/[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 87 by nigel, Sat Feb 24 21:41:21 2007 UTC revision 556 by ph10, Tue Oct 26 11:06:44 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
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 72  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 96  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 115  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 154  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 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 groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "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"
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      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 235  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 271  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 285  static const unsigned char digitab[] = Line 483  static const unsigned char digitab[] =
483    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 319  static const unsigned char ebcdic_charta Line 517  static const unsigned char ebcdic_charta
517    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
518    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
519    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
520    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
522    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 346  static const unsigned char ebcdic_charta Line 544  static const unsigned char ebcdic_charta
544  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
545    
546  static BOOL  static BOOL
547    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
548      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
549    
550    
551    
552    /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578  /*************************************************  /*************************************************
# Line 357  static BOOL Line 581  static BOOL
581    
582  /* 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
583  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
584  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
585  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
586  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,
587    ptr is pointing at the \. On exit, it is on the final character of the escape
588    sequence.
589    
590  Arguments:  Arguments:
591    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 370  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
# Line 388  ptr--;                            /* Set Line 614  ptr--;                            /* Set
614    
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
618  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 406  else if ((i = escapes[c - 0x48]) != 0) Line 632  else if ((i = escapes[c - 0x48]) != 0)
632  else  else
633    {    {
634    const uschar *oldptr;    const uschar *oldptr;
635      BOOL braced, negated;
636    
637    switch (c)    switch (c)
638      {      {
639      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649        /* \g must be followed by one of a number of specific things:
650    
651        (1) A number, either plain or braced. If positive, it is an absolute
652        backreference. If negative, it is a relative backreference. This is a Perl
653        5.10 feature.
654    
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665        case CHAR_g:
666        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675          {
676          const uschar *p;
677          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680            {
681            c = -ESC_k;
682            break;
683            }
684          braced = TRUE;
685          ptr++;
686          }
687        else braced = FALSE;
688    
689        if (ptr[1] == CHAR_MINUS)
690          {
691          negated = TRUE;
692          ptr++;
693          }
694        else negated = FALSE;
695    
696        c = 0;
697        while ((digitab[ptr[1]] & ctype_digit) != 0)
698          c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707          {
708          *errorcodeptr = ERR57;
709          break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718        if (negated)
719          {
720          if (c > bracount)
721            {
722            *errorcodeptr = ERR15;
723            break;
724            }
725          c = bracount - (c - 1);
726          }
727    
728        c = -(ESC_REF + c);
729        break;
730    
731      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
732      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
733      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 431  else Line 740  else
740      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 452  else Line 766  else
766      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 460  else Line 774  else
774        }        }
775    
776      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
777      larger first octal digit. */      larger first octal digit. The original code used just to take the least
778        significant 8 bits of octal numbers (I think this is what early Perls used
779      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
780      c -= '0';      than 3 octal digits. */
781      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
782          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
783      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
784        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785            c = c * 8 + *(++ptr) - CHAR_0;
786        if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
789      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
790      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 483  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 511  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
842    
843      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
844        This coding is ASCII-specific, but then the whole concept of \cx is
845        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
     if (c >= 'a' && c <= 'z') c -= 32;  
857      c ^= 0x40;      c ^= 0x40;
858  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
867      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 563  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 603  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
940      }      }
941    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 639  top = _pcre_utt_size; Line 967  top = _pcre_utt_size;
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top) >> 1;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0)    if (c == 0)
972      {      {
973      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 682  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 723  int max = -1; Line 1051  int max = -1;
1051  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 733  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 763  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094    *  Subroutine for finding forward reference      *
1095    *************************************************/
1096    
1097    /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100    subpatterns, and counting them. If it finds a named pattern that matches the
1101    name it is given, it returns its number. Alternatively, if the name is NULL, it
1102    returns when it reaches a given numbered subpattern. We know that if (?P< is
1103    encountered, the name will be terminated by '>' because that is checked in the
1104    first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107    Arguments:
1108      ptrptr       address of the current character pointer (updated)
1109      cd           compile background data
1110      name         name to seek, or NULL if seeking a numbered subpattern
1111      lorn         name length, or subpattern number if name is NULL
1112      xmode        TRUE if we are in /x mode
1113      utf8         TRUE if we are in UTF-8 mode
1114      count        pointer to the current capturing subpattern number (updated)
1115    
1116    Returns:       the number of the named subpattern, or -1 if not found
1117    */
1118    
1119    static int
1120    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1121      BOOL xmode, BOOL utf8, int *count)
1122    {
1123    uschar *ptr = *ptrptr;
1124    int start_count = *count;
1125    int hwm_count = start_count;
1126    BOOL dup_parens = FALSE;
1127    
1128    /* If the first character is a parenthesis, check on the type of group we are
1129    dealing with. The very first call may not start with a parenthesis. */
1130    
1131    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1132      {
1133      /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1134    
1135      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1136    
1137      /* Handle a normal, unnamed capturing parenthesis. */
1138    
1139      else if (ptr[1] != CHAR_QUESTION_MARK)
1140        {
1141        *count += 1;
1142        if (name == NULL && *count == lorn) return *count;
1143        ptr++;
1144        }
1145    
1146      /* All cases now have (? at the start. Remember when we are in a group
1147      where the parenthesis numbers are duplicated. */
1148    
1149      else if (ptr[2] == CHAR_VERTICAL_LINE)
1150        {
1151        ptr += 3;
1152        dup_parens = TRUE;
1153        }
1154    
1155      /* Handle comments; all characters are allowed until a ket is reached. */
1156    
1157      else if (ptr[2] == CHAR_NUMBER_SIGN)
1158        {
1159        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1160        goto FAIL_EXIT;
1161        }
1162    
1163      /* Handle a condition. If it is an assertion, just carry on so that it
1164      is processed as normal. If not, skip to the closing parenthesis of the
1165      condition (there can't be any nested parens). */
1166    
1167      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1168        {
1169        ptr += 2;
1170        if (ptr[1] != CHAR_QUESTION_MARK)
1171          {
1172          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1173          if (*ptr != 0) ptr++;
1174          }
1175        }
1176    
1177      /* Start with (? but not a condition. */
1178    
1179      else
1180        {
1181        ptr += 2;
1182        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1183    
1184        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1185    
1186        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1187            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1188          {
1189          int term;
1190          const uschar *thisname;
1191          *count += 1;
1192          if (name == NULL && *count == lorn) return *count;
1193          term = *ptr++;
1194          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1195          thisname = ptr;
1196          while (*ptr != term) ptr++;
1197          if (name != NULL && lorn == ptr - thisname &&
1198              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1199            return *count;
1200          term++;
1201          }
1202        }
1203      }
1204    
1205    /* Past any initial parenthesis handling, scan for parentheses or vertical
1206    bars. */
1207    
1208    for (; *ptr != 0; ptr++)
1209      {
1210      /* Skip over backslashed characters and also entire \Q...\E */
1211    
1212      if (*ptr == CHAR_BACKSLASH)
1213        {
1214        if (*(++ptr) == 0) goto FAIL_EXIT;
1215        if (*ptr == CHAR_Q) for (;;)
1216          {
1217          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1218          if (*ptr == 0) goto FAIL_EXIT;
1219          if (*(++ptr) == CHAR_E) break;
1220          }
1221        continue;
1222        }
1223    
1224      /* Skip over character classes; this logic must be similar to the way they
1225      are handled for real. If the first character is '^', skip it. Also, if the
1226      first few characters (either before or after ^) are \Q\E or \E we skip them
1227      too. This makes for compatibility with Perl. Note the use of STR macros to
1228      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1229    
1230      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1231        {
1232        BOOL negate_class = FALSE;
1233        for (;;)
1234          {
1235          if (ptr[1] == CHAR_BACKSLASH)
1236            {
1237            if (ptr[2] == CHAR_E)
1238              ptr+= 2;
1239            else if (strncmp((const char *)ptr+2,
1240                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1241              ptr += 4;
1242            else
1243              break;
1244            }
1245          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1246            {
1247            negate_class = TRUE;
1248            ptr++;
1249            }
1250          else break;
1251          }
1252    
1253        /* If the next character is ']', it is a data character that must be
1254        skipped, except in JavaScript compatibility mode. */
1255    
1256        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1257            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1258          ptr++;
1259    
1260        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1261          {
1262          if (*ptr == 0) return -1;
1263          if (*ptr == CHAR_BACKSLASH)
1264            {
1265            if (*(++ptr) == 0) goto FAIL_EXIT;
1266            if (*ptr == CHAR_Q) for (;;)
1267              {
1268              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1269              if (*ptr == 0) goto FAIL_EXIT;
1270              if (*(++ptr) == CHAR_E) break;
1271              }
1272            continue;
1273            }
1274          }
1275        continue;
1276        }
1277    
1278      /* Skip comments in /x mode */
1279    
1280      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1281        {
1282        ptr++;
1283        while (*ptr != 0)
1284          {
1285          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1286          ptr++;
1287    #ifdef SUPPORT_UTF8
1288          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1289    #endif
1290          }
1291        if (*ptr == 0) goto FAIL_EXIT;
1292        continue;
1293        }
1294    
1295      /* Check for the special metacharacters */
1296    
1297      if (*ptr == CHAR_LEFT_PARENTHESIS)
1298        {
1299        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1300        if (rc > 0) return rc;
1301        if (*ptr == 0) goto FAIL_EXIT;
1302        }
1303    
1304      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1305        {
1306        if (dup_parens && *count < hwm_count) *count = hwm_count;
1307        goto FAIL_EXIT;
1308        }
1309    
1310      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1311        {
1312        if (*count > hwm_count) hwm_count = *count;
1313        *count = start_count;
1314        }
1315      }
1316    
1317    FAIL_EXIT:
1318    *ptrptr = ptr;
1319    return -1;
1320    }
1321    
1322    
1323    
1324    
1325    /*************************************************
1326    *       Find forward referenced subpattern       *
1327    *************************************************/
1328    
1329    /* This function scans along a pattern's text looking for capturing
1330    subpatterns, and counting them. If it finds a named pattern that matches the
1331    name it is given, it returns its number. Alternatively, if the name is NULL, it
1332    returns when it reaches a given numbered subpattern. This is used for forward
1333    references to subpatterns. We used to be able to start this scan from the
1334    current compiling point, using the current count value from cd->bracount, and
1335    do it all in a single loop, but the addition of the possibility of duplicate
1336    subpattern numbers means that we have to scan from the very start, in order to
1337    take account of such duplicates, and to use a recursive function to keep track
1338    of the different types of group.
1339    
1340    Arguments:
1341      cd           compile background data
1342      name         name to seek, or NULL if seeking a numbered subpattern
1343      lorn         name length, or subpattern number if name is NULL
1344      xmode        TRUE if we are in /x mode
1345      utf8         TRUE if we are in UTF-8 mode
1346    
1347    Returns:       the number of the found subpattern, or -1 if not found
1348    */
1349    
1350    static int
1351    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1352      BOOL utf8)
1353    {
1354    uschar *ptr = (uschar *)cd->start_pattern;
1355    int count = 0;
1356    int rc;
1357    
1358    /* If the pattern does not start with an opening parenthesis, the first call
1359    to find_parens_sub() will scan right to the end (if necessary). However, if it
1360    does start with a parenthesis, find_parens_sub() will return when it hits the
1361    matching closing parens. That is why we have to have a loop. */
1362    
1363    for (;;)
1364      {
1365      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1366      if (rc > 0 || *ptr++ == 0) break;
1367      }
1368    
1369    return rc;
1370    }
1371    
1372    
1373    
1374    
1375    /*************************************************
1376  *      Find first significant op code            *  *      Find first significant op code            *
1377  *************************************************/  *************************************************/
1378    
# Line 811  for (;;) Line 1421  for (;;)
1421    
1422      case OP_CALLOUT:      case OP_CALLOUT:
1423      case OP_CREF:      case OP_CREF:
1424      case OP_BRANUMBER:      case OP_NCREF:
1425        case OP_RREF:
1426        case OP_NRREF:
1427        case OP_DEF:
1428      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1429      break;      break;
1430    
# Line 826  for (;;) Line 1439  for (;;)
1439    
1440    
1441  /*************************************************  /*************************************************
1442  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1443  *************************************************/  *************************************************/
1444    
1445  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1446  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1447  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1448    temporarily terminated with OP_END when this function is called.
1449    
1450    This function is called when a backward assertion is encountered, so that if it
1451    fails, the error message can point to the correct place in the pattern.
1452    However, we cannot do this when the assertion contains subroutine calls,
1453    because they can be forward references. We solve this by remembering this case
1454    and doing the check at the end; a flag specifies which mode we are running in.
1455    
1456  Arguments:  Arguments:
1457    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1458    options  the compiling options    options  the compiling options
1459      atend    TRUE if called when the pattern is complete
1460      cd       the "compile data" structure
1461    
1462  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1463                 or -1 if there is no fixed length,
1464               or -2 if \C was encountered               or -2 if \C was encountered
1465                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1466  */  */
1467    
1468  static int  static int
1469  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1470  {  {
1471  int length = -1;  int length = -1;
1472    
# Line 855  branch, check the length against that of Line 1479  branch, check the length against that of
1479  for (;;)  for (;;)
1480    {    {
1481    int d;    int d;
1482      uschar *ce, *cs;
1483    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1484    switch (op)    switch (op)
1485      {      {
1486        case OP_CBRA:
1487      case OP_BRA:      case OP_BRA:
1488      case OP_ONCE:      case OP_ONCE:
1489      case OP_COND:      case OP_COND:
1490      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1491      if (d < 0) return d;      if (d < 0) return d;
1492      branchlength += d;      branchlength += d;
1493      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 886  for (;;) Line 1510  for (;;)
1510      branchlength = 0;      branchlength = 0;
1511      break;      break;
1512    
1513        /* A true recursion implies not fixed length, but a subroutine call may
1514        be OK. If the subroutine is a forward reference, we can't deal with
1515        it until the end of the pattern, so return -3. */
1516    
1517        case OP_RECURSE:
1518        if (!atend) return -3;
1519        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1520        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1521        if (cc > cs && cc < ce) return -1;                /* Recursion */
1522        d = find_fixedlength(cs + 2, options, atend, cd);
1523        if (d < 0) return d;
1524        branchlength += d;
1525        cc += 1 + LINK_SIZE;
1526        break;
1527    
1528      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1529    
1530      case OP_ASSERT:      case OP_ASSERT:
# Line 898  for (;;) Line 1537  for (;;)
1537      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1538    
1539      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1540      case OP_CREF:      case OP_CREF:
1541        case OP_NCREF:
1542        case OP_RREF:
1543        case OP_NRREF:
1544        case OP_DEF:
1545      case OP_OPT:      case OP_OPT:
1546      case OP_CALLOUT:      case OP_CALLOUT:
1547      case OP_SOD:      case OP_SOD:
1548      case OP_SOM:      case OP_SOM:
1549        case OP_SET_SOM:
1550      case OP_EOD:      case OP_EOD:
1551      case OP_EODN:      case OP_EODN:
1552      case OP_CIRC:      case OP_CIRC:
# Line 917  for (;;) Line 1560  for (;;)
1560    
1561      case OP_CHAR:      case OP_CHAR:
1562      case OP_CHARNC:      case OP_CHARNC:
1563        case OP_NOT:
1564      branchlength++;      branchlength++;
1565      cc += 2;      cc += 2;
1566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1567      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1568        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1569  #endif  #endif
1570      break;      break;
1571    
# Line 934  for (;;) Line 1576  for (;;)
1576      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1577      cc += 4;      cc += 4;
1578  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1579      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1580        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1581  #endif  #endif
1582      break;      break;
1583    
1584      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1585      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1586        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1587      cc += 4;      cc += 4;
1588      break;      break;
1589    
# Line 960  for (;;) Line 1601  for (;;)
1601      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1602      case OP_WORDCHAR:      case OP_WORDCHAR:
1603      case OP_ANY:      case OP_ANY:
1604        case OP_ALLANY:
1605      branchlength++;      branchlength++;
1606      cc++;      cc++;
1607      break;      break;
# Line 1014  for (;;) Line 1656  for (;;)
1656    
1657    
1658  /*************************************************  /*************************************************
1659  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1660  *************************************************/  *************************************************/
1661    
1662  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1663  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1664    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1665    so that it can be called from pcre_study() when finding the minimum matching
1666    length.
1667    
1668  Arguments:  Arguments:
1669    code        points to start of expression    code        points to start of expression
1670    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1671    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1672    
1673  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
1674  */  */
1675    
1676  static const uschar *  const uschar *
1677  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1678  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1679  for (;;)  for (;;)
1680    {    {
1681    register int c = *code;    register int c = *code;
1682    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1683    else if (c > OP_BRA)  
1684      /* XCLASS is used for classes that cannot be represented just by a bit
1685      map. This includes negated single high-valued characters. The length in
1686      the table is zero; the actual length is stored in the compiled code. */
1687    
1688      if (c == OP_XCLASS) code += GET(code, 1);
1689    
1690      /* Handle recursion */
1691    
1692      else if (c == OP_REVERSE)
1693      {      {
1694      int n = c - OP_BRA;      if (number < 0) return (uschar *)code;
1695      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      code += _pcre_OP_lengths[c];
     if (n == number) return (uschar *)code;  
     code += _pcre_OP_lengths[OP_BRA];  
1696      }      }
1697    else  
1698      /* Handle capturing bracket */
1699    
1700      else if (c == OP_CBRA)
1701      {      {
1702        int n = GET2(code, 1+LINK_SIZE);
1703        if (n == number) return (uschar *)code;
1704      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1705        }
1706    
1707  #ifdef SUPPORT_UTF8    /* Otherwise, we can get the item's length from the table, except that for
1708      repeated character types, we have to test for \p and \P, which have an extra
1709      /* In UTF-8 mode, opcodes that are followed by a character may be followed    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1710      by a multi-byte character. The length in the table is a minimum, so we have    must add in its length. */
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
1711    
1712      if (utf8) switch(c)    else
1713        {
1714        switch(c)
1715          {
1716          case OP_TYPESTAR:
1717          case OP_TYPEMINSTAR:
1718          case OP_TYPEPLUS:
1719          case OP_TYPEMINPLUS:
1720          case OP_TYPEQUERY:
1721          case OP_TYPEMINQUERY:
1722          case OP_TYPEPOSSTAR:
1723          case OP_TYPEPOSPLUS:
1724          case OP_TYPEPOSQUERY:
1725          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1726          break;
1727    
1728          case OP_TYPEUPTO:
1729          case OP_TYPEMINUPTO:
1730          case OP_TYPEEXACT:
1731          case OP_TYPEPOSUPTO:
1732          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1733          break;
1734    
1735          case OP_MARK:
1736          case OP_PRUNE_ARG:
1737          case OP_SKIP_ARG:
1738          code += code[1];
1739          break;
1740    
1741          case OP_THEN_ARG:
1742          code += code[1+LINK_SIZE];
1743          break;
1744          }
1745    
1746        /* Add in the fixed length from the table */
1747    
1748        code += _pcre_OP_lengths[c];
1749    
1750      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1751      a multi-byte character. The length in the table is a minimum, so we have to
1752      arrange to skip the extra bytes. */
1753    
1754    #ifdef SUPPORT_UTF8
1755        if (utf8) switch(c)
1756        {        {
1757        case OP_CHAR:        case OP_CHAR:
1758        case OP_CHARNC:        case OP_CHARNC:
1759        case OP_EXACT:        case OP_EXACT:
1760        case OP_UPTO:        case OP_UPTO:
1761        case OP_MINUPTO:        case OP_MINUPTO:
1762          case OP_POSUPTO:
1763        case OP_STAR:        case OP_STAR:
1764        case OP_MINSTAR:        case OP_MINSTAR:
1765          case OP_POSSTAR:
1766        case OP_PLUS:        case OP_PLUS:
1767        case OP_MINPLUS:        case OP_MINPLUS:
1768          case OP_POSPLUS:
1769        case OP_QUERY:        case OP_QUERY:
1770        case OP_MINQUERY:        case OP_MINQUERY:
1771        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1772        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1773        break;        break;
1774        }        }
1775    #else
1776        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1777  #endif  #endif
1778      }      }
1779    }    }
# Line 1105  Returns:      pointer to the opcode for Line 1798  Returns:      pointer to the opcode for
1798  static const uschar *  static const uschar *
1799  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1800  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1801  for (;;)  for (;;)
1802    {    {
1803    register int c = *code;    register int c = *code;
1804    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1805    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1806    else if (c > OP_BRA)  
1807      {    /* XCLASS is used for classes that cannot be represented just by a bit
1808      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1809      }    the table is zero; the actual length is stored in the compiled code. */
1810    
1811      if (c == OP_XCLASS) code += GET(code, 1);
1812    
1813      /* Otherwise, we can get the item's length from the table, except that for
1814      repeated character types, we have to test for \p and \P, which have an extra
1815      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1816      must add in its length. */
1817    
1818    else    else
1819      {      {
1820      code += _pcre_OP_lengths[c];      switch(c)
1821          {
1822          case OP_TYPESTAR:
1823          case OP_TYPEMINSTAR:
1824          case OP_TYPEPLUS:
1825          case OP_TYPEMINPLUS:
1826          case OP_TYPEQUERY:
1827          case OP_TYPEMINQUERY:
1828          case OP_TYPEPOSSTAR:
1829          case OP_TYPEPOSPLUS:
1830          case OP_TYPEPOSQUERY:
1831          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1832          break;
1833    
1834  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1835          case OP_TYPEUPTO:
1836          case OP_TYPEMINUPTO:
1837          case OP_TYPEEXACT:
1838          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1839          break;
1840    
1841          case OP_MARK:
1842          case OP_PRUNE_ARG:
1843          case OP_SKIP_ARG:
1844          code += code[1];
1845          break;
1846    
1847          case OP_THEN_ARG:
1848          code += code[1+LINK_SIZE];
1849          break;
1850          }
1851    
1852        /* Add in the fixed length from the table */
1853    
1854        code += _pcre_OP_lengths[c];
1855    
1856      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1857      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1858      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1859    
1860    #ifdef SUPPORT_UTF8
1861      if (utf8) switch(c)      if (utf8) switch(c)
1862        {        {
1863        case OP_CHAR:        case OP_CHAR:
# Line 1136  for (;;) Line 1865  for (;;)
1865        case OP_EXACT:        case OP_EXACT:
1866        case OP_UPTO:        case OP_UPTO:
1867        case OP_MINUPTO:        case OP_MINUPTO:
1868          case OP_POSUPTO:
1869        case OP_STAR:        case OP_STAR:
1870        case OP_MINSTAR:        case OP_MINSTAR:
1871          case OP_POSSTAR:
1872        case OP_PLUS:        case OP_PLUS:
1873        case OP_MINPLUS:        case OP_MINPLUS:
1874          case OP_POSPLUS:
1875        case OP_QUERY:        case OP_QUERY:
1876        case OP_MINQUERY:        case OP_MINQUERY:
1877        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1878        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1879        break;        break;
1880        }        }
1881    #else
1882        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1883  #endif  #endif
1884      }      }
1885    }    }
# Line 1165  for (;;) Line 1892  for (;;)
1892  *************************************************/  *************************************************/
1893    
1894  /* 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
1895  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()
1896  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
1897  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
1898  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1899    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1900    bracket whose current branch will already have been scanned.
1901    
1902  Arguments:  Arguments:
1903    code        points to start of search    code        points to start of search
1904    endcode     points to where to stop    endcode     points to where to stop
1905    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1906      cd          contains pointers to tables etc.
1907    
1908  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1909  */  */
1910    
1911  static BOOL  static BOOL
1912  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1913      compile_data *cd)
1914  {  {
1915  register int c;  register int c;
1916  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1917       code < endcode;       code < endcode;
1918       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1919    {    {
# Line 1190  for (code = first_significant_code(code Line 1921  for (code = first_significant_code(code
1921    
1922    c = *code;    c = *code;
1923    
1924    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1925      first_significant_code() with a TRUE final argument. */
1926    
1927      if (c == OP_ASSERT)
1928      {      {
1929      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1930      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1931        continue;
1932        }
1933    
1934      /* Groups with zero repeats can of course be empty; skip them. */
1935    
1936      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1937        {
1938        code += _pcre_OP_lengths[c];
1939        do code += GET(code, 1); while (*code == OP_ALT);
1940        c = *code;
1941        continue;
1942        }
1943    
1944      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1945      implies a subroutine call, we can scan it. */
1946    
1947      empty_branch = FALSE;    if (c == OP_RECURSE)
1948        {
1949        BOOL empty_branch = FALSE;
1950        const uschar *scode = cd->start_code + GET(code, 1);
1951        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1952      do      do
1953        {        {
1954        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1955            {
1956          empty_branch = TRUE;          empty_branch = TRUE;
1957            break;
1958            }
1959          scode += GET(scode, 1);
1960          }
1961        while (*scode == OP_ALT);
1962        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1963        continue;
1964        }
1965    
1966      /* For other groups, scan the branches. */
1967    
1968      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1969        {
1970        BOOL empty_branch;
1971        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1972    
1973        /* If a conditional group has only one branch, there is a second, implied,
1974        empty branch, so just skip over the conditional, because it could be empty.
1975        Otherwise, scan the individual branches of the group. */
1976    
1977        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1978        code += GET(code, 1);        code += GET(code, 1);
1979        else
1980          {
1981          empty_branch = FALSE;
1982          do
1983            {
1984            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1985              empty_branch = TRUE;
1986            code += GET(code, 1);
1987            }
1988          while (*code == OP_ALT);
1989          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1990        }        }
1991      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1992      c = *code;      c = *code;
1993        continue;
1994      }      }
1995    
1996    else switch (c)    /* Handle the other opcodes */
1997    
1998      switch (c)
1999      {      {
2000      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2001        cannot be represented just by a bit map. This includes negated single
2002        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2003        actual length is stored in the compiled code, so we must update "code"
2004        here. */
2005    
2006  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2007      case OP_XCLASS:      case OP_XCLASS:
2008      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2009      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2010  #endif  #endif
2011    
# Line 1260  for (code = first_significant_code(code Line 2049  for (code = first_significant_code(code
2049      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2050      case OP_WORDCHAR:      case OP_WORDCHAR:
2051      case OP_ANY:      case OP_ANY:
2052        case OP_ALLANY:
2053      case OP_ANYBYTE:      case OP_ANYBYTE:
2054      case OP_CHAR:      case OP_CHAR:
2055      case OP_CHARNC:      case OP_CHARNC:
2056      case OP_NOT:      case OP_NOT:
2057      case OP_PLUS:      case OP_PLUS:
2058      case OP_MINPLUS:      case OP_MINPLUS:
2059        case OP_POSPLUS:
2060      case OP_EXACT:      case OP_EXACT:
2061      case OP_NOTPLUS:      case OP_NOTPLUS:
2062      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2063        case OP_NOTPOSPLUS:
2064      case OP_NOTEXACT:      case OP_NOTEXACT:
2065      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2066      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2067        case OP_TYPEPOSPLUS:
2068      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2069      return FALSE;      return FALSE;
2070    
2071        /* These are going to continue, as they may be empty, but we have to
2072        fudge the length for the \p and \P cases. */
2073    
2074        case OP_TYPESTAR:
2075        case OP_TYPEMINSTAR:
2076        case OP_TYPEPOSSTAR:
2077        case OP_TYPEQUERY:
2078        case OP_TYPEMINQUERY:
2079        case OP_TYPEPOSQUERY:
2080        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2081        break;
2082    
2083        /* Same for these */
2084    
2085        case OP_TYPEUPTO:
2086        case OP_TYPEMINUPTO:
2087        case OP_TYPEPOSUPTO:
2088        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2089        break;
2090    
2091      /* End of branch */      /* End of branch */
2092    
2093      case OP_KET:      case OP_KET:
# Line 1283  for (code = first_significant_code(code Line 2096  for (code = first_significant_code(code
2096      case OP_ALT:      case OP_ALT:
2097      return TRUE;      return TRUE;
2098    
2099      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2100      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2101    
2102  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2103      case OP_STAR:      case OP_STAR:
2104      case OP_MINSTAR:      case OP_MINSTAR:
2105        case OP_POSSTAR:
2106      case OP_QUERY:      case OP_QUERY:
2107      case OP_MINQUERY:      case OP_MINQUERY:
2108        case OP_POSQUERY:
2109        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2110        break;
2111    
2112      case OP_UPTO:      case OP_UPTO:
2113      case OP_MINUPTO:      case OP_MINUPTO:
2114      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2115        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2116      break;      break;
2117  #endif  #endif
2118    
2119        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2120        string. */
2121    
2122        case OP_MARK:
2123        case OP_PRUNE_ARG:
2124        case OP_SKIP_ARG:
2125        code += code[1];
2126        break;
2127    
2128        case OP_THEN_ARG:
2129        code += code[1+LINK_SIZE];
2130        break;
2131    
2132        /* None of the remaining opcodes are required to match a character. */
2133    
2134        default:
2135        break;
2136      }      }
2137    }    }
2138    
# Line 1318  Arguments: Line 2155  Arguments:
2155    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2156    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2157    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2158      cd          pointers to tables etc
2159    
2160  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2161  */  */
2162    
2163  static BOOL  static BOOL
2164  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2165    BOOL utf8)    BOOL utf8, compile_data *cd)
2166  {  {
2167  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2168    {    {
2169    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2170        return FALSE;
2171    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2172    }    }
2173  return TRUE;  return TRUE;
# Line 1341  return TRUE; Line 2180  return TRUE;
2180  *************************************************/  *************************************************/
2181    
2182  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2183  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
2184  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2185  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2186    
2187    Originally, this function only recognized a sequence of letters between the
2188    terminators, but it seems that Perl recognizes any sequence of characters,
2189    though of course unknown POSIX names are subsequently rejected. Perl gives an
2190    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2191    didn't consider this to be a POSIX class. Likewise for [:1234:].
2192    
2193    The problem in trying to be exactly like Perl is in the handling of escapes. We
2194    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2195    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2196    below handles the special case of \], but does not try to do any other escape
2197    processing. This makes it different from Perl for cases such as [:l\ower:]
2198    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2199    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2200    I think.
2201    
2202  Argument:  Arguments:
2203    ptr      pointer to the initial [    ptr      pointer to the initial [
2204    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2205    
2206  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2207  */  */
2208    
2209  static BOOL  static BOOL
2210  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2211  {  {
2212  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2213  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2214  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2215    {    {
2216    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2217    return TRUE;      {
2218        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2219        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2220          {
2221          *endptr = ptr;
2222          return TRUE;
2223          }
2224        }
2225    }    }
2226  return FALSE;  return FALSE;
2227  }  }
# Line 1388  Returns:     a value representing the na Line 2246  Returns:     a value representing the na
2246  static int  static int
2247  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2248  {  {
2249    const char *pn = posix_names;
2250  register int yield = 0;  register int yield = 0;
2251  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2252    {    {
2253    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2254      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2255      pn += posix_name_lengths[yield] + 1;
2256    yield++;    yield++;
2257    }    }
2258  return -1;  return -1;
# Line 1407  return -1; Line 2267  return -1;
2267  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2268  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2269  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
2270  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
2271  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
2272  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
2273  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
2274  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2275    OP_END.
2276    
2277    This function has been extended with the possibility of forward references for
2278    recursions and subroutine calls. It must also check the list of such references
2279    for the group we are dealing with. If it finds that one of the recursions in
2280    the current group is on this list, it adjusts the offset in the list, not the
2281    value in the reference (which is a group number).
2282    
2283  Arguments:  Arguments:
2284    group      points to the start of the group    group      points to the start of the group
2285    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2286    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2287    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2288      save_hwm   the hwm forward reference pointer at the start of the group
2289    
2290  Returns:     nothing  Returns:     nothing
2291  */  */
2292    
2293  static void  static void
2294  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2295      uschar *save_hwm)
2296  {  {
2297  uschar *ptr = group;  uschar *ptr = group;
2298    
2299  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2300    {    {
2301    int offset = GET(ptr, 1);    int offset;
2302    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2303    
2304      /* See if this recursion is on the forward reference list. If so, adjust the
2305      reference. */
2306    
2307      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2308        {
2309        offset = GET(hc, 0);
2310        if (cd->start_code + offset == ptr + 1)
2311          {
2312          PUT(hc, 0, offset + adjust);
2313          break;
2314          }
2315        }
2316    
2317      /* Otherwise, adjust the recursion offset if it's after the start of this
2318      group. */
2319    
2320      if (hc >= cd->hwm)
2321        {
2322        offset = GET(ptr, 1);
2323        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2324        }
2325    
2326    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2327    }    }
2328  }  }
# Line 1456  auto_callout(uschar *code, const uschar Line 2349  auto_callout(uschar *code, const uschar
2349  {  {
2350  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2351  *code++ = 255;  *code++ = 255;
2352  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2353  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2354  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2355  }  }
2356    
# Line 1482  Returns:             nothing Line 2375  Returns:             nothing
2375  static void  static void
2376  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2377  {  {
2378  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2379  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2380  }  }
2381    
# Line 1508  Yield:        TRUE when range returned; Line 2401  Yield:        TRUE when range returned;
2401  */  */
2402    
2403  static BOOL  static BOOL
2404  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2405      unsigned int *odptr)
2406  {  {
2407  int c, othercase, next;  unsigned int c, othercase, next;
2408    
2409  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2410    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2411    
2412  if (c > d) return FALSE;  if (c > d) return FALSE;
2413    
# Line 1522  next = othercase + 1; Line 2416  next = othercase + 1;
2416    
2417  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2418    {    {
2419    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2420    next++;    next++;
2421    }    }
2422    
# Line 1531  for (++c; c <= d; c++) Line 2425  for (++c; c <= d; c++)
2425    
2426  return TRUE;  return TRUE;
2427  }  }
2428  #endif  /* SUPPORT_UCP */  
2429    
2430    
2431  /*************************************************  /*************************************************
2432  *           Compile one branch                   *  *        Check a character and a property        *
2433  *************************************************/  *************************************************/
2434    
2435  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called by check_auto_possessive() when a property item
2436  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2437    
2438  Arguments:  Arguments:
2439    optionsptr     pointer to the option bits    c            the character
2440    brackets       points to number of extracting brackets used    ptype        the property type
2441    codeptr        points to the pointer to the current code point    pdata        the data for the type
2442    ptrptr         points to the current pattern pointer    negated      TRUE if it's a negated property (\P or \p{^)
   errorcodeptr   points to error code variable  
   firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)  
   reqbyteptr     set to the last literal character required, else < 0  
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
2443    
2444  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2445  */  */
2446    
2447  static BOOL  static BOOL
2448  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(int c, int ptype, int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2449  {  {
2450  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2451  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2452  int bravalue = 0;    {
2453  int greedy_default, greedy_non_default;    case PT_LAMP:
2454  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2455  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2456  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2457  int condcount = 0;  
2458  int options = *optionsptr;    case PT_GC:
2459  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2460  register int c;  
2461  register uschar *code = *codeptr;    case PT_PC:
2462  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2463  BOOL inescq = FALSE;  
2464  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2465  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2466  const uschar *tempptr;  
2467  uschar *previous = NULL;    /* These are specials */
2468  uschar *previous_callout = NULL;  
2469  uschar classbits[32];    case PT_ALNUM:
2470      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2471  #ifdef SUPPORT_UTF8            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2472  BOOL class_utf8;  
2473  BOOL utf8 = (options & PCRE_UTF8) != 0;    case PT_SPACE:    /* Perl space */
2474  uschar *class_utf8data;    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2475  uschar utf8_char[6];            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2476  #else            == negated;
2477  BOOL utf8 = FALSE;  
2478  #endif    case PT_PXSPACE:  /* POSIX space */
2479      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2480              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2481              c == CHAR_FF || c == CHAR_CR)
2482              == negated;
2483    
2484      case PT_WORD:
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2486              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2487              c == CHAR_UNDERSCORE) == negated;
2488      }
2489    return FALSE;
2490    }
2491    #endif  /* SUPPORT_UCP */
2492    
 /* Set up the default and non-default settings for greediness */  
2493    
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2494    
2495  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /*************************************************
2496  matching encountered yet". It gets changed to REQ_NONE if we hit something that  *     Check if auto-possessifying is possible    *
2497  matches a non-fixed char first char; reqbyte just remains unset if we never  *************************************************/
 find one.  
2498    
2499  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* This function is called for unlimited repeats of certain items, to see
2500  to take the zero repeat into account. This is implemented by setting them to  whether the next thing could possibly match the repeated item. If not, it makes
2501  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  sense to automatically possessify the repeated item.
 item types that can be repeated set these backoff variables appropriately. */  
2502    
2503  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Arguments:
2504      previous      pointer to the repeated opcode
2505      utf8          TRUE in UTF-8 mode
2506      ptr           next character in pattern
2507      options       options bits
2508      cd            contains pointers to tables etc.
2509    
2510  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  Returns:        TRUE if possessifying is wanted
2511  according to the current setting of the caseless flag. REQ_CASELESS is a bit  */
 value > 255. It is added into the firstbyte or reqbyte variables to record the  
 case status of the value. This is used only for ASCII characters. */  
2512    
2513  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  static BOOL
2514    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2515      int options, compile_data *cd)
2516    {
2517    int c, next;
2518    int op_code = *previous++;
2519    
2520  /* Switch on next character until the end of the branch */  /* Skip whitespace and comments in extended mode */
2521    
2522  for (;; ptr++)  if ((options & PCRE_EXTENDED) != 0)
2523    {    {
2524    BOOL negate_class;    for (;;)
   BOOL possessive_quantifier;  
   BOOL is_quantifier;  
   int class_charcount;  
   int class_lastchar;  
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
   
   /* Next byte in the pattern */  
   
   c = *ptr;  
   
   /* If in \Q...\E, check for the end; if not, we have a literal */  
   
   if (inescq && c != 0)  
2525      {      {
2526      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2527        {      if (*ptr == CHAR_NUMBER_SIGN)
       inescq = FALSE;  
       ptr++;  
       continue;  
       }  
     else  
2528        {        {
2529        if (previous_callout != NULL)        ptr++;
2530          {        while (*ptr != 0)
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
2531          {          {
2532          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2533          code = auto_callout(code, ptr, cd);          ptr++;
2534    #ifdef SUPPORT_UTF8
2535            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2536    #endif
2537          }          }
       goto NORMAL_CHAR;  
2538        }        }
2539        else break;
2540      }      }
2541      }
2542    
2543    /* Fill in length of a previous callout, except when the next thing is  /* If the next item is one that we can handle, get its value. A non-negative
2544    a quantifier. */  value is a character, a negative value is an escape value. */
2545    
2546    is_quantifier = c == '*' || c == '+' || c == '?' ||  if (*ptr == CHAR_BACKSLASH)
2547      (c == '{' && is_counted_repeat(ptr+1));    {
2548      int temperrorcode = 0;
2549      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2550      if (temperrorcode != 0) return FALSE;
2551      ptr++;    /* Point after the escape sequence */
2552      }
2553    
2554    if (!is_quantifier && previous_callout != NULL &&  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2555         after_manual_callout-- <= 0)    {
2556      {  #ifdef SUPPORT_UTF8
2557      complete_callout(previous_callout, ptr, cd);    if (utf8) { GETCHARINC(next, ptr); } else
2558      previous_callout = NULL;  #endif
2559      }    next = *ptr++;
2560      }
2561    
2562    /* In extended mode, skip white space and comments */  else return FALSE;
2563    
2564    if ((options & PCRE_EXTENDED) != 0)  /* Skip whitespace and comments in extended mode */
2565    
2566    if ((options & PCRE_EXTENDED) != 0)
2567      {
2568      for (;;)
2569      {      {
2570      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2571      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2572        {        {
2573        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
2574        on the Macintosh. */        while (*ptr != 0)
2575        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
2576        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2577            ptr++;
2578    #ifdef SUPPORT_UTF8
2579            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2580    #endif
2581            }
2582        }        }
2583        else break;
2584      }      }
2585      }
2586    
2587    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
2588    
2589    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2590      {    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2591      previous_callout = code;      return FALSE;
     code = auto_callout(code, ptr, cd);  
     }  
2592    
2593    switch(c)  /* Now compare the next item with the previous opcode. First, handle cases when
2594      {  the next item is a character. */
     /* The branch terminates at end of string, |, or ). */  
2595    
2596      case 0:  if (next >= 0) switch(op_code)
2597      case '|':    {
2598      case ')':    case OP_CHAR:
2599      *firstbyteptr = firstbyte;  #ifdef SUPPORT_UTF8
2600      *reqbyteptr = reqbyte;    GETCHARTEST(c, previous);
2601      *codeptr = code;  #else
2602      *ptrptr = ptr;    c = *previous;
2603      return TRUE;  #endif
2604      return c != next;
2605    
2606      /* Handle single-character metacharacters. In multiline mode, ^ disables    /* For CHARNC (caseless character) we must check the other case. If we have
2607      the setting of any following char as a first character. */    Unicode property support, we can use it to test the other case of
2608      high-valued characters. */
2609    
2610      case '^':    case OP_CHARNC:
2611      if ((options & PCRE_MULTILINE) != 0)  #ifdef SUPPORT_UTF8
2612        {    GETCHARTEST(c, previous);
2613        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  #else
2614        }    c = *previous;
2615      previous = NULL;  #endif
2616      *code++ = OP_CIRC;    if (c == next) return FALSE;
2617      break;  #ifdef SUPPORT_UTF8
2618      if (utf8)
2619        {
2620        unsigned int othercase;
2621        if (next < 128) othercase = cd->fcc[next]; else
2622    #ifdef SUPPORT_UCP
2623        othercase = UCD_OTHERCASE((unsigned int)next);
2624    #else
2625        othercase = NOTACHAR;
2626    #endif
2627        return (unsigned int)c != othercase;
2628        }
2629      else
2630    #endif  /* SUPPORT_UTF8 */
2631      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2632    
2633      case '$':    /* For OP_NOT, its data is always a single-byte character. */
     previous = NULL;  
     *code++ = OP_DOLL;  
     break;  
2634    
2635      /* There can never be a first char if '.' is first, whatever happens about    case OP_NOT:
2636      repeats. The value of reqbyte doesn't change either. */    if ((c = *previous) == next) return TRUE;
2637      if ((options & PCRE_CASELESS) == 0) return FALSE;
2638    #ifdef SUPPORT_UTF8
2639      if (utf8)
2640        {
2641        unsigned int othercase;
2642        if (next < 128) othercase = cd->fcc[next]; else
2643    #ifdef SUPPORT_UCP
2644        othercase = UCD_OTHERCASE(next);
2645    #else
2646        othercase = NOTACHAR;
2647    #endif
2648        return (unsigned int)c == othercase;
2649        }
2650      else
2651    #endif  /* SUPPORT_UTF8 */
2652      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2653    
2654      case '.':    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2655      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
2656    
2657      /* Character classes. If the included characters are all < 256, we build a    case OP_DIGIT:
2658      32-byte bitmap of the permitted characters, except in the special case    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
     where there is only one such character. For negated classes, we build the  
     map as usual, then invert it at the end. However, we use a different opcode  
     so that data characters > 255 can be handled correctly.  
2659    
2660      If the class contains characters outside the 0-255 range, a different    case OP_NOT_DIGIT:
2661      opcode is compiled. It may optionally have a bit map for characters < 256,    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
     but those above are are explicitly listed afterwards. A flag byte tells  
     whether the bitmap is present, and whether this is a negated class or not.  
     */  
2662    
2663      case '[':    case OP_WHITESPACE:
2664      previous = code;    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2665    
2666      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if    case OP_NOT_WHITESPACE:
2667      they are encountered at the top level, so we'll do that too. */    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2668    
2669      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    case OP_WORDCHAR:
2670          check_posix_syntax(ptr, &tempptr, cd))    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
       }  
2671    
2672      /* If the first character is '^', set the negation flag and skip it. */    case OP_NOT_WORDCHAR:
2673      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2674    
2675      if ((c = *(++ptr)) == '^')    case OP_HSPACE:
2676        {    case OP_NOT_HSPACE:
2677        negate_class = TRUE;    switch(next)
2678        c = *(++ptr);      {
2679        }      case 0x09:
2680      else      case 0x20:
2681        {      case 0xa0:
2682        negate_class = FALSE;      case 0x1680:
2683        }      case 0x180e:
2684        case 0x2000:
2685        case 0x2001:
2686        case 0x2002:
2687        case 0x2003:
2688        case 0x2004:
2689        case 0x2005:
2690        case 0x2006:
2691        case 0x2007:
2692        case 0x2008:
2693        case 0x2009:
2694        case 0x200A:
2695        case 0x202f:
2696        case 0x205f:
2697        case 0x3000:
2698        return op_code == OP_NOT_HSPACE;
2699        default:
2700        return op_code != OP_NOT_HSPACE;
2701        }
2702    
2703      /* Keep a count of chars with values < 256 so that we can optimize the case    case OP_ANYNL:
2704      of just a single character (as long as it's < 256). For higher valued UTF-8    case OP_VSPACE:
2705      characters, we don't yet do any optimization. */    case OP_NOT_VSPACE:
2706      switch(next)
2707        {
2708        case 0x0a:
2709        case 0x0b:
2710        case 0x0c:
2711        case 0x0d:
2712        case 0x85:
2713        case 0x2028:
2714        case 0x2029:
2715        return op_code == OP_NOT_VSPACE;
2716        default:
2717        return op_code != OP_NOT_VSPACE;
2718        }
2719    
2720      class_charcount = 0;  #ifdef SUPPORT_UCP
2721      class_lastchar = -1;    case OP_PROP:
2722      return check_char_prop(next, previous[0], previous[1], FALSE);
2723    
2724  #ifdef SUPPORT_UTF8    case OP_NOTPROP:
2725      class_utf8 = FALSE;                       /* No chars >= 256 */    return check_char_prop(next, previous[0], previous[1], TRUE);
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
2726  #endif  #endif
2727    
2728      /* Initialize the 32-char bit map to all zeros. We have to build the    default:
2729      map in a temporary bit of store, in case the class contains only 1    return FALSE;
2730      character (< 256), because in that case the compiled code doesn't use the    }
     bit map. */  
2731    
     memset(classbits, 0, 32 * sizeof(uschar));  
2732    
2733      /* Process characters until ] is reached. By writing this as a "do" it  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2734      means that an initial ] is taken as a data character. The first pass  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2735      through the regex checked the overall syntax, so we don't need to be very  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2736      strict here. At the start of the loop, c contains the first byte of the  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2737      character. */  replaced by OP_PROP codes when PCRE_UCP is set. */
2738    
2739      do  switch(op_code)
2740        {    {
2741      case OP_CHAR:
2742      case OP_CHARNC:
2743  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2744        if (utf8 && c > 127)    GETCHARTEST(c, previous);
2745          {                           /* Braces are required because the */  #else
2746          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */    c = *previous;
         }  
2747  #endif  #endif
2748      switch(-next)
2749        {
2750        case ESC_d:
2751        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2752    
2753        /* Inside \Q...\E everything is literal except \E */      case ESC_D:
2754        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
       if (inescq)  
         {  
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
   
       /* Handle POSIX class names. Perl allows a negation extension of the  
       form [:^name:]. A square bracket that doesn't match the syntax is  
       treated as a literal. We also recognize the POSIX constructions  
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
2755    
2756        if (c == '[' &&      case ESC_s:
2757            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, taboffset, tabopt;  
         register const uschar *cbits = cd->cbits;  
         uschar pbits[32];  
2758    
2759          if (ptr[1] != ':')      case ESC_S:
2760            {      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2761            *errorcodeptr = ERR31;  
2762            goto FAILED;      case ESC_w:
2763            }      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2764    
2765        case ESC_W:
2766        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2767    
2768        case ESC_h:
2769        case ESC_H:
2770        switch(c)
2771          {
2772          case 0x09:
2773          case 0x20:
2774          case 0xa0:
2775          case 0x1680:
2776          case 0x180e:
2777          case 0x2000:
2778          case 0x2001:
2779          case 0x2002:
2780          case 0x2003:
2781          case 0x2004:
2782          case 0x2005:
2783          case 0x2006:
2784          case 0x2007:
2785          case 0x2008:
2786          case 0x2009:
2787          case 0x200A:
2788          case 0x202f:
2789          case 0x205f:
2790          case 0x3000:
2791          return -next != ESC_h;
2792          default:
2793          return -next == ESC_h;
2794          }
2795    
2796          ptr += 2;      case ESC_v:
2797          if (*ptr == '^')      case ESC_V:
2798            {      switch(c)
2799            local_negate = TRUE;        {
2800            ptr++;        case 0x0a:
2801            }        case 0x0b:
2802          case 0x0c:
2803          case 0x0d:
2804          case 0x85:
2805          case 0x2028:
2806          case 0x2029:
2807          return -next != ESC_v;
2808          default:
2809          return -next == ESC_v;
2810          }
2811    
2812          posix_class = check_posix_name(ptr, tempptr - ptr);      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2813          if (posix_class < 0)      their substitutions and process them. The result will always be either
2814            {      -ESC_p or -ESC_P. Then fall through to process those values. */
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
2815    
2816          /* If matching is caseless, upper and lower are converted to  #ifdef SUPPORT_UCP
2817          alpha. This relies on the fact that the class table starts with      case ESC_du:
2818          alpha, lower, upper as the first 3 entries. */      case ESC_DU:
2819        case ESC_wu:
2820        case ESC_WU:
2821        case ESC_su:
2822        case ESC_SU:
2823          {
2824          int temperrorcode = 0;
2825          ptr = substitutes[-next - ESC_DU];
2826          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2827          if (temperrorcode != 0) return FALSE;
2828          ptr++;    /* For compatibility */
2829          }
2830        /* Fall through */
2831    
2832          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)      case ESC_p:
2833            posix_class = 0;      case ESC_P:
2834          {
2835          int ptype, pdata, errorcodeptr;
2836          BOOL negated;
2837    
2838          /* We build the bit map for the POSIX class in a chunk of local store        ptr--;      /* Make ptr point at the p or P */
2839          because we may be adding and subtracting from it, and we don't want to        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2840          subtract bits that may be in the main map already. At the end we or the        if (ptype < 0) return FALSE;
2841          result into the bit map that is being built. */        ptr++;      /* Point past the final curly ket */
2842    
2843          /* If the property item is optional, we have to give up. (When generated
2844          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2845          to the original \d etc. At this point, ptr will point to a zero byte. */
2846    
2847          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2848            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2849              return FALSE;
2850    
2851          posix_class *= 3;        /* Do the property check. */
2852    
2853          /* Copy in the first table (always present) */        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2854          }
2855    #endif
2856    
2857          memcpy(pbits, cbits + posix_class_maps[posix_class],      default:
2858            32 * sizeof(uschar));      return FALSE;
2859        }
2860    
2861          /* If there is a second table, add or remove it as required. */    /* In principle, support for Unicode properties should be integrated here as
2862      well. It means re-organizing the above code so as to get hold of the property
2863      values before switching on the op-code. However, I wonder how many patterns
2864      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2865      these op-codes are never generated.) */
2866    
2867          taboffset = posix_class_maps[posix_class + 1];    case OP_DIGIT:
2868          tabopt = posix_class_maps[posix_class + 2];    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2869             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2870    
2871          if (taboffset >= 0)    case OP_NOT_DIGIT:
2872            {    return next == -ESC_d;
           if (tabopt >= 0)  
             for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];  
           else  
             for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];  
           }  
2873    
2874          /* Not see if we need to remove any special characters. An option    case OP_WHITESPACE:
2875          value of 1 removes vertical space and 2 removes underscore. */    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2876    
2877          if (tabopt < 0) tabopt = -tabopt;    case OP_NOT_WHITESPACE:
2878          if (tabopt == 1) pbits[1] &= ~0x3c;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
           else if (tabopt == 2) pbits[11] &= 0x7f;  
2879    
2880          /* Add the POSIX table or its complement into the main table that is    case OP_HSPACE:
2881          being built and we are done. */    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2882             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2883    
2884          if (local_negate)    case OP_NOT_HSPACE:
2885            for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];    return next == -ESC_h;
         else  
           for (c = 0; c < 32; c++) classbits[c] |= pbits[c];  
2886    
2887          ptr = tempptr + 1;    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2888          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */    case OP_ANYNL:
2889          continue;    /* End of POSIX syntax handling */    case OP_VSPACE:
2890          }    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2891    
2892        /* Backslash may introduce a single character, or it may introduce one    case OP_NOT_VSPACE:
2893        of the specials, which just set a flag. Escaped items are checked for    return next == -ESC_v || next == -ESC_R;
       validity in the pre-compiling pass. The sequence \b is a special case.  
       Inside a class (and only there) it is treated as backspace. Elsewhere  
       it marks a word boundary. Other escapes have preset maps ready to  
       or into the one we are building. We assume they have more than one  
       character in them, so set class_charcount bigger than one. */  
2894    
2895        if (c == '\\')    case OP_WORDCHAR:
2896          {    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2897          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);           next == -ESC_v || next == -ESC_R;
2898    
2899          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */    case OP_NOT_WORDCHAR:
2900          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */    return next == -ESC_w || next == -ESC_d;
         else if (-c == ESC_Q)            /* Handle start of quoted string */  
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
2901    
2902          if (c < 0)    default:
2903            {    return FALSE;
2904            register const uschar *cbits = cd->cbits;    }
           class_charcount += 2;     /* Greater than 1 is what matters */  
           switch (-c)  
             {  
             case ESC_d:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];  
             continue;  
2905    
2906              case ESC_D:  /* Control does not reach here */
2907              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];  }
             continue;  
2908    
             case ESC_w:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];  
             continue;  
2909    
             case ESC_W:  
             for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];  
             continue;  
2910    
2911              case ESC_s:  /*************************************************
2912              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  *           Compile one branch                   *
2913              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  *************************************************/
             continue;  
2914    
2915              case ESC_S:  /* Scan the pattern, compiling it into the a vector. If the options are
2916              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];  changed during the branch, the pointer is used to change the external options
2917              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  bits. This function is used during the pre-compile phase when we are trying
2918              continue;  to find out the amount of memory needed, as well as during the real compile
2919    phase. The value of lengthptr distinguishes the two phases.
2920    
2921  #ifdef SUPPORT_UCP  Arguments:
2922              case ESC_p:    optionsptr     pointer to the option bits
2923              case ESC_P:    codeptr        points to the pointer to the current code point
2924                {    ptrptr         points to the current pattern pointer
2925                BOOL negated;    errorcodeptr   points to error code variable
2926                int pdata;    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2927                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);    reqbyteptr     set to the last literal character required, else < 0
2928                if (ptype < 0) goto FAILED;    bcptr          points to current branch chain
2929                class_utf8 = TRUE;    cd             contains pointers to tables etc.
2930                *class_utf8data++ = ((-c == ESC_p) != negated)?    lengthptr      NULL during the real compile phase
2931                  XCL_PROP : XCL_NOTPROP;                   points to length accumulator during pre-compile phase
               *class_utf8data++ = ptype;  
               *class_utf8data++ = pdata;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
             continue;  
 #endif  
2932    
2933              /* Unrecognized escapes are faulted if PCRE is running in its  Returns:         TRUE on success
2934              strict mode. By default, for compatibility with Perl, they are                   FALSE, with *errorcodeptr set non-zero on error
2935              treated as literals. */  */
2936    
2937              default:  static BOOL
2938              if ((options & PCRE_EXTRA) != 0)  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2939                {    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2940      compile_data *cd, int *lengthptr)
2941    {
2942    int repeat_type, op_type;
2943    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2944    int bravalue = 0;
2945    int greedy_default, greedy_non_default;
2946    int firstbyte, reqbyte;
2947    int zeroreqbyte, zerofirstbyte;
2948    int req_caseopt, reqvary, tempreqvary;
2949    int options = *optionsptr;
2950    int after_manual_callout = 0;
2951    int length_prevgroup = 0;
2952    register int c;
2953    register uschar *code = *codeptr;
2954    uschar *last_code = code;
2955    uschar *orig_code = code;
2956    uschar *tempcode;
2957    BOOL inescq = FALSE;
2958    BOOL groupsetfirstbyte = FALSE;
2959    const uschar *ptr = *ptrptr;
2960    const uschar *tempptr;
2961    const uschar *nestptr = NULL;
2962    uschar *previous = NULL;
2963    uschar *previous_callout = NULL;
2964    uschar *save_hwm = NULL;
2965    uschar classbits[32];
2966    
2967    #ifdef SUPPORT_UTF8
2968    BOOL class_utf8;
2969    BOOL utf8 = (options & PCRE_UTF8) != 0;
2970    uschar *class_utf8data;
2971    uschar *class_utf8data_base;
2972    uschar utf8_char[6];
2973    #else
2974    BOOL utf8 = FALSE;
2975    uschar *utf8_char = NULL;
2976    #endif
2977    
2978    #ifdef PCRE_DEBUG
2979    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2980    #endif
2981    
2982    /* Set up the default and non-default settings for greediness */
2983    
2984    greedy_default = ((options & PCRE_UNGREEDY) != 0);
2985    greedy_non_default = greedy_default ^ 1;
2986    
2987    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2988    matching encountered yet". It gets changed to REQ_NONE if we hit something that
2989    matches a non-fixed char first char; reqbyte just remains unset if we never
2990    find one.
2991    
2992    When we hit a repeat whose minimum is zero, we may have to adjust these values
2993    to take the zero repeat into account. This is implemented by setting them to
2994    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2995    item types that can be repeated set these backoff variables appropriately. */
2996    
2997    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
2998    
2999    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
3000    according to the current setting of the caseless flag. REQ_CASELESS is a bit
3001    value > 255. It is added into the firstbyte or reqbyte variables to record the
3002    case status of the value. This is used only for ASCII characters. */
3003    
3004    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
3005    
3006    /* Switch on next character until the end of the branch */
3007    
3008    for (;; ptr++)
3009      {
3010      BOOL negate_class;
3011      BOOL should_flip_negation;
3012      BOOL possessive_quantifier;
3013      BOOL is_quantifier;
3014      BOOL is_recurse;
3015      BOOL reset_bracount;
3016      int class_charcount;
3017      int class_lastchar;
3018      int newoptions;
3019      int recno;
3020      int refsign;
3021      int skipbytes;
3022      int subreqbyte;
3023      int subfirstbyte;
3024      int terminator;
3025      int mclength;
3026      uschar mcbuffer[8];
3027    
3028      /* Get next byte in the pattern */
3029    
3030      c = *ptr;
3031    
3032      /* If we are at the end of a nested substitution, revert to the outer level
3033      string. Nesting only happens one level deep. */
3034    
3035      if (c == 0 && nestptr != NULL)
3036        {
3037        ptr = nestptr;
3038        nestptr = NULL;
3039        c = *ptr;
3040        }
3041    
3042      /* If we are in the pre-compile phase, accumulate the length used for the
3043      previous cycle of this loop. */
3044    
3045      if (lengthptr != NULL)
3046        {
3047    #ifdef PCRE_DEBUG
3048        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3049    #endif
3050        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3051          {
3052          *errorcodeptr = ERR52;
3053          goto FAILED;
3054          }
3055    
3056        /* There is at least one situation where code goes backwards: this is the
3057        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3058        the class is simply eliminated. However, it is created first, so we have to
3059        allow memory for it. Therefore, don't ever reduce the length at this point.
3060        */
3061    
3062        if (code < last_code) code = last_code;
3063    
3064        /* Paranoid check for integer overflow */
3065    
3066        if (OFLOW_MAX - *lengthptr < code - last_code)
3067          {
3068          *errorcodeptr = ERR20;
3069          goto FAILED;
3070          }
3071    
3072        *lengthptr += (int)(code - last_code);
3073        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3074    
3075        /* If "previous" is set and it is not at the start of the work space, move
3076        it back to there, in order to avoid filling up the work space. Otherwise,
3077        if "previous" is NULL, reset the current code pointer to the start. */
3078    
3079        if (previous != NULL)
3080          {
3081          if (previous > orig_code)
3082            {
3083            memmove(orig_code, previous, code - previous);
3084            code -= previous - orig_code;
3085            previous = orig_code;
3086            }
3087          }
3088        else code = orig_code;
3089    
3090        /* Remember where this code item starts so we can pick up the length
3091        next time round. */
3092    
3093        last_code = code;
3094        }
3095    
3096      /* In the real compile phase, just check the workspace used by the forward
3097      reference list. */
3098    
3099      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3100        {
3101        *errorcodeptr = ERR52;
3102        goto FAILED;
3103        }
3104    
3105      /* If in \Q...\E, check for the end; if not, we have a literal */
3106    
3107      if (inescq && c != 0)
3108        {
3109        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3110          {
3111          inescq = FALSE;
3112          ptr++;
3113          continue;
3114          }
3115        else
3116          {
3117          if (previous_callout != NULL)
3118            {
3119            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3120              complete_callout(previous_callout, ptr, cd);
3121            previous_callout = NULL;
3122            }
3123          if ((options & PCRE_AUTO_CALLOUT) != 0)
3124            {
3125            previous_callout = code;
3126            code = auto_callout(code, ptr, cd);
3127            }
3128          goto NORMAL_CHAR;
3129          }
3130        }
3131    
3132      /* Fill in length of a previous callout, except when the next thing is
3133      a quantifier. */
3134    
3135      is_quantifier =
3136        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3137        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3138    
3139      if (!is_quantifier && previous_callout != NULL &&
3140           after_manual_callout-- <= 0)
3141        {
3142        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3143          complete_callout(previous_callout, ptr, cd);
3144        previous_callout = NULL;
3145        }
3146    
3147      /* In extended mode, skip white space and comments */
3148    
3149      if ((options & PCRE_EXTENDED) != 0)
3150        {
3151        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3152        if (c == CHAR_NUMBER_SIGN)
3153          {
3154          ptr++;
3155          while (*ptr != 0)
3156            {
3157            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3158            ptr++;
3159    #ifdef SUPPORT_UTF8
3160            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3161    #endif
3162            }
3163          if (*ptr != 0) continue;
3164    
3165          /* Else fall through to handle end of string */
3166          c = 0;
3167          }
3168        }
3169    
3170      /* No auto callout for quantifiers. */
3171    
3172      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3173        {
3174        previous_callout = code;
3175        code = auto_callout(code, ptr, cd);
3176        }
3177    
3178      switch(c)
3179        {
3180        /* ===================================================================*/
3181        case 0:                        /* The branch terminates at string end */
3182        case CHAR_VERTICAL_LINE:       /* or | or ) */
3183        case CHAR_RIGHT_PARENTHESIS:
3184        *firstbyteptr = firstbyte;
3185        *reqbyteptr = reqbyte;
3186        *codeptr = code;
3187        *ptrptr = ptr;
3188        if (lengthptr != NULL)
3189          {
3190          if (OFLOW_MAX - *lengthptr < code - last_code)
3191            {
3192            *errorcodeptr = ERR20;
3193            goto FAILED;
3194            }
3195          *lengthptr += (int)(code - last_code);   /* To include callout length */
3196          DPRINTF((">> end branch\n"));
3197          }
3198        return TRUE;
3199    
3200    
3201        /* ===================================================================*/
3202        /* Handle single-character metacharacters. In multiline mode, ^ disables
3203        the setting of any following char as a first character. */
3204    
3205        case CHAR_CIRCUMFLEX_ACCENT:
3206        if ((options & PCRE_MULTILINE) != 0)
3207          {
3208          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3209          }
3210        previous = NULL;
3211        *code++ = OP_CIRC;
3212        break;
3213    
3214        case CHAR_DOLLAR_SIGN:
3215        previous = NULL;
3216        *code++ = OP_DOLL;
3217        break;
3218    
3219        /* There can never be a first char if '.' is first, whatever happens about
3220        repeats. The value of reqbyte doesn't change either. */
3221    
3222        case CHAR_DOT:
3223        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3224        zerofirstbyte = firstbyte;
3225        zeroreqbyte = reqbyte;
3226        previous = code;
3227        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3228        break;
3229    
3230    
3231        /* ===================================================================*/
3232        /* Character classes. If the included characters are all < 256, we build a
3233        32-byte bitmap of the permitted characters, except in the special case
3234        where there is only one such character. For negated classes, we build the
3235        map as usual, then invert it at the end. However, we use a different opcode
3236        so that data characters > 255 can be handled correctly.
3237    
3238        If the class contains characters outside the 0-255 range, a different
3239        opcode is compiled. It may optionally have a bit map for characters < 256,
3240        but those above are are explicitly listed afterwards. A flag byte tells
3241        whether the bitmap is present, and whether this is a negated class or not.
3242    
3243        In JavaScript compatibility mode, an isolated ']' causes an error. In
3244        default (Perl) mode, it is treated as a data character. */
3245    
3246        case CHAR_RIGHT_SQUARE_BRACKET:
3247        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3248          {
3249          *errorcodeptr = ERR64;
3250          goto FAILED;
3251          }
3252        goto NORMAL_CHAR;
3253    
3254        case CHAR_LEFT_SQUARE_BRACKET:
3255        previous = code;
3256    
3257        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3258        they are encountered at the top level, so we'll do that too. */
3259    
3260        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3261             ptr[1] == CHAR_EQUALS_SIGN) &&
3262            check_posix_syntax(ptr, &tempptr))
3263          {
3264          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3265          goto FAILED;
3266          }
3267    
3268        /* If the first character is '^', set the negation flag and skip it. Also,
3269        if the first few characters (either before or after ^) are \Q\E or \E we
3270        skip them too. This makes for compatibility with Perl. */
3271    
3272        negate_class = FALSE;
3273        for (;;)
3274          {
3275          c = *(++ptr);
3276          if (c == CHAR_BACKSLASH)
3277            {
3278            if (ptr[1] == CHAR_E)
3279              ptr++;
3280            else if (strncmp((const char *)ptr+1,
3281                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3282              ptr += 3;
3283            else
3284              break;
3285            }
3286          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3287            negate_class = TRUE;
3288          else break;
3289          }
3290    
3291        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3292        an initial ']' is taken as a data character -- the code below handles
3293        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3294        [^] must match any character, so generate OP_ALLANY. */
3295    
3296        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3297            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3298          {
3299          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3300          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3301          zerofirstbyte = firstbyte;
3302          break;
3303          }
3304    
3305        /* If a class contains a negative special such as \S, we need to flip the
3306        negation flag at the end, so that support for characters > 255 works
3307        correctly (they are all included in the class). */
3308    
3309        should_flip_negation = FALSE;
3310    
3311        /* Keep a count of chars with values < 256 so that we can optimize the case
3312        of just a single character (as long as it's < 256). However, For higher
3313        valued UTF-8 characters, we don't yet do any optimization. */
3314    
3315        class_charcount = 0;
3316        class_lastchar = -1;
3317    
3318        /* Initialize the 32-char bit map to all zeros. We build the map in a
3319        temporary bit of memory, in case the class contains only 1 character (less
3320        than 256), because in that case the compiled code doesn't use the bit map.
3321        */
3322    
3323        memset(classbits, 0, 32 * sizeof(uschar));
3324    
3325    #ifdef SUPPORT_UTF8
3326        class_utf8 = FALSE;                       /* No chars >= 256 */
3327        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3328        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3329    #endif
3330    
3331        /* Process characters until ] is reached. By writing this as a "do" it
3332        means that an initial ] is taken as a data character. At the start of the
3333        loop, c contains the first byte of the character. */
3334    
3335        if (c != 0) do
3336          {
3337          const uschar *oldptr;
3338    
3339    #ifdef SUPPORT_UTF8
3340          if (utf8 && c > 127)
3341            {                           /* Braces are required because the */
3342            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3343            }
3344    
3345          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3346          data and reset the pointer. This is so that very large classes that
3347          contain a zillion UTF-8 characters no longer overwrite the work space
3348          (which is on the stack). */
3349    
3350          if (lengthptr != NULL)
3351            {
3352            *lengthptr += class_utf8data - class_utf8data_base;
3353            class_utf8data = class_utf8data_base;
3354            }
3355    
3356    #endif
3357    
3358          /* Inside \Q...\E everything is literal except \E */
3359    
3360          if (inescq)
3361            {
3362            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3363              {
3364              inescq = FALSE;                   /* Reset literal state */
3365              ptr++;                            /* Skip the 'E' */
3366              continue;                         /* Carry on with next */
3367              }
3368            goto CHECK_RANGE;                   /* Could be range if \E follows */
3369            }
3370    
3371          /* Handle POSIX class names. Perl allows a negation extension of the
3372          form [:^name:]. A square bracket that doesn't match the syntax is
3373          treated as a literal. We also recognize the POSIX constructions
3374          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3375          5.6 and 5.8 do. */
3376    
3377          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3378              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3379               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3380            {
3381            BOOL local_negate = FALSE;
3382            int posix_class, taboffset, tabopt;
3383            register const uschar *cbits = cd->cbits;
3384            uschar pbits[32];
3385    
3386            if (ptr[1] != CHAR_COLON)
3387              {
3388              *errorcodeptr = ERR31;
3389              goto FAILED;
3390              }
3391    
3392            ptr += 2;
3393            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3394              {
3395              local_negate = TRUE;
3396              should_flip_negation = TRUE;  /* Note negative special */
3397              ptr++;
3398              }
3399    
3400            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3401            if (posix_class < 0)
3402              {
3403              *errorcodeptr = ERR30;
3404              goto FAILED;
3405              }
3406    
3407            /* If matching is caseless, upper and lower are converted to
3408            alpha. This relies on the fact that the class table starts with
3409            alpha, lower, upper as the first 3 entries. */
3410    
3411            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3412              posix_class = 0;
3413    
3414            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3415            different escape sequences that use Unicode properties. */
3416    
3417    #ifdef SUPPORT_UCP
3418            if ((options & PCRE_UCP) != 0)
3419              {
3420              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3421              if (posix_substitutes[pc] != NULL)
3422                {
3423                nestptr = tempptr + 1;
3424                ptr = posix_substitutes[pc] - 1;
3425                continue;
3426                }
3427              }
3428    #endif
3429            /* In the non-UCP case, we build the bit map for the POSIX class in a
3430            chunk of local store because we may be adding and subtracting from it,
3431            and we don't want to subtract bits that may be in the main map already.
3432            At the end we or the result into the bit map that is being built. */
3433    
3434            posix_class *= 3;
3435    
3436            /* Copy in the first table (always present) */
3437    
3438            memcpy(pbits, cbits + posix_class_maps[posix_class],
3439              32 * sizeof(uschar));
3440    
3441            /* If there is a second table, add or remove it as required. */
3442    
3443            taboffset = posix_class_maps[posix_class + 1];
3444            tabopt = posix_class_maps[posix_class + 2];
3445    
3446            if (taboffset >= 0)
3447              {
3448              if (tabopt >= 0)
3449                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3450              else
3451                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3452              }
3453    
3454            /* Not see if we need to remove any special characters. An option
3455            value of 1 removes vertical space and 2 removes underscore. */
3456    
3457            if (tabopt < 0) tabopt = -tabopt;
3458            if (tabopt == 1) pbits[1] &= ~0x3c;
3459              else if (tabopt == 2) pbits[11] &= 0x7f;
3460    
3461            /* Add the POSIX table or its complement into the main table that is
3462            being built and we are done. */
3463    
3464            if (local_negate)
3465              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3466            else
3467              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3468    
3469            ptr = tempptr + 1;
3470            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3471            continue;    /* End of POSIX syntax handling */
3472            }
3473    
3474          /* Backslash may introduce a single character, or it may introduce one
3475          of the specials, which just set a flag. The sequence \b is a special
3476          case. Inside a class (and only there) it is treated as backspace. We
3477          assume that other escapes have more than one character in them, so set
3478          class_charcount bigger than one. Unrecognized escapes fall through and
3479          are either treated as literal characters (by default), or are faulted if
3480          PCRE_EXTRA is set. */
3481    
3482          if (c == CHAR_BACKSLASH)
3483            {
3484            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3485            if (*errorcodeptr != 0) goto FAILED;
3486    
3487            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3488            else if (-c == ESC_Q)            /* Handle start of quoted string */
3489              {
3490              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3491                {
3492                ptr += 2; /* avoid empty string */
3493                }
3494              else inescq = TRUE;
3495              continue;
3496              }
3497            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3498    
3499            if (c < 0)
3500              {
3501              register const uschar *cbits = cd->cbits;
3502              class_charcount += 2;     /* Greater than 1 is what matters */
3503    
3504              switch (-c)
3505                {
3506    #ifdef SUPPORT_UCP
3507                case ESC_du:     /* These are the values given for \d etc */
3508                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3509                case ESC_wu:     /* escape sequence with an appropriate \p */
3510                case ESC_WU:     /* or \P to test Unicode properties instead */
3511                case ESC_su:     /* of the default ASCII testing. */
3512                case ESC_SU:
3513                nestptr = ptr;
3514                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3515                class_charcount -= 2;                /* Undo! */
3516                continue;
3517    #endif
3518                case ESC_d:
3519                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3520                continue;
3521    
3522                case ESC_D:
3523                should_flip_negation = TRUE;
3524                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3525                continue;
3526    
3527                case ESC_w:
3528                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
3529                continue;
3530    
3531                case ESC_W:
3532                should_flip_negation = TRUE;
3533                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3534                continue;
3535    
3536                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3537                if it was previously set by something earlier in the character
3538                class. */
3539    
3540                case ESC_s:
3541                classbits[0] |= cbits[cbit_space];
3542                classbits[1] |= cbits[cbit_space+1] & ~0x08;
3543                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3544                continue;
3545    
3546                case ESC_S:
3547                should_flip_negation = TRUE;
3548                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3549                classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3550                continue;
3551    
3552                case ESC_h:
3553                SETBIT(classbits, 0x09); /* VT */
3554                SETBIT(classbits, 0x20); /* SPACE */
3555                SETBIT(classbits, 0xa0); /* NSBP */
3556    #ifdef SUPPORT_UTF8
3557                if (utf8)
3558                  {
3559                  class_utf8 = TRUE;
3560                  *class_utf8data++ = XCL_SINGLE;
3561                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3562                  *class_utf8data++ = XCL_SINGLE;
3563                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3564                  *class_utf8data++ = XCL_RANGE;
3565                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3566                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3567                  *class_utf8data++ = XCL_SINGLE;
3568                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3569                  *class_utf8data++ = XCL_SINGLE;
3570                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3571                  *class_utf8data++ = XCL_SINGLE;
3572                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3573                  }
3574    #endif
3575                continue;
3576    
3577                case ESC_H:
3578                for (c = 0; c < 32; c++)
3579                  {
3580                  int x = 0xff;
3581                  switch (c)
3582                    {
3583                    case 0x09/8: x ^= 1 << (0x09%8); break;
3584                    case 0x20/8: x ^= 1 << (0x20%8); break;
3585                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3586                    default: break;
3587                    }
3588                  classbits[c] |= x;
3589                  }
3590    
3591    #ifdef SUPPORT_UTF8
3592                if (utf8)
3593                  {
3594                  class_utf8 = TRUE;
3595                  *class_utf8data++ = XCL_RANGE;
3596                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3597                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3598                  *class_utf8data++ = XCL_RANGE;
3599                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3600                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3601                  *class_utf8data++ = XCL_RANGE;
3602                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3603                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3604                  *class_utf8data++ = XCL_RANGE;
3605                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3606                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3607                  *class_utf8data++ = XCL_RANGE;
3608                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3609                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3610                  *class_utf8data++ = XCL_RANGE;
3611                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3612                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3613                  *class_utf8data++ = XCL_RANGE;
3614                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3615                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3616                  }
3617    #endif
3618                continue;
3619    
3620                case ESC_v:
3621                SETBIT(classbits, 0x0a); /* LF */
3622                SETBIT(classbits, 0x0b); /* VT */
3623                SETBIT(classbits, 0x0c); /* FF */
3624                SETBIT(classbits, 0x0d); /* CR */
3625                SETBIT(classbits, 0x85); /* NEL */
3626    #ifdef SUPPORT_UTF8
3627                if (utf8)
3628                  {
3629                  class_utf8 = TRUE;
3630                  *class_utf8data++ = XCL_RANGE;
3631                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3632                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3633                  }
3634    #endif
3635                continue;
3636    
3637                case ESC_V:
3638                for (c = 0; c < 32; c++)
3639                  {
3640                  int x = 0xff;
3641                  switch (c)
3642                    {
3643                    case 0x0a/8: x ^= 1 << (0x0a%8);
3644                                 x ^= 1 << (0x0b%8);
3645                                 x ^= 1 << (0x0c%8);
3646                                 x ^= 1 << (0x0d%8);
3647                                 break;
3648                    case 0x85/8: x ^= 1 << (0x85%8); break;
3649                    default: break;
3650                    }
3651                  classbits[c] |= x;
3652                  }
3653    
3654    #ifdef SUPPORT_UTF8
3655                if (utf8)
3656                  {
3657                  class_utf8 = TRUE;
3658                  *class_utf8data++ = XCL_RANGE;
3659                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3660                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3661                  *class_utf8data++ = XCL_RANGE;
3662                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3663                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3664                  }
3665    #endif
3666                continue;
3667    
3668    #ifdef SUPPORT_UCP
3669                case ESC_p:
3670                case ESC_P:
3671                  {
3672                  BOOL negated;
3673                  int pdata;
3674