/[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 578 by ph10, Tue Nov 23 15:34:55 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      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 235  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 271  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 285  static const unsigned char digitab[] = Line 484  static const unsigned char digitab[] =
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
490    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 518  static const unsigned char ebcdic_charta
518    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
519    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
520    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
521    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
523    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
524    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 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
549      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553    /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579  /*************************************************  /*************************************************
# Line 357  static BOOL Line 582  static BOOL
582    
583  /* 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
584  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
585  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
586  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
587  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,
588    ptr is pointing at the \. On exit, it is on the final character of the escape
589    sequence.
590    
591  Arguments:  Arguments:
592    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 370  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
# Line 388  ptr--; /* Set Line 615  ptr--; /* Set
615    
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
619  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 406  else if ((i = escapes[c - 0x48]) != 0) Line 633  else if ((i = escapes[c - 0x48]) != 0)
633  else  else
634    {    {
635    const uschar *oldptr;    const uschar *oldptr;
636      BOOL braced, negated;
637    
638    switch (c)    switch (c)
639      {      {
640      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650        /* \g must be followed by one of a number of specific things:
651    
652        (1) A number, either plain or braced. If positive, it is an absolute
653        backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676          {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685          braced = TRUE;
686          ptr++;
687          }
688        else braced = FALSE;
689    
690        if (ptr[1] == CHAR_MINUS)
691          {
692          negated = TRUE;
693          ptr++;
694          }
695        else negated = FALSE;
696    
697        c = 0;
698        while ((digitab[ptr[1]] & ctype_digit) != 0)
699          c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708          {
709          *errorcodeptr = ERR57;
710          break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717          }
718    
719        if (negated)
720          {
721          if (c > bracount)
722            {
723            *errorcodeptr = ERR15;
724            break;
725            }
726          c = bracount - (c - 1);
727          }
728    
729        c = -(ESC_REF + c);
730        break;
731    
732      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
733      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
734      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 431  else Line 741  else
741      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 452  else Line 767  else
767      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 460  else Line 775  else
775        }        }
776    
777      /* \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
778      larger first octal digit. */      larger first octal digit. The original code used just to take the least
779        significant 8 bits of octal numbers (I think this is what early Perls used
780      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
781      c -= '0';      than 3 octal digits. */
782      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
783          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
784      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
785        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786            c = c * 8 + *(++ptr) - CHAR_0;
787        if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
790      /* \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
791      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 483  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 511  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845        An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
859      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
860          break;
861  #if !EBCDIC    /* ASCII coding */        }
862      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
873      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 563  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 603  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 639  top = _pcre_utt_size; Line 973  top = _pcre_utt_size;
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top) >> 1;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0)    if (c == 0)
978      {      {
979      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 682  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 723  int max = -1; Line 1057  int max = -1;
1057  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 733  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 763  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100    *  Subroutine for finding forward reference      *
1101    *************************************************/
1102    
1103    /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106    subpatterns, and counting them. If it finds a named pattern that matches the
1107    name it is given, it returns its number. Alternatively, if the name is NULL, it
1108    returns when it reaches a given numbered subpattern. Recursion is used to keep
1109    track of subpatterns that reset the capturing group numbers - the (?| feature.
1110    
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124    Arguments:
1125      ptrptr       address of the current character pointer (updated)
1126      cd           compile background data
1127      name         name to seek, or NULL if seeking a numbered subpattern
1128      lorn         name length, or subpattern number if name is NULL
1129      xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133    Returns:       the number of the named subpattern, or -1 if not found
1134    */
1135    
1136    static int
1137    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1138      BOOL xmode, BOOL utf8, int *count)
1139    {
1140    uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145    /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149      {
1150      /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229      /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231      if (*ptr == CHAR_BACKSLASH)
1232        {
1233        if (*(++ptr) == 0) goto FAIL_EXIT;
1234        if (*ptr == CHAR_Q) for (;;)
1235          {
1236          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237          if (*ptr == 0) goto FAIL_EXIT;
1238          if (*(++ptr) == CHAR_E) break;
1239          }
1240        continue;
1241        }
1242    
1243      /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250        {
1251        BOOL negate_class = FALSE;
1252        for (;;)
1253          {
1254          if (ptr[1] == CHAR_BACKSLASH)
1255            {
1256            if (ptr[2] == CHAR_E)
1257              ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280          {
1281          if (*ptr == 0) return -1;
1282          if (*ptr == CHAR_BACKSLASH)
1283            {
1284            if (*(++ptr) == 0) goto FAIL_EXIT;
1285            if (*ptr == CHAR_Q) for (;;)
1286              {
1287              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288              if (*ptr == 0) goto FAIL_EXIT;
1289              if (*(++ptr) == CHAR_E) break;
1290              }
1291            continue;
1292            }
1293          }
1294        continue;
1295        }
1296    
1297      /* Skip comments in /x mode */
1298    
1299      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300        {
1301        ptr++;
1302        while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311        continue;
1312        }
1313    
1314      /* Check for the special metacharacters */
1315    
1316      if (*ptr == CHAR_LEFT_PARENTHESIS)
1317        {
1318        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319        if (rc > 0) return rc;
1320        if (*ptr == 0) goto FAIL_EXIT;
1321        }
1322    
1323      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324        {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    FAIL_EXIT:
1337    *ptrptr = ptr;
1338    return -1;
1339    }
1340    
1341    
1342    
1343    
1344    /*************************************************
1345    *       Find forward referenced subpattern       *
1346    *************************************************/
1347    
1348    /* This function scans along a pattern's text looking for capturing
1349    subpatterns, and counting them. If it finds a named pattern that matches the
1350    name it is given, it returns its number. Alternatively, if the name is NULL, it
1351    returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    for (;;)
1383      {
1384      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385      if (rc > 0 || *ptr++ == 0) break;
1386      }
1387    
1388    return rc;
1389    }
1390    
1391    
1392    
1393    
1394    /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
1397    
# Line 811  for (;;) Line 1440  for (;;)
1440    
1441      case OP_CALLOUT:      case OP_CALLOUT:
1442      case OP_CREF:      case OP_CREF:
1443      case OP_BRANUMBER:      case OP_NCREF:
1444        case OP_RREF:
1445        case OP_NRREF:
1446        case OP_DEF:
1447      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1448      break;      break;
1449    
# Line 826  for (;;) Line 1458  for (;;)
1458    
1459    
1460  /*************************************************  /*************************************************
1461  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1462  *************************************************/  *************************************************/
1463    
1464  /* 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,
1465  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.
1466  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
1467    temporarily terminated with OP_END when this function is called.
1468    
1469    This function is called when a backward assertion is encountered, so that if it
1470    fails, the error message can point to the correct place in the pattern.
1471    However, we cannot do this when the assertion contains subroutine calls,
1472    because they can be forward references. We solve this by remembering this case
1473    and doing the check at the end; a flag specifies which mode we are running in.
1474    
1475  Arguments:  Arguments:
1476    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1477    options  the compiling options    options  the compiling options
1478      atend    TRUE if called when the pattern is complete
1479      cd       the "compile data" structure
1480    
1481  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1482                 or -1 if there is no fixed length,
1483               or -2 if \C was encountered               or -2 if \C was encountered
1484                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1485  */  */
1486    
1487  static int  static int
1488  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1489  {  {
1490  int length = -1;  int length = -1;
1491    
# Line 855  branch, check the length against that of Line 1498  branch, check the length against that of
1498  for (;;)  for (;;)
1499    {    {
1500    int d;    int d;
1501      uschar *ce, *cs;
1502    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1503    switch (op)    switch (op)
1504      {      {
1505        case OP_CBRA:
1506      case OP_BRA:      case OP_BRA:
1507      case OP_ONCE:      case OP_ONCE:
1508      case OP_COND:      case OP_COND:
1509      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1510      if (d < 0) return d;      if (d < 0) return d;
1511      branchlength += d;      branchlength += d;
1512      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 886  for (;;) Line 1529  for (;;)
1529      branchlength = 0;      branchlength = 0;
1530      break;      break;
1531    
1532        /* A true recursion implies not fixed length, but a subroutine call may
1533        be OK. If the subroutine is a forward reference, we can't deal with
1534        it until the end of the pattern, so return -3. */
1535    
1536        case OP_RECURSE:
1537        if (!atend) return -3;
1538        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1539        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1540        if (cc > cs && cc < ce) return -1;                /* Recursion */
1541        d = find_fixedlength(cs + 2, options, atend, cd);
1542        if (d < 0) return d;
1543        branchlength += d;
1544        cc += 1 + LINK_SIZE;
1545        break;
1546    
1547      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1548    
1549      case OP_ASSERT:      case OP_ASSERT:
# Line 898  for (;;) Line 1556  for (;;)
1556      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1557    
1558      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1559      case OP_CREF:      case OP_CREF:
1560        case OP_NCREF:
1561        case OP_RREF:
1562        case OP_NRREF:
1563        case OP_DEF:
1564      case OP_OPT:      case OP_OPT:
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
# Line 917  for (;;) Line 1579  for (;;)
1579    
1580      case OP_CHAR:      case OP_CHAR:
1581      case OP_CHARNC:      case OP_CHARNC:
1582        case OP_NOT:
1583      branchlength++;      branchlength++;
1584      cc += 2;      cc += 2;
1585  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1586      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1587        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1588  #endif  #endif
1589      break;      break;
1590    
# Line 934  for (;;) Line 1595  for (;;)
1595      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1596      cc += 4;      cc += 4;
1597  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1598      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1599        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1600  #endif  #endif
1601      break;      break;
1602    
1603      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1604      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1605        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1606      cc += 4;      cc += 4;
1607      break;      break;
1608    
# Line 960  for (;;) Line 1620  for (;;)
1620      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1621      case OP_WORDCHAR:      case OP_WORDCHAR:
1622      case OP_ANY:      case OP_ANY:
1623        case OP_ALLANY:
1624      branchlength++;      branchlength++;
1625      cc++;      cc++;
1626      break;      break;
# Line 1014  for (;;) Line 1675  for (;;)
1675    
1676    
1677  /*************************************************  /*************************************************
1678  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1679  *************************************************/  *************************************************/
1680    
1681  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1682  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1683    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1684    so that it can be called from pcre_study() when finding the minimum matching
1685    length.
1686    
1687  Arguments:  Arguments:
1688    code        points to start of expression    code        points to start of expression
1689    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1690    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1691    
1692  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
1693  */  */
1694    
1695  static const uschar *  const uschar *
1696  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1697  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1698  for (;;)  for (;;)
1699    {    {
1700    register int c = *code;    register int c = *code;
1701    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1702    else if (c > OP_BRA)  
1703      /* XCLASS is used for classes that cannot be represented just by a bit
1704      map. This includes negated single high-valued characters. The length in
1705      the table is zero; the actual length is stored in the compiled code. */
1706    
1707      if (c == OP_XCLASS) code += GET(code, 1);
1708    
1709      /* Handle recursion */
1710    
1711      else if (c == OP_REVERSE)
1712      {      {
1713      int n = c - OP_BRA;      if (number < 0) return (uschar *)code;
1714      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];  
1715      }      }
1716    else  
1717      /* Handle capturing bracket */
1718    
1719      else if (c == OP_CBRA)
1720      {      {
1721        int n = GET2(code, 1+LINK_SIZE);
1722        if (n == number) return (uschar *)code;
1723      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1724        }
1725    
1726  #ifdef SUPPORT_UTF8    /* Otherwise, we can get the item's length from the table, except that for
1727      repeated character types, we have to test for \p and \P, which have an extra
1728      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731      /* In UTF-8 mode, opcodes that are followed by a character may be followed    else
1732      by a multi-byte character. The length in the table is a minimum, so we have      {
1733      to scan along to skip the extra bytes. All opcodes are less than 128, so we      switch(c)
1734      can use relatively efficient code. */        {
1735          case OP_TYPESTAR:
1736          case OP_TYPEMINSTAR:
1737          case OP_TYPEPLUS:
1738          case OP_TYPEMINPLUS:
1739          case OP_TYPEQUERY:
1740          case OP_TYPEMINQUERY:
1741          case OP_TYPEPOSSTAR:
1742          case OP_TYPEPOSPLUS:
1743          case OP_TYPEPOSQUERY:
1744          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1745          break;
1746    
1747          case OP_TYPEUPTO:
1748          case OP_TYPEMINUPTO:
1749          case OP_TYPEEXACT:
1750          case OP_TYPEPOSUPTO:
1751          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752          break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          code += code[1];
1758          break;
1759    
1760          case OP_THEN_ARG:
1761          code += code[1+LINK_SIZE];
1762          break;
1763          }
1764    
1765        /* Add in the fixed length from the table */
1766    
1767        code += _pcre_OP_lengths[c];
1768    
1769      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1770      a multi-byte character. The length in the table is a minimum, so we have to
1771      arrange to skip the extra bytes. */
1772    
1773    #ifdef SUPPORT_UTF8
1774      if (utf8) switch(c)      if (utf8) switch(c)
1775        {        {
1776        case OP_CHAR:        case OP_CHAR:
# Line 1064  for (;;) Line 1778  for (;;)
1778        case OP_EXACT:        case OP_EXACT:
1779        case OP_UPTO:        case OP_UPTO:
1780        case OP_MINUPTO:        case OP_MINUPTO:
1781          case OP_POSUPTO:
1782        case OP_STAR:        case OP_STAR:
1783        case OP_MINSTAR:        case OP_MINSTAR:
1784          case OP_POSSTAR:
1785        case OP_PLUS:        case OP_PLUS:
1786        case OP_MINPLUS:        case OP_MINPLUS:
1787          case OP_POSPLUS:
1788        case OP_QUERY:        case OP_QUERY:
1789        case OP_MINQUERY:        case OP_MINQUERY:
1790        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1791        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;  
1792        break;        break;
1793        }        }
1794    #else
1795        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1796  #endif  #endif
1797      }      }
1798    }    }
# Line 1105  Returns: pointer to the opcode for Line 1817  Returns: pointer to the opcode for
1817  static const uschar *  static const uschar *
1818  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1819  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1820  for (;;)  for (;;)
1821    {    {
1822    register int c = *code;    register int c = *code;
1823    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1824    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1825    else if (c > OP_BRA)  
1826      {    /* XCLASS is used for classes that cannot be represented just by a bit
1827      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1828      }    the table is zero; the actual length is stored in the compiled code. */
1829    
1830      if (c == OP_XCLASS) code += GET(code, 1);
1831    
1832      /* Otherwise, we can get the item's length from the table, except that for
1833      repeated character types, we have to test for \p and \P, which have an extra
1834      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1835      must add in its length. */
1836    
1837    else    else
1838      {      {
1839      code += _pcre_OP_lengths[c];      switch(c)
1840          {
1841          case OP_TYPESTAR:
1842          case OP_TYPEMINSTAR:
1843          case OP_TYPEPLUS:
1844          case OP_TYPEMINPLUS:
1845          case OP_TYPEQUERY:
1846          case OP_TYPEMINQUERY:
1847          case OP_TYPEPOSSTAR:
1848          case OP_TYPEPOSPLUS:
1849          case OP_TYPEPOSQUERY:
1850          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1851          break;
1852    
1853  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1854          case OP_TYPEUPTO:
1855          case OP_TYPEMINUPTO:
1856          case OP_TYPEEXACT:
1857          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1858          break;
1859    
1860          case OP_MARK:
1861          case OP_PRUNE_ARG:
1862          case OP_SKIP_ARG:
1863          code += code[1];
1864          break;
1865    
1866          case OP_THEN_ARG:
1867          code += code[1+LINK_SIZE];
1868          break;
1869          }
1870    
1871        /* Add in the fixed length from the table */
1872    
1873        code += _pcre_OP_lengths[c];
1874    
1875      /* 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
1876      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
1877      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. */  
1878    
1879    #ifdef SUPPORT_UTF8
1880      if (utf8) switch(c)      if (utf8) switch(c)
1881        {        {
1882        case OP_CHAR:        case OP_CHAR:
# Line 1136  for (;;) Line 1884  for (;;)
1884        case OP_EXACT:        case OP_EXACT:
1885        case OP_UPTO:        case OP_UPTO:
1886        case OP_MINUPTO:        case OP_MINUPTO:
1887          case OP_POSUPTO:
1888        case OP_STAR:        case OP_STAR:
1889        case OP_MINSTAR:        case OP_MINSTAR:
1890          case OP_POSSTAR:
1891        case OP_PLUS:        case OP_PLUS:
1892        case OP_MINPLUS:        case OP_MINPLUS:
1893          case OP_POSPLUS:
1894        case OP_QUERY:        case OP_QUERY:
1895        case OP_MINQUERY:        case OP_MINQUERY:
1896        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1897        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;  
1898        break;        break;
1899        }        }
1900    #else
1901        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1902  #endif  #endif
1903      }      }
1904    }    }
# Line 1165  for (;;) Line 1911  for (;;)
1911  *************************************************/  *************************************************/
1912    
1913  /* 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
1914  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()
1915  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
1916  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
1917  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1918    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1919    bracket whose current branch will already have been scanned.
1920    
1921  Arguments:  Arguments:
1922    code        points to start of search    code        points to start of search
1923    endcode     points to where to stop    endcode     points to where to stop
1924    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1925      cd          contains pointers to tables etc.
1926    
1927  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1928  */  */
1929    
1930  static BOOL  static BOOL
1931  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1932      compile_data *cd)
1933  {  {
1934  register int c;  register int c;
1935  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);
1936       code < endcode;       code < endcode;
1937       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1938    {    {
# Line 1190  for (code = first_significant_code(code Line 1940  for (code = first_significant_code(code
1940    
1941    c = *code;    c = *code;
1942    
1943    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1944      first_significant_code() with a TRUE final argument. */
1945    
1946      if (c == OP_ASSERT)
1947      {      {
1948      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1949      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1950        continue;
1951        }
1952    
1953      /* Groups with zero repeats can of course be empty; skip them. */
1954    
1955      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1956        {
1957        code += _pcre_OP_lengths[c];
1958        do code += GET(code, 1); while (*code == OP_ALT);
1959        c = *code;
1960        continue;
1961        }
1962    
1963      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1964      implies a subroutine call, we can scan it. */
1965    
1966      empty_branch = FALSE;    if (c == OP_RECURSE)
1967        {
1968        BOOL empty_branch = FALSE;
1969        const uschar *scode = cd->start_code + GET(code, 1);
1970        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1971      do      do
1972        {        {
1973        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1974            {
1975          empty_branch = TRUE;          empty_branch = TRUE;
1976            break;
1977            }
1978          scode += GET(scode, 1);
1979          }
1980        while (*scode == OP_ALT);
1981        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1982        continue;
1983        }
1984    
1985      /* For other groups, scan the branches. */
1986    
1987      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1988        {
1989        BOOL empty_branch;
1990        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1991    
1992        /* If a conditional group has only one branch, there is a second, implied,
1993        empty branch, so just skip over the conditional, because it could be empty.
1994        Otherwise, scan the individual branches of the group. */
1995    
1996        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1997        code += GET(code, 1);        code += GET(code, 1);
1998        else
1999          {
2000          empty_branch = FALSE;
2001          do
2002            {
2003            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2004              empty_branch = TRUE;
2005            code += GET(code, 1);
2006            }
2007          while (*code == OP_ALT);
2008          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2009        }        }
2010      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2011      c = *code;      c = *code;
2012        continue;
2013      }      }
2014    
2015    else switch (c)    /* Handle the other opcodes */
2016    
2017      switch (c)
2018      {      {
2019      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2020        cannot be represented just by a bit map. This includes negated single
2021        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2022        actual length is stored in the compiled code, so we must update "code"
2023        here. */
2024    
2025  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2026      case OP_XCLASS:      case OP_XCLASS:
2027      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2028      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2029  #endif  #endif
2030    
# Line 1260  for (code = first_significant_code(code Line 2068  for (code = first_significant_code(code
2068      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2069      case OP_WORDCHAR:      case OP_WORDCHAR:
2070      case OP_ANY:      case OP_ANY:
2071        case OP_ALLANY:
2072      case OP_ANYBYTE:      case OP_ANYBYTE:
2073      case OP_CHAR:      case OP_CHAR:
2074      case OP_CHARNC:      case OP_CHARNC:
2075      case OP_NOT:      case OP_NOT:
2076      case OP_PLUS:      case OP_PLUS:
2077      case OP_MINPLUS:      case OP_MINPLUS:
2078        case OP_POSPLUS:
2079      case OP_EXACT:      case OP_EXACT:
2080      case OP_NOTPLUS:      case OP_NOTPLUS:
2081      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2082        case OP_NOTPOSPLUS:
2083      case OP_NOTEXACT:      case OP_NOTEXACT:
2084      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2085      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2086        case OP_TYPEPOSPLUS:
2087      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2088      return FALSE;      return FALSE;
2089    
2090        /* These are going to continue, as they may be empty, but we have to
2091        fudge the length for the \p and \P cases. */
2092    
2093        case OP_TYPESTAR:
2094        case OP_TYPEMINSTAR:
2095        case OP_TYPEPOSSTAR:
2096        case OP_TYPEQUERY:
2097        case OP_TYPEMINQUERY:
2098        case OP_TYPEPOSQUERY:
2099        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2100        break;
2101    
2102        /* Same for these */
2103    
2104        case OP_TYPEUPTO:
2105        case OP_TYPEMINUPTO:
2106        case OP_TYPEPOSUPTO:
2107        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2108        break;
2109    
2110      /* End of branch */      /* End of branch */
2111    
2112      case OP_KET:      case OP_KET:
# Line 1283  for (code = first_significant_code(code Line 2115  for (code = first_significant_code(code
2115      case OP_ALT:      case OP_ALT:
2116      return TRUE;      return TRUE;
2117    
2118      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2119      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2120    
2121  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2122      case OP_STAR:      case OP_STAR:
2123      case OP_MINSTAR:      case OP_MINSTAR:
2124        case OP_POSSTAR:
2125      case OP_QUERY:      case OP_QUERY:
2126      case OP_MINQUERY:      case OP_MINQUERY:
2127        case OP_POSQUERY:
2128        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2129        break;
2130    
2131      case OP_UPTO:      case OP_UPTO:
2132      case OP_MINUPTO:      case OP_MINUPTO:
2133      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2134        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2135      break;      break;
2136  #endif  #endif
2137    
2138        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2139        string. */
2140    
2141        case OP_MARK:
2142        case OP_PRUNE_ARG:
2143        case OP_SKIP_ARG:
2144        code += code[1];
2145        break;
2146    
2147        case OP_THEN_ARG:
2148        code += code[1+LINK_SIZE];
2149        break;
2150    
2151        /* None of the remaining opcodes are required to match a character. */
2152    
2153        default:
2154        break;
2155      }      }
2156    }    }
2157    
# Line 1318  Arguments: Line 2174  Arguments:
2174    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2175    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2176    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2177      cd          pointers to tables etc
2178    
2179  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2180  */  */
2181    
2182  static BOOL  static BOOL
2183  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2184    BOOL utf8)    BOOL utf8, compile_data *cd)
2185  {  {
2186  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2187    {    {
2188    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2189        return FALSE;
2190    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2191    }    }
2192  return TRUE;  return TRUE;
# Line 1341  return TRUE; Line 2199  return TRUE;
2199  *************************************************/  *************************************************/
2200    
2201  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2202  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
2203  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2204  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2205    
2206    Originally, this function only recognized a sequence of letters between the
2207    terminators, but it seems that Perl recognizes any sequence of characters,
2208    though of course unknown POSIX names are subsequently rejected. Perl gives an
2209    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2210    didn't consider this to be a POSIX class. Likewise for [:1234:].
2211    
2212    The problem in trying to be exactly like Perl is in the handling of escapes. We
2213    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2214    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2215    below handles the special case of \], but does not try to do any other escape
2216    processing. This makes it different from Perl for cases such as [:l\ower:]
2217    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2218    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2219    I think.
2220    
2221  Argument:  Arguments:
2222    ptr      pointer to the initial [    ptr      pointer to the initial [
2223    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2224    
2225  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2226  */  */
2227    
2228  static BOOL  static BOOL
2229  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2230  {  {
2231  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2232  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2233  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2234    {    {
2235    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2236    return TRUE;      {
2237        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2238        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2239          {
2240          *endptr = ptr;
2241          return TRUE;
2242          }
2243        }
2244    }    }
2245  return FALSE;  return FALSE;
2246  }  }
# Line 1388  Returns: a value representing the na Line 2265  Returns: a value representing the na
2265  static int  static int
2266  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2267  {  {
2268    const char *pn = posix_names;
2269  register int yield = 0;  register int yield = 0;
2270  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2271    {    {
2272    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2273      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2274      pn += posix_name_lengths[yield] + 1;
2275    yield++;    yield++;
2276    }    }
2277  return -1;  return -1;
# Line 1407  return -1; Line 2286  return -1;
2286  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2287  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2288  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
2289  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
2290  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
2291  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
2292  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
2293  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2294    OP_END.
2295    
2296    This function has been extended with the possibility of forward references for
2297    recursions and subroutine calls. It must also check the list of such references
2298    for the group we are dealing with. If it finds that one of the recursions in
2299    the current group is on this list, it adjusts the offset in the list, not the
2300    value in the reference (which is a group number).
2301    
2302  Arguments:  Arguments:
2303    group      points to the start of the group    group      points to the start of the group
2304    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2305    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2306    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2307      save_hwm   the hwm forward reference pointer at the start of the group
2308    
2309  Returns:     nothing  Returns:     nothing
2310  */  */
2311    
2312  static void  static void
2313  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2314      uschar *save_hwm)
2315  {  {
2316  uschar *ptr = group;  uschar *ptr = group;
2317    
2318  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2319    {    {
2320    int offset = GET(ptr, 1);    int offset;
2321    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2322    
2323      /* See if this recursion is on the forward reference list. If so, adjust the
2324      reference. */
2325    
2326      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2327        {
2328        offset = GET(hc, 0);
2329        if (cd->start_code + offset == ptr + 1)
2330          {
2331          PUT(hc, 0, offset + adjust);
2332          break;
2333          }
2334        }
2335    
2336      /* Otherwise, adjust the recursion offset if it's after the start of this
2337      group. */
2338    
2339      if (hc >= cd->hwm)
2340        {
2341        offset = GET(ptr, 1);
2342        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2343        }
2344    
2345    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2346    }    }
2347  }  }
# Line 1456  auto_callout(uschar *code, const uschar Line 2368  auto_callout(uschar *code, const uschar
2368  {  {
2369  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2370  *code++ = 255;  *code++ = 255;
2371  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2372  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2373  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2374  }  }
2375    
# Line 1482  Returns: nothing Line 2394  Returns: nothing
2394  static void  static void
2395  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2396  {  {
2397  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2398  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2399  }  }
2400    
# Line 1508  Yield: TRUE when range returned; Line 2420  Yield: TRUE when range returned;
2420  */  */
2421    
2422  static BOOL  static BOOL
2423  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2424      unsigned int *odptr)
2425  {  {
2426  int c, othercase, next;  unsigned int c, othercase, next;
2427    
2428  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2429    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2430    
2431  if (c > d) return FALSE;  if (c > d) return FALSE;
2432    
# Line 1522  next = othercase + 1; Line 2435  next = othercase + 1;
2435    
2436  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2437    {    {
2438    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2439    next++;    next++;
2440    }    }
2441    
# Line 1531  for (++c; c <= d; c++) Line 2444  for (++c; c <= d; c++)
2444    
2445  return TRUE;  return TRUE;
2446  }  }
2447  #endif  /* SUPPORT_UCP */  
2448    
2449    
2450  /*************************************************  /*************************************************
2451  *           Compile one branch                   *  *        Check a character and a property        *
2452  *************************************************/  *************************************************/
2453    
2454  /* 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
2455  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2456    
2457  Arguments:  Arguments:
2458    optionsptr     pointer to the option bits    c            the character
2459    brackets       points to number of extracting brackets used    ptype        the property type
2460    codeptr        points to the pointer to the current code point    pdata        the data for the type
2461    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.  
2462    
2463  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2464  */  */
2465    
2466  static BOOL  static BOOL
2467  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)  
2468  {  {
2469  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2470  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2471  int bravalue = 0;    {
2472  int greedy_default, greedy_non_default;    case PT_LAMP:
2473  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2474  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2475  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2476  int condcount = 0;  
2477  int options = *optionsptr;    case PT_GC:
2478  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2479  register int c;  
2480  register uschar *code = *codeptr;    case PT_PC:
2481  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2482  BOOL inescq = FALSE;  
2483  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2484  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2485  const uschar *tempptr;  
2486  uschar *previous = NULL;    /* These are specials */
2487  uschar *previous_callout = NULL;  
2488  uschar classbits[32];    case PT_ALNUM:
2489      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2490  #ifdef SUPPORT_UTF8            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2491  BOOL class_utf8;  
2492  BOOL utf8 = (options & PCRE_UTF8) != 0;    case PT_SPACE:    /* Perl space */
2493  uschar *class_utf8data;    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2494  uschar utf8_char[6];            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2495  #else            == negated;
2496  BOOL utf8 = FALSE;  
2497  #endif    case PT_PXSPACE:  /* POSIX space */
2498      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2499              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2500              c == CHAR_FF || c == CHAR_CR)
2501              == negated;
2502    
2503      case PT_WORD:
2504      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2505              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2506              c == CHAR_UNDERSCORE) == negated;
2507      }
2508    return FALSE;
2509    }
2510    #endif  /* SUPPORT_UCP */
2511    
 /* Set up the default and non-default settings for greediness */  
2512    
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2513    
2514  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /*************************************************
2515  matching encountered yet". It gets changed to REQ_NONE if we hit something that  *     Check if auto-possessifying is possible    *
2516  matches a non-fixed char first char; reqbyte just remains unset if we never  *************************************************/
 find one.  
2517    
2518  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
2519  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
2520  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. */  
2521    
2522  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Arguments:
2523      previous      pointer to the repeated opcode
2524      utf8          TRUE in UTF-8 mode
2525      ptr           next character in pattern
2526      options       options bits
2527      cd            contains pointers to tables etc.
2528    
2529  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  Returns:        TRUE if possessifying is wanted
2530  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. */  
2531    
2532  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  static BOOL
2533    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2534      int options, compile_data *cd)
2535    {
2536    int c, next;
2537    int op_code = *previous++;
2538    
2539  /* Switch on next character until the end of the branch */  /* Skip whitespace and comments in extended mode */
2540    
2541  for (;; ptr++)  if ((options & PCRE_EXTENDED) != 0)
2542    {    {
2543    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)  
2544      {      {
2545      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2546        {      if (*ptr == CHAR_NUMBER_SIGN)
       inescq = FALSE;  
       ptr++;  
       continue;  
       }  
     else  
2547        {        {
2548        if (previous_callout != NULL)        ptr++;
2549          {        while (*ptr != 0)
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
2550          {          {
2551          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2552          code = auto_callout(code, ptr, cd);          ptr++;
2553    #ifdef SUPPORT_UTF8
2554            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2555    #endif
2556          }          }
       goto NORMAL_CHAR;  
2557        }        }
2558        else break;
2559      }      }
2560      }
2561    
2562    /* 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
2563    a quantifier. */  value is a character, a negative value is an escape value. */
2564    
2565    is_quantifier = c == '*' || c == '+' || c == '?' ||  if (*ptr == CHAR_BACKSLASH)
2566      (c == '{' && is_counted_repeat(ptr+1));    {
2567      int temperrorcode = 0;
2568      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2569      if (temperrorcode != 0) return FALSE;
2570      ptr++;    /* Point after the escape sequence */
2571      }
2572    
2573    if (!is_quantifier && previous_callout != NULL &&  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2574         after_manual_callout-- <= 0)    {
2575      {  #ifdef SUPPORT_UTF8
2576      complete_callout(previous_callout, ptr, cd);    if (utf8) { GETCHARINC(next, ptr); } else
2577      previous_callout = NULL;  #endif
2578      }    next = *ptr++;
2579      }
2580    
2581    /* In extended mode, skip white space and comments */  else return FALSE;
2582    
2583    if ((options & PCRE_EXTENDED) != 0)  /* Skip whitespace and comments in extended mode */
2584    
2585    if ((options & PCRE_EXTENDED) != 0)
2586      {
2587      for (;;)
2588      {      {
2589      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2590      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2591        {        {
2592        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
2593        on the Macintosh. */        while (*ptr != 0)
2594        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
2595        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2596            ptr++;
2597    #ifdef SUPPORT_UTF8
2598            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2599    #endif
2600            }
2601        }        }
2602        else break;
2603      }      }
2604      }
2605    
2606    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
2607    
2608    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2609      {    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2610      previous_callout = code;      return FALSE;
     code = auto_callout(code, ptr, cd);  
     }  
2611    
2612    switch(c)  /* Now compare the next item with the previous opcode. First, handle cases when
2613      {  the next item is a character. */
     /* The branch terminates at end of string, |, or ). */  
2614    
2615      case 0:  if (next >= 0) switch(op_code)
2616      case '|':    {
2617      case ')':    case OP_CHAR:
2618      *firstbyteptr = firstbyte;  #ifdef SUPPORT_UTF8
2619      *reqbyteptr = reqbyte;    GETCHARTEST(c, previous);
2620      *codeptr = code;  #else
2621      *ptrptr = ptr;    c = *previous;
2622      return TRUE;  #endif
2623      return c != next;
2624    
2625      /* Handle single-character metacharacters. In multiline mode, ^ disables    /* For CHARNC (caseless character) we must check the other case. If we have
2626      the setting of any following char as a first character. */    Unicode property support, we can use it to test the other case of
2627      high-valued characters. */
2628    
2629      case '^':    case OP_CHARNC:
2630      if ((options & PCRE_MULTILINE) != 0)  #ifdef SUPPORT_UTF8
2631        {    GETCHARTEST(c, previous);
2632        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  #else
2633        }    c = *previous;
2634      previous = NULL;  #endif
2635      *code++ = OP_CIRC;    if (c == next) return FALSE;
2636      break;  #ifdef SUPPORT_UTF8
2637      if (utf8)
2638        {
2639        unsigned int othercase;
2640        if (next < 128) othercase = cd->fcc[next]; else
2641    #ifdef SUPPORT_UCP
2642        othercase = UCD_OTHERCASE((unsigned int)next);
2643    #else
2644        othercase = NOTACHAR;
2645    #endif
2646        return (unsigned int)c != othercase;
2647        }
2648      else
2649    #endif  /* SUPPORT_UTF8 */
2650      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2651    
2652      case '$':    /* For OP_NOT, its data is always a single-byte character. */
     previous = NULL;  
     *code++ = OP_DOLL;  
     break;  
2653    
2654      /* There can never be a first char if '.' is first, whatever happens about    case OP_NOT:
2655      repeats. The value of reqbyte doesn't change either. */    if ((c = *previous) == next) return TRUE;
2656      if ((options & PCRE_CASELESS) == 0) return FALSE;
2657    #ifdef SUPPORT_UTF8
2658      if (utf8)
2659        {
2660        unsigned int othercase;
2661        if (next < 128) othercase = cd->fcc[next]; else
2662    #ifdef SUPPORT_UCP
2663        othercase = UCD_OTHERCASE(next);
2664    #else
2665        othercase = NOTACHAR;
2666    #endif
2667        return (unsigned int)c == othercase;
2668        }
2669      else
2670    #endif  /* SUPPORT_UTF8 */
2671      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2672    
2673      case '.':    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2674      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;  
2675    
2676      /* Character classes. If the included characters are all < 256, we build a    case OP_DIGIT:
2677      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.  
2678    
2679      If the class contains characters outside the 0-255 range, a different    case OP_NOT_DIGIT:
2680      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.  
     */  
2681    
2682      case '[':    case OP_WHITESPACE:
2683      previous = code;    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2684    
2685      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if    case OP_NOT_WHITESPACE:
2686      they are encountered at the top level, so we'll do that too. */    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2687    
2688      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    case OP_WORDCHAR:
2689          check_posix_syntax(ptr, &tempptr, cd))    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
       }  
2690    
2691      /* If the first character is '^', set the negation flag and skip it. */    case OP_NOT_WORDCHAR:
2692      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2693    
2694      if ((c = *(++ptr)) == '^')    case OP_HSPACE:
2695        {    case OP_NOT_HSPACE:
2696        negate_class = TRUE;    switch(next)
2697        c = *(++ptr);      {
2698        }      case 0x09:
2699      else      case 0x20:
2700        {      case 0xa0:
2701        negate_class = FALSE;      case 0x1680:
2702        }      case 0x180e:
2703        case 0x2000:
2704        case 0x2001:
2705        case 0x2002:
2706        case 0x2003:
2707        case 0x2004:
2708        case 0x2005:
2709        case 0x2006:
2710        case 0x2007:
2711        case 0x2008:
2712        case 0x2009:
2713        case 0x200A:
2714        case 0x202f:
2715        case 0x205f:
2716        case 0x3000:
2717        return op_code == OP_NOT_HSPACE;
2718        default:
2719        return op_code != OP_NOT_HSPACE;
2720        }
2721    
2722      /* Keep a count of chars with values < 256 so that we can optimize the case    case OP_ANYNL:
2723      of just a single character (as long as it's < 256). For higher valued UTF-8    case OP_VSPACE:
2724      characters, we don't yet do any optimization. */    case OP_NOT_VSPACE:
2725      switch(next)
2726        {
2727        case 0x0a:
2728        case 0x0b:
2729        case 0x0c:
2730        case 0x0d:
2731        case 0x85:
2732        case 0x2028:
2733        case 0x2029:
2734        return op_code == OP_NOT_VSPACE;
2735        default:
2736        return op_code != OP_NOT_VSPACE;
2737        }
2738    
2739      class_charcount = 0;  #ifdef SUPPORT_UCP
2740      class_lastchar = -1;    case OP_PROP:
2741      return check_char_prop(next, previous[0], previous[1], FALSE);
2742    
2743  #ifdef SUPPORT_UTF8    case OP_NOTPROP:
2744      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 */  
2745  #endif  #endif
2746    
2747      /* Initialize the 32-char bit map to all zeros. We have to build the    default:
2748      map in a temporary bit of store, in case the class contains only 1    return FALSE;
2749      character (< 256), because in that case the compiled code doesn't use the    }
     bit map. */  
2750    
     memset(classbits, 0, 32 * sizeof(uschar));  
2751    
2752      /* 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
2753      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
2754      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
2755      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
2756      character. */  replaced by OP_PROP codes when PCRE_UCP is set. */
2757    
2758      do  switch(op_code)
2759        {    {
2760      case OP_CHAR:
2761      case OP_CHARNC:
2762  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2763        if (utf8 && c > 127)    GETCHARTEST(c, previous);
2764          {                           /* Braces are required because the */  #else
2765          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */    c = *previous;
         }  
2766  #endif  #endif
2767      switch(-next)
2768        {
2769        case ESC_d:
2770        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2771    
2772        /* Inside \Q...\E everything is literal except \E */      case ESC_D:
2773        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. */  
2774    
2775        if (c == '[' &&      case ESC_s:
2776            (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];  
2777    
2778          if (ptr[1] != ':')      case ESC_S:
2779            {      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2780            *errorcodeptr = ERR31;  
2781            goto FAILED;      case ESC_w:
2782            }      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2783    
2784        case ESC_W:
2785        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2786    
2787        case ESC_h:
2788        case ESC_H:
2789        switch(c)
2790          {
2791          case 0x09:
2792          case 0x20:
2793          case 0xa0:
2794          case 0x1680:
2795          case 0x180e:
2796          case 0x2000:
2797          case 0x2001:
2798          case 0x2002:
2799          case 0x2003:
2800          case 0x2004:
2801          case 0x2005:
2802          case 0x2006:
2803          case 0x2007:
2804          case 0x2008:
2805          case 0x2009:
2806          case 0x200A:
2807          case 0x202f:
2808          case 0x205f:
2809          case 0x3000:
2810          return -next != ESC_h;
2811          default:
2812          return -next == ESC_h;
2813          }
2814    
2815          ptr += 2;      case ESC_v:
2816          if (*ptr == '^')      case ESC_V:
2817            {      switch(c)
2818            local_negate = TRUE;        {
2819            ptr++;        case 0x0a:
2820            }        case 0x0b:
2821          case 0x0c:
2822          case 0x0d:
2823          case 0x85:
2824          case 0x2028:
2825          case 0x2029:
2826          return -next != ESC_v;
2827          default:
2828          return -next == ESC_v;
2829          }
2830    
2831          posix_class = check_posix_name(ptr, tempptr - ptr);      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2832          if (posix_class < 0)      their substitutions and process them. The result will always be either
2833            {      -ESC_p or -ESC_P. Then fall through to process those values. */
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
2834    
2835          /* If matching is caseless, upper and lower are converted to  #ifdef SUPPORT_UCP
2836          alpha. This relies on the fact that the class table starts with      case ESC_du:
2837          alpha, lower, upper as the first 3 entries. */      case ESC_DU:
2838        case ESC_wu:
2839        case ESC_WU:
2840        case ESC_su:
2841        case ESC_SU:
2842          {
2843          int temperrorcode = 0;
2844          ptr = substitutes[-next - ESC_DU];
2845          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2846          if (temperrorcode != 0) return FALSE;
2847          ptr++;    /* For compatibility */
2848          }
2849        /* Fall through */
2850    
2851          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)      case ESC_p:
2852            posix_class = 0;      case ESC_P:
2853          {
2854          int ptype, pdata, errorcodeptr;
2855          BOOL negated;
2856    
2857          /* We build the bit map for the POSIX class in a chunk of local store        ptr--;      /* Make ptr point at the p or P */
2858          because we may be adding and subtracting from it, and we don't want to        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2859          subtract bits that may be in the main map already. At the end we or the        if (ptype < 0) return FALSE;
2860          result into the bit map that is being built. */        ptr++;      /* Point past the final curly ket */
2861    
2862          /* If the property item is optional, we have to give up. (When generated
2863          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2864          to the original \d etc. At this point, ptr will point to a zero byte. */
2865    
2866          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2867            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2868              return FALSE;
2869    
2870          posix_class *= 3;        /* Do the property check. */
2871    
2872          /* Copy in the first table (always present) */        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2873          }
2874    #endif
2875    
2876          memcpy(pbits, cbits + posix_class_maps[posix_class],      default:
2877            32 * sizeof(uschar));      return FALSE;
2878        }
2879    
2880          /* If there is a second table, add or remove it as required. */    /* In principle, support for Unicode properties should be integrated here as
2881      well. It means re-organizing the above code so as to get hold of the property
2882      values before switching on the op-code. However, I wonder how many patterns
2883      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2884      these op-codes are never generated.) */
2885    
2886          taboffset = posix_class_maps[posix_class + 1];    case OP_DIGIT:
2887          tabopt = posix_class_maps[posix_class + 2];    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2888             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2889    
2890          if (taboffset >= 0)    case OP_NOT_DIGIT:
2891            {    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];  
           }  
2892    
2893          /* Not see if we need to remove any special characters. An option    case OP_WHITESPACE:
2894          value of 1 removes vertical space and 2 removes underscore. */    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2895    
2896          if (tabopt < 0) tabopt = -tabopt;    case OP_NOT_WHITESPACE:
2897          if (tabopt == 1) pbits[1] &= ~0x3c;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
           else if (tabopt == 2) pbits[11] &= 0x7f;  
2898    
2899          /* Add the POSIX table or its complement into the main table that is    case OP_HSPACE:
2900          being built and we are done. */    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2901             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2902    
2903          if (local_negate)    case OP_NOT_HSPACE:
2904            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];  
2905    
2906          ptr = tempptr + 1;    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2907          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */    case OP_ANYNL:
2908          continue;    /* End of POSIX syntax handling */    case OP_VSPACE:
2909          }    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2910    
2911        /* Backslash may introduce a single character, or it may introduce one    case OP_NOT_VSPACE:
2912        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. */  
2913    
2914        if (c == '\\')    case OP_WORDCHAR:
2915          {    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2916          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);           next == -ESC_v || next == -ESC_R;
2917    
2918          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */    case OP_NOT_WORDCHAR:
2919          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;  
           }  
2920    
2921          if (c < 0)    default:
2922            {    return FALSE;
2923            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;  
2924    
2925              case ESC_D:  /* Control does not reach here */
2926              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];  }
             continue;  
2927    
             case ESC_w:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];  
             continue;  
2928    
             case ESC_W:  
             for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];  
             continue;  
2929    
2930              case ESC_s:  /*************************************************
2931              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  *           Compile one branch                   *
2932              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  *************************************************/
             continue;  
2933    
2934              case ESC_S:  /* Scan the pattern, compiling it into the a vector. If the options are
2935              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
2936              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
2937              continue;  to find out the amount of memory needed, as well as during the real compile
2938    phase. The value of lengthptr distinguishes the two phases.
2939    
2940  #ifdef SUPPORT_UCP  Arguments:
2941              case ESC_p:    optionsptr     pointer to the option bits
2942              case ESC_P:    codeptr        points to the pointer to the current code point
2943                {    ptrptr         points to the current pattern pointer
2944                BOOL negated;    errorcodeptr   points to error code variable
2945                int pdata;    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2946                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);    reqbyteptr     set to the last literal character required, else < 0
2947                if (ptype < 0) goto FAILED;    bcptr          points to current branch chain
2948                class_utf8 = TRUE;    cd             contains pointers to tables etc.
2949                *class_utf8data++ = ((-c == ESC_p) != negated)?    lengthptr      NULL during the real compile phase
2950                  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  
2951    
2952              /* Unrecognized escapes are faulted if PCRE is running in its  Returns:         TRUE on success
2953              strict mode. By default, for compatibility with Perl, they are                   FALSE, with *errorcodeptr set non-zero on error
2954              treated as literals. */  */
2955    
2956              default:  static BOOL
2957              if ((options & PCRE_EXTRA) != 0)  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2958                {    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2959                *errorcodeptr = ERR7;    compile_data *cd, int *lengthptr)
2960                goto FAILED;  {
2961    int repeat_type, op_type;
2962    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2963    int bravalue = 0;
2964    int greedy_default, greedy_non_default;
2965    int firstbyte, reqbyte;
2966    int zeroreqbyte, zerofirstbyte;
2967    int req_caseopt, reqvary, tempreqvary;
2968    int options = *optionsptr;
2969    int after_manual_callout = 0;
2970    int length_prevgroup = 0;
2971    register int c;
2972    register uschar *code = *codeptr;
2973    uschar *last_code = code;
2974    uschar *orig_code = code;
2975    uschar *tempcode;
2976    BOOL inescq = FALSE;
2977    BOOL groupsetfirstbyte = FALSE;
2978    const uschar *ptr = *ptrptr;
2979    const uschar *tempptr;
2980    const uschar *nestptr = NULL;
2981    uschar *previous = NULL;
2982    uschar *previous_callout = NULL;
2983    uschar *save_hwm = NULL;
2984    uschar classbits[32];
2985    
2986    #ifdef SUPPORT_UTF8
2987    BOOL class_utf8;
2988    BOOL utf8 = (options & PCRE_UTF8) != 0;
2989    uschar *class_utf8data;
2990    uschar *class_utf8data_base;
2991    uschar utf8_char[6];
2992    #else
2993    BOOL utf8 = FALSE;
2994    uschar *utf8_char = NULL;
2995    #endif
2996    
2997    #ifdef PCRE_DEBUG
2998    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2999    #endif
3000    
3001    /* Set up the default and non-default settings for greediness */
3002    
3003    greedy_default = ((options & PCRE_UNGREEDY) != 0);
3004    greedy_non_default = greedy_default ^ 1;
3005    
3006    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3007    matching encountered yet". It gets changed to REQ_NONE if we hit something that
3008    matches a non-fixed char first char; reqbyte just remains unset if we never
3009    find one.
3010    
3011    When we hit a repeat whose minimum is zero, we may have to adjust these values
3012    to take the zero repeat into account. This is implemented by setting them to
3013    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
3014    item types that can be repeated set these backoff variables appropriately. */
3015    
3016    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
3017    
3018    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
3019    according to the current setting of the caseless flag. REQ_CASELESS is a bit
3020    value > 255. It is added into the firstbyte or reqbyte variables to record the
3021    case status of the value. This is used only for ASCII characters. */
3022    
3023    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
3024    
3025    /* Switch on next character until the end of the branch */
3026    
3027    for (;; ptr++)
3028      {
3029      BOOL negate_class;
3030      BOOL should_flip_negation;
3031      BOOL possessive_quantifier;
3032      BOOL is_quantifier;
3033      BOOL is_recurse;
3034      BOOL reset_bracount;
3035      int class_charcount;
3036      int class_lastchar;
3037      int newoptions;
3038      int recno;
3039      int refsign;
3040      int skipbytes;
3041      int subreqbyte;
3042      int subfirstbyte;
3043      int terminator;
3044      int mclength;
3045      uschar mcbuffer[8];
3046    
3047      /* Get next byte in the pattern */
3048    
3049      c = *ptr;
3050    
3051      /* If we are at the end of a nested substitution, revert to the outer level
3052      string. Nesting only happens one level deep. */
3053    
3054      if (c == 0 && nestptr != NULL)
3055        {
3056        ptr = nestptr;
3057        nestptr = NULL;
3058        c = *ptr;
3059        }
3060    
3061      /* If we are in the pre-compile phase, accumulate the length used for the
3062      previous cycle of this loop. */
3063    
3064      if (lengthptr != NULL)
3065        {
3066    #ifdef PCRE_DEBUG
3067        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3068    #endif
3069        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3070          {
3071          *errorcodeptr = ERR52;
3072          goto FAILED;
3073          }
3074    
3075        /* There is at least one situation where code goes backwards: this is the
3076        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3077        the class is simply eliminated. However, it is created first, so we have to
3078        allow memory for it. Therefore, don't ever reduce the length at this point.
3079        */
3080    
3081        if (code < last_code) code = last_code;
3082    
3083        /* Paranoid check for integer overflow */
3084    
3085        if (OFLOW_MAX - *lengthptr < code - last_code)
3086          {
3087          *errorcodeptr = ERR20;
3088          goto FAILED;
3089          }
3090    
3091        *lengthptr += (int)(code - last_code);
3092        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3093    
3094        /* If "previous" is set and it is not at the start of the work space, move
3095        it back to there, in order to avoid filling up the work space. Otherwise,
3096        if "previous" is NULL, reset the current code pointer to the start. */
3097    
3098        if (previous != NULL)
3099          {
3100          if (previous > orig_code)
3101            {
3102            memmove(orig_code, previous, code - previous);
3103            code -= previous - orig_code;
3104            previous = orig_code;
3105            }
3106          }
3107        else code = orig_code;
3108    
3109        /* Remember where this code item starts so we can pick up the length
3110        next time round. */
3111    
3112        last_code = code;
3113        }
3114    
3115      /* In the real compile phase, just check the workspace used by the forward
3116      reference list. */
3117    
3118      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3119        {
3120        *errorcodeptr = ERR52;
3121        goto FAILED;
3122        }
3123    
3124      /* If in \Q...\E, check for the end; if not, we have a literal */
3125    
3126      if (inescq && c != 0)
3127        {
3128        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3129          {
3130          inescq = FALSE;
3131          ptr++;
3132          continue;
3133          }
3134        else
3135          {
3136          if (previous_callout != NULL)
3137            {
3138            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3139              complete_callout(previous_callout, ptr, cd);
3140            previous_callout = NULL;
3141            }
3142          if ((options & PCRE_AUTO_CALLOUT) != 0)
3143            {
3144            previous_callout = code;
3145            code = auto_callout(code, ptr, cd);
3146            }
3147          goto NORMAL_CHAR;
3148          }
3149        }
3150    
3151      /* Fill in length of a previous callout, except when the next thing is
3152      a quantifier. */
3153    
3154      is_quantifier =
3155        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3156        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3157    
3158      if (!is_quantifier && previous_callout != NULL &&
3159           after_manual_callout-- <= 0)
3160        {
3161        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3162          complete_callout(previous_callout, ptr, cd);
3163        previous_callout = NULL;
3164        }
3165    
3166      /* In extended mode, skip white space and comments */
3167    
3168      if ((options & PCRE_EXTENDED) != 0)
3169        {
3170        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3171        if (c == CHAR_NUMBER_SIGN)
3172          {
3173          ptr++;
3174          while (*ptr != 0)
3175            {
3176            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3177            ptr++;
3178    #ifdef SUPPORT_UTF8
3179            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3180    #endif
3181            }
3182          if (*ptr != 0) continue;
3183    
3184          /* Else fall through to handle end of string */
3185          c = 0;
3186          }
3187        }
3188    
3189      /* No auto callout for quantifiers. */
3190    
3191      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3192        {
3193        previous_callout = code;
3194        code = auto_callout(code, ptr, cd);
3195        }
3196    
3197      switch(c)
3198        {
3199        /* ===================================================================*/
3200        case 0:                        /* The branch terminates at string end */
3201        case CHAR_VERTICAL_LINE:       /* or | or ) */
3202        case CHAR_RIGHT_PARENTHESIS:
3203        *firstbyteptr = firstbyte;
3204        *reqbyteptr = reqbyte;
3205        *codeptr = code;
3206        *ptrptr = ptr;
3207        if (lengthptr != NULL)
3208          {
3209          if (OFLOW_MAX - *lengthptr < code - last_code)
3210            {
3211            *errorcodeptr = ERR20;
3212            goto FAILED;
3213            }
3214          *lengthptr += (int)(code - last_code);   /* To include callout length */
3215          DPRINTF((">> end branch\n"));
3216          }
3217        return TRUE;
3218    
3219    
3220        /* ===================================================================*/
3221        /* Handle single-character metacharacters. In multiline mode, ^ disables
3222        the setting of any following char as a first character. */
3223    
3224        case CHAR_CIRCUMFLEX_ACCENT:
3225        if ((options & PCRE_MULTILINE) != 0)
3226          {
3227          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3228          }
3229        previous = NULL;
3230        *code++ = OP_CIRC;
3231        break;
3232    
3233        case CHAR_DOLLAR_SIGN:
3234        previous = NULL;
3235        *code++ = OP_DOLL;
3236        break;
3237    
3238        /* There can never be a first char if '.' is first, whatever happens about
3239        repeats. The value of reqbyte doesn't change either. */
3240    
3241        case CHAR_DOT:
3242        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3243        zerofirstbyte = firstbyte;
3244        zeroreqbyte = reqbyte;
3245        previous = code;
3246        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3247        break;
3248    
3249    
3250        /* ===================================================================*/
3251        /* Character classes. If the included characters are all < 256, we build a
3252        32-byte bitmap of the permitted characters, except in the special case
3253        where there is only one such character. For negated classes, we build the
3254        map as usual, then invert it at the end. However, we use a different opcode
3255        so that data characters > 255 can be handled correctly.
3256    
3257        If the class contains characters outside the 0-255 range, a different
3258        opcode is compiled. It may optionally have a bit map for characters < 256,
3259        but those above are are explicitly listed afterwards. A flag byte tells
3260        whether the bitmap is present, and whether this is a negated class or not.
3261    
3262        In JavaScript compatibility mode, an isolated ']' causes an error. In
3263        default (Perl) mode, it is treated as a data character. */
3264    
3265        case CHAR_RIGHT_SQUARE_BRACKET:
3266        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3267          {
3268          *errorcodeptr = ERR64;
3269          goto FAILED;
3270          }
3271        goto NORMAL_CHAR;
3272    
3273        case CHAR_LEFT_SQUARE_BRACKET:
3274        previous = code;
3275    
3276        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3277        they are encountered at the top level, so we'll do that too. */
3278    
3279        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3280             ptr[1] == CHAR_EQUALS_SIGN) &&
3281            check_posix_syntax(ptr, &tempptr))
3282          {
3283          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3284          goto FAILED;
3285          }
3286    
3287        /* If the first character is '^', set the negation flag and skip it. Also,
3288        if the first few characters (either before or after ^) are \Q\E or \E we
3289        skip them too. This makes for compatibility with Perl. */
3290    
3291        negate_class = FALSE;
3292        for (;;)
3293          {
3294          c = *(++ptr);
3295          if (c == CHAR_BACKSLASH)
3296            {
3297            if (ptr[1] == CHAR_E)
3298              ptr++;
3299            else if (strncmp((const char *)ptr+1,
3300                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3301              ptr += 3;
3302            else
3303              break;
3304            }
3305          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3306            negate_class = TRUE;
3307          else break;
3308          }
3309    
3310        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3311        an initial ']' is taken as a data character -- the code below handles
3312        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3313        [^] must match any character, so generate OP_ALLANY. */
3314    
3315        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3316            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3317          {
3318          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3319          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3320          zerofirstbyte = firstbyte;
3321          break;
3322          }
3323    
3324        /* If a class contains a negative special such as \S, we need to flip the
3325        negation flag at the end, so that support for characters > 255 works
3326        correctly (they are all included in the class). */
3327    
3328        should_flip_negation = FALSE;
3329    
3330        /* Keep a count of chars with values < 256 so that we can optimize the case
3331        of just a single character (as long as it's < 256). However, For higher
3332        valued UTF-8 characters, we don't yet do any optimization. */
3333    
3334        class_charcount = 0;
3335        class_lastchar = -1;
3336    
3337        /* Initialize the 32-char bit map to all zeros. We build the map in a
3338        temporary bit of memory, in case the class contains only 1 character (less
3339        than 256), because in that case the compiled code doesn't use the bit map.
3340        */
3341    
3342        memset(classbits, 0, 32 * sizeof(uschar));
3343    
3344    #ifdef SUPPORT_UTF8
3345        class_utf8 = FALSE;                       /* No chars >= 256 */
3346        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3347        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3348    #endif
3349    
3350        /* Process characters until ] is reached. By writing this as a "do" it
3351        means that an initial ] is taken as a data character. At the start of the
3352        loop, c contains the first byte of the character. */
3353    
3354        if (c != 0) do
3355          {
3356          const uschar *oldptr;
3357    
3358    #ifdef SUPPORT_UTF8
3359          if (utf8 && c > 127)
3360            {                           /* Braces are required because the */
3361            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3362            }
3363    
3364          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3365          data and reset the pointer. This is so that very large classes that
3366          contain a zillion UTF-8 characters no longer overwrite the work space
3367          (which is on the stack). */
3368    
3369          if (lengthptr != NULL)
3370            {
3371            *lengthptr += class_utf8data - class_utf8data_base;
3372            class_utf8data = class_utf8data_base;
3373            }
3374    
3375    #endif
3376    
3377          /* Inside \Q...\E everything is literal except \E */
3378    
3379          if (inescq)
3380            {
3381            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3382              {
3383              inescq = FALSE;                   /* Reset literal state */
3384              ptr++;                            /* Skip the 'E' */
3385              continue;                         /* Carry on with next */
3386              }
3387            goto CHECK_RANGE;                   /* Could be range if \E follows */
3388            }
3389    
3390          /* Handle POSIX class names. Perl allows a negation extension of the
3391          form [:^name:]. A square bracket that doesn't match the syntax is
3392          treated as a literal. We also recognize the POSIX constructions
3393          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3394          5.6 and 5.8 do. */
3395    
3396          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3397              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3398               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3399            {
3400            BOOL local_negate = FALSE;
3401            int posix_class, taboffset, tabopt;
3402            register const uschar *cbits = cd->cbits;
3403            uschar pbits[32];
3404    
3405            if (ptr[1] != CHAR_COLON)
3406              {
3407              *errorcodeptr = ERR31;
3408              goto FAILED;
3409              }
3410    
3411            ptr += 2;
3412            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3413              {
3414              local_negate = TRUE;
3415              should_flip_negation = TRUE;  /* Note negative special */
3416              ptr++;
3417              }
3418    
3419            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3420            if (posix_class < 0)
3421              {
3422              *errorcodeptr = ERR30;
3423              goto FAILED;
3424              }
3425    
3426            /* If matching is caseless, upper and lower are converted to
3427            alpha. This relies on the fact that the class table starts with
3428            alpha, lower, upper as the first 3 entries. */
3429    
3430            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3431              posix_class = 0;
3432    
3433            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3434            different escape sequences that use Unicode properties. */
3435    
3436    #ifdef SUPPORT_UCP
3437            if ((options & PCRE_UCP) != 0)
3438              {
3439              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3440              if (posix_substitutes[pc] != NULL)
3441                {
3442                nestptr = tempptr + 1;
3443                ptr = posix_substitutes[pc] - 1;
3444                continue;
3445                }
3446              }
3447    #endif
3448            /* In the non-UCP case, we build the bit map for the POSIX class in a
3449            chunk of local store because we may be adding and subtracting from it,
3450            and we don't want to subtract bits that may be in the main map already.
3451            At the end we or the result into the bit map that is being built. */
3452    
3453            posix_class *= 3;
3454    
3455            /* Copy in the first table (always present) */
3456    
3457            memcpy(pbits, cbits + posix_class_maps[posix_class],
3458              32 * sizeof(uschar));
3459    
3460            /* If there is a second table, add or remove it as required. */
3461    
3462            taboffset = posix_class_maps[posix_class + 1];
3463            tabopt = posix_class_maps[posix_class + 2];
3464    
3465            if (taboffset >= 0)
3466              {
3467              if (tabopt >= 0)
3468                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3469              else
3470                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3471              }
3472    
3473            /* Not see if we need to remove any special characters. An option
3474            value of 1 removes vertical space and 2 removes underscore. */
3475    
3476            if (tabopt < 0) tabopt = -tabopt;
3477            if (tabopt == 1) pbits[1] &= ~0x3c;
3478              else if (tabopt == 2) pbits[11] &= 0x7f;
3479    
3480            /* Add the POSIX table or its complement into the main table that is
3481            being built and we are done. */
3482    
3483            if (local_negate)
3484              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3485            else
3486              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3487    
3488            ptr = tempptr + 1;
3489            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3490            continue;    /* End of POSIX syntax handling */
3491            }
3492    
3493          /* Backslash may introduce a single character, or it may introduce one
3494          of the specials, which just set a flag. The sequence \b is a special
3495          case. Inside a class (and only there) it is treated as backspace. We
3496          assume that other escapes have more than one character in them, so set
3497          class_charcount bigger than one. Unrecognized escapes fall through and
3498          are either treated as literal characters (by default), or are faulted if
3499          PCRE_EXTRA is set. */
3500    
3501          if (c == CHAR_BACKSLASH)
3502            {
3503            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3504            if (*errorcodeptr != 0) goto FAILED;
3505    
3506            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3507            else if (-c == ESC_Q)            /* Handle start of quoted string */
3508              {
3509              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3510                {
3511                ptr += 2; /* avoid empty string */
3512                }
3513              else inescq = TRUE;
3514              continue;
3515              }
3516            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3517    
3518            if (c < 0)
3519              {
3520              register const uschar *cbits = cd->cbits;
3521              class_charcount += 2;     /* Greater than 1 is what matters */
3522    
3523              switch (-c)
3524                {
3525    #ifdef SUPPORT_UCP
3526                case ESC_du:     /* These are the values given for \d etc */
3527                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3528                case ESC_wu:     /* escape sequence with an appropriate \p */
3529                case ESC_WU:     /* or \P to test Unicode properties instead */
3530                case ESC_su:     /* of the default ASCII testing. */
3531                case ESC_SU:
3532                nestptr = ptr;
3533                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3534                class_charcount -= 2;                /* Undo! */
3535                continue;
3536    #endif
3537                case ESC_d:
3538                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3539                continue;
3540    
3541                case ESC_D:
3542                should_flip_negation = TRUE;
3543                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3544                continue;
3545    
3546                case ESC_w:
3547                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
3548                continue;
3549    
3550                case ESC_W:
3551                should_flip_negation = TRUE;
3552                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3553                continue;
3554    
3555                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3556                if it was previously set by something earlier in the character
3557                class. */
3558    
3559                case ESC_s:
3560                classbits[0] |= cbits[cbit_space];
3561                classbits[1] |= cbits[cbit_space+1] & ~0x08;
3562                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3563                continue;
3564    
3565                case ESC_S:
3566                should_flip_negation = TRUE;
3567                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3568                classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3569                continue;
3570    
3571                case ESC_h:
3572                SETBIT(classbits, 0x09); /* VT */
3573                SETBIT(classbits, 0x20); /* SPACE */
3574                SETBIT(classbits, 0xa0); /* NSBP */
3575    #ifdef SUPPORT_UTF8
3576                if (utf8)
3577                  {
3578                  class_utf8 = TRUE;
3579                  *class_utf8data++ = XCL_SINGLE;
3580                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3581                  *class_utf8data++ = XCL_SINGLE;
3582                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3583                  *class_utf8data++ = XCL_RANGE;
3584                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3585                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3586                  *class_utf8data++ = XCL_SINGLE;
3587                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3588                  *class_utf8data++ = XCL_SINGLE;
3589                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3590                  *class_utf8data++ = XCL_SINGLE;
3591                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3592                  }
3593    #endif
3594                continue;
3595    
3596                case ESC_H:
3597                for (c = 0; c < 32; c++)
3598                  {
3599                  int x = 0xff;
3600                  switch (c)
3601                    {
3602                    case 0x09/8: x ^= 1 << (0x09%8); break;
3603                    case 0x20/8: x ^= 1 << (0x20%8); break;
3604                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3605                    default: break;
3606                    }
3607                  classbits[c] |= x;
3608                  }
3609    
3610    #ifdef SUPPORT_UTF8
3611                if (utf8)
3612                  {
3613                  class_utf8 = TRUE;
3614                  *class_utf8data++ = XCL_RANGE;
3615                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3616                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3617                  *class_utf8data++ = XCL_RANGE;
3618                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3619                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3620                  *class_utf8data++ = XCL_RANGE;
3621                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3622                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3623                  *class_utf8data++ = XCL_RANGE;
3624                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3625                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3626                  *class_utf8data++ = XCL_RANGE;
3627                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3628                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3629                  *class_utf8data++ = XCL_RANGE;
3630                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3631                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3632                  *class_utf8data++ = XCL_RANGE;
3633                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3634                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3635                  }
3636    #endif
3637                continue;
3638    
3639                case ESC_v:
3640                SETBIT(classbits, 0x0a); /* LF */
3641                SETBIT(classbits, 0x0b); /* VT */
3642                SETBIT(classbits, 0x0c); /* FF */
3643                SETBIT(classbits, 0x0d); /* CR */
3644                SETBIT(classbits, 0x85); /* NEL */
3645    #ifdef SUPPORT_UTF8
3646                if (utf8)
3647                  {
3648                  class_utf8 = TRUE;
3649                  *class_utf8data++ = XCL_RANGE;
3650                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3651                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3652                  }
3653    #endif
3654                continue;
3655    
3656                case ESC_V:
3657                for (c = 0; c < 32; c++)
3658                  {
3659                  int x = 0xff;
3660                  switch (c)
3661                    {
3662                    case 0x0a/8: x ^= 1 << (0x0a%8);
3663                                 x ^= 1 << (0x0b%8);
3664                                 x ^= 1 << (0x0c%8);
3665                                 x ^= 1 << (0x0d%8);
3666                                 break;
3667                    case 0x85/8: x ^= 1 << (0x85%8); break;
3668                    default: break;
3669                    }
3670                  classbits[c] |= x;
3671                  }
3672    
3673    #ifdef SUPPORT_UTF8
3674                if (utf8)
3675                  {
3676                  class_utf8 = TRUE;
3677                  *class_utf8data++ = XCL_RANGE;
3678                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3679                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3680                  *class_utf8data++ = XCL_RANGE;
3681                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3682                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3683                  }
3684    #endif
3685                continue;
3686    
3687    #ifdef SUPPORT_UCP
3688                case ESC_p:
3689                case ESC_P:
3690                  {
3691                  BOOL negated;
3692                  int pdata;
3693                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3694                  if (ptype < 0) goto FAILED;
3695                  class_utf8 = TRUE;
3696                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3697                    XCL_PROP : XCL_NOTPROP;
3698                  *class_utf8data++ = ptype;
3699                  *class_utf8data++ = pdata;
3700                  class_charcount -= 2;   /* Not a < 256 character */
3701                  continue;
3702                  }
3703    #endif
3704                /* Unrecognized escapes are faulted if PCRE is running in its
3705                strict mode. By default, for compatibility with Perl, they are
3706                treated as literals. */
3707    
3708                default:
3709                if ((options & PCRE_EXTRA) != 0)
3710                  {
3711                  *errorcodeptr = ERR7;
3712                  goto FAILED;
3713                }                }
             c = *ptr;              /* The final character */  
3714              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3715                c = *ptr;              /* Get the final character and fall through */
3716                break;
3717              }              }
3718            }            }
3719    
3720          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
3721          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3722    
3723          }   /* End of backslash handling */          }   /* End of backslash handling */
3724    
3725        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3726        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3727        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3728          entirely. The code for handling \Q and \E is messy. */
3729    
3730          CHECK_RANGE:
3731          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3732            {
3733            inescq = FALSE;
3734            ptr += 2;
3735            }
3736    
3737          oldptr = ptr;
3738    
3739          /* Remember \r or \n */
3740    
3741        if (ptr[1] == '-' && ptr[2] != ']')        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3742    
3743          /* Check for range */
3744    
3745          if (!inescq && ptr[1] == CHAR_MINUS)