876    
2877      case ESC_w:      case ESC_w:
2878      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2879    
2880      case ESC_W:      case ESC_W:
2881      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2882    
2883        case ESC_h:
2884        case ESC_H:
2885        switch(c)
2886          {
2887          case 0x09:
2888          case 0x20:
2889          case 0xa0:
2890          case 0x1680:
2891          case 0x180e:
2892          case 0x2000:
2893          case 0x2001:
2894          case 0x2002:
2895          case 0x2003:
2896          case 0x2004:
2897          case 0x2005:
2898          case 0x2006:
2899          case 0x2007:
2900          case 0x2008:
2901          case 0x2009:
2902          case 0x200A:
2903          case 0x202f:
2904          case 0x205f:
2905          case 0x3000:
2906          return -next != ESC_h;
2907          default:
2908          return -next == ESC_h;
2909          }
2910    
2911        case ESC_v:
2912        case ESC_V:
2913        switch(c)
2914          {
2915          case 0x0a:
2916          case 0x0b:
2917          case 0x0c:
2918          case 0x0d:
2919          case 0x85:
2920          case 0x2028:
2921          case 0x2029:
2922          return -next != ESC_v;
2923          default:
2924          return -next == ESC_v;
2925          }
2926    
2927        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2928        their substitutions and process them. The result will always be either
2929        -ESC_p or -ESC_P. Then fall through to process those values. */
2930    
2931    #ifdef SUPPORT_UCP
2932        case ESC_du:
2933        case ESC_DU:
2934        case ESC_wu:
2935        case ESC_WU:
2936        case ESC_su:
2937        case ESC_SU:
2938          {
2939          int temperrorcode = 0;
2940          ptr = substitutes[-next - ESC_DU];
2941          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2942          if (temperrorcode != 0) return FALSE;
2943          ptr++;    /* For compatibility */
2944          }
2945        /* Fall through */
2946    
2947        case ESC_p:
2948        case ESC_P:
2949          {
2950          int ptype, pdata, errorcodeptr;
2951          BOOL negated;
2952    
2953          ptr--;      /* Make ptr point at the p or P */
2954          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2955          if (ptype < 0) return FALSE;
2956          ptr++;      /* Point past the final curly ket */
2957    
2958          /* If the property item is optional, we have to give up. (When generated
2959          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2960          to the original \d etc. At this point, ptr will point to a zero byte. */
2961    
2962          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2963            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2964              return FALSE;
2965    
2966          /* Do the property check. */
2967    
2968          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2969          }
2970    #endif
2971    
2972      default:      default:
2973      return FALSE;      return FALSE;
2974      }      }
2975    
2976      /* In principle, support for Unicode properties should be integrated here as
2977      well. It means re-organizing the above code so as to get hold of the property
2978      values before switching on the op-code. However, I wonder how many patterns
2979      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2980      these op-codes are never generated.) */
2981    
2982    case OP_DIGIT:    case OP_DIGIT:
2983    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2984             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2985    
2986    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2987    return next == -ESC_d;    return next == -ESC_d;
2988    
2989    case OP_WHITESPACE:    case OP_WHITESPACE:
2990    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2991    
2992    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2993    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2994    
2995      case OP_HSPACE:
2996      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2997             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2998    
2999      case OP_NOT_HSPACE:
3000      return next == -ESC_h;
3001    
3002      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3003      case OP_ANYNL:
3004      case OP_VSPACE:
3005      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3006    
3007      case OP_NOT_VSPACE:
3008      return next == -ESC_v || next == -ESC_R;
3009    
3010    case OP_WORDCHAR:    case OP_WORDCHAR:
3011    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3012             next == -ESC_v || next == -ESC_R;
3013    
3014    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3015    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2008  Arguments: Line 3041  Arguments:
3041    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3042    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3043    bcptr          points to current branch chain    bcptr          points to current branch chain
3044      cond_depth     conditional nesting depth
3045    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3046    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3047                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2019  Returns: TRUE on success Line 3053  Returns: TRUE on success
3053  static BOOL  static BOOL
3054  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3055    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3056    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3057  {  {
3058  int repeat_type, op_type;  int repeat_type, op_type;
3059  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2028  int greedy_default, greedy_non_default; Line 3062  int greedy_default, greedy_non_default;
3062  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3063  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3064  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3065  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3066  int after_manual_callout = 0;  int after_manual_callout = 0;
3067  int length_prevgroup = 0;  int length_prevgroup = 0;
3068  register int c;  register int c;
# Line 2040  BOOL inescq = FALSE; Line 3074  BOOL inescq = FALSE;
3074  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3075  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3076  const uschar *tempptr;  const uschar *tempptr;
3077    const uschar *nestptr = NULL;
3078  uschar *previous = NULL;  uschar *previous = NULL;
3079  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3080  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3081  uschar classbits[32];  uschar classbits[32];
3082    
3083    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3084    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3085    dynamically as we process the pattern. */
3086    
3087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3088  BOOL class_utf8;  BOOL class_utf8;
3089  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3090  uschar *class_utf8data;  uschar *class_utf8data;
3091    uschar *class_utf8data_base;
3092  uschar utf8_char[6];  uschar utf8_char[6];
3093  #else  #else
3094  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3095  #endif  #endif
3096    
3097  #ifdef DEBUG  #ifdef PCRE_DEBUG
3098  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3099  #endif  #endif
3100    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 3127  req_caseopt = ((options & PCRE_CASELESS)
3127  for (;; ptr++)  for (;; ptr++)
3128    {    {
3129    BOOL negate_class;    BOOL negate_class;
3130      BOOL should_flip_negation;
3131    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3132    BOOL is_quantifier;    BOOL is_quantifier;
3133    BOOL is_recurse;    BOOL is_recurse;
3134      BOOL reset_bracount;
3135    int class_charcount;    int class_charcount;
3136    int class_lastchar;    int class_lastchar;
3137    int newoptions;    int newoptions;
3138    int recno;    int recno;
3139      int refsign;
3140    int skipbytes;    int skipbytes;
3141    int subreqbyte;    int subreqbyte;
3142    int subfirstbyte;    int subfirstbyte;
# Line 2106  for (;; ptr++) Line 3148  for (;; ptr++)
3148    
3149    c = *ptr;    c = *ptr;
3150    
3151      /* If we are at the end of a nested substitution, revert to the outer level
3152      string. Nesting only happens one level deep. */
3153    
3154      if (c == 0 && nestptr != NULL)
3155        {
3156        ptr = nestptr;
3157        nestptr = NULL;
3158        c = *ptr;
3159        }
3160    
3161    /* 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
3162    previous cycle of this loop. */    previous cycle of this loop. */
3163    
3164    if (lengthptr != NULL)    if (lengthptr != NULL)
3165      {      {
3166  #ifdef DEBUG  #ifdef PCRE_DEBUG
3167      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3168  #endif  #endif
3169      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3170        {        {
3171        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3172        goto FAILED;        goto FAILED;
# Line 2127  for (;; ptr++) Line 3179  for (;; ptr++)
3179      */      */
3180    
3181      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3182      *lengthptr += code - last_code;  
3183        /* Paranoid check for integer overflow */
3184    
3185        if (OFLOW_MAX - *lengthptr < code - last_code)
3186          {
3187          *errorcodeptr = ERR20;
3188          goto FAILED;
3189          }
3190    
3191        *lengthptr += (int)(code - last_code);
3192      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));
3193    
3194      /* 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 3215  for (;; ptr++)
3215    /* 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
3216    reference list. */    reference list. */
3217    
3218    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3219      {      {
3220      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3221      goto FAILED;      goto FAILED;
# Line 2164  for (;; ptr++) Line 3225  for (;; ptr++)
3225    
3226    if (inescq && c != 0)    if (inescq && c != 0)
3227      {      {
3228      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3229        {        {
3230        inescq = FALSE;        inescq = FALSE;
3231        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 3251  for (;; ptr++)
3251    /* 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
3252    a quantifier. */    a quantifier. */
3253    
3254    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3255      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3256        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3257    
3258    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3259         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2201  for (;; ptr++) Line 3263  for (;; ptr++)
3263      previous_callout = NULL;      previous_callout = NULL;
3264      }      }
3265    
3266    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3267    
3268    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3269      {      {
3270      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3271      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3272        {        {
3273        while (*(++ptr) != 0)        ptr++;
3274          while (*ptr != 0)
3275          {          {
3276          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3277            ptr++;
3278    #ifdef SUPPORT_UTF8
3279            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3280    #endif
3281          }          }
3282        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3283    
# Line 2231  for (;; ptr++) Line 3298  for (;; ptr++)
3298      {      {
3299      /* ===================================================================*/      /* ===================================================================*/
3300      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3301      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3302      case ')':      case CHAR_RIGHT_PARENTHESIS:
3303      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3304      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3305      *codeptr = code;      *codeptr = code;
3306      *ptrptr = ptr;      *ptrptr = ptr;
3307      if (lengthptr != NULL)      if (lengthptr != NULL)
3308        {        {
3309        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3310            {
3311            *errorcodeptr = ERR20;
3312            goto FAILED;
3313            }
3314          *lengthptr += (int)(code - last_code);   /* To include callout length */
3315        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3316        }        }
3317      return TRUE;      return TRUE;
# Line 2249  for (;; ptr++) Line 3321  for (;; ptr++)
3321      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3322      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3323    
3324      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3325        previous = NULL;
3326      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3327        {        {
3328        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3329          *code++ = OP_CIRCM;
3330        }        }
3331      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3332      break;      break;
3333    
3334      case '$':      case CHAR_DOLLAR_SIGN:
3335      previous = NULL;      previous = NULL;
3336      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3337      break;      break;
3338    
3339      /* 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
3340      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3341    
3342      case '.':      case CHAR_DOT:
3343      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3344      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3345      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3346      previous = code;      previous = code;
3347      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3348      break;      break;
3349    
3350    
# Line 2286  for (;; ptr++) Line 3359  for (;; ptr++)
3359      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,
3360      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3361      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.
     */  
3362    
3363      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3364        default (Perl) mode, it is treated as a data character. */
3365    
3366        case CHAR_RIGHT_SQUARE_BRACKET:
3367        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3368          {
3369          *errorcodeptr = ERR64;
3370          goto FAILED;
3371          }
3372        goto NORMAL_CHAR;
3373    
3374        case CHAR_LEFT_SQUARE_BRACKET:
3375      previous = code;      previous = code;
3376    
3377      /* 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
3378      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. */
3379    
3380      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3381          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3382            check_posix_syntax(ptr, &tempptr))
3383        {        {
3384        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3385        goto FAILED;        goto FAILED;
3386        }        }
3387    
3388      /* 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,
3389        if the first few characters (either before or after ^) are \Q\E or \E we
3390        skip them too. This makes for compatibility with Perl. */
3391    
3392      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3393        for (;;)
3394        {        {
       negate_class = TRUE;  
3395        c = *(++ptr);        c = *(++ptr);
3396          if (c == CHAR_BACKSLASH)
3397            {
3398            if (ptr[1] == CHAR_E)
3399              ptr++;
3400            else if (strncmp((const char *)ptr+1,
3401                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3402              ptr += 3;
3403            else
3404              break;
3405            }
3406          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3407            negate_class = TRUE;
3408          else break;
3409        }        }
3410      else  
3411        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3412        an initial ']' is taken as a data character -- the code below handles
3413        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3414        [^] must match any character, so generate OP_ALLANY. */
3415    
3416        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3417            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3418        {        {
3419        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3420          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3421          zerofirstbyte = firstbyte;
3422          break;
3423        }        }
3424    
3425        /* If a class contains a negative special such as \S, we need to flip the
3426        negation flag at the end, so that support for characters > 255 works
3427        correctly (they are all included in the class). */
3428    
3429        should_flip_negation = FALSE;
3430    
3431      /* 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
3432      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
3433      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 3445  for (;; ptr++)
3445  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3446      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3447      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3448        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3449  #endif  #endif
3450    
3451      /* 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 3461  for (;; ptr++)
3461          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3462          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3463          }          }
3464    
3465          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3466          data and reset the pointer. This is so that very large classes that
3467          contain a zillion UTF-8 characters no longer overwrite the work space
3468          (which is on the stack). */
3469    
3470          if (lengthptr != NULL)
3471            {
3472            *lengthptr += class_utf8data - class_utf8data_base;
3473            class_utf8data = class_utf8data_base;
3474            }
3475    
3476  #endif  #endif
3477    
3478        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3479    
3480        if (inescq)        if (inescq)
3481          {          {
3482          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3483            {            {
3484            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3485            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3494  for (;; ptr++)
3494        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3495        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3496    
3497        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3498            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3499            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3500          {          {
3501          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3502          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3503          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3504          uschar pbits[32];          uschar pbits[32];
3505    
3506          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3507            {            {
3508            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3509            goto FAILED;            goto FAILED;
3510            }            }
3511    
3512          ptr += 2;          ptr += 2;
3513          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3514            {            {
3515            local_negate = TRUE;            local_negate = TRUE;
3516              should_flip_negation = TRUE;  /* Note negative special */
3517            ptr++;            ptr++;
3518            }            }
3519    
3520          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3521          if (posix_class < 0)          if (posix_class < 0)
3522            {            {
3523            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2402  for (;; ptr++) Line 3531  for (;; ptr++)
3531          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3532            posix_class = 0;            posix_class = 0;
3533    
3534          /* 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
3535          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3536          subtract bits that may be in the main map already. At the end we or the  
3537          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3538            if ((options & PCRE_UCP) != 0)
3539              {
3540              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3541              if (posix_substitutes[pc] != NULL)
3542                {
3543                nestptr = tempptr + 1;
3544                ptr = posix_substitutes[pc] - 1;
3545                continue;
3546                }
3547              }
3548    #endif
3549            /* In the non-UCP case, we build the bit map for the POSIX class in a
3550            chunk of local store because we may be adding and subtracting from it,
3551            and we don't want to subtract bits that may be in the main map already.
3552            At the end we or the result into the bit map that is being built. */
3553    
3554          posix_class *= 3;          posix_class *= 3;
3555    
# Line 2449  for (;; ptr++) Line 3593  for (;; ptr++)
3593    
3594        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3595        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
3596        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
3597        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
3598        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
3599        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3600          PCRE_EXTRA is set. */
3601    
3602        if (c == '\\')        if (c == CHAR_BACKSLASH)
3603          {          {
3604          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3605          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3606    
3607          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 */  
3608          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3609            {            {
3610            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3611              {              {
3612              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3613              }              }
3614            else inescq = TRUE;            else inescq = TRUE;
3615            continue;            continue;
3616            }            }
3617            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3618    
3619          if (c < 0)          if (c < 0)
3620            {            {
3621            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3622            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3623    
3624            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3625              {              {
3626    #ifdef SUPPORT_UCP
3627                case ESC_du:     /* These are the values given for \d etc */
3628                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3629                case ESC_wu:     /* escape sequence with an appropriate \p */
3630                case ESC_WU:     /* or \P to test Unicode properties instead */
3631                case ESC_su:     /* of the default ASCII testing. */
3632                case ESC_SU:
3633                nestptr = ptr;
3634                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3635                class_charcount -= 2;                /* Undo! */
3636                continue;
3637    #endif
3638              case ESC_d:              case ESC_d:
3639              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3640              continue;              continue;
3641    
3642              case ESC_D:              case ESC_D:
3643                should_flip_negation = TRUE;
3644              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3645              continue;              continue;
3646    
# Line 2494  for (;; ptr++) Line 3649  for (;; ptr++)
3649              continue;              continue;
3650    
3651              case ESC_W:              case ESC_W:
3652                should_flip_negation = TRUE;
3653              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3654              continue;              continue;
3655    
3656                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3657                if it was previously set by something earlier in the character
3658                class. */
3659    
3660              case ESC_s:              case ESC_s:
3661              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3662              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3663                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3664              continue;              continue;
3665    
3666              case ESC_S:              case ESC_S:
3667                should_flip_negation = TRUE;
3668              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3669              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3670              continue;              continue;
3671    
3672              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3673                SETBIT(classbits, 0x09); /* VT */
3674                SETBIT(classbits, 0x20); /* SPACE */
3675                SETBIT(classbits, 0xa0); /* NSBP */
3676    #ifdef SUPPORT_UTF8
3677                if (utf8)
3678                  {
3679                  class_utf8 = TRUE;
3680                  *class_utf8data++ = XCL_SINGLE;
3681                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3682                  *class_utf8data++ = XCL_SINGLE;
3683                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3684                  *class_utf8data++ = XCL_RANGE;
3685                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3686                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3687                  *class_utf8data++ = XCL_SINGLE;
3688                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3689                  *class_utf8data++ = XCL_SINGLE;
3690                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3691                  *class_utf8data++ = XCL_SINGLE;
3692                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3693                  }
3694    #endif
3695              continue;              continue;
3696    
3697              default:    /* Not recognized; fall through */              case ESC_H:
3698              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3699              }                {
3700                  int x = 0xff;
3701                  switch (c)
3702                    {
3703                    case 0x09/8: x ^= 1 << (0x09%8); break;
3704                    case 0x20/8: x ^= 1 << (0x20%8); break;
3705                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3706                    default: break;
3707                    }
3708                  classbits[c] |= x;
3709                  }
3710    
3711    #ifdef SUPPORT_UTF8
3712                if (utf8)
3713                  {
3714                  class_utf8 = TRUE;
3715                  *class_utf8data++ = XCL_RANGE;
3716                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3717                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3718                  *class_utf8data++ = XCL_RANGE;
3719                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3720                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3721                  *class_utf8data++ = XCL_RANGE;
3722                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3723                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3724                  *class_utf8data++ = XCL_RANGE;
3725                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3726                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3727                  *class_utf8data++ = XCL_RANGE;
3728                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3729                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3730                  *class_utf8data++ = XCL_RANGE;
3731                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3732                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3733                  *class_utf8data++ = XCL_RANGE;
3734                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3735                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3736                  }
3737    #endif
3738                continue;
3739    
3740            /* In the pre-compile phase, just do the recognition. */              case ESC_v:
3741                SETBIT(classbits, 0x0a); /* LF */
3742                SETBIT(classbits, 0x0b); /* VT */
3743                SETBIT(classbits, 0x0c); /* FF */
3744                SETBIT(classbits, 0x0d); /* CR */
3745                SETBIT(classbits, 0x85); /* NEL */
3746    #ifdef SUPPORT_UTF8
3747                if (utf8)
3748                  {
3749                  class_utf8 = TRUE;
3750                  *class_utf8data++ = XCL_RANGE;
3751                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3752                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3753                  }
3754    #endif
3755                continue;
3756    
3757            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_V:
3758                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              for (c = 0; c < 32; c++)
3759                  {
3760                  int x = 0xff;
3761                  switch (c)
3762                    {
3763                    case 0x0a/8: x ^= 1 << (0x0a%8);
3764                                 x ^= 1 << (0x0b%8);
3765                                 x ^= 1 << (0x0c%8);
3766                                 x ^= 1 << (0x0d%8);
3767                                 break;
3768                    case 0x85/8: x ^= 1 << (0x85%8); break;
3769                    default: break;
3770                    }
3771                  classbits[c] |= x;
3772                  }
3773    
3774            /* We need to deal with \P and \p in both phases. */  #ifdef SUPPORT_UTF8
3775                if (utf8)
3776                  {
3777                  class_utf8 = TRUE;
3778                  *class_utf8data++ = XCL_RANGE;
3779                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3780                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3781                  *class_utf8data++ = XCL_RANGE;
3782                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3783                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3784                  }
3785    #endif
3786                continue;
3787    
3788  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3789            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3790              {              case ESC_P:
3791              BOOL negated;                {
3792              int pdata;                BOOL negated;
3793              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3794              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3795              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3796              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3797                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3798              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3799              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3800              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3801              continue;                class_charcount -= 2;   /* Not a < 256 character */
3802              }                continue;
3803                  }
3804  #endif  #endif
3805            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3806            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3807            treated as literals. */              treated as literals. */
3808    
3809            if ((options & PCRE_EXTRA) != 0)              default:
3810              {              if ((options & PCRE_EXTRA) != 0)
3811              *errorcodeptr = ERR7;                {
3812              goto FAILED;                *errorcodeptr = ERR7;
3813                  goto FAILED;
3814                  }
3815                class_charcount -= 2;  /* Undo the default count from above */
3816                c = *ptr;              /* Get the final character and fall through */
3817                break;
3818              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3819            }            }
3820    
3821          /* 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 3829  for (;; ptr++)
3829        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3830    
3831        CHECK_RANGE:        CHECK_RANGE:
3832        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3833          {          {
3834          inescq = FALSE;          inescq = FALSE;
3835          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3837  for (;; ptr++)
3837    
3838        oldptr = ptr;        oldptr = ptr;
3839    
3840        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3841    
3842          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3843    
3844          /* Check for range */
3845    
3846          if (!inescq && ptr[1] == CHAR_MINUS)
3847          {          {
3848          int d;          int d;
3849          ptr += 2;          ptr += 2;
3850          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3851    
3852          /* 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
3853          mode. */          mode. */
3854    
3855          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3856            {            {
3857            ptr += 2;            ptr += 2;
3858            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3859                { ptr += 2; continue; }
3860            inescq = TRUE;            inescq = TRUE;
3861            break;            break;
3862            }            }
3863    
3864          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3865            {            {
3866            ptr = oldptr;            ptr = oldptr;
3867            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3880  for (;; ptr++)
3880          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
3881          in such circumstances. */          in such circumstances. */
3882    
3883          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3884            {            {
3885            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3886            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3887    
3888            /* \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 */  
3889    
3890            if (d < 0)            if (d < 0)
3891              {              {
3892              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  
3893                {                {
3894                ptr = oldptr;                ptr = oldptr;
3895                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3908  for (;; ptr++)
3908    
3909          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3910    
3911            /* Remember \r or \n */
3912    
3913            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3914    
3915          /* 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
3916          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3917          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 3934  for (;; ptr++)
3934              unsigned int origd = d;              unsigned int origd = d;
3935              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3936                {                {
3937       &nbs