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

Diff of /code/trunk/pcre_compile.c

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

revision 91 by nigel, Sat Feb 24 21:41:34 2007 UTC revision 746 by ph10, Tue Nov 15 15:07:02 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) 1997-2006 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  #define NLBLOCK cd            /* The block containing newline information */  #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #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  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93    #define COMPILE_WORK_SIZE (4096)
94    
95  #define BRASTACK_SIZE 200  /* 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
# Line 73  are simple data values; negative values Line 103  are simple data values; negative values
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  #if !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,      0, -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,      0,      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 97  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,      0,     '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,     0,      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 116  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 155  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  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  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  static const char *error_texts[] = {  #define STRING(a)  # a
313    "no error",  #define XSTRING(s) STRING(s)
314    "\\ at end of pattern",  
315    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
316    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
317    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
318    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    it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    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",    "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",    "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",    "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 after (?P",    "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 32 characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum 10,000)",    "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\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397      "inconsistent NEWLINE options\0"
398      "\\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 241  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  #if !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 277  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 291  static const unsigned char digitab[] = Line 485  static const unsigned char digitab[] =
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
490    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
491    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 325  static const unsigned char ebcdic_charta Line 519  static const unsigned char ebcdic_charta
519    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
520    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
522    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
524    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
525    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 352  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, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *);      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 363  static BOOL Line 616  static BOOL
616    
617  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
618  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
619  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
620  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
621  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
622    ptr is pointing at the \. On exit, it is on the final character of the escape
623    sequence.
624    
625  Arguments:  Arguments:
626    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 376  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 394  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  #if !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 412  else if ((i = escapes[c - 0x48]) != 0) Line 667  else if ((i = escapes[c - 0x48]) != 0)
667  else  else
668    {    {
669    const uschar *oldptr;    const uschar *oldptr;
670      BOOL braced, negated;
671    
672    switch (c)    switch (c)
673      {      {
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:
     case 'N':  
     case 'u':  
     case 'U':  
679      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
680      break;      break;
681    
682        case CHAR_u:
683        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684          {
685          /* In JavaScript, \u must be followed by four hexadecimal numbers.
686          Otherwise it is a lowercase u letter. */
687          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            c = 0;
691            for (i = 0; i < 4; ++i)
692              {
693              register int cc = *(++ptr);
694    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
695              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
696              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
697    #else           /* EBCDIC coding */
698              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
699              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
700    #endif
701              }
702            }
703          }
704        else
705          *errorcodeptr = ERR37;
706        break;
707    
708        case CHAR_U:
709        /* In JavaScript, \U is an uppercase U letter. */
710        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
711        break;
712    
713        /* In a character class, \g is just a literal "g". Outside a character
714        class, \g must be followed by one of a number of specific things:
715    
716        (1) A number, either plain or braced. If positive, it is an absolute
717        backreference. If negative, it is a relative backreference. This is a Perl
718        5.10 feature.
719    
720        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
721        is part of Perl's movement towards a unified syntax for back references. As
722        this is synonymous with \k{name}, we fudge it up by pretending it really
723        was \k.
724    
725        (3) For Oniguruma compatibility we also support \g followed by a name or a
726        number either in angle brackets or in single quotes. However, these are
727        (possibly recursive) subroutine calls, _not_ backreferences. Just return
728        the -ESC_g code (cf \k). */
729    
730        case CHAR_g:
731        if (isclass) break;
732        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
733          {
734          c = -ESC_g;
735          break;
736          }
737    
738        /* Handle the Perl-compatible cases */
739    
740        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
741          {
742          const uschar *p;
743          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
744            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
745          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
746            {
747            c = -ESC_k;
748            break;
749            }
750          braced = TRUE;
751          ptr++;
752          }
753        else braced = FALSE;
754    
755        if (ptr[1] == CHAR_MINUS)
756          {
757          negated = TRUE;
758          ptr++;
759          }
760        else negated = FALSE;
761    
762        c = 0;
763        while ((digitab[ptr[1]] & ctype_digit) != 0)
764          c = c * 10 + *(++ptr) - CHAR_0;
765    
766        if (c < 0)   /* Integer overflow */
767          {
768          *errorcodeptr = ERR61;
769          break;
770          }
771    
772        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
773          {
774          *errorcodeptr = ERR57;
775          break;
776          }
777    
778        if (c == 0)
779          {
780          *errorcodeptr = ERR58;
781          break;
782          }
783    
784        if (negated)
785          {
786          if (c > bracount)
787            {
788            *errorcodeptr = ERR15;
789            break;
790            }
791          c = bracount - (c - 1);
792          }
793    
794        c = -(ESC_REF + c);
795        break;
796    
797      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
798      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
799      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 437  else Line 806  else
806      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
807      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
808    
809      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:
810      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
811    
812      if (!isclass)      if (!isclass)
813        {        {
814        oldptr = ptr;        oldptr = ptr;
815        c -= '0';        c -= CHAR_0;
816        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
817          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
818          if (c < 0)    /* Integer overflow */
819            {
820            *errorcodeptr = ERR61;
821            break;
822            }
823        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
824          {          {
825          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 458  else Line 832  else
832      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.
833      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
834    
835      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
836        {        {
837        ptr--;        ptr--;
838        c = 0;        c = 0;
# Line 471  else Line 845  else
845      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
846      than 3 octal digits. */      than 3 octal digits. */
847    
848      case '0':      case CHAR_0:
849      c -= '0';      c -= CHAR_0;
850      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
851          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
852      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
853      break;      break;
854    
# Line 482  else Line 856  else
856      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
857      treated as a data character. */      treated as a data character. */
858    
859      case 'x':      case CHAR_x:
860      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
861          {
862          /* In JavaScript, \x must be followed by two hexadecimal numbers.
863          Otherwise it is a lowercase x letter. */
864          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
865            {
866            c = 0;
867            for (i = 0; i < 2; ++i)
868              {
869              register int cc = *(++ptr);
870    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
872              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873    #else           /* EBCDIC coding */
874              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
875              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876    #endif
877              }
878            }
879          break;
880          }
881    
882        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
883        {        {
884        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
885        int count = 0;        int count = 0;
# Line 492  else Line 888  else
888        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
889          {          {
890          register int cc = *pt++;          register int cc = *pt++;
891          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
892          count++;          count++;
893    
894  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
895          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
896          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
897  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
898          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
899          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
900  #endif  #endif
901          }          }
902    
903        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
904          {          {
905          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
906          ptr = pt;          ptr = pt;
# Line 520  else Line 916  else
916      c = 0;      c = 0;
917      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
918        {        {
919        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
920        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
921  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
922        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
923        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
924  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
925        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
926        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
927  #endif  #endif
928        }        }
929      break;      break;
930    
931      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
932        An error is given if the byte following \c is not an ASCII character. This
933        coding is ASCII-specific, but then the whole concept of \cx is
934        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
935    
936      case 'c':      case CHAR_c:
937      c = *(++ptr);      c = *(++ptr);
938      if (c == 0)      if (c == 0)
939        {        {
940        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
941        return 0;        break;
942        }        }
943    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
944      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
945      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
946      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
947          break;
948  #if !EBCDIC    /* ASCII coding */        }
949      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
950      c ^= 0x40;      c ^= 0x40;
951  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
952      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
953      c ^= 0xC0;      c ^= 0xC0;
954  #endif  #endif
955      break;      break;
956    
957      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
958      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
959      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
960      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
961      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
962    
963      default:      default:
964      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 572  else Line 971  else
971      }      }
972    }    }
973    
974    /* Perl supports \N{name} for character names, as well as plain \N for "not
975    newline". PCRE does not support \N{name}. However, it does support
976    quantification such as \N{2,3}. */
977    
978    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
979         !is_counted_repeat(ptr+2))
980      *errorcodeptr = ERR37;
981    
982    /* If PCRE_UCP is set, we change the values for \d etc. */
983    
984    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
985      c -= (ESC_DU - ESC_D);
986    
987    /* Set the pointer to the final character before returning. */
988    
989  *ptrptr = ptr;  *ptrptr = ptr;
990  return c;  return c;
991  }  }
# Line 612  if (c == 0) goto ERROR_RETURN; Line 1026  if (c == 0) goto ERROR_RETURN;
1026  /* \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
1027  negation. */  negation. */
1028    
1029  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1030    {    {
1031    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1032      {      {
1033      *negptr = TRUE;      *negptr = TRUE;
1034      ptr++;      ptr++;
1035      }      }
1036    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
1037      {      {
1038      c = *(++ptr);      c = *(++ptr);
1039      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1040      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1041      name[i] = c;      name[i] = c;
1042      }      }
1043    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1044    name[i] = 0;    name[i] = 0;
1045    }    }
1046    
# Line 648  top = _pcre_utt_size; Line 1062  top = _pcre_utt_size;
1062  while (bot < top)  while (bot < top)
1063    {    {
1064    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1065    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1066    if (c == 0)    if (c == 0)
1067      {      {
1068      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 672  return -1; Line 1086  return -1;
1086    
1087    
1088  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1089  *         Read repeat counts                     *  *         Read repeat counts                     *
1090  *************************************************/  *************************************************/
1091    
# Line 732  int max = -1; Line 1113  int max = -1;
1113  /* 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
1114  an integer overflow. */  an integer overflow. */
1115    
1116  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1117  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1118    {    {
1119    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 742  if (min < 0 || min > 65535) Line 1123  if (min < 0 || min > 65535)
1123  /* 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.
1124  Also, max must not be less than min. */  Also, max must not be less than min. */
1125    
1126  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1127    {    {
1128    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1129      {      {
1130      max = 0;      max = 0;
1131      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1132      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1133        {        {
1134        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 772  return p; Line 1153  return p;
1153    
1154    
1155  /*************************************************  /*************************************************
1156  *     Find forward referenced named subpattern   *  *  Subroutine for finding forward reference      *
1157  *************************************************/  *************************************************/
1158    
1159  /* This function scans along a pattern looking for capturing subpatterns, and  /* This recursive function is called only from find_parens() below. The
1160  counting them. If it finds a named pattern that matches the name it is given,  top-level call starts at the beginning of the pattern. All other calls must
1161  it returns its number. This is used for forward references to named  start at a parenthesis. It scans along a pattern's text looking for capturing
1162  subpatterns. We know that if (?P< is encountered, the name will be terminated  subpatterns, and counting them. If it finds a named pattern that matches the
1163  by '>' because that is checked in the first pass.  name it is given, it returns its number. Alternatively, if the name is NULL, it
1164    returns when it reaches a given numbered subpattern. Recursion is used to keep
1165    track of subpatterns that reset the capturing group numbers - the (?| feature.
1166    
1167    This function was originally called only from the second pass, in which we know
1168    that if (?< or (?' or (?P< is encountered, the name will be correctly
1169    terminated because that is checked in the first pass. There is now one call to
1170    this function in the first pass, to check for a recursive back reference by
1171    name (so that we can make the whole group atomic). In this case, we need check
1172    only up to the current position in the pattern, and that is still OK because
1173    and previous occurrences will have been checked. To make this work, the test
1174    for "end of pattern" is a check against cd->end_pattern in the main loop,
1175    instead of looking for a binary zero. This means that the special first-pass
1176    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1177    processing items within the loop are OK, because afterwards the main loop will
1178    terminate.)
1179    
1180  Arguments:  Arguments:
1181    pointer      current position in the pattern    ptrptr       address of the current character pointer (updated)
1182    count        current count of capturing parens    cd           compile background data
1183    name         name to seek    name         name to seek, or NULL if seeking a numbered subpattern
1184    namelen      name length    lorn         name length, or subpattern number if name is NULL
1185      xmode        TRUE if we are in /x mode
1186      utf8         TRUE if we are in UTF-8 mode
1187      count        pointer to the current capturing subpattern number (updated)
1188    
1189  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1190  */  */
1191    
1192  static int  static int
1193  find_named_parens(const uschar *ptr, int count, const uschar *name, int namelen)  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1194      BOOL xmode, BOOL utf8, int *count)
1195  {  {
1196  const uschar *thisname;  uschar *ptr = *ptrptr;
1197  for (; *ptr != 0; ptr++)  int start_count = *count;
1198    int hwm_count = start_count;
1199    BOOL dup_parens = FALSE;
1200    
1201    /* If the first character is a parenthesis, check on the type of group we are
1202    dealing with. The very first call may not start with a parenthesis. */
1203    
1204    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1205    {    {
1206    if (*ptr == '\\' && ptr[1] != 0) { ptr++; continue; }    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
   if (*ptr != '(') continue;  
   if (ptr[1] != '?') { count++; continue; }  
   if (ptr[2] == '(') { ptr += 2; continue; }  
   if (ptr[2] != 'P' || ptr[3] != '<') continue;  
   count++;  
   ptr += 4;  
   thisname = ptr;  
   while (*ptr != '>') ptr++;  
   if (namelen == ptr - thisname && strncmp(name, thisname, namelen) == 0)  
     return count;  
   }  
 return -1;  
 }  
1207    
1208      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1209    
1210      /* Handle a normal, unnamed capturing parenthesis. */
1211    
1212  /*************************************************    else if (ptr[1] != CHAR_QUESTION_MARK)
1213  *      Find first significant op code            *      {
1214  *************************************************/      *count += 1;
1215        if (name == NULL && *count == lorn) return *count;
1216        ptr++;
1217        }
1218    
1219  /* This is called by several functions that scan a compiled expression looking    /* All cases now have (? at the start. Remember when we are in a group
1220  for a fixed first character, or an anchoring op code etc. It skips over things    where the parenthesis numbers are duplicated. */
 that do not influence this. For some calls, a change of option is important.  
 For some calls, it makes sense to skip negative forward and all backward  
 assertions, and also the \b assertion; for others it does not.  
1221    
1222  Arguments:    else if (ptr[2] == CHAR_VERTICAL_LINE)
1223    code         pointer to the start of the group      {
1224    options      pointer to external options      ptr += 3;
1225    optbit       the option bit whose changing is significant, or      dup_parens = TRUE;
1226                   zero if none are      }
   skipassert   TRUE if certain assertions are to be skipped  
1227    
1228  Returns:       pointer to the first significant opcode    /* Handle comments; all characters are allowed until a ket is reached. */
 */  
1229    
1230  static const uschar*    else if (ptr[2] == CHAR_NUMBER_SIGN)
 first_significant_code(const uschar *code, int *options, int optbit,  
   BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
1231      {      {
1232      case OP_OPT:      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1233      if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))      goto FAIL_EXIT;
1234        *options = (int)code[1];      }
     code += 2;  
     break;  
1235    
1236      case OP_ASSERT_NOT:    /* Handle a condition. If it is an assertion, just carry on so that it
1237      case OP_ASSERTBACK:    is processed as normal. If not, skip to the closing parenthesis of the
1238      case OP_ASSERTBACK_NOT:    condition (there can't be any nested parens). */
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1239    
1240      case OP_WORD_BOUNDARY:    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1241      case OP_NOT_WORD_BOUNDARY:      {
1242      if (!skipassert) return code;      ptr += 2;
1243      /* Fall through */      if (ptr[1] != CHAR_QUESTION_MARK)
1244          {
1245          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1246          if (*ptr != 0) ptr++;
1247          }
1248        }
1249    
1250      case OP_CALLOUT:    /* Start with (? but not a condition. */
     case OP_CREF:  
     case OP_BRANUMBER:  
     code += _pcre_OP_lengths[*code];  
     break;  
1251    
1252      default:    else
1253      return code;      {
1254        ptr += 2;
1255        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1256    
1257        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1258    
1259        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1260            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1261          {
1262          int term;
1263          const uschar *thisname;
1264          *count += 1;
1265          if (name == NULL && *count == lorn) return *count;
1266          term = *ptr++;
1267          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1268          thisname = ptr;
1269          while (*ptr != term) ptr++;
1270          if (name != NULL && lorn == ptr - thisname &&
1271              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1272            return *count;
1273          term++;
1274          }
1275      }      }
1276    }    }
 /* Control never reaches here */  
 }  
1277    
1278    /* Past any initial parenthesis handling, scan for parentheses or vertical
1279    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1280    first-pass call when this value is temporarily adjusted to stop at the current
1281    position. So DO NOT change this to a test for binary zero. */
1282    
1283    for (; ptr < cd->end_pattern; ptr++)
1284      {
1285      /* Skip over backslashed characters and also entire \Q...\E */
1286    
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 != CHAR_BACKSLASH) {};
1293          if (*ptr == 0) goto FAIL_EXIT;
1294          if (*(++ptr) == CHAR_E) break;
1295          }
1296        continue;
1297        }
1298    
1299  /*************************************************    /* Skip over character classes; this logic must be similar to the way they
1300  *        Find the fixed length of a pattern      *    are handled for real. If the first character is '^', skip it. Also, if the
1301  *************************************************/    first few characters (either before or after ^) are \Q\E or \E we skip them
1302      too. This makes for compatibility with Perl. Note the use of STR macros to
1303      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1304    
1305  /* Scan a pattern and compute the fixed length of subject that will match it,    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1306  if the length is fixed. This is needed for dealing with backward assertions.      {
1307  In UTF8 mode, the result is in characters rather than bytes.      BOOL negate_class = FALSE;
1308        for (;;)
1309          {
1310          if (ptr[1] == CHAR_BACKSLASH)
1311            {
1312            if (ptr[2] == CHAR_E)
1313              ptr+= 2;
1314            else if (strncmp((const char *)ptr+2,
1315                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1316              ptr += 4;
1317            else
1318              break;
1319            }
1320          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1321            {
1322            negate_class = TRUE;
1323            ptr++;
1324            }
1325          else break;
1326          }
1327    
1328  Arguments:      /* If the next character is ']', it is a data character that must be
1329    code     points to the start of the pattern (the bracket)      skipped, except in JavaScript compatibility mode. */
   options  the compiling options  
1330    
1331  Returns:   the fixed length, or -1 if there is no fixed length,      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1332               or -2 if \C was encountered          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1333  */        ptr++;
1334    
1335        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1336          {
1337          if (*ptr == 0) return -1;
1338          if (*ptr == CHAR_BACKSLASH)
1339            {
1340            if (*(++ptr) == 0) goto FAIL_EXIT;
1341            if (*ptr == CHAR_Q) for (;;)
1342              {
1343              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1344              if (*ptr == 0) goto FAIL_EXIT;
1345              if (*(++ptr) == CHAR_E) break;
1346              }
1347            continue;
1348            }
1349          }
1350        continue;
1351        }
1352    
1353      /* Skip comments in /x mode */
1354    
1355      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1356        {
1357        ptr++;
1358        while (*ptr != 0)
1359          {
1360          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1361          ptr++;
1362    #ifdef SUPPORT_UTF8
1363          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1364    #endif
1365          }
1366        if (*ptr == 0) goto FAIL_EXIT;
1367        continue;
1368        }
1369    
1370      /* Check for the special metacharacters */
1371    
1372      if (*ptr == CHAR_LEFT_PARENTHESIS)
1373        {
1374        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1375        if (rc > 0) return rc;
1376        if (*ptr == 0) goto FAIL_EXIT;
1377        }
1378    
1379      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1380        {
1381        if (dup_parens && *count < hwm_count) *count = hwm_count;
1382        goto FAIL_EXIT;
1383        }
1384    
1385      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1386        {
1387        if (*count > hwm_count) hwm_count = *count;
1388        *count = start_count;
1389        }
1390      }
1391    
1392    FAIL_EXIT:
1393    *ptrptr = ptr;
1394    return -1;
1395    }
1396    
1397    
1398    
1399    
1400    /*************************************************
1401    *       Find forward referenced subpattern       *
1402    *************************************************/
1403    
1404    /* This function scans along a pattern's text looking for capturing
1405    subpatterns, and counting them. If it finds a named pattern that matches the
1406    name it is given, it returns its number. Alternatively, if the name is NULL, it
1407    returns when it reaches a given numbered subpattern. This is used for forward
1408    references to subpatterns. We used to be able to start this scan from the
1409    current compiling point, using the current count value from cd->bracount, and
1410    do it all in a single loop, but the addition of the possibility of duplicate
1411    subpattern numbers means that we have to scan from the very start, in order to
1412    take account of such duplicates, and to use a recursive function to keep track
1413    of the different types of group.
1414    
1415    Arguments:
1416      cd           compile background data
1417      name         name to seek, or NULL if seeking a numbered subpattern
1418      lorn         name length, or subpattern number if name is NULL
1419      xmode        TRUE if we are in /x mode
1420      utf8         TRUE if we are in UTF-8 mode
1421    
1422    Returns:       the number of the found subpattern, or -1 if not found
1423    */
1424    
1425    static int
1426    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427      BOOL utf8)
1428    {
1429    uschar *ptr = (uschar *)cd->start_pattern;
1430    int count = 0;
1431    int rc;
1432    
1433    /* If the pattern does not start with an opening parenthesis, the first call
1434    to find_parens_sub() will scan right to the end (if necessary). However, if it
1435    does start with a parenthesis, find_parens_sub() will return when it hits the
1436    matching closing parens. That is why we have to have a loop. */
1437    
1438    for (;;)
1439      {
1440      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441      if (rc > 0 || *ptr++ == 0) break;
1442      }
1443    
1444    return rc;
1445    }
1446    
1447    
1448    
1449    
1450    /*************************************************
1451    *      Find first significant op code            *
1452    *************************************************/
1453    
1454    /* This is called by several functions that scan a compiled expression looking
1455    for a fixed first character, or an anchoring op code etc. It skips over things
1456    that do not influence this. For some calls, it makes sense to skip negative
1457    forward and all backward assertions, and also the \b assertion; for others it
1458    does not.
1459    
1460    Arguments:
1461      code         pointer to the start of the group
1462      skipassert   TRUE if certain assertions are to be skipped
1463    
1464    Returns:       pointer to the first significant opcode
1465    */
1466    
1467    static const uschar*
1468    first_significant_code(const uschar *code, BOOL skipassert)
1469    {
1470    for (;;)
1471      {
1472      switch ((int)*code)
1473        {
1474        case OP_ASSERT_NOT:
1475        case OP_ASSERTBACK:
1476        case OP_ASSERTBACK_NOT:
1477        if (!skipassert) return code;
1478        do code += GET(code, 1); while (*code == OP_ALT);
1479        code += _pcre_OP_lengths[*code];
1480        break;
1481    
1482        case OP_WORD_BOUNDARY:
1483        case OP_NOT_WORD_BOUNDARY:
1484        if (!skipassert) return code;
1485        /* Fall through */
1486    
1487        case OP_CALLOUT:
1488        case OP_CREF:
1489        case OP_NCREF:
1490        case OP_RREF:
1491        case OP_NRREF:
1492        case OP_DEF:
1493        code += _pcre_OP_lengths[*code];
1494        break;
1495    
1496        default:
1497        return code;
1498        }
1499      }
1500    /* Control never reaches here */
1501    }
1502    
1503    
1504    
1505    
1506    /*************************************************
1507    *        Find the fixed length of a branch       *
1508    *************************************************/
1509    
1510    /* Scan a branch and compute the fixed length of subject that will match it,
1511    if the length is fixed. This is needed for dealing with backward assertions.
1512    In UTF8 mode, the result is in characters rather than bytes. The branch is
1513    temporarily terminated with OP_END when this function is called.
1514    
1515    This function is called when a backward assertion is encountered, so that if it
1516    fails, the error message can point to the correct place in the pattern.
1517    However, we cannot do this when the assertion contains subroutine calls,
1518    because they can be forward references. We solve this by remembering this case
1519    and doing the check at the end; a flag specifies which mode we are running in.
1520    
1521    Arguments:
1522      code     points to the start of the pattern (the bracket)
1523      utf8     TRUE in UTF-8 mode
1524      atend    TRUE if called when the pattern is complete
1525      cd       the "compile data" structure
1526    
1527    Returns:   the fixed length,
1528                 or -1 if there is no fixed length,
1529                 or -2 if \C was encountered
1530                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1531    */
1532    
1533  static int  static int
1534  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1535  {  {
1536  int length = -1;  int length = -1;
1537    
# Line 906  branch, check the length against that of Line 1544  branch, check the length against that of
1544  for (;;)  for (;;)
1545    {    {
1546    int d;    int d;
1547      uschar *ce, *cs;
1548    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1549    switch (op)    switch (op)
1550      {      {
1551        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1552        OP_BRA (normal non-capturing bracket) because the other variants of these
1553        opcodes are all concerned with unlimited repeated groups, which of course
1554        are not of fixed length. They will cause a -1 response from the default
1555        case of this switch. */
1556    
1557        case OP_CBRA:
1558      case OP_BRA:      case OP_BRA:
1559      case OP_ONCE:      case OP_ONCE:
1560        case OP_ONCE_NC:
1561      case OP_COND:      case OP_COND:
1562      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1563      if (d < 0) return d;      if (d < 0) return d;
1564      branchlength += d;      branchlength += d;
1565      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 923  for (;;) Line 1568  for (;;)
1568    
1569      /* 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
1570      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
1571      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.
1572        Note that we must not include the OP_KETRxxx opcodes here, because they
1573        all imply an unlimited repeat. */
1574    
1575      case OP_ALT:      case OP_ALT:
1576      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1577      case OP_END:      case OP_END:
1578      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1579        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 937  for (;;) Line 1582  for (;;)
1582      branchlength = 0;      branchlength = 0;
1583      break;      break;
1584    
1585        /* A true recursion implies not fixed length, but a subroutine call may
1586        be OK. If the subroutine is a forward reference, we can't deal with
1587        it until the end of the pattern, so return -3. */
1588    
1589        case OP_RECURSE:
1590        if (!atend) return -3;
1591        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1592        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1593        if (cc > cs && cc < ce) return -1;                /* Recursion */
1594        d = find_fixedlength(cs + 2, utf8, atend, cd);
1595        if (d < 0) return d;
1596        branchlength += d;
1597        cc += 1 + LINK_SIZE;
1598        break;
1599    
1600      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1601    
1602      case OP_ASSERT:      case OP_ASSERT:
# Line 949  for (;;) Line 1609  for (;;)
1609      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1610    
1611      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1612      case OP_CREF:      case OP_CREF:
1613      case OP_OPT:      case OP_NCREF:
1614        case OP_RREF:
1615        case OP_NRREF:
1616        case OP_DEF:
1617      case OP_CALLOUT:      case OP_CALLOUT:
1618      case OP_SOD:      case OP_SOD:
1619      case OP_SOM:      case OP_SOM:
1620        case OP_SET_SOM:
1621      case OP_EOD:      case OP_EOD:
1622      case OP_EODN:      case OP_EODN:
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625      case OP_DOLL:      case OP_DOLL:
1626        case OP_DOLLM:
1627      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1628      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1629      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 967  for (;;) Line 1632  for (;;)
1632      /* Handle literal characters */      /* Handle literal characters */
1633    
1634      case OP_CHAR:      case OP_CHAR:
1635      case OP_CHARNC:      case OP_CHARI:
1636      case OP_NOT:      case OP_NOT:
1637        case OP_NOTI:
1638      branchlength++;      branchlength++;
1639      cc += 2;      cc += 2;
1640  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1641      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1642  #endif  #endif
1643      break;      break;
1644    
# Line 983  for (;;) Line 1646  for (;;)
1646      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1647    
1648      case OP_EXACT:      case OP_EXACT:
1649        case OP_EXACTI:
1650      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1651      cc += 4;      cc += 4;
1652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1653      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1654  #endif  #endif
1655      break;      break;
1656    
1657      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1658      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1659        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1660      cc += 4;      cc += 4;
1661      break;      break;
1662    
# Line 1012  for (;;) Line 1674  for (;;)
1674      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1675      case OP_WORDCHAR:      case OP_WORDCHAR:
1676      case OP_ANY:      case OP_ANY:
1677        case OP_ALLANY:
1678      branchlength++;      branchlength++;
1679      cc++;      cc++;
1680      break;      break;
# Line 1066  for (;;) Line 1729  for (;;)
1729    
1730    
1731  /*************************************************  /*************************************************
1732  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1733  *************************************************/  *************************************************/
1734    
1735  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1736  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1737    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1738    so that it can be called from pcre_study() when finding the minimum matching
1739    length.
1740    
1741  Arguments:  Arguments:
1742    code        points to start of expression    code        points to start of expression
1743    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1744    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1745    
1746  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
1747  */  */
1748    
1749  static const uschar *  const uschar *
1750  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1751  {  {
1752  for (;;)  for (;;)
1753    {    {
1754    register int c = *code;    register int c = *code;
1755    
1756    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1757    
1758    /* 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 1094  for (;;) Line 1761  for (;;)
1761    
1762    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1763    
1764    /* Handle bracketed group */    /* Handle recursion */
1765    
1766      else if (c == OP_REVERSE)
1767        {
1768        if (number < 0) return (uschar *)code;
1769        code += _pcre_OP_lengths[c];
1770        }
1771    
1772      /* Handle capturing bracket */
1773    
1774    else if (c > OP_BRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1775               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1776      {      {
1777      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1778      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1779      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1780      }      }
1781    
1782    /* 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
1783    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
1784    The length in the table is a minimum, so we have to scan along to skip the    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1785    extra bytes. All opcodes are less than 128, so we can use relatively    must add in its length. */
   efficient code. */  
1786    
1787    else    else
1788      {      {
1789        switch(c)
1790          {
1791          case OP_TYPESTAR:
1792          case OP_TYPEMINSTAR:
1793          case OP_TYPEPLUS:
1794          case OP_TYPEMINPLUS:
1795          case OP_TYPEQUERY:
1796          case OP_TYPEMINQUERY:
1797          case OP_TYPEPOSSTAR:
1798          case OP_TYPEPOSPLUS:
1799          case OP_TYPEPOSQUERY:
1800          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1801          break;
1802    
1803          case OP_TYPEUPTO:
1804          case OP_TYPEMINUPTO:
1805          case OP_TYPEEXACT:
1806          case OP_TYPEPOSUPTO:
1807          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1808          break;
1809    
1810          case OP_MARK:
1811          case OP_PRUNE_ARG:
1812          case OP_SKIP_ARG:
1813          code += code[1];
1814          break;
1815    
1816          case OP_THEN_ARG:
1817          code += code[1];
1818          break;
1819          }
1820    
1821        /* Add in the fixed length from the table */
1822    
1823      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1824    
1825      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1826      a multi-byte character. The length in the table is a minimum, so we have to
1827      arrange to skip the extra bytes. */
1828    
1829    #ifdef SUPPORT_UTF8
1830      if (utf8) switch(c)      if (utf8) switch(c)
1831        {        {
1832        case OP_CHAR:        case OP_CHAR:
1833        case OP_CHARNC:        case OP_CHARI:
1834        case OP_EXACT:        case OP_EXACT:
1835          case OP_EXACTI:
1836        case OP_UPTO:        case OP_UPTO:
1837          case OP_UPTOI:
1838        case OP_MINUPTO:        case OP_MINUPTO:
1839          case OP_MINUPTOI:
1840          case OP_POSUPTO:
1841          case OP_POSUPTOI:
1842        case OP_STAR:        case OP_STAR:
1843          case OP_STARI:
1844        case OP_MINSTAR:        case OP_MINSTAR:
1845          case OP_MINSTARI:
1846          case OP_POSSTAR:
1847          case OP_POSSTARI:
1848        case OP_PLUS:        case OP_PLUS:
1849          case OP_PLUSI:
1850        case OP_MINPLUS:        case OP_MINPLUS:
1851          case OP_MINPLUSI:
1852          case OP_POSPLUS:
1853          case OP_POSPLUSI:
1854        case OP_QUERY:        case OP_QUERY:
1855          case OP_QUERYI:
1856        case OP_MINQUERY:        case OP_MINQUERY:
1857        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1858          case OP_POSQUERY:
1859          case OP_POSQUERYI:
1860          if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1861        break;        break;
1862        }        }
1863    #else
1864        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1865    #endif
1866      }      }
1867    }    }
1868  }  }
# Line 1164  for (;;) Line 1898  for (;;)
1898    
1899    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1900    
1901    /* All bracketed groups have the same length. */    /* Otherwise, we can get the item's length from the table, except that for
1902      repeated character types, we have to test for \p and \P, which have an extra
1903      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1904      must add in its length. */
1905    
1906    else if (c > OP_BRA)    else
1907      {      {
1908      code += _pcre_OP_lengths[OP_BRA];      switch(c)
1909      }        {
1910          case OP_TYPESTAR:
1911          case OP_TYPEMINSTAR:
1912          case OP_TYPEPLUS:
1913          case OP_TYPEMINPLUS:
1914          case OP_TYPEQUERY:
1915          case OP_TYPEMINQUERY:
1916          case OP_TYPEPOSSTAR:
1917          case OP_TYPEPOSPLUS:
1918          case OP_TYPEPOSQUERY:
1919          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1920          break;
1921    
1922    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes        case OP_TYPEPOSUPTO:
1923    that are followed by a character may be followed by a multi-byte character.        case OP_TYPEUPTO:
1924    The length in the table is a minimum, so we have to scan along to skip the        case OP_TYPEMINUPTO:
1925    extra bytes. All opcodes are less than 128, so we can use relatively        case OP_TYPEEXACT:
1926    efficient code. */        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1927          break;
1928    
1929          case OP_MARK:
1930          case OP_PRUNE_ARG:
1931          case OP_SKIP_ARG:
1932          code += code[1];
1933          break;
1934    
1935          case OP_THEN_ARG:
1936          code += code[1];
1937          break;
1938          }
1939    
1940        /* Add in the fixed length from the table */
1941    
   else  
     {  
1942      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1943    
1944        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1945        by a multi-byte character. The length in the table is a minimum, so we have
1946        to arrange to skip the extra bytes. */
1947    
1948    #ifdef SUPPORT_UTF8
1949      if (utf8) switch(c)      if (utf8) switch(c)
1950        {        {
1951        case OP_CHAR:        case OP_CHAR:
1952        case OP_CHARNC:        case OP_CHARI:
1953        case OP_EXACT:        case OP_EXACT:
1954          case OP_EXACTI:
1955        case OP_UPTO:        case OP_UPTO:
1956          case OP_UPTOI:
1957        case OP_MINUPTO:        case OP_MINUPTO:
1958          case OP_MINUPTOI:
1959          case OP_POSUPTO:
1960          case OP_POSUPTOI:
1961        case OP_STAR:        case OP_STAR:
1962          case OP_STARI:
1963        case OP_MINSTAR:        case OP_MINSTAR:
1964          case OP_MINSTARI:
1965          case OP_POSSTAR:
1966          case OP_POSSTARI:
1967        case OP_PLUS:        case OP_PLUS:
1968          case OP_PLUSI:
1969        case OP_MINPLUS:        case OP_MINPLUS:
1970          case OP_MINPLUSI:
1971          case OP_POSPLUS:
1972          case OP_POSPLUSI:
1973        case OP_QUERY:        case OP_QUERY:
1974          case OP_QUERYI:
1975        case OP_MINQUERY:        case OP_MINQUERY:
1976        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1977          case OP_POSQUERY:
1978          case OP_POSQUERYI:
1979          if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1980        break;        break;
1981        }        }
1982    #else
1983        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1984    #endif
1985      }      }
1986    }    }
1987  }  }
# Line 1207  for (;;) Line 1993  for (;;)
1993  *************************************************/  *************************************************/
1994    
1995  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
1996  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1997  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
1998  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
1999  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2000    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2001    bracket whose current branch will already have been scanned.
2002    
2003  Arguments:  Arguments:
2004    code        points to start of search    code        points to start of search
2005    endcode     points to where to stop    endcode     points to where to stop
2006    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2007      cd          contains pointers to tables etc.
2008    
2009  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2010  */  */
2011    
2012  static BOOL  static BOOL
2013  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2014      compile_data *cd)
2015  {  {
2016  register int c;  register int c;
2017  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2018       code < endcode;       code < endcode;
2019       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2020    {    {
2021    const uschar *ccode;    const uschar *ccode;
2022    
2023    c = *code;    c = *code;
2024    
2025    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2026      first_significant_code() with a TRUE final argument. */
2027    
2028      if (c == OP_ASSERT)
2029      {      {
2030        do code += GET(code, 1); while (*code == OP_ALT);
2031        c = *code;
2032        continue;
2033        }
2034    
2035      /* For a recursion/subroutine call, if its end has been reached, which
2036      implies a backward reference subroutine call, we can scan it. If it's a
2037      forward reference subroutine call, we can't. To detect forward reference
2038      we have to scan up the list that is kept in the workspace. This function is
2039      called only when doing the real compile, not during the pre-compile that
2040      measures the size of the compiled pattern. */
2041    
2042      if (c == OP_RECURSE)
2043        {
2044        const uschar *scode;
2045      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2046    
2047      /* Scan a closed bracket */      /* Test for forward reference */
2048    
2049        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2050          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2051    
2052        /* Not a forward reference, test for completed backward reference */
2053    
2054      empty_branch = FALSE;      empty_branch = FALSE;
2055        scode = cd->start_code + GET(code, 1);
2056        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2057    
2058        /* Completed backwards reference */
2059    
2060      do      do
2061        {        {
2062        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2063            {
2064          empty_branch = TRUE;          empty_branch = TRUE;
2065            break;
2066            }
2067          scode += GET(scode, 1);
2068          }
2069        while (*scode == OP_ALT);
2070    
2071        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2072        continue;
2073        }
2074    
2075      /* Groups with zero repeats can of course be empty; skip them. */
2076    
2077      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2078          c == OP_BRAPOSZERO)
2079        {
2080        code += _pcre_OP_lengths[c];
2081        do code += GET(code, 1); while (*code == OP_ALT);
2082        c = *code;
2083        continue;
2084        }
2085    
2086      /* A nested group that is already marked as "could be empty" can just be
2087      skipped. */
2088    
2089      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2090          c == OP_SCBRA || c == OP_SCBRAPOS)
2091        {
2092        do code += GET(code, 1); while (*code == OP_ALT);
2093        c = *code;
2094        continue;
2095        }
2096    
2097      /* For other groups, scan the branches. */
2098    
2099      if (c == OP_BRA  || c == OP_BRAPOS ||
2100          c == OP_CBRA || c == OP_CBRAPOS ||
2101          c == OP_ONCE || c == OP_ONCE_NC ||
2102          c == OP_COND)
2103        {
2104        BOOL empty_branch;
2105        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2106    
2107        /* If a conditional group has only one branch, there is a second, implied,
2108        empty branch, so just skip over the conditional, because it could be empty.
2109        Otherwise, scan the individual branches of the group. */
2110    
2111        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2112        code += GET(code, 1);        code += GET(code, 1);
2113        else
2114          {
2115          empty_branch = FALSE;
2116          do
2117            {
2118            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2119              empty_branch = TRUE;
2120            code += GET(code, 1);
2121            }
2122          while (*code == OP_ALT);
2123          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2124        }        }
2125      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2126      c = *code;      c = *code;
2127        continue;
2128      }      }
2129    
2130    else switch (c)    /* Handle the other opcodes */
2131    
2132      switch (c)
2133      {      {
2134      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2135        cannot be represented just by a bit map. This includes negated single
2136        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2137        actual length is stored in the compiled code, so we must update "code"
2138        here. */
2139    
2140  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2141      case OP_XCLASS:      case OP_XCLASS:
2142      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2143      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2144  #endif  #endif
2145    
# Line 1302  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2184      case OP_WORDCHAR:      case OP_WORDCHAR:
2185      case OP_ANY:      case OP_ANY:
2186        case OP_ALLANY:
2187      case OP_ANYBYTE:      case OP_ANYBYTE:
2188      case OP_CHAR:      case OP_CHAR:
2189      case OP_CHARNC:      case OP_CHARI:
2190      case OP_NOT:      case OP_NOT:
2191        case OP_NOTI:
2192      case OP_PLUS:      case OP_PLUS:
2193      case OP_MINPLUS:      case OP_MINPLUS:
2194        case OP_POSPLUS:
2195      case OP_EXACT:      case OP_EXACT:
2196      case OP_NOTPLUS:      case OP_NOTPLUS:
2197      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2198        case OP_NOTPOSPLUS:
2199      case OP_NOTEXACT:      case OP_NOTEXACT:
2200      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2201      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2202        case OP_TYPEPOSPLUS:
2203      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2204      return FALSE;      return FALSE;
2205    
2206        /* These are going to continue, as they may be empty, but we have to
2207        fudge the length for the \p and \P cases. */
2208    
2209        case OP_TYPESTAR:
2210        case OP_TYPEMINSTAR:
2211        case OP_TYPEPOSSTAR:
2212        case OP_TYPEQUERY:
2213        case OP_TYPEMINQUERY:
2214        case OP_TYPEPOSQUERY:
2215        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2216        break;
2217    
2218        /* Same for these */
2219    
2220        case OP_TYPEUPTO:
2221        case OP_TYPEMINUPTO:
2222        case OP_TYPEPOSUPTO:
2223        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2224        break;
2225    
2226      /* End of branch */      /* End of branch */
2227    
2228      case OP_KET:      case OP_KET:
2229      case OP_KETRMAX:      case OP_KETRMAX:
2230      case OP_KETRMIN:      case OP_KETRMIN:
2231        case OP_KETRPOS:
2232      case OP_ALT:      case OP_ALT:
2233      return TRUE;      return TRUE;
2234    
2235      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2236      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2237    
2238  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2239      case OP_STAR:      case OP_STAR:
2240        case OP_STARI:
2241      case OP_MINSTAR:      case OP_MINSTAR:
2242        case OP_MINSTARI:
2243        case OP_POSSTAR:
2244        case OP_POSSTARI:
2245      case OP_QUERY:      case OP_QUERY:
2246        case OP_QUERYI:
2247      case OP_MINQUERY:      case OP_MINQUERY:
2248        case OP_MINQUERYI:
2249        case OP_POSQUERY:
2250        case OP_POSQUERYI:
2251        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2252        break;
2253    
2254      case OP_UPTO:      case OP_UPTO:
2255        case OP_UPTOI:
2256      case OP_MINUPTO:      case OP_MINUPTO:
2257      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2258        case OP_POSUPTO:
2259        case OP_POSUPTOI:
2260        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2261      break;      break;
2262  #endif  #endif
2263    
2264        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2265        string. */
2266    
2267        case OP_MARK:
2268        case OP_PRUNE_ARG:
2269        case OP_SKIP_ARG:
2270        code += code[1];
2271        break;
2272    
2273        case OP_THEN_ARG:
2274        code += code[1];
2275        break;
2276    
2277        /* None of the remaining opcodes are required to match a character. */
2278    
2279        default:
2280        break;
2281      }      }
2282    }    }
2283    
# Line 1354  return TRUE; Line 2294  return TRUE;
2294  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
2295  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,
2296  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.
2297    This function is called only during the real compile, not during the
2298    pre-compile.
2299    
2300  Arguments:  Arguments:
2301    code        points to start of the recursion    code        points to start of the recursion
2302    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2303    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2304    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2305      cd          pointers to tables etc
2306    
2307  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2308  */  */
2309    
2310  static BOOL  static BOOL
2311  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2312    BOOL utf8)    BOOL utf8, compile_data *cd)
2313  {  {
2314  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2315    {    {
2316    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2317        return FALSE;
2318    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2319    }    }
2320  return TRUE;  return TRUE;
# Line 1383  return TRUE; Line 2327  return TRUE;
2327  *************************************************/  *************************************************/
2328    
2329  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2330  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
2331  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2332  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2333    
2334    Originally, this function only recognized a sequence of letters between the
2335    terminators, but it seems that Perl recognizes any sequence of characters,
2336    though of course unknown POSIX names are subsequently rejected. Perl gives an
2337    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2338    didn't consider this to be a POSIX class. Likewise for [:1234:].
2339    
2340    The problem in trying to be exactly like Perl is in the handling of escapes. We
2341    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2342    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2343    below handles the special case of \], but does not try to do any other escape
2344    processing. This makes it different from Perl for cases such as [:l\ower:]
2345    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2346    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2347    I think.
2348    
2349    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2350    It seems that the appearance of a nested POSIX class supersedes an apparent
2351    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2352    a digit.
2353    
2354    In Perl, unescaped square brackets may also appear as part of class names. For
2355    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2356    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2357    seem right at all. PCRE does not allow closing square brackets in POSIX class
2358    names.
2359    
2360  Argument:  Arguments:
2361    ptr      pointer to the initial [    ptr      pointer to the initial [
2362    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2363    
2364  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2365  */  */
2366    
2367  static BOOL  static BOOL
2368  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2369  {  {
2370  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2371  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2372  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2373    {    {
2374    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2375    return TRUE;      ptr++;
2376      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2377      else
2378        {
2379        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2380          {
2381          *endptr = ptr;
2382          return TRUE;
2383          }
2384        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2385             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2386              ptr[1] == CHAR_EQUALS_SIGN) &&
2387            check_posix_syntax(ptr, endptr))
2388          return FALSE;
2389        }
2390    }    }
2391  return FALSE;  return FALSE;
2392  }  }
# Line 1430  Returns:     a value representing the na Line 2411  Returns:     a value representing the na
2411  static int  static int
2412  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2413  {  {
2414    const char *pn = posix_names;
2415  register int yield = 0;  register int yield = 0;
2416  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2417    {    {
2418    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2419      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2420      pn += posix_name_lengths[yield] + 1;
2421    yield++;    yield++;
2422    }    }
2423  return -1;  return -1;
# Line 1449  return -1; Line 2432  return -1;
2432  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2433  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2434  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
2435  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
2436  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
2437  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
2438  offsets adjusted. That is the job of this function. Before it is called, the  have their offsets adjusted. That one of the jobs of this function. Before it
2439  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2440    OP_END.
2441    
2442    This function has been extended with the possibility of forward references for
2443    recursions and subroutine calls. It must also check the list of such references
2444    for the group we are dealing with. If it finds that one of the recursions in
2445    the current group is on this list, it adjusts the offset in the list, not the
2446    value in the reference (which is a group number).
2447    
2448  Arguments:  Arguments:
2449    group      points to the start of the group    group      points to the start of the group
2450    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2451    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2452    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2453      save_hwm   the hwm forward reference pointer at the start of the group
2454    
2455  Returns:     nothing  Returns:     nothing
2456  */  */
2457    
2458  static void  static void
2459  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2460      uschar *save_hwm)
2461  {  {
2462  uschar *ptr = group;  uschar *ptr = group;
2463    
2464  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2465    {    {
2466    int offset = GET(ptr, 1);    int offset;
2467    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2468    ptr += 1 + LINK_SIZE;  
2469      /* See if this recursion is on the forward reference list. If so, adjust the
2470      reference. */
2471    
2472      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2473        {
2474        offset = GET(hc, 0);
2475        if (cd->start_code + offset == ptr + 1)
2476          {
2477          PUT(hc, 0, offset + adjust);
2478          break;
2479          }
2480        }
2481    
2482      /* Otherwise, adjust the recursion offset if it's after the start of this
2483      group. */
2484    
2485      if (hc >= cd->hwm)
2486        {
2487        offset = GET(ptr, 1);
2488        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2489        }
2490    
2491      ptr += 1 + LINK_SIZE;
2492    }    }
2493  }  }
2494    
# Line 1498  auto_callout(uschar *code, const uschar Line 2514  auto_callout(uschar *code, const uschar
2514  {  {
2515  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2516  *code++ = 255;  *code++ = 255;
2517  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2518  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2519  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2520  }  }
2521    
# Line 1524  Returns:             nothing Line 2540  Returns:             nothing
2540  static void  static void
2541  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2542  {  {
2543  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2544  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2545  }  }
2546    
# Line 1550  Yield:        TRUE when range returned; Line 2566  Yield:        TRUE when range returned;
2566  */  */
2567    
2568  static BOOL  static BOOL
2569  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2570      unsigned int *odptr)
2571  {  {
2572  int c, othercase, next;  unsigned int c, othercase, next;
2573    
2574  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2575    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2576    
2577  if (c > d) return FALSE;  if (c > d) return FALSE;
2578    
# Line 1564  next = othercase + 1; Line 2581  next = othercase + 1;
2581    
2582  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2583    {    {
2584    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2585    next++;    next++;
2586    }    }
2587    
# Line 1573  for (++c; c <= d; c++) Line 2590  for (++c; c <= d; c++)
2590    
2591  return TRUE;  return TRUE;
2592  }  }
2593    
2594    
2595    
2596    /*************************************************
2597    *        Check a character and a property        *
2598    *************************************************/
2599    
2600    /* This function is called by check_auto_possessive() when a property item
2601    is adjacent to a fixed character.
2602    
2603    Arguments:
2604      c            the character
2605      ptype        the property type
2606      pdata        the data for the type
2607      negated      TRUE if it's a negated property (\P or \p{^)
2608    
2609    Returns:       TRUE if auto-possessifying is OK
2610    */
2611    
2612    static BOOL
2613    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2614    {
2615    const ucd_record *prop = GET_UCD(c);
2616    switch(ptype)
2617      {
2618      case PT_LAMP:
2619      return (prop->chartype == ucp_Lu ||
2620              prop->chartype == ucp_Ll ||
2621              prop->chartype == ucp_Lt) == negated;
2622    
2623      case PT_GC:
2624      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2625    
2626      case PT_PC:
2627      return (pdata == prop->chartype) == negated;
2628    
2629      case PT_SC:
2630      return (pdata == prop->script) == negated;
2631    
2632      /* These are specials */
2633    
2634      case PT_ALNUM:
2635      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2636              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2637    
2638      case PT_SPACE:    /* Perl space */
2639      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2640              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2641              == negated;
2642    
2643      case PT_PXSPACE:  /* POSIX space */
2644      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2645              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2646              c == CHAR_FF || c == CHAR_CR)
2647              == negated;
2648    
2649      case PT_WORD:
2650      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2651              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2652              c == CHAR_UNDERSCORE) == negated;
2653      }
2654    return FALSE;
2655    }
2656  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2657    
2658    
2659    
2660    /*************************************************
2661    *     Check if auto-possessifying is possible    *
2662    *************************************************/
2663    
2664    /* This function is called for unlimited repeats of certain items, to see
2665    whether the next thing could possibly match the repeated item. If not, it makes
2666    sense to automatically possessify the repeated item.
2667    
2668    Arguments:
2669      previous      pointer to the repeated opcode
2670      utf8          TRUE in UTF-8 mode
2671      ptr           next character in pattern
2672      options       options bits
2673      cd            contains pointers to tables etc.
2674    
2675    Returns:        TRUE if possessifying is wanted
2676    */
2677    
2678    static BOOL
2679    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2680      int options, compile_data *cd)
2681    {
2682    int c, next;
2683    int op_code = *previous++;
2684    
2685    /* Skip whitespace and comments in extended mode */
2686    
2687    if ((options & PCRE_EXTENDED) != 0)
2688      {
2689      for (;;)
2690        {
2691        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2692        if (*ptr == CHAR_NUMBER_SIGN)
2693          {
2694          ptr++;
2695          while (*ptr != 0)
2696            {
2697            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2698            ptr++;
2699    #ifdef SUPPORT_UTF8
2700            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2701    #endif
2702            }
2703          }
2704        else break;
2705        }
2706      }
2707    
2708    /* If the next item is one that we can handle, get its value. A non-negative
2709    value is a character, a negative value is an escape value. */
2710    
2711    if (*ptr == CHAR_BACKSLASH)
2712      {
2713      int temperrorcode = 0;
2714      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2715      if (temperrorcode != 0) return FALSE;
2716      ptr++;    /* Point after the escape sequence */
2717      }
2718    
2719    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2720      {
2721    #ifdef SUPPORT_UTF8
2722      if (utf8) { GETCHARINC(next, ptr); } else
2723    #endif
2724      next = *ptr++;
2725      }
2726    
2727    else return FALSE;
2728    
2729    /* Skip whitespace and comments in extended mode */
2730    
2731    if ((options & PCRE_EXTENDED) != 0)
2732      {
2733      for (;;)
2734        {
2735        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2736        if (*ptr == CHAR_NUMBER_SIGN)
2737          {
2738          ptr++;
2739          while (*ptr != 0)
2740            {
2741            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2742            ptr++;
2743    #ifdef SUPPORT_UTF8
2744            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2745    #endif
2746            }
2747          }
2748        else break;
2749        }
2750      }
2751    
2752    /* If the next thing is itself optional, we have to give up. */
2753    
2754    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2755      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2756        return FALSE;
2757    
2758    /* Now compare the next item with the previous opcode. First, handle cases when
2759    the next item is a character. */
2760    
2761    if (next >= 0) switch(op_code)
2762      {
2763      case OP_CHAR:
2764    #ifdef SUPPORT_UTF8
2765      GETCHARTEST(c, previous);
2766    #else
2767      c = *previous;
2768    #endif
2769      return c != next;
2770    
2771      /* For CHARI (caseless character) we must check the other case. If we have
2772      Unicode property support, we can use it to test the other case of
2773      high-valued characters. */
2774    
2775      case OP_CHARI:
2776    #ifdef SUPPORT_UTF8
2777      GETCHARTEST(c, previous);
2778    #else
2779      c = *previous;
2780    #endif
2781      if (c == next) return FALSE;
2782    #ifdef SUPPORT_UTF8
2783      if (utf8)
2784        {
2785        unsigned int othercase;
2786        if (next < 128) othercase = cd->fcc[next]; else
2787    #ifdef SUPPORT_UCP
2788        othercase = UCD_OTHERCASE((unsigned int)next);
2789    #else
2790        othercase = NOTACHAR;
2791    #endif
2792        return (unsigned int)c != othercase;
2793        }
2794      else
2795    #endif  /* SUPPORT_UTF8 */
2796      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2797    
2798      /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2799      opcodes are not used for multi-byte characters, because they are coded using
2800      an XCLASS instead. */
2801    
2802      case OP_NOT:
2803      return (c = *previous) == next;
2804    
2805      case OP_NOTI:
2806      if ((c = *previous) == next) return TRUE;
2807    #ifdef SUPPORT_UTF8
2808      if (utf8)
2809        {
2810        unsigned int othercase;
2811        if (next < 128) othercase = cd->fcc[next]; else
2812    #ifdef SUPPORT_UCP
2813        othercase = UCD_OTHERCASE(next);
2814    #else
2815        othercase = NOTACHAR;
2816    #endif
2817        return (unsigned int)c == othercase;
2818        }
2819      else
2820    #endif  /* SUPPORT_UTF8 */
2821      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2822    
2823      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2824      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2825    
2826      case OP_DIGIT:
2827      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2828    
2829      case OP_NOT_DIGIT:
2830      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2831    
2832      case OP_WHITESPACE:
2833      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2834    
2835      case OP_NOT_WHITESPACE:
2836      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2837    
2838      case OP_WORDCHAR:
2839      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2840    
2841      case OP_NOT_WORDCHAR:
2842      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2843    
2844      case OP_HSPACE:
2845      case OP_NOT_HSPACE:
2846      switch(next)
2847        {
2848        case 0x09:
2849        case 0x20:
2850        case 0xa0:
2851        case 0x1680:
2852        case 0x180e:
2853        case 0x2000:
2854        case 0x2001:
2855        case 0x2002:
2856        case 0x2003:
2857        case 0x2004:
2858        case 0x2005:
2859        case 0x2006:
2860        case 0x2007:
2861        case 0x2008:
2862        case 0x2009:
2863        case 0x200A:
2864        case 0x202f:
2865        case 0x205f:
2866        case 0x3000:
2867        return op_code == OP_NOT_HSPACE;
2868        default:
2869        return op_code != OP_NOT_HSPACE;
2870        }
2871    
2872      case OP_ANYNL:
2873      case OP_VSPACE:
2874      case OP_NOT_VSPACE:
2875      switch(next)
2876        {
2877        case 0x0a:
2878        case 0x0b:
2879        case 0x0c:
2880        case 0x0d:
2881        case 0x85:
2882        case 0x2028:
2883        case 0x2029:
2884        return op_code == OP_NOT_VSPACE;
2885        default:
2886        return op_code != OP_NOT_VSPACE;
2887        }
2888    
2889    #ifdef SUPPORT_UCP
2890      case OP_PROP:
2891      return check_char_prop(next, previous[0], previous[1], FALSE);
2892    
2893      case OP_NOTPROP:
2894      return check_char_prop(next, previous[0], previous[1], TRUE);
2895    #endif
2896    
2897      default:
2898      return FALSE;
2899      }
2900    
2901    
2902    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2903    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2904    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2905    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2906    replaced by OP_PROP codes when PCRE_UCP is set. */
2907    
2908    switch(op_code)
2909      {
2910      case OP_CHAR:
2911      case OP_CHARI:
2912    #ifdef SUPPORT_UTF8
2913      GETCHARTEST(c, previous);
2914    #else
2915      c = *previous;
2916    #endif
2917      switch(-next)
2918        {
2919        case ESC_d:
2920        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2921    
2922        case ESC_D:
2923        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2924    
2925        case ESC_s:
2926        return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2927    
2928        case ESC_S:
2929        return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2930    
2931        case ESC_w:
2932        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2933    
2934        case ESC_W:
2935        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2936    
2937        case ESC_h:
2938        case ESC_H:
2939        switch(c)
2940          {
2941          case 0x09:
2942          case 0x20:
2943          case 0xa0:
2944          case 0x1680:
2945          case 0x180e:
2946          case 0x2000:
2947          case 0x2001:
2948          case 0x2002:
2949          case 0x2003:
2950          case 0x2004:
2951          case 0x2005:
2952          case 0x2006:
2953          case 0x2007:
2954          case 0x2008:
2955          case 0x2009:
2956          case 0x200A:
2957          case 0x202f:
2958          case 0x205f:
2959          case 0x3000:
2960          return -next != ESC_h;
2961          default:
2962          return -next == ESC_h;
2963          }
2964    
2965        case ESC_v:
2966        case ESC_V:
2967        switch(c)
2968          {
2969          case 0x0a:
2970          case 0x0b:
2971          case 0x0c:
2972          case 0x0d:
2973          case 0x85:
2974          case 0x2028:
2975          case 0x2029:
2976          return -next != ESC_v;
2977          default:
2978          return -next == ESC_v;
2979          }
2980    
2981        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2982        their substitutions and process them. The result will always be either
2983        -ESC_p or -ESC_P. Then fall through to process those values. */
2984    
2985    #ifdef SUPPORT_UCP
2986        case ESC_du:
2987        case ESC_DU:
2988        case ESC_wu:
2989        case ESC_WU:
2990        case ESC_su:
2991        case ESC_SU:
2992          {
2993          int temperrorcode = 0;
2994          ptr = substitutes[-next - ESC_DU];
2995          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2996          if (temperrorcode != 0) return FALSE;
2997          ptr++;    /* For compatibility */
2998          }
2999        /* Fall through */
3000    
3001        case ESC_p:
3002        case ESC_P:
3003          {
3004          int ptype, pdata, errorcodeptr;
3005          BOOL negated;
3006    
3007          ptr--;      /* Make ptr point at the p or P */
3008          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3009          if (ptype < 0) return FALSE;
3010          ptr++;      /* Point past the final curly ket */
3011    
3012          /* If the property item is optional, we have to give up. (When generated
3013          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3014          to the original \d etc. At this point, ptr will point to a zero byte. */
3015    
3016          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3017            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3018              return FALSE;
3019    
3020          /* Do the property check. */
3021    
3022          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3023          }
3024    #endif
3025    
3026        default:
3027        return FALSE;
3028        }
3029    
3030      /* In principle, support for Unicode properties should be integrated here as
3031      well. It means re-organizing the above code so as to get hold of the property
3032      values before switching on the op-code. However, I wonder how many patterns
3033      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3034      these op-codes are never generated.) */
3035    
3036      case OP_DIGIT:
3037      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3038             next == -ESC_h || next == -ESC_v || next == -ESC_R;
3039    
3040      case OP_NOT_DIGIT:
3041      return next == -ESC_d;
3042    
3043      case OP_WHITESPACE:
3044      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3045    
3046      case OP_NOT_WHITESPACE:
3047      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3048    
3049      case OP_HSPACE:
3050      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3051             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3052    
3053      case OP_NOT_HSPACE:
3054      return next == -ESC_h;
3055    
3056      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3057      case OP_ANYNL:
3058      case OP_VSPACE:
3059      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3060    
3061      case OP_NOT_VSPACE:
3062      return next == -ESC_v || next == -ESC_R;
3063    
3064      case OP_WORDCHAR:
3065      return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3066             next == -ESC_v || next == -ESC_R;
3067    
3068      case OP_NOT_WORDCHAR:
3069      return next == -ESC_w || next == -ESC_d;
3070    
3071      default:
3072      return FALSE;
3073      }
3074    
3075    /* Control does not reach here */
3076    }
3077    
3078    
3079    
3080  /*************************************************  /*************************************************
3081  *           Compile one branch                   *  *           Compile one branch                   *
3082  *************************************************/  *************************************************/
3083    
3084  /* Scan the pattern, compiling it into the code vector. If the options are  /* Scan the pattern, compiling it into the a vector. If the options are
3085  changed during the branch, the pointer is used to change the external options  changed during the branch, the pointer is used to change the external options
3086  bits.  bits. This function is used during the pre-compile phase when we are trying
3087    to find out the amount of memory needed, as well as during the real compile
3088    phase. The value of lengthptr distinguishes the two phases.
3089    
3090  Arguments:  Arguments:
3091    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
3092    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3093    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3094    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3095    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3096    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3097    bcptr          points to current branch chain    bcptr          points to current branch chain
3098      cond_depth     conditional nesting depth
3099    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3100      lengthptr      NULL during the real compile phase
3101                     points to length accumulator during pre-compile phase
3102    
3103  Returns:         TRUE on success  Returns:         TRUE on success
3104                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
3105  */  */
3106    
3107  static BOOL  static BOOL
3108  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3109    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3110    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    int cond_depth, compile_data *cd, int *lengthptr)
3111  {  {
3112  int repeat_type, op_type;  int repeat_type, op_type;
3113  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1611  int greedy_default, greedy_non_default; Line 3116  int greedy_default, greedy_non_default;
3116  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3117  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3118  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3119  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3120  int after_manual_callout = 0;  int after_manual_callout = 0;
3121    int length_prevgroup = 0;
3122  register int c;  register int c;
3123  register uschar *code = *codeptr;  register uschar *code = *codeptr;
3124    uschar *last_code = code;
3125    uschar *orig_code = code;
3126  uschar *tempcode;  uschar *tempcode;
3127  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3128  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3129  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3130  const uschar *tempptr;  const uschar *tempptr;
3131    const uschar *nestptr = NULL;
3132  uschar *previous = NULL;  uschar *previous = NULL;
3133  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3134    uschar *save_hwm = NULL;
3135  uschar classbits[32];  uschar classbits[32];
3136    
3137    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3138    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3139    dynamically as we process the pattern. */
3140    
3141  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3142  BOOL class_utf8;  BOOL class_utf8;
3143  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3144  uschar *class_utf8data;  uschar *class_utf8data;
3145    uschar *class_utf8data_base;
3146  uschar utf8_char[6];  uschar utf8_char[6];
3147  #else  #else
3148  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
3149  #endif  #endif
3150    
3151    #ifdef PCRE_DEBUG
3152    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3153    #endif
3154    
3155  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
3156    
3157  greedy_default = ((options & PCRE_UNGREEDY) != 0);  greedy_default = ((options & PCRE_UNGREEDY) != 0);
# Line 1662  req_caseopt = ((options & PCRE_CASELESS) Line 3181  req_caseopt = ((options & PCRE_CASELESS)
3181  for (;; ptr++)  for (;; ptr++)
3182    {    {
3183    BOOL negate_class;    BOOL negate_class;
3184      BOOL should_flip_negation;
3185    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3186    BOOL is_quantifier;    BOOL is_quantifier;
3187      BOOL is_recurse;
3188      BOOL reset_bracount;
3189    int class_charcount;    int class_charcount;
3190    int class_lastchar;    int class_lastchar;
3191    int newoptions;    int newoptions;
3192    int recno;    int recno;
3193      int refsign;
3194    int skipbytes;    int skipbytes;
3195    int subreqbyte;    int subreqbyte;
3196    int subfirstbyte;    int subfirstbyte;
3197      int terminator;
3198    int mclength;    int mclength;
3199      int tempbracount;
3200    uschar mcbuffer[8];    uschar mcbuffer[8];
3201    
3202    /* Next byte in the pattern */    /* Get next byte in the pattern */
3203    
3204    c = *ptr;    c = *ptr;
3205    
3206      /* If we are at the end of a nested substitution, revert to the outer level
3207      string. Nesting only happens one level deep. */
3208    
3209      if (c == 0 && nestptr != NULL)
3210        {
3211        ptr = nestptr;
3212        nestptr = NULL;
3213        c = *ptr;
3214        }
3215    
3216      /* If we are in the pre-compile phase, accumulate the length used for the
3217      previous cycle of this loop. */
3218    
3219      if (lengthptr != NULL)
3220        {
3221    #ifdef PCRE_DEBUG
3222        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3223    #endif
3224        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3225          {
3226          *errorcodeptr = ERR52;
3227          goto FAILED;
3228          }
3229    
3230        /* There is at least one situation where code goes backwards: this is the
3231        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3232        the class is simply eliminated. However, it is created first, so we have to
3233        allow memory for it. Therefore, don't ever reduce the length at this point.
3234        */
3235    
3236        if (code < last_code) code = last_code;
3237    
3238        /* Paranoid check for integer overflow */
3239    
3240        if (OFLOW_MAX - *lengthptr < code - last_code)
3241          {
3242          *errorcodeptr = ERR20;
3243          goto FAILED;
3244          }
3245    
3246        *lengthptr += (int)(code - last_code);
3247        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3248    
3249        /* If "previous" is set and it is not at the start of the work space, move
3250        it back to there, in order to avoid filling up the work space. Otherwise,
3251        if "previous" is NULL, reset the current code pointer to the start. */
3252    
3253        if (previous != NULL)
3254          {
3255          if (previous > orig_code)
3256            {
3257            memmove(orig_code, previous, code - previous);
3258            code -= previous - orig_code;
3259            previous = orig_code;
3260            }
3261          }
3262        else code = orig_code;
3263    
3264        /* Remember where this code item starts so we can pick up the length
3265        next time round. */
3266    
3267        last_code = code;
3268        }
3269    
3270      /* In the real compile phase, just check the workspace used by the forward
3271      reference list. */
3272    
3273      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3274        {
3275        *errorcodeptr = ERR52;
3276        goto FAILED;
3277        }
3278    
3279    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
3280    
3281    if (inescq && c != 0)    if (inescq && c != 0)
3282      {      {
3283      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3284        {        {
3285        inescq = FALSE;        inescq = FALSE;
3286        ptr++;        ptr++;
# Line 1692  for (;; ptr++) Line 3290  for (;; ptr++)
3290        {        {
3291        if (previous_callout != NULL)        if (previous_callout != NULL)
3292          {          {
3293          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3294              complete_callout(previous_callout, ptr, cd);
3295          previous_callout = NULL;          previous_callout = NULL;
3296          }          }
3297        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1707  for (;; ptr++) Line 3306  for (;; ptr++)
3306    /* 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
3307    a quantifier. */    a quantifier. */
3308    
3309    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3310      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3311        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3312    
3313    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3314         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
3315      {      {
3316      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3317          complete_callout(previous_callout, ptr, cd);
3318      previous_callout = NULL;      previous_callout = NULL;
3319      }      }
3320    
3321    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3322    
3323    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3324      {      {
3325      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3326      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3327        {        {
3328        while (*(++ptr) != 0) if (IS_NEWLINE(ptr)) break;        ptr++;
3329        if (*ptr != 0)        while (*ptr != 0)
3330          {          {
3331          ptr += cd->nllen - 1;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3332          continue;          ptr++;
3333    #ifdef SUPPORT_UTF8
3334            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3335    #endif
3336          }          }
3337          if (*ptr != 0) continue;
3338    
3339        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
3340        c = 0;        c = 0;
3341        }        }
# Line 1745  for (;; ptr++) Line 3351  for (;; ptr++)
3351    
3352    switch(c)    switch(c)
3353      {      {
3354      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
3355        case 0:                        /* The branch terminates at string end */
3356      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
3357      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
3358      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3359      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3360      *codeptr = code;      *codeptr = code;
3361      *ptrptr = ptr;      *ptrptr = ptr;
3362        if (lengthptr != NULL)
3363          {
3364          if (OFLOW_MAX - *lengthptr < code - last_code)
3365            {
3366            *errorcodeptr = ERR20;
3367            goto FAILED;
3368            }
3369          *lengthptr += (int)(code - last_code);   /* To include callout length */
3370          DPRINTF((">> end branch\n"));
3371          }
3372      return TRUE;      return TRUE;
3373    
3374    
3375        /* ===================================================================*/
3376      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3377      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3378    
3379      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3380        previous = NULL;
3381      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3382        {        {
3383        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3384          *code++ = OP_CIRCM;
3385        }        }
3386      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3387      break;      break;
3388    
3389      case '$':      case CHAR_DOLLAR_SIGN:
3390      previous = NULL;      previous = NULL;
3391      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3392      break;      break;
3393    
3394      /* 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
3395      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3396    
3397      case '.':      case CHAR_DOT:
3398      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3399      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3400      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3401      previous = code;      previous = code;
3402      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3403      break;      break;
3404    
3405    
3406        /* ===================================================================*/
3407      /* Character classes. If the included characters are all < 256, we build a      /* Character classes. If the included characters are all < 256, we build a
3408      32-byte bitmap of the permitted characters, except in the special case      32-byte bitmap of the permitted characters, except in the special case
3409      where there is only one such character. For negated classes, we build the      where there is only one such character. For negated classes, we build the
# Line 1794  for (;; ptr++) Line 3414  for (;; ptr++)
3414      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,
3415      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3416      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.
     */  
3417    
3418      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3419        default (Perl) mode, it is treated as a data character. */
3420    
3421        case CHAR_RIGHT_SQUARE_BRACKET:
3422        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3423          {
3424          *errorcodeptr = ERR64;
3425          goto FAILED;
3426          }
3427        goto NORMAL_CHAR;
3428    
3429        case CHAR_LEFT_SQUARE_BRACKET:
3430      previous = code;      previous = code;
3431    
3432      /* 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
3433      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. */
3434    
3435      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3436          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3437            check_posix_syntax(ptr, &tempptr))
3438        {        {
3439        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3440        goto FAILED;        goto FAILED;
3441        }        }
3442    
3443      /* 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,
3444        if the first few characters (either before or after ^) are \Q\E or \E we
3445        skip them too. This makes for compatibility with Perl. */
3446    
3447      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3448        for (;;)
3449        {        {
       negate_class = TRUE;  
3450        c = *(++ptr);        c = *(++ptr);
3451          if (c == CHAR_BACKSLASH)
3452            {
3453            if (ptr[1] == CHAR_E)
3454              ptr++;
3455            else if (strncmp((const char *)ptr+1,
3456                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3457              ptr += 3;
3458            else
3459              break;
3460            }
3461          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3462            negate_class = TRUE;
3463          else break;
3464        }        }
3465      else  
3466        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3467        an initial ']' is taken as a data character -- the code below handles
3468        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3469        [^] must match any character, so generate OP_ALLANY. */
3470    
3471        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3472            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3473        {        {
3474        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3475          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3476          zerofirstbyte = firstbyte;
3477          break;
3478        }        }
3479    
3480        /* If a class contains a negative special such as \S, we need to flip the
3481        negation flag at the end, so that support for characters > 255 works
3482        correctly (they are all included in the class). */
3483    
3484        should_flip_negation = FALSE;
3485    
3486      /* 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
3487      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
3488      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3489    
3490      class_charcount = 0;      class_charcount = 0;
3491      class_lastchar = -1;      class_lastchar = -1;
3492    
3493        /* Initialize the 32-char bit map to all zeros. We build the map in a
3494        temporary bit of memory, in case the class contains only 1 character (less
3495        than 256), because in that case the compiled code doesn't use the bit map.
3496        */
3497    
3498        memset(classbits, 0, 32 * sizeof(uschar));
3499    
3500  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3501      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3502      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3503        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3504  #endif  #endif
3505    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
3506      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3507      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
3508      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3509    
3510      do      if (c != 0) do
3511        {        {
3512          const uschar *oldptr;
3513    
3514  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3515        if (utf8 && c > 127)        if (utf8 && c > 127)
3516          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3517          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3518          }          }
3519    
3520          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3521          data and reset the pointer. This is so that very large classes that
3522          contain a zillion UTF-8 characters no longer overwrite the work space
3523          (which is on the stack). */
3524    
3525          if (lengthptr != NULL)
3526            {
3527            *lengthptr += class_utf8data - class_utf8data_base;
3528            class_utf8data = class_utf8data_base;
3529            }
3530    
3531  #endif  #endif
3532    
3533        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3534    
3535        if (inescq)        if (inescq)
3536          {          {
3537          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3538            {            {
3539            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3540            ptr++;            ptr++;                            /* Skip the 'E' */
3541            continue;            continue;                         /* Carry on with next */
3542            }            }
3543          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3544          }          }
3545    
3546        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1874  for (;; ptr++) Line 3549  for (;; ptr++)
3549        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3550        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3551    
3552        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3553            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3554            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3555          {          {
3556          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3557          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3558          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3559          uschar pbits[32];          uschar pbits[32];
3560    
3561          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3562            {            {
3563            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3564            goto FAILED;            goto FAILED;
3565            }            }
3566    
3567          ptr += 2;          ptr += 2;
3568          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3569            {            {
3570            local_negate = TRUE;            local_negate = TRUE;
3571              should_flip_negation = TRUE;  /* Note negative special */
3572            ptr++;            ptr++;
3573            }            }
3574    
3575          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3576          if (posix_class < 0)          if (posix_class < 0)
3577            {            {
3578            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 1910  for (;; ptr++) Line 3586  for (;; ptr++)
3586          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3587            posix_class = 0;            posix_class = 0;
3588    
3589          /* 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
3590          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3591          subtract bits that may be in the main map already. At the end we or the  
3592          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3593            if ((options & PCRE_UCP) != 0)
3594              {
3595              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3596              if (posix_substitutes[pc] != NULL)
3597                {
3598                nestptr = tempptr + 1;
3599                ptr = posix_substitutes[pc] - 1;
3600                continue;
3601                }
3602              }
3603    #endif
3604            /* In the non-UCP case, we build the bit map for the POSIX class in a
3605            chunk of local store because we may be adding and subtracting from it,
3606            and we don't want to subtract bits that may be in the main map already.
3607            At the end we or the result into the bit map that is being built. */
3608    
3609          posix_class *= 3;          posix_class *= 3;
3610    
# Line 1956  for (;; ptr++) Line 3647  for (;; ptr++)
3647          }          }
3648    
3649        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3650        of the specials, which just set a flag. Escaped items are checked for        of the specials, which just set a flag. The sequence \b is a special
3651        validity in the pre-compiling pass. The sequence \b is a special case.        case. Inside a class (and only there) it is treated as backspace. We
3652        Inside a class (and only there) it is treated as backspace. Elsewhere        assume that other escapes have more than one character in them, so set
3653        it marks a word boundary. Other escapes have preset maps ready to        class_charcount bigger than one. Unrecognized escapes fall through and
3654        or into the one we are building. We assume they have more than one        are either treated as literal characters (by default), or are faulted if
3655        character in them, so set class_charcount bigger than one. */        PCRE_EXTRA is set. */
3656    
3657        if (c == '\\')        if (c == CHAR_BACKSLASH)
3658          {          {
3659          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3660            if (*errorcodeptr != 0) goto FAILED;
3661    
3662          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 */  
3663          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3664            {            {
3665            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3666              {              {
3667              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3668              }              }
3669            else inescq = TRUE;            else inescq = TRUE;
3670            continue;            continue;
3671            }            }
3672            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3673    
3674          if (c < 0)          if (c < 0)
3675            {            {
3676            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3677            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3678    
3679            switch (-c)            switch (-c)
3680              {              {
3681    #ifdef SUPPORT_UCP
3682                case ESC_du:     /* These are the values given for \d etc */
3683                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3684                case ESC_wu:     /* escape sequence with an appropriate \p */
3685                case ESC_WU:     /* or \P to test Unicode properties instead */
3686                case ESC_su:     /* of the default ASCII testing. */
3687                case ESC_SU:
3688                nestptr = ptr;
3689                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3690                class_charcount -= 2;                /* Undo! */
3691                continue;
3692    #endif
3693              case ESC_d:              case ESC_d:
3694              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3695              continue;              continue;
3696    
3697              case ESC_D:              case ESC_D:
3698                should_flip_negation = TRUE;
3699              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3700              continue;              continue;
3701    
# Line 1998  for (;; ptr++) Line 3704  for (;; ptr++)
3704              continue;              continue;
3705    
3706              case ESC_W:              case ESC_W:
3707                should_flip_negation = TRUE;
3708              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3709              continue;              continue;
3710    
3711                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3712                if it was previously set by something earlier in the character
3713                class. */
3714    
3715              case ESC_s:              case ESC_s:
3716              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3717              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3718                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3719              continue;              continue;
3720    
3721              case ESC_S:              case ESC_S:
3722                should_flip_negation = TRUE;
3723              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3724              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3725              continue;              continue;
3726    
3727                case ESC_h:
3728                SETBIT(classbits, 0x09); /* VT */
3729                SETBIT(classbits, 0x20); /* SPACE */
3730                SETBIT(classbits, 0xa0); /* NSBP */
3731    #ifdef SUPPORT_UTF8
3732                if (utf8)
3733                  {
3734                  class_utf8 = TRUE;
3735                  *class_utf8data++ = XCL_SINGLE;
3736                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3737                  *class_utf8data++ = XCL_SINGLE;
3738                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3739                  *class_utf8data++ = XCL_RANGE;
3740                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3741                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3742                  *class_utf8data++ = XCL_SINGLE;
3743                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3744                  *class_utf8data++ = XCL_SINGLE;
3745                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3746                  *class_utf8data++ = XCL_SINGLE;
3747                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3748                  }
3749    #endif
3750                continue;
3751    
3752                case ESC_H:
3753                for (c = 0; c < 32; c++)
3754                  {
3755                  int x = 0xff;
3756                  switch (c)
3757                    {
3758                    case 0x09/8: x ^= 1 << (0x09%8); break;
3759                    case 0x20/8: x ^= 1 << (0x20%8); break;
3760                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3761                    default: break;
3762                    }
3763                  classbits[c] |= x;
3764                  }
3765    
3766    #ifdef SUPPORT_UTF8
3767                if (utf8)
3768                  {
3769                  class_utf8 = TRUE;
3770                  *class_utf8data++ = XCL_RANGE;
3771                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3772                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3773                  *class_utf8data++ = XCL_RANGE;
3774                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3775                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3776                  *class_utf8data++ = XCL_RANGE;
3777                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3778                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3779                  *class_utf8data++ = XCL_RANGE;
3780                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3781                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3782                  *class_utf8data++ = XCL_RANGE;
3783                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3784                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3785                  *class_utf8data++ = XCL_RANGE;
3786                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3787                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3788                  *class_utf8data++ = XCL_RANGE;
3789                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3790                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3791                  }
3792    #endif
3793                continue;
3794    
3795                case ESC_v:
3796                SETBIT(classbits, 0x0a); /* LF */
3797                SETBIT(classbits, 0x0b); /* VT */
3798                SETBIT(classbits, 0x0c); /* FF */
3799                SETBIT(classbits, 0x0d); /* CR */
3800                SETBIT(classbits, 0x85); /* NEL */
3801    #ifdef SUPPORT_UTF8
3802                if (utf8)
3803                  {
3804                  class_utf8 = TRUE;
3805                  *class_utf8data++ = XCL_RANGE;
3806                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3807                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3808                  }
3809    #endif
3810                continue;
3811    
3812                case ESC_V:
3813                for (c = 0; c < 32; c++)
3814                  {
3815                  int x = 0xff;
3816                  switch (c)
3817                    {
3818                    case 0x0a/8: x ^= 1 << (0x0a%8);
3819                                 x ^= 1 << (0x0b%8);
3820                                 x ^= 1 << (0x0c%8);
3821                                 x ^= 1 << (0x0d%8);
3822                                 break;
3823                    case 0x85/8: x ^= 1 << (0x85%8); break;
3824                    default: break;
3825                    }
3826                  classbits[c] |= x;
3827                  }
3828    
3829    #ifdef SUPPORT_UTF8
3830                if (utf8)
3831                  {
3832                  class_utf8 = TRUE;
3833                  *class_utf8data++ = XCL_RANGE;
3834                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3835                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3836                  *class_utf8data++ = XCL_RANGE;
3837                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3838                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3839                  }
3840    #endif
3841                continue;
3842    
3843  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3844              case ESC_p:              case ESC_p:
3845              case ESC_P:              case ESC_P:
# Line 2025  for (;; ptr++) Line 3854  for (;; ptr++)
3854                *class_utf8data++ = ptype;                *class_utf8data++ = ptype;
3855                *class_utf8data++ = pdata;                *class_utf8data++ = pdata;
3856                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
3857                  continue;
3858                }                }
             continue;  
3859  #endif  #endif
   
3860              /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3861              strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3862              treated as literals. */              treated as literals. */
# Line 2039  for (;; ptr++) Line 3867  for (;; ptr++)
3867                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
3868                goto FAILED;                goto FAILED;
3869                }                }
             c = *ptr;              /* The final character */  
3870              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3871                c = *ptr;              /* Get the final character and fall through */
3872                break;
3873              }              }
3874            }            }
3875    
3876          /* 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
3877          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3878    
3879          }   /* End of backslash handling */          }   /* End of backslash handling */
3880    
3881        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3882        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3883        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3884          entirely. The code for handling \Q and \E is messy. */
3885    
3886          CHECK_RANGE:
3887          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3888            {
3889            inescq = FALSE;
3890            ptr += 2;
3891            }
3892    
3893          oldptr = ptr;
3894    
3895        if (ptr[1] == '-' && ptr[2] != ']')        /* Remember \r or \n */
3896    
3897          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3898    
3899          /* Check for range */
3900    
3901          if (!inescq && ptr[1] == CHAR_MINUS)
3902          {          {
3903          int d;          int d;
3904          ptr += 2;          ptr += 2;
3905            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3906    
3907            /* If we hit \Q (not followed by \E) at this point, go into escaped
3908            mode. */
3909    
3910            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3911              {
3912              ptr += 2;
3913              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3914                { ptr += 2; continue; }
3915              inescq = TRUE;
3916              break;
3917              }
3918    
3919            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3920              {
3921              ptr = oldptr;
3922              goto LONE_SINGLE_CHARACTER;
3923              }
3924    
3925  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3926          if (utf8)          if (utf8)
# Line 2071  for (;; ptr++) Line 3935  for (;; ptr++)
3935          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
3936          in such circumstances. */          in such circumstances. */
3937    
3938          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3939            {            {
3940            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3941            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3942    
3943            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; any other special means the '-' was literal */
           was literal */  
3944    
3945            if (d < 0)            if (d < 0)
3946              {              {
3947              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X'; else  
3948                {                {
3949                ptr = oldptr - 2;                ptr = oldptr;
3950                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3951                }                }
3952              }              }
3953            }            }
3954    
3955          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3956          the pre-pass. Optimize one-character ranges */          one-character ranges */
3957    
3958            if (d < c)
3959              {
3960              *errorcodeptr = ERR8;
3961              goto FAILED;
3962              }
3963    
3964          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3965    
3966            /* Remember \r or \n */
3967    
3968            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3969    
3970          /* 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
3971          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3972          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2112  for (;; ptr++) Line 3984  for (;; ptr++)
3984  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3985            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3986              {              {
3987              int occ, ocd;              unsigned int occ, ocd;
3988              int cc = c;              unsigned int cc = c;
3989              int origd = d;              unsigned int origd = d;
3990              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3991                {                {
3992                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3993                      ocd <= (unsigned int)d)
3994                    continue;                          /* Skip embedded ranges */
3995    
3996                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3997                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3998                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3999                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
4000                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
4001                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
4002                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
4003                      occ <= (unsigned int)d + 1)      /* always shorter than    */
4004                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
4005                  d = ocd;                  d = ocd;
4006                  continue;                  continue;
# Line 2172  for (;; ptr++) Line 4048  for (;; ptr++)
4048          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
4049          for partial ranges without UCP support. */          for partial ranges without UCP support. */
4050    
4051          for (; c <= d; c++)          class_charcount += d - c + 1;
4052            class_lastchar = d;
4053    
4054            /* We can save a bit of time by skipping this in the pre-compile. */
4055    
4056            if (lengthptr == NULL) for (; c <= d; c++)
4057            {            {
4058            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4059            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2180  for (;; ptr++) Line 4061  for (;; ptr++)
4061              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
4062              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
4063              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
4064            }            }
4065    
4066          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2205  for (;; ptr++) Line 4084  for (;; ptr++)
4084  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4085          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4086            {            {
4087            int othercase;            unsigned int othercase;
4088            if ((othercase = _pcre_ucp_othercase(c)) >= 0)            if ((othercase = UCD_OTHERCASE(c)) != c)
4089              {              {
4090              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4091              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2231  for (;; ptr++) Line 4110  for (;; ptr++)
4110          }          }
4111        }        }
4112    
4113      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4114      loop. This "while" is the end of the "do" above. */      If we are at the end of an internal nested string, revert to the outer
4115        string. */
4116    
4117        while (((c = *(++ptr)) != 0 ||
4118               (nestptr != NULL &&
4119                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4120               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4121    
4122      while ((c = *(++ptr)) != ']' || inescq);      /* Check for missing terminating ']' */
4123    
4124        if (c == 0)
4125          {
4126          *errorcodeptr = ERR6;
4127          goto FAILED;
4128          }
4129    
4130      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4131      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
4132      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
4133      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4134      single-bytes only. This is an historical hangover. Maybe one day we can  
4135      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4136        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4137        operate on single-bytes characters only. This is an historical hangover.
4138        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4139    
4140      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4141      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4142      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4143      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4144      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4145      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4146    
4147  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4148      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4149            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4150  #else  #else
4151      if (class_charcount == 1)      if (class_charcount == 1)
4152  #endif  #endif
4153        {        {
4154        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4155    
4156        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4157    
4158        if (negate_class)        if (negate_class)
4159          {          {
4160          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4161          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4162          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4163          *code++ = class_lastchar;          *code++ = class_lastchar;
4164          break;          break;
4165          }          }
# Line 2297  for (;; ptr++) Line 4189  for (;; ptr++)
4189      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4190    
4191      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4192      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4193      we can omit the bitmap. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
4194        characters > 255 are in the class, so any that were explicitly given as
4195        well can be ignored. If (when there are explicit characters > 255 that must
4196        be listed) there are no characters < 256, we can omit the bitmap in the
4197        actual compiled code. */
4198    
4199  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4200      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4201        {        {
4202        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4203        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4204        code += LINK_SIZE;        code += LINK_SIZE;
4205        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
4206    
4207        /* If the map is required, install it, and move on to the end of        /* If the map is required, move up the extra data to make room for it;
4208        the extra data */        otherwise just move the code pointer to the end of the extra data. */
4209    
4210        if (class_charcount > 0)        if (class_charcount > 0)
4211          {          {
4212