/[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 552 by ph10, Wed Oct 13 10:15:41 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93    #define COMPILE_WORK_SIZE (4096)
94    
95  #define BRASTACK_SIZE 200  /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 154  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  static const char *error_texts[] = {  #define STRING(a)  # a
313    "no error",  #define XSTRING(s) STRING(s)
314    "\\ at end of pattern",  
315    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
316    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
317    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
318    are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320    it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
381    "two named groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "octal value is greater than \\377 (not in UTF-8 mode)\0"
392      "internal error: overran compiling workspace\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"
397      "inconsistent NEWLINE options\0"
398      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399      "a numbered reference must not be zero\0"
400      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 235  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 271  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 285  static const unsigned char digitab[] = Line 483  static const unsigned char digitab[] =
483    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 319  static const unsigned char ebcdic_charta Line 517  static const unsigned char ebcdic_charta
517    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
518    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
519    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
520    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
522    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 346  static const unsigned char ebcdic_charta Line 544  static const unsigned char ebcdic_charta
544  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
545    
546  static BOOL  static BOOL
547    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
548      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
549    
550    
551    
552    /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578  /*************************************************  /*************************************************
# Line 357  static BOOL Line 581  static BOOL
581    
582  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
583  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
584  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
585  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
586  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
587    ptr is pointing at the \. On exit, it is on the final character of the escape
588    sequence.
589    
590  Arguments:  Arguments:
591    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 370  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
# Line 388  ptr--;                            /* Set Line 614  ptr--;                            /* Set
614    
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
618  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 406  else if ((i = escapes[c - 0x48]) != 0) Line 632  else if ((i = escapes[c - 0x48]) != 0)
632  else  else
633    {    {
634    const uschar *oldptr;    const uschar *oldptr;
635      BOOL braced, negated;
636    
637    switch (c)    switch (c)
638      {      {
639      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649        /* \g must be followed by one of a number of specific things:
650    
651        (1) A number, either plain or braced. If positive, it is an absolute
652        backreference. If negative, it is a relative backreference. This is a Perl
653        5.10 feature.
654    
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665        case CHAR_g:
666        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675          {
676          const uschar *p;
677          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680            {
681            c = -ESC_k;
682            break;
683            }
684          braced = TRUE;
685          ptr++;
686          }
687        else braced = FALSE;
688    
689        if (ptr[1] == CHAR_MINUS)
690          {
691          negated = TRUE;
692          ptr++;
693          }
694        else negated = FALSE;
695    
696        c = 0;
697        while ((digitab[ptr[1]] & ctype_digit) != 0)
698          c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707          {
708          *errorcodeptr = ERR57;
709          break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718        if (negated)
719          {
720          if (c > bracount)
721            {
722            *errorcodeptr = ERR15;
723            break;
724            }
725          c = bracount - (c - 1);
726          }
727    
728        c = -(ESC_REF + c);
729        break;
730    
731      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
732      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
733      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 431  else Line 740  else
740      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 452  else Line 766  else
766      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 460  else Line 774  else
774        }        }
775    
776      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
777      larger first octal digit. */      larger first octal digit. The original code used just to take the least
778        significant 8 bits of octal numbers (I think this is what early Perls used
779      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
780      c -= '0';      than 3 octal digits. */
781      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
782          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
783      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
784        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785            c = c * 8 + *(++ptr) - CHAR_0;
786        if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
789      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
790      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 483  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 511  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
842    
843      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
844        This coding is ASCII-specific, but then the whole concept of \cx is
845        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
     if (c >= 'a' && c <= 'z') c -= 32;  
857      c ^= 0x40;      c ^= 0x40;
858  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
867      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 563  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 603  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
940      }      }
941    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 639  top = _pcre_utt_size; Line 967  top = _pcre_utt_size;
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top) >> 1;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0)    if (c == 0)
972      {      {
973      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 682  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 723  int max = -1; Line 1051  int max = -1;
1051  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 733  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 763  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094    *  Subroutine for finding forward reference      *
1095    *************************************************/
1096    
1097    /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100    subpatterns, and counting them. If it finds a named pattern that matches the
1101    name it is given, it returns its number. Alternatively, if the name is NULL, it
1102    returns when it reaches a given numbered subpattern. We know that if (?P< is
1103    encountered, the name will be terminated by '>' because that is checked in the
1104    first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107    Arguments:
1108      ptrptr       address of the current character pointer (updated)
1109      cd           compile background data
1110      name         name to seek, or NULL if seeking a numbered subpattern
1111      lorn         name length, or subpattern number if name is NULL
1112      xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115    Returns:       the number of the named subpattern, or -1 if not found
1116    */
1117    
1118    static int
1119    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1120      BOOL xmode, int *count)
1121    {
1122    uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127    /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131      {
1132      /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1133    
1134      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1135    
1136      /* Handle a normal, unnamed capturing parenthesis. */
1137    
1138      else if (ptr[1] != CHAR_QUESTION_MARK)
1139        {
1140        *count += 1;
1141        if (name == NULL && *count == lorn) return *count;
1142        ptr++;
1143        }
1144    
1145      /* All cases now have (? at the start. Remember when we are in a group
1146      where the parenthesis numbers are duplicated. */
1147    
1148      else if (ptr[2] == CHAR_VERTICAL_LINE)
1149        {
1150        ptr += 3;
1151        dup_parens = TRUE;
1152        }
1153    
1154      /* Handle comments; all characters are allowed until a ket is reached. */
1155    
1156      else if (ptr[2] == CHAR_NUMBER_SIGN)
1157        {
1158        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1159        goto FAIL_EXIT;
1160        }
1161    
1162      /* Handle a condition. If it is an assertion, just carry on so that it
1163      is processed as normal. If not, skip to the closing parenthesis of the
1164      condition (there can't be any nested parens). */
1165    
1166      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1167        {
1168        ptr += 2;
1169        if (ptr[1] != CHAR_QUESTION_MARK)
1170          {
1171          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1172          if (*ptr != 0) ptr++;
1173          }
1174        }
1175    
1176      /* Start with (? but not a condition. */
1177    
1178      else
1179        {
1180        ptr += 2;
1181        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1182    
1183        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1184    
1185        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1186            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1187          {
1188          int term;
1189          const uschar *thisname;
1190          *count += 1;
1191          if (name == NULL && *count == lorn) return *count;
1192          term = *ptr++;
1193          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1194          thisname = ptr;
1195          while (*ptr != term) ptr++;
1196          if (name != NULL && lorn == ptr - thisname &&
1197              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1198            return *count;
1199          term++;
1200          }
1201        }
1202      }
1203    
1204    /* Past any initial parenthesis handling, scan for parentheses or vertical
1205    bars. */
1206    
1207    for (; *ptr != 0; ptr++)
1208      {
1209      /* Skip over backslashed characters and also entire \Q...\E */
1210    
1211      if (*ptr == CHAR_BACKSLASH)
1212        {
1213        if (*(++ptr) == 0) goto FAIL_EXIT;
1214        if (*ptr == CHAR_Q) for (;;)
1215          {
1216          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1217          if (*ptr == 0) goto FAIL_EXIT;
1218          if (*(++ptr) == CHAR_E) break;
1219          }
1220        continue;
1221        }
1222    
1223      /* Skip over character classes; this logic must be similar to the way they
1224      are handled for real. If the first character is '^', skip it. Also, if the
1225      first few characters (either before or after ^) are \Q\E or \E we skip them
1226      too. This makes for compatibility with Perl. Note the use of STR macros to
1227      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1228    
1229      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1230        {
1231        BOOL negate_class = FALSE;
1232        for (;;)
1233          {
1234          if (ptr[1] == CHAR_BACKSLASH)
1235            {
1236            if (ptr[2] == CHAR_E)
1237              ptr+= 2;
1238            else if (strncmp((const char *)ptr+2,
1239                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1240              ptr += 4;
1241            else
1242              break;
1243            }
1244          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1245            {
1246            negate_class = TRUE;
1247            ptr++;
1248            }
1249          else break;
1250          }
1251    
1252        /* If the next character is ']', it is a data character that must be
1253        skipped, except in JavaScript compatibility mode. */
1254    
1255        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1256            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1257          ptr++;
1258    
1259        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1260          {
1261          if (*ptr == 0) return -1;
1262          if (*ptr == CHAR_BACKSLASH)
1263            {
1264            if (*(++ptr) == 0) goto FAIL_EXIT;
1265            if (*ptr == CHAR_Q) for (;;)
1266              {
1267              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1268              if (*ptr == 0) goto FAIL_EXIT;
1269              if (*(++ptr) == CHAR_E) break;
1270              }
1271            continue;
1272            }
1273          }
1274        continue;
1275        }
1276    
1277      /* Skip comments in /x mode */
1278    
1279      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1280        {
1281        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1282        if (*ptr == 0) goto FAIL_EXIT;
1283        continue;
1284        }
1285    
1286      /* Check for the special metacharacters */
1287    
1288      if (*ptr == CHAR_LEFT_PARENTHESIS)
1289        {
1290        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1291        if (rc > 0) return rc;
1292        if (*ptr == 0) goto FAIL_EXIT;
1293        }
1294    
1295      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1296        {
1297        if (dup_parens && *count < hwm_count) *count = hwm_count;
1298        goto FAIL_EXIT;
1299        }
1300    
1301      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1302        {
1303        if (*count > hwm_count) hwm_count = *count;
1304        *count = start_count;
1305        }
1306      }
1307    
1308    FAIL_EXIT:
1309    *ptrptr = ptr;
1310    return -1;
1311    }
1312    
1313    
1314    
1315    
1316    /*************************************************
1317    *       Find forward referenced subpattern       *
1318    *************************************************/
1319    
1320    /* This function scans along a pattern's text looking for capturing
1321    subpatterns, and counting them. If it finds a named pattern that matches the
1322    name it is given, it returns its number. Alternatively, if the name is NULL, it
1323    returns when it reaches a given numbered subpattern. This is used for forward
1324    references to subpatterns. We used to be able to start this scan from the
1325    current compiling point, using the current count value from cd->bracount, and
1326    do it all in a single loop, but the addition of the possibility of duplicate
1327    subpattern numbers means that we have to scan from the very start, in order to
1328    take account of such duplicates, and to use a recursive function to keep track
1329    of the different types of group.
1330    
1331    Arguments:
1332      cd           compile background data
1333      name         name to seek, or NULL if seeking a numbered subpattern
1334      lorn         name length, or subpattern number if name is NULL
1335      xmode        TRUE if we are in /x mode
1336    
1337    Returns:       the number of the found subpattern, or -1 if not found
1338    */
1339    
1340    static int
1341    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1342    {
1343    uschar *ptr = (uschar *)cd->start_pattern;
1344    int count = 0;
1345    int rc;
1346    
1347    /* If the pattern does not start with an opening parenthesis, the first call
1348    to find_parens_sub() will scan right to the end (if necessary). However, if it
1349    does start with a parenthesis, find_parens_sub() will return when it hits the
1350    matching closing parens. That is why we have to have a loop. */
1351    
1352    for (;;)
1353      {
1354      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1355      if (rc > 0 || *ptr++ == 0) break;
1356      }
1357    
1358    return rc;
1359    }
1360    
1361    
1362    
1363    
1364    /*************************************************
1365  *      Find first significant op code            *  *      Find first significant op code            *
1366  *************************************************/  *************************************************/
1367    
# Line 811  for (;;) Line 1410  for (;;)
1410    
1411      case OP_CALLOUT:      case OP_CALLOUT:
1412      case OP_CREF:      case OP_CREF:
1413      case OP_BRANUMBER:      case OP_NCREF:
1414        case OP_RREF:
1415        case OP_NRREF:
1416        case OP_DEF:
1417      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1418      break;      break;
1419    
# Line 826  for (;;) Line 1428  for (;;)
1428    
1429    
1430  /*************************************************  /*************************************************
1431  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1432  *************************************************/  *************************************************/
1433    
1434  /* 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,
1435  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.
1436  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
1437    temporarily terminated with OP_END when this function is called.
1438    
1439    This function is called when a backward assertion is encountered, so that if it
1440    fails, the error message can point to the correct place in the pattern.
1441    However, we cannot do this when the assertion contains subroutine calls,
1442    because they can be forward references. We solve this by remembering this case
1443    and doing the check at the end; a flag specifies which mode we are running in.
1444    
1445  Arguments:  Arguments:
1446    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1447    options  the compiling options    options  the compiling options
1448      atend    TRUE if called when the pattern is complete
1449      cd       the "compile data" structure
1450    
1451  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1452                 or -1 if there is no fixed length,
1453               or -2 if \C was encountered               or -2 if \C was encountered
1454                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1455  */  */
1456    
1457  static int  static int
1458  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1459  {  {
1460  int length = -1;  int length = -1;
1461    
# Line 855  branch, check the length against that of Line 1468  branch, check the length against that of
1468  for (;;)  for (;;)
1469    {    {
1470    int d;    int d;
1471      uschar *ce, *cs;
1472    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1473    switch (op)    switch (op)
1474      {      {
1475        case OP_CBRA:
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478      case OP_COND:      case OP_COND:
1479      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1480      if (d < 0) return d;      if (d < 0) return d;
1481      branchlength += d;      branchlength += d;
1482      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 886  for (;;) Line 1499  for (;;)
1499      branchlength = 0;      branchlength = 0;
1500      break;      break;
1501    
1502        /* A true recursion implies not fixed length, but a subroutine call may
1503        be OK. If the subroutine is a forward reference, we can't deal with
1504        it until the end of the pattern, so return -3. */
1505    
1506        case OP_RECURSE:
1507        if (!atend) return -3;
1508        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1509        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1510        if (cc > cs && cc < ce) return -1;                /* Recursion */
1511        d = find_fixedlength(cs + 2, options, atend, cd);
1512        if (d < 0) return d;
1513        branchlength += d;
1514        cc += 1 + LINK_SIZE;
1515        break;
1516    
1517      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1518    
1519      case OP_ASSERT:      case OP_ASSERT:
# Line 898  for (;;) Line 1526  for (;;)
1526      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1527    
1528      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1529      case OP_CREF:      case OP_CREF:
1530        case OP_NCREF:
1531        case OP_RREF:
1532        case OP_NRREF:
1533        case OP_DEF:
1534      case OP_OPT:      case OP_OPT:
1535      case OP_CALLOUT:      case OP_CALLOUT:
1536      case OP_SOD:      case OP_SOD:
1537      case OP_SOM:      case OP_SOM:
1538        case OP_SET_SOM:
1539      case OP_EOD:      case OP_EOD:
1540      case OP_EODN:      case OP_EODN:
1541      case OP_CIRC:      case OP_CIRC:
# Line 917  for (;;) Line 1549  for (;;)
1549    
1550      case OP_CHAR:      case OP_CHAR:
1551      case OP_CHARNC:      case OP_CHARNC:
1552        case OP_NOT:
1553      branchlength++;      branchlength++;
1554      cc += 2;      cc += 2;
1555  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1556      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1557        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1558  #endif  #endif
1559      break;      break;
1560    
# Line 934  for (;;) Line 1565  for (;;)
1565      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1566      cc += 4;      cc += 4;
1567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1568      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1569        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1570  #endif  #endif
1571      break;      break;
1572    
1573      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1574      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1575        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1576      cc += 4;      cc += 4;
1577      break;      break;
1578    
# Line 960  for (;;) Line 1590  for (;;)
1590      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1591      case OP_WORDCHAR:      case OP_WORDCHAR:
1592      case OP_ANY:      case OP_ANY:
1593        case OP_ALLANY:
1594      branchlength++;      branchlength++;
1595      cc++;      cc++;
1596      break;      break;
# Line 1014  for (;;) Line 1645  for (;;)
1645    
1646    
1647  /*************************************************  /*************************************************
1648  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1649  *************************************************/  *************************************************/
1650    
1651  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1652  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1653    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1654    so that it can be called from pcre_study() when finding the minimum matching
1655    length.
1656    
1657  Arguments:  Arguments:
1658    code        points to start of expression    code        points to start of expression
1659    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1660    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1661    
1662  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
1663  */  */
1664    
1665  static const uschar *  const uschar *
1666  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1667  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1668  for (;;)  for (;;)
1669    {    {
1670    register int c = *code;    register int c = *code;
1671    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1672    else if (c > OP_BRA)  
1673      /* XCLASS is used for classes that cannot be represented just by a bit
1674      map. This includes negated single high-valued characters. The length in
1675      the table is zero; the actual length is stored in the compiled code. */
1676    
1677      if (c == OP_XCLASS) code += GET(code, 1);
1678    
1679      /* Handle recursion */
1680    
1681      else if (c == OP_REVERSE)
1682        {
1683        if (number < 0) return (uschar *)code;
1684        code += _pcre_OP_lengths[c];
1685        }
1686    
1687      /* Handle capturing bracket */
1688    
1689      else if (c == OP_CBRA)
1690      {      {
1691      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1692      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1693      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1694      }      }
1695    
1696      /* Otherwise, we can get the item's length from the table, except that for
1697      repeated character types, we have to test for \p and \P, which have an extra
1698      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1699      must add in its length. */
1700    
1701    else    else
1702      {      {
1703      code += _pcre_OP_lengths[c];      switch(c)
1704          {
1705          case OP_TYPESTAR:
1706          case OP_TYPEMINSTAR:
1707          case OP_TYPEPLUS:
1708          case OP_TYPEMINPLUS:
1709          case OP_TYPEQUERY:
1710          case OP_TYPEMINQUERY:
1711          case OP_TYPEPOSSTAR:
1712          case OP_TYPEPOSPLUS:
1713          case OP_TYPEPOSQUERY:
1714          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1715          break;
1716    
1717  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1718          case OP_TYPEMINUPTO:
1719          case OP_TYPEEXACT:
1720          case OP_TYPEPOSUPTO:
1721          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722          break;
1723    
1724      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1725      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1726      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1727      can use relatively efficient code. */        code += code[1];
1728          break;
1729    
1730      if (utf8) switch(c)        case OP_THEN_ARG:
1731          code += code[1+LINK_SIZE];
1732          break;
1733          }
1734    
1735        /* Add in the fixed length from the table */
1736    
1737        code += _pcre_OP_lengths[c];
1738    
1739      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1740      a multi-byte character. The length in the table is a minimum, so we have to
1741      arrange to skip the extra bytes. */
1742    
1743    #ifdef SUPPORT_UTF8
1744        if (utf8) switch(c)
1745        {        {
1746        case OP_CHAR:        case OP_CHAR:
1747        case OP_CHARNC:        case OP_CHARNC:
1748        case OP_EXACT:        case OP_EXACT:
1749        case OP_UPTO:        case OP_UPTO:
1750        case OP_MINUPTO:        case OP_MINUPTO:
1751          case OP_POSUPTO:
1752        case OP_STAR:        case OP_STAR:
1753        case OP_MINSTAR:        case OP_MINSTAR:
1754          case OP_POSSTAR:
1755        case OP_PLUS:        case OP_PLUS:
1756        case OP_MINPLUS:        case OP_MINPLUS:
1757          case OP_POSPLUS:
1758        case OP_QUERY:        case OP_QUERY:
1759        case OP_MINQUERY:        case OP_MINQUERY:
1760        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1761        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;  
1762        break;        break;
1763        }        }
1764    #else
1765        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1766  #endif  #endif
1767      }      }
1768    }    }
# Line 1105  Returns:      pointer to the opcode for Line 1787  Returns:      pointer to the opcode for
1787  static const uschar *  static const uschar *
1788  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1789  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1790  for (;;)  for (;;)
1791    {    {
1792    register int c = *code;    register int c = *code;
1793    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1794    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1795    else if (c > OP_BRA)  
1796      {    /* XCLASS is used for classes that cannot be represented just by a bit
1797      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1798      }    the table is zero; the actual length is stored in the compiled code. */
1799    
1800      if (c == OP_XCLASS) code += GET(code, 1);
1801    
1802      /* Otherwise, we can get the item's length from the table, except that for
1803      repeated character types, we have to test for \p and \P, which have an extra
1804      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1805      must add in its length. */
1806    
1807    else    else
1808      {      {
1809      code += _pcre_OP_lengths[c];      switch(c)
1810          {
1811          case OP_TYPESTAR:
1812          case OP_TYPEMINSTAR:
1813          case OP_TYPEPLUS:
1814          case OP_TYPEMINPLUS:
1815          case OP_TYPEQUERY:
1816          case OP_TYPEMINQUERY:
1817          case OP_TYPEPOSSTAR:
1818          case OP_TYPEPOSPLUS:
1819          case OP_TYPEPOSQUERY:
1820          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1821          break;
1822    
1823  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1824          case OP_TYPEUPTO:
1825          case OP_TYPEMINUPTO:
1826          case OP_TYPEEXACT:
1827          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1828          break;
1829    
1830          case OP_MARK:
1831          case OP_PRUNE_ARG:
1832          case OP_SKIP_ARG:
1833          code += code[1];
1834          break;
1835    
1836          case OP_THEN_ARG:
1837          code += code[1+LINK_SIZE];
1838          break;
1839          }
1840    
1841        /* Add in the fixed length from the table */
1842    
1843        code += _pcre_OP_lengths[c];
1844    
1845      /* 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
1846      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
1847      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. */  
1848    
1849    #ifdef SUPPORT_UTF8
1850      if (utf8) switch(c)      if (utf8) switch(c)
1851        {        {
1852        case OP_CHAR:        case OP_CHAR:
# Line 1136  for (;;) Line 1854  for (;;)
1854        case OP_EXACT:        case OP_EXACT:
1855        case OP_UPTO:        case OP_UPTO:
1856        case OP_MINUPTO:        case OP_MINUPTO:
1857          case OP_POSUPTO:
1858        case OP_STAR:        case OP_STAR:
1859        case OP_MINSTAR:        case OP_MINSTAR:
1860          case OP_POSSTAR:
1861        case OP_PLUS:        case OP_PLUS:
1862        case OP_MINPLUS:        case OP_MINPLUS:
1863          case OP_POSPLUS:
1864        case OP_QUERY:        case OP_QUERY:
1865        case OP_MINQUERY:        case OP_MINQUERY:
1866        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1867        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;  
1868        break;        break;
1869        }        }
1870    #else
1871        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1872  #endif  #endif
1873      }      }
1874    }    }
# Line 1165  for (;;) Line 1881  for (;;)
1881  *************************************************/  *************************************************/
1882    
1883  /* 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
1884  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()
1885  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
1886  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
1887  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1888    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1889    bracket whose current branch will already have been scanned.
1890    
1891  Arguments:  Arguments:
1892    code        points to start of search    code        points to start of search
1893    endcode     points to where to stop    endcode     points to where to stop
1894    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1895      cd          contains pointers to tables etc.
1896    
1897  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1898  */  */
1899    
1900  static BOOL  static BOOL
1901  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1902      compile_data *cd)
1903  {  {
1904  register int c;  register int c;
1905  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);
1906       code < endcode;       code < endcode;
1907       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1908    {    {
# Line 1190  for (code = first_significant_code(code Line 1910  for (code = first_significant_code(code
1910    
1911    c = *code;    c = *code;
1912    
1913    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1914      first_significant_code() with a TRUE final argument. */
1915    
1916      if (c == OP_ASSERT)
1917      {      {
1918      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1919      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1920        continue;
1921        }
1922    
1923      /* Groups with zero repeats can of course be empty; skip them. */
1924    
1925      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1926        {
1927        code += _pcre_OP_lengths[c];
1928        do code += GET(code, 1); while (*code == OP_ALT);
1929        c = *code;
1930        continue;
1931        }
1932    
1933      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1934      implies a subroutine call, we can scan it. */
1935    
1936      empty_branch = FALSE;    if (c == OP_RECURSE)
1937        {
1938        BOOL empty_branch = FALSE;
1939        const uschar *scode = cd->start_code + GET(code, 1);
1940        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1941      do      do
1942        {        {
1943        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1944            {
1945          empty_branch = TRUE;          empty_branch = TRUE;
1946            break;
1947            }
1948          scode += GET(scode, 1);
1949          }
1950        while (*scode == OP_ALT);
1951        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1952        continue;
1953        }
1954    
1955      /* For other groups, scan the branches. */
1956    
1957      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1958        {
1959        BOOL empty_branch;
1960        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1961    
1962        /* If a conditional group has only one branch, there is a second, implied,
1963        empty branch, so just skip over the conditional, because it could be empty.
1964        Otherwise, scan the individual branches of the group. */
1965    
1966        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1967        code += GET(code, 1);        code += GET(code, 1);
1968        else
1969          {
1970          empty_branch = FALSE;
1971          do
1972            {
1973            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1974              empty_branch = TRUE;
1975            code += GET(code, 1);
1976            }
1977          while (*code == OP_ALT);
1978          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1979        }        }
1980      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1981      c = *code;      c = *code;
1982        continue;
1983      }      }
1984    
1985    else switch (c)    /* Handle the other opcodes */
1986    
1987      switch (c)
1988      {      {
1989      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1990        cannot be represented just by a bit map. This includes negated single
1991        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1992        actual length is stored in the compiled code, so we must update "code"
1993        here. */
1994    
1995  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1996      case OP_XCLASS:      case OP_XCLASS:
1997      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1998      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1999  #endif  #endif
2000    
# Line 1260  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2039      case OP_WORDCHAR:      case OP_WORDCHAR:
2040      case OP_ANY:      case OP_ANY:
2041        case OP_ALLANY:
2042      case OP_ANYBYTE:      case OP_ANYBYTE:
2043      case OP_CHAR:      case OP_CHAR:
2044      case OP_CHARNC:      case OP_CHARNC:
2045      case OP_NOT:      case OP_NOT:
2046      case OP_PLUS:      case OP_PLUS:
2047      case OP_MINPLUS:      case OP_MINPLUS:
2048        case OP_POSPLUS:
2049      case OP_EXACT:      case OP_EXACT:
2050      case OP_NOTPLUS:      case OP_NOTPLUS:
2051      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2052        case OP_NOTPOSPLUS:
2053      case OP_NOTEXACT:      case OP_NOTEXACT:
2054      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2055      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2056        case OP_TYPEPOSPLUS:
2057      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2058      return FALSE;      return FALSE;
2059    
2060        /* These are going to continue, as they may be empty, but we have to
2061        fudge the length for the \p and \P cases. */
2062    
2063        case OP_TYPESTAR:
2064        case OP_TYPEMINSTAR:
2065        case OP_TYPEPOSSTAR:
2066        case OP_TYPEQUERY:
2067        case OP_TYPEMINQUERY:
2068        case OP_TYPEPOSQUERY:
2069        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2070        break;
2071    
2072        /* Same for these */
2073    
2074        case OP_TYPEUPTO:
2075        case OP_TYPEMINUPTO:
2076        case OP_TYPEPOSUPTO:
2077        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2078        break;
2079    
2080      /* End of branch */      /* End of branch */
2081    
2082      case OP_KET:      case OP_KET:
# Line 1283  for (code = first_significant_code(code Line 2085  for (code = first_significant_code(code
2085      case OP_ALT:      case OP_ALT:
2086      return TRUE;      return TRUE;
2087    
2088      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2089      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2090    
2091  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2092      case OP_STAR:      case OP_STAR:
2093      case OP_MINSTAR:      case OP_MINSTAR:
2094        case OP_POSSTAR:
2095      case OP_QUERY:      case OP_QUERY:
2096      case OP_MINQUERY:      case OP_MINQUERY:
2097        case OP_POSQUERY:
2098        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2099        break;
2100    
2101      case OP_UPTO:      case OP_UPTO:
2102      case OP_MINUPTO:      case OP_MINUPTO:
2103      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2104        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2105      break;      break;
2106  #endif  #endif
2107    
2108        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2109        string. */
2110    
2111        case OP_MARK:
2112        case OP_PRUNE_ARG:
2113        case OP_SKIP_ARG:
2114        code += code[1];
2115        break;
2116    
2117        case OP_THEN_ARG:
2118        code += code[1+LINK_SIZE];
2119        break;
2120    
2121        /* None of the remaining opcodes are required to match a character. */
2122    
2123        default:
2124        break;
2125      }      }
2126    }    }
2127    
# Line 1318  Arguments: Line 2144  Arguments:
2144    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2145    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2146    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2147      cd          pointers to tables etc
2148    
2149  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2150  */  */
2151    
2152  static BOOL  static BOOL
2153  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2154    BOOL utf8)    BOOL utf8, compile_data *cd)
2155  {  {
2156  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2157    {    {
2158    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2159        return FALSE;
2160    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2161    }    }
2162  return TRUE;  return TRUE;
# Line 1341  return TRUE; Line 2169  return TRUE;
2169  *************************************************/  *************************************************/
2170    
2171  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2172  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
2173  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2174  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2175    
2176    Originally, this function only recognized a sequence of letters between the
2177    terminators, but it seems that Perl recognizes any sequence of characters,
2178    though of course unknown POSIX names are subsequently rejected. Perl gives an
2179    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2180    didn't consider this to be a POSIX class. Likewise for [:1234:].
2181    
2182    The problem in trying to be exactly like Perl is in the handling of escapes. We
2183    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2184    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2185    below handles the special case of \], but does not try to do any other escape
2186    processing. This makes it different from Perl for cases such as [:l\ower:]
2187    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2188    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2189    I think.
2190    
2191  Argument:  Arguments:
2192    ptr      pointer to the initial [    ptr      pointer to the initial [
2193    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2194    
2195  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2196  */  */
2197    
2198  static BOOL  static BOOL
2199  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2200  {  {
2201  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2202  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2203  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2204    {    {
2205    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2206    return TRUE;      {
2207        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2208        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2209          {
2210          *endptr = ptr;
2211          return TRUE;
2212          }
2213        }
2214    }    }
2215  return FALSE;  return FALSE;
2216  }  }
# Line 1388  Returns:     a value representing the na Line 2235  Returns:     a value representing the na
2235  static int  static int
2236  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2237  {  {
2238    const char *pn = posix_names;
2239  register int yield = 0;  register int yield = 0;
2240  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2241    {    {
2242    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2243      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2244      pn += posix_name_lengths[yield] + 1;
2245    yield++;    yield++;
2246    }    }
2247  return -1;  return -1;
# Line 1407  return -1; Line 2256  return -1;
2256  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2257  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2258  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
2259  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
2260  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
2261  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
2262  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
2263  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2264    OP_END.
2265    
2266    This function has been extended with the possibility of forward references for
2267    recursions and subroutine calls. It must also check the list of such references
2268    for the group we are dealing with. If it finds that one of the recursions in
2269    the current group is on this list, it adjusts the offset in the list, not the
2270    value in the reference (which is a group number).
2271    
2272  Arguments:  Arguments:
2273    group      points to the start of the group    group      points to the start of the group
2274    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2275    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2276    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2277      save_hwm   the hwm forward reference pointer at the start of the group
2278    
2279  Returns:     nothing  Returns:     nothing
2280  */  */
2281    
2282  static void  static void
2283  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2284      uschar *save_hwm)
2285  {  {
2286  uschar *ptr = group;  uschar *ptr = group;
2287    
2288  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2289    {    {
2290    int offset = GET(ptr, 1);    int offset;
2291    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2292    
2293      /* See if this recursion is on the forward reference list. If so, adjust the
2294      reference. */
2295    
2296      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2297        {
2298        offset = GET(hc, 0);
2299        if (cd->start_code + offset == ptr + 1)
2300          {
2301          PUT(hc, 0, offset + adjust);
2302          break;
2303          }
2304        }
2305    
2306      /* Otherwise, adjust the recursion offset if it's after the start of this
2307      group. */
2308    
2309      if (hc >= cd->hwm)
2310        {
2311        offset = GET(ptr, 1);
2312        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2313        }
2314    
2315    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2316    }    }
2317  }  }
# Line 1456  auto_callout(uschar *code, const uschar Line 2338  auto_callout(uschar *code, const uschar
2338  {  {
2339  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2340  *code++ = 255;  *code++ = 255;
2341  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2342  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2343  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2344  }  }
2345    
# Line 1482  Returns:             nothing Line 2364  Returns:             nothing
2364  static void  static void
2365  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2366  {  {
2367  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2368  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2369  }  }
2370    
# Line 1508  Yield:        TRUE when range returned; Line 2390  Yield:        TRUE when range returned;
2390  */  */
2391    
2392  static BOOL  static BOOL
2393  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2394      unsigned int *odptr)
2395  {  {
2396  int c, othercase, next;  unsigned int c, othercase, next;
2397    
2398  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2399    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2400    
2401  if (c > d) return FALSE;  if (c > d) return FALSE;
2402    
# Line 1522  next = othercase + 1; Line 2405  next = othercase + 1;
2405    
2406  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2407    {    {
2408    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2409    next++;    next++;
2410    }    }
2411    
# Line 1531  for (++c; c <= d; c++) Line 2414  for (++c; c <= d; c++)
2414    
2415  return TRUE;  return TRUE;
2416  }  }
2417  #endif  /* SUPPORT_UCP */  
2418    
2419    
2420  /*************************************************  /*************************************************
2421  *           Compile one branch                   *  *        Check a character and a property        *
2422  *************************************************/  *************************************************/
2423    
2424  /* 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
2425  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2426    
2427  Arguments:  Arguments:
2428    optionsptr     pointer to the option bits    c            the character
2429    brackets       points to number of extracting brackets used    ptype        the property type
2430    codeptr        points to the pointer to the current code point    pdata        the data for the type
2431    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.  
2432    
2433  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2434  */  */
2435    
2436  static BOOL  static BOOL
2437  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)  
2438  {  {
2439  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2440  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2441  int bravalue = 0;    {
2442  int greedy_default, greedy_non_default;    case PT_LAMP:
2443  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2444  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2445  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2446  int condcount = 0;  
2447  int options = *optionsptr;    case PT_GC:
2448  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2449  register int c;  
2450  register uschar *code = *codeptr;    case PT_PC:
2451  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2452  BOOL inescq = FALSE;  
2453  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2454  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2455  const uschar *tempptr;  
2456  uschar *previous = NULL;    /* These are specials */
2457  uschar *previous_callout = NULL;  
2458  uschar classbits[32];    case PT_ALNUM:
2459      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2460              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2461    
2462      case PT_SPACE:    /* Perl space */
2463      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2464              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2465              == negated;
2466    
2467      case PT_PXSPACE:  /* POSIX space */
2468      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2469              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2470              c == CHAR_FF || c == CHAR_CR)
2471              == negated;
2472    
2473      case PT_WORD:
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2475              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2476              c == CHAR_UNDERSCORE) == negated;
2477      }
2478    return FALSE;
2479    }
2480    #endif  /* SUPPORT_UCP */
2481    
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
 #endif  
2482    
 /* Set up the default and non-default settings for greediness */  
2483    
2484  greedy_default = ((options & PCRE_UNGREEDY) != 0);  /*************************************************
2485  greedy_non_default = greedy_default ^ 1;  *     Check if auto-possessifying is possible    *
2486    *************************************************/
2487    
2488  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* This function is called for unlimited repeats of certain items, to see
2489  matching encountered yet". It gets changed to REQ_NONE if we hit something that  whether the next thing could possibly match the repeated item. If not, it makes
2490  matches a non-fixed char first char; reqbyte just remains unset if we never  sense to automatically possessify the repeated item.
 find one.  
2491    
2492  When we hit a repeat whose minimum is zero, we may have to adjust these values  Arguments:
2493  to take the zero repeat into account. This is implemented by setting them to    previous      pointer to the repeated opcode
2494  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual    utf8          TRUE in UTF-8 mode
2495  item types that can be repeated set these backoff variables appropriately. */    ptr           next character in pattern
2496      options       options bits
2497      cd            contains pointers to tables etc.
2498    
2499  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Returns:        TRUE if possessifying is wanted
2500    */
2501    
2502  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  static BOOL
2503  according to the current setting of the caseless flag. REQ_CASELESS is a bit  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2504  value > 255. It is added into the firstbyte or reqbyte variables to record the    int options, compile_data *cd)
2505  case status of the value. This is used only for ASCII characters. */  {
2506    int c, next;
2507    int op_code = *previous++;
2508    
2509  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  /* Skip whitespace and comments in extended mode */
2510    
2511  /* Switch on next character until the end of the branch */  if ((options & PCRE_EXTENDED) != 0)
2512      {
2513      for (;;)
2514        {
2515        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2516        if (*ptr == CHAR_NUMBER_SIGN)
2517          {
2518          while (*(++ptr) != 0)
2519            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2520          }
2521        else break;
2522        }
2523      }
2524    
2525  for (;; ptr++)  /* If the next item is one that we can handle, get its value. A non-negative
2526    value is a character, a negative value is an escape value. */
2527    
2528    if (*ptr == CHAR_BACKSLASH)
2529    {    {
2530    BOOL negate_class;    int temperrorcode = 0;
2531    BOOL possessive_quantifier;    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2532    BOOL is_quantifier;    if (temperrorcode != 0) return FALSE;
2533    int class_charcount;    ptr++;    /* Point after the escape sequence */
2534    int class_lastchar;    }
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
2535    
2536    /* Next byte in the pattern */  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2537      {
2538    #ifdef SUPPORT_UTF8
2539      if (utf8) { GETCHARINC(next, ptr); } else
2540    #endif
2541      next = *ptr++;
2542      }
2543    
2544    c = *ptr;  else return FALSE;
2545    
2546    /* If in \Q...\E, check for the end; if not, we have a literal */  /* Skip whitespace and comments in extended mode */
2547    
2548    if (inescq && c != 0)  if ((options & PCRE_EXTENDED) != 0)
2549      {
2550      for (;;)
2551      {      {
2552      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2553        {      if (*ptr == CHAR_NUMBER_SIGN)
       inescq = FALSE;  
       ptr++;  
       continue;  
       }  
     else  
2554        {        {
2555        if (previous_callout != NULL)        while (*(++ptr) != 0)
2556          {          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
         {  
         previous_callout = code;  
         code = auto_callout(code, ptr, cd);  
         }  
       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 thing is itself optional, we have to give up. */
   a quantifier. */  
2563    
2564    is_quantifier = c == '*' || c == '+' || c == '?' ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2565      (c == '{' && is_counted_repeat(ptr+1));    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2566        return FALSE;
2567    
2568    if (!is_quantifier && previous_callout != NULL &&  /* Now compare the next item with the previous opcode. First, handle cases when
2569         after_manual_callout-- <= 0)  the next item is a character. */
     {  
     complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
2570    
2571    /* In extended mode, skip white space and comments */  if (next >= 0) switch(op_code)
2572      {
2573      case OP_CHAR:
2574    #ifdef SUPPORT_UTF8
2575      GETCHARTEST(c, previous);
2576    #else
2577      c = *previous;
2578    #endif
2579      return c != next;
2580    
2581    if ((options & PCRE_EXTENDED) != 0)    /* For CHARNC (caseless character) we must check the other case. If we have
2582      Unicode property support, we can use it to test the other case of
2583      high-valued characters. */
2584    
2585      case OP_CHARNC:
2586    #ifdef SUPPORT_UTF8
2587      GETCHARTEST(c, previous);
2588    #else
2589      c = *previous;
2590    #endif
2591      if (c == next) return FALSE;
2592    #ifdef SUPPORT_UTF8
2593      if (utf8)
2594      {      {
2595      if ((cd->ctypes[c] & ctype_space) != 0) continue;      unsigned int othercase;
2596      if (c == '#')      if (next < 128) othercase = cd->fcc[next]; else
2597        {  #ifdef SUPPORT_UCP
2598        /* The space before the ; is to avoid a warning on a silly compiler      othercase = UCD_OTHERCASE((unsigned int)next);
2599        on the Macintosh. */  #else
2600        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;      othercase = NOTACHAR;
2601        if (c != 0) continue;   /* Else fall through to handle end of string */  #endif
2602        }      return (unsigned int)c != othercase;
2603      }      }
2604      else
2605    #endif  /* SUPPORT_UTF8 */
2606      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2607    
2608    /* No auto callout for quantifiers. */    /* For OP_NOT, its data is always a single-byte character. */
2609    
2610    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    case OP_NOT:
2611      if ((c = *previous) == next) return TRUE;
2612      if ((options & PCRE_CASELESS) == 0) return FALSE;
2613    #ifdef SUPPORT_UTF8
2614      if (utf8)
2615      {      {
2616      previous_callout = code;      unsigned int othercase;
2617      code = auto_callout(code, ptr, cd);      if (next < 128) othercase = cd->fcc[next]; else
2618    #ifdef SUPPORT_UCP
2619        othercase = UCD_OTHERCASE(next);
2620    #else
2621        othercase = NOTACHAR;
2622    #endif
2623        return (unsigned int)c == othercase;
2624      }      }
2625      else
2626    #endif  /* SUPPORT_UTF8 */
2627      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2628    
2629    switch(c)    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2630      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2631    
2632      case OP_DIGIT:
2633      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2634    
2635      case OP_NOT_DIGIT:
2636      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2637    
2638      case OP_WHITESPACE:
2639      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2640    
2641      case OP_NOT_WHITESPACE:
2642      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2643    
2644      case OP_WORDCHAR:
2645      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2646    
2647      case OP_NOT_WORDCHAR:
2648      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2649    
2650      case OP_HSPACE:
2651      case OP_NOT_HSPACE:
2652      switch(next)
2653      {      {
2654      /* The branch terminates at end of string, |, or ). */      case 0x09:
2655        case 0x20:
2656        case 0xa0:
2657        case 0x1680:
2658        case 0x180e:
2659        case 0x2000:
2660        case 0x2001:
2661        case 0x2002:
2662        case 0x2003:
2663        case 0x2004:
2664        case 0x2005:
2665        case 0x2006:
2666        case 0x2007:
2667        case 0x2008:
2668        case 0x2009:
2669        case 0x200A:
2670        case 0x202f:
2671        case 0x205f:
2672        case 0x3000:
2673        return op_code == OP_NOT_HSPACE;
2674        default:
2675        return op_code != OP_NOT_HSPACE;
2676        }
2677    
2678      case 0:    case OP_ANYNL:
2679      case '|':    case OP_VSPACE:
2680      case ')':    case OP_NOT_VSPACE:
2681      *firstbyteptr = firstbyte;    switch(next)
2682      *reqbyteptr = reqbyte;      {
2683      *codeptr = code;      case 0x0a:
2684      *ptrptr = ptr;      case 0x0b:
2685      return TRUE;      case 0x0c:
2686        case 0x0d:
2687        case 0x85:
2688        case 0x2028:
2689        case 0x2029:
2690        return op_code == OP_NOT_VSPACE;
2691        default:
2692        return op_code != OP_NOT_VSPACE;
2693        }
2694    
2695      /* Handle single-character metacharacters. In multiline mode, ^ disables  #ifdef SUPPORT_UCP
2696      the setting of any following char as a first character. */    case OP_PROP:
2697      return check_char_prop(next, previous[0], previous[1], FALSE);
2698    
2699      case '^':    case OP_NOTPROP:
2700      if ((options & PCRE_MULTILINE) != 0)    return check_char_prop(next, previous[0], previous[1], TRUE);
2701        {  #endif
       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
       }  
     previous = NULL;  
     *code++ = OP_CIRC;  
     break;  
2702    
2703      case '$':    default:
2704      previous = NULL;    return FALSE;
2705      *code++ = OP_DOLL;    }
     break;  
2706    
     /* There can never be a first char if '.' is first, whatever happens about  
     repeats. The value of reqbyte doesn't change either. */  
2707    
2708      case '.':  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2709      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2710      zerofirstbyte = firstbyte;  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2711      zeroreqbyte = reqbyte;  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2712      previous = code;  replaced by OP_PROP codes when PCRE_UCP is set. */
     *code++ = OP_ANY;  
     break;  
2713    
2714      /* Character classes. If the included characters are all < 256, we build a  switch(op_code)
2715      32-byte bitmap of the permitted characters, except in the special case    {
2716      where there is only one such character. For negated classes, we build the    case OP_CHAR:
2717      map as usual, then invert it at the end. However, we use a different opcode    case OP_CHARNC:
2718      so that data characters > 255 can be handled correctly.  #ifdef SUPPORT_UTF8
2719      GETCHARTEST(c, previous);
2720    #else
2721      c = *previous;
2722    #endif
2723      switch(-next)
2724        {
2725        case ESC_d:
2726        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2727    
2728      If the class contains characters outside the 0-255 range, a different      case ESC_D:
2729      opcode is compiled. It may optionally have a bit map for characters < 256,      return c <= 127 && (cd->ctypes[c] & 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.  
     */  
2730    
2731      case '[':      case ESC_s:
2732      previous = code;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2733    
2734      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case ESC_S:
2735      they are encountered at the top level, so we'll do that too. */      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2736    
2737        case ESC_w:
2738        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2739    
2740        case ESC_W:
2741        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2742    
2743        case ESC_h:
2744        case ESC_H:
2745        switch(c)
2746          {
2747          case 0x09:
2748          case 0x20:
2749          case 0xa0:
2750          case 0x1680:
2751          case 0x180e:
2752          case 0x2000:
2753          case 0x2001:
2754          case 0x2002:
2755          case 0x2003:
2756          case 0x2004:
2757          case 0x2005:
2758          case 0x2006:
2759          case 0x2007:
2760          case 0x2008:
2761          case 0x2009:
2762          case 0x200A:
2763          case 0x202f:
2764          case 0x205f:
2765          case 0x3000:
2766          return -next != ESC_h;
2767          default:
2768          return -next == ESC_h;
2769          }
2770    
2771      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case ESC_v:
2772          check_posix_syntax(ptr, &tempptr, cd))      case ESC_V:
2773        {      switch(c)
2774        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        {
2775        goto FAILED;        case 0x0a:
2776          case 0x0b:
2777          case 0x0c:
2778          case 0x0d:
2779          case 0x85:
2780          case 0x2028:
2781          case 0x2029:
2782          return -next != ESC_v;
2783          default:
2784          return -next == ESC_v;
2785        }        }
2786    
2787      /* If the first character is '^', set the negation flag and skip it. */      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2788        their substitutions and process them. The result will always be either
2789        -ESC_p or -ESC_P. Then fall through to process those values. */
2790    
2791      if ((c = *(++ptr)) == '^')  #ifdef SUPPORT_UCP
2792        {      case ESC_du:
2793        negate_class = TRUE;      case ESC_DU:
2794        c = *(++ptr);      case ESC_wu:
2795        case ESC_WU:
2796        case ESC_su:
2797        case ESC_SU:
2798          {
2799          int temperrorcode = 0;
2800          ptr = substitutes[-next - ESC_DU];
2801          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2802          if (temperrorcode != 0) return FALSE;
2803          ptr++;    /* For compatibility */
2804        }        }
2805      else      /* Fall through */
2806    
2807        case ESC_p:
2808        case ESC_P:
2809        {        {
2810        negate_class = FALSE;        int ptype, pdata, errorcodeptr;
2811        }        BOOL negated;
2812    
2813      /* Keep a count of chars with values < 256 so that we can optimize the case        ptr--;      /* Make ptr point at the p or P */
2814      of just a single character (as long as it's < 256). For higher valued UTF-8        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2815      characters, we don't yet do any optimization. */        if (ptype < 0) return FALSE;
2816          ptr++;      /* Point past the final curly ket */
2817    
2818          /* If the property item is optional, we have to give up. (When generated
2819          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2820          to the original \d etc. At this point, ptr will point to a zero byte. */
2821    
2822          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2823            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2824              return FALSE;
2825    
2826      class_charcount = 0;        /* Do the property check. */
     class_lastchar = -1;  
2827    
2828  #ifdef SUPPORT_UTF8        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2829      class_utf8 = FALSE;                       /* No chars >= 256 */        }
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
2830  #endif  #endif
2831    
2832      /* Initialize the 32-char bit map to all zeros. We have to build the      default:
2833      map in a temporary bit of store, in case the class contains only 1      return FALSE;
2834      character (< 256), because in that case the compiled code doesn't use the      }
     bit map. */  
2835    
2836      memset(classbits, 0, 32 * sizeof(uschar));    /* In principle, support for Unicode properties should be integrated here as
2837      well. It means re-organizing the above code so as to get hold of the property
2838      values before switching on the op-code. However, I wonder how many patterns
2839      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2840      these op-codes are never generated.) */
2841    
2842      /* Process characters until ] is reached. By writing this as a "do" it    case OP_DIGIT:
2843      means that an initial ] is taken as a data character. The first pass    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2844      through the regex checked the overall syntax, so we don't need to be very           next == -ESC_h || next == -ESC_v || next == -ESC_R;
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
2845    
2846      do    case OP_NOT_DIGIT:
2847        {    return next == -ESC_d;
 #ifdef SUPPORT_UTF8  
       if (utf8 && c > 127)  
         {                           /* Braces are required because the */  
         GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */  
         }  
 #endif  
2848    
2849        /* Inside \Q...\E everything is literal except \E */    case OP_WHITESPACE:
2850      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2851    
2852        if (inescq)    case OP_NOT_WHITESPACE:
2853          {    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
2854    
2855        /* Handle POSIX class names. Perl allows a negation extension of the    case OP_HSPACE:
2856        form [:^name:]. A square bracket that doesn't match the syntax is    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2857        treated as a literal. We also recognize the POSIX constructions           next == -ESC_w || next == -ESC_v || next == -ESC_R;
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
2858    
2859        if (c == '[' &&    case OP_NOT_HSPACE:
2860            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    return next == -ESC_h;
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, taboffset, tabopt;  
         register const uschar *cbits = cd->cbits;  
         uschar pbits[32];  
2861    
2862          if (ptr[1] != ':')    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2863            {    case OP_ANYNL:
2864            *errorcodeptr = ERR31;    case OP_VSPACE:
2865            goto FAILED;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
           }  
2866    
2867          ptr += 2;    case OP_NOT_VSPACE:
2868          if (*ptr == '^')    return next == -ESC_v || next == -ESC_R;
           {  
           local_negate = TRUE;  
           ptr++;  
           }  
2869    
2870          posix_class = check_posix_name(ptr, tempptr - ptr);    case OP_WORDCHAR:
2871          if (posix_class < 0)    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2872            {           next == -ESC_v || next == -ESC_R;
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
2873    
2874          /* If matching is caseless, upper and lower are converted to    case OP_NOT_WORDCHAR:
2875          alpha. This relies on the fact that the class table starts with    return next == -ESC_w || next == -ESC_d;
         alpha, lower, upper as the first 3 entries. */  
2876    
2877          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)    default:
2878            posix_class = 0;    return FALSE;
2879      }
2880    
2881          /* We build the bit map for the POSIX class in a chunk of local store  /* Control does not reach here */
2882          because we may be adding and subtracting from it, and we don't want to  }
         subtract bits that may be in the main map already. At the end we or the  
         result into the bit map that is being built. */  
2883    
         posix_class *= 3;  
2884    
         /* Copy in the first table (always present) */  
2885    
2886          memcpy(pbits, cbits + posix_class_maps[posix_class],  /*************************************************
2887            32 * sizeof(uschar));  *           Compile one branch                   *
2888    *************************************************/
2889    
2890          /* If there is a second table, add or remove it as required. */  /* Scan the pattern, compiling it into the a vector. If the options are
2891    changed during the branch, the pointer is used to change the external options
2892    bits. This function is used during the pre-compile phase when we are trying
2893    to find out the amount of memory needed, as well as during the real compile
2894    phase. The value of lengthptr distinguishes the two phases.
2895    
2896          taboffset = posix_class_maps[posix_class + 1];  Arguments:
2897          tabopt = posix_class_maps[posix_class + 2];    optionsptr     pointer to the option bits
2898      codeptr        points to the pointer to the current code point
2899      ptrptr         points to the current pattern pointer
2900      errorcodeptr   points to error code variable
2901      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2902      reqbyteptr     set to the last literal character required, else < 0
2903      bcptr          points to current branch chain
2904      cd             contains pointers to tables etc.
2905      lengthptr      NULL during the real compile phase
2906                     points to length accumulator during pre-compile phase
2907    
2908          if (taboffset >= 0)  Returns:         TRUE on success
2909            {                   FALSE, with *errorcodeptr set non-zero on error
2910            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];  
           }  
2911    
2912          /* Not see if we need to remove any special characters. An option  static BOOL
2913          value of 1 removes vertical space and 2 removes underscore. */  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2914      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2915      compile_data *cd, int *lengthptr)
2916    {
2917    int repeat_type, op_type;
2918    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2919    int bravalue = 0;
2920    int greedy_default, greedy_non_default;
2921    int firstbyte, reqbyte;
2922    int zeroreqbyte, zerofirstbyte;
2923    int req_caseopt, reqvary, tempreqvary;
2924    int options = *optionsptr;
2925    int after_manual_callout = 0;
2926    int length_prevgroup = 0;
2927    register int c;
2928    register uschar *code = *codeptr;
2929    uschar *last_code = code;
2930    uschar *orig_code = code;
2931    uschar *tempcode;
2932    BOOL inescq = FALSE;
2933    BOOL groupsetfirstbyte = FALSE;
2934    const uschar *ptr = *ptrptr;
2935    const uschar *tempptr;
2936    const uschar *nestptr = NULL;
2937    uschar *previous = NULL;
2938    uschar *previous_callout = NULL;
2939    uschar *save_hwm = NULL;
2940    uschar classbits[32];
2941    
2942          if (tabopt < 0) tabopt = -tabopt;  #ifdef SUPPORT_UTF8
2943          if (tabopt == 1) pbits[1] &= ~0x3c;  BOOL class_utf8;
2944            else if (tabopt == 2) pbits[11] &= 0x7f;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2945    uschar *class_utf8data;
2946    uschar *class_utf8data_base;
2947    uschar utf8_char[6];
2948    #else
2949    BOOL utf8 = FALSE;
2950    uschar *utf8_char = NULL;
2951    #endif
2952    
2953          /* Add the POSIX table or its complement into the main table that is  #ifdef PCRE_DEBUG
2954          being built and we are done. */  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2955    #endif
2956    
2957          if (local_negate)  /* Set up the default and non-default settings for greediness */
           for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];  
         else  
           for (c = 0; c < 32; c++) classbits[c] |= pbits[c];  
2958    
2959          ptr = tempptr + 1;  greedy_default = ((options & PCRE_UNGREEDY) != 0);
2960          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */  greedy_non_default = greedy_default ^ 1;
         continue;    /* End of POSIX syntax handling */  
         }  
2961    
2962        /* Backslash may introduce a single character, or it may introduce one  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2963        of the specials, which just set a flag. Escaped items are checked for  matching encountered yet". It gets changed to REQ_NONE if we hit something that
2964        validity in the pre-compiling pass. The sequence \b is a special case.  matches a non-fixed char first char; reqbyte just remains unset if we never
2965        Inside a class (and only there) it is treated as backspace. Elsewhere  find one.
       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. */  
2966    
2967        if (c == '\\')  When we hit a repeat whose minimum is zero, we may have to adjust these values
2968          {  to take the zero repeat into account. This is implemented by setting them to
2969          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2970    item types that can be repeated set these backoff variables appropriately. */
2971    
2972          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         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;  
           }  
2973    
2974          if (c < 0)  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
2975            {  according to the current setting of the caseless flag. REQ_CASELESS is a bit
2976            register const uschar *cbits = cd->cbits;  value > 255. It is added into the firstbyte or reqbyte variables to record the
2977            class_charcount += 2;     /* Greater than 1 is what matters */  case status of the value. This is used only for ASCII characters. */
           switch (-c)  
             {  
             case ESC_d:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];  
             continue;  
2978    
2979              case ESC_D:  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
             for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];  
             continue;  
2980    
2981              case ESC_w:  /* Switch on next character until the end of the branch */
2982              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];  
2983              continue;  for (;; ptr++)
2984      {
2985      BOOL negate_class;
2986      BOOL should_flip_negation;
2987      BOOL possessive_quantifier;
2988      BOOL is_quantifier;
2989      BOOL is_recurse;
2990      BOOL reset_bracount;
2991      int class_charcount;
2992      int class_lastchar;
2993      int newoptions;
2994      int recno;
2995      int refsign;
2996      int skipbytes;
2997      int subreqbyte;
2998      int subfirstbyte;
2999      int terminator;
3000      int mclength;
3001      uschar mcbuffer[8];
3002    
3003      /* Get next byte in the pattern */
3004    
3005      c = *ptr;
3006    
3007      /* If we are at the end of a nested substitution, revert to the outer level
3008      string. Nesting only happens one level deep. */
3009    
3010      if (c == 0 && nestptr != NULL)
3011        {
3012        ptr = nestptr;
3013        nestptr = NULL;
3014        c = *ptr;
3015        }
3016    
3017      /* If we are in the pre-compile phase, accumulate the length used for the
3018      previous cycle of this loop. */
3019    
3020      if (lengthptr != NULL)
3021        {
3022    #ifdef PCRE_DEBUG
3023        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3024    #endif
3025        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3026          {
3027          *errorcodeptr = ERR52;
3028          goto FAILED;
3029          }
3030    
3031        /* There is at least one situation where code goes backwards: this is the
3032        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3033        the class is simply eliminated. However, it is created first, so we have to
3034        allow memory for it. Therefore, don't ever reduce the length at this point.
3035        */
3036    
3037        if (code < last_code) code = last_code;
3038    
3039        /* Paranoid check for integer overflow */
3040    
3041        if (OFLOW_MAX - *lengthptr < code - last_code)
3042          {
3043          *errorcodeptr = ERR20;
3044          goto FAILED;
3045          }
3046    
3047        *lengthptr += (int)(code - last_code);
3048        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3049    
3050        /* If "previous" is set and it is not at the start of the work space, move
3051        it back to there, in order to avoid filling up the work space. Otherwise,
3052        if "previous" is NULL, reset the current code pointer to the start. */
3053    
3054        if (previous != NULL)
3055          {
3056          if (previous > orig_code)
3057            {
3058            memmove(orig_code, previous, code - previous);
3059            code -= previous - orig_code;
3060            previous = orig_code;
3061            }
3062          }
3063        else code = orig_code;
3064    
3065        /* Remember where this code item starts so we can pick up the length
3066        next time round. */
3067    
3068        last_code = code;
3069        }
3070    
3071      /* In the real compile phase, just check the workspace used by the forward
3072      reference list. */
3073    
3074      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3075        {
3076        *errorcodeptr = ERR52;
3077        goto FAILED;
3078        }
3079    
3080      /* If in \Q...\E, check for the end; if not, we have a literal */
3081    
3082      if (inescq && c != 0)
3083        {
3084        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3085          {
3086          inescq = FALSE;
3087          ptr++;
3088          continue;
3089          }
3090        else
3091          {
3092          if (previous_callout != NULL)
3093            {
3094            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3095              complete_callout(previous_callout, ptr, cd);
3096            previous_callout = NULL;
3097            }
3098          if ((options & PCRE_AUTO_CALLOUT) != 0)
3099            {
3100            previous_callout = code;
3101            code = auto_callout(code, ptr, cd);
3102            }
3103          goto NORMAL_CHAR;
3104          }
3105        }
3106    
3107      /* Fill in length of a previous callout, except when the next thing is
3108      a quantifier. */
3109    
3110      is_quantifier =
3111        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3112        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3113    
3114      if (!is_quantifier && previous_callout != NULL &&
3115           after_manual_callout-- <= 0)
3116        {
3117        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3118          complete_callout(previous_callout, ptr, cd);
3119        previous_callout = NULL;
3120        }
3121    
3122      /* In extended mode, skip white space and comments */
3123    
3124      if ((options & PCRE_EXTENDED) != 0)
3125        {
3126        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3127        if (c == CHAR_NUMBER_SIGN)
3128          {
3129          while (*(++ptr) != 0)
3130            {
3131            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3132            }
3133          if (*ptr != 0) continue;
3134    
3135          /* Else fall through to handle end of string */
3136          c = 0;
3137          }
3138        }
3139    
3140      /* No auto callout for quantifiers. */
3141    
3142      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3143        {
3144        previous_callout = code;
3145        code = auto_callout(code, ptr, cd);
3146        }
3147    
3148      switch(c)
3149        {
3150        /* ===================================================================*/
3151        case 0:                        /* The branch terminates at string end */
3152        case CHAR_VERTICAL_LINE:       /* or | or ) */
3153        case CHAR_RIGHT_PARENTHESIS:
3154        *firstbyteptr = firstbyte;
3155        *reqbyteptr = reqbyte;
3156        *codeptr = code;
3157        *ptrptr = ptr;
3158        if (lengthptr != NULL)
3159          {
3160          if (OFLOW_MAX - *lengthptr < code - last_code)
3161            {
3162            *errorcodeptr = ERR20;
3163            goto FAILED;
3164            }
3165          *lengthptr += (int)(code - last_code);   /* To include callout length */
3166          DPRINTF((">> end branch\n"));
3167          }
3168        return TRUE;
3169    
3170    
3171        /* ===================================================================*/
3172        /* Handle single-character metacharacters. In multiline mode, ^ disables
3173        the setting of any following char as a first character. */
3174    
3175        case CHAR_CIRCUMFLEX_ACCENT:
3176        if ((options & PCRE_MULTILINE) != 0)
3177          {
3178          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3179          }
3180        previous = NULL;
3181        *code++ = OP_CIRC;
3182        break;
3183    
3184        case CHAR_DOLLAR_SIGN:
3185        previous = NULL;
3186        *code++ = OP_DOLL;
3187        break;
3188    
3189        /* There can never be a first char if '.' is first, whatever happens about
3190        repeats. The value of reqbyte doesn't change either. */
3191    
3192        case CHAR_DOT:
3193        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3194        zerofirstbyte = firstbyte;
3195        zeroreqbyte = reqbyte;
3196        previous = code;
3197        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3198        break;
3199    
3200    
3201        /* ===================================================================*/
3202        /* Character classes. If the included characters are all < 256, we build a
3203        32-byte bitmap of the permitted characters, except in the special case
3204        where there is only one such character. For negated classes, we build the
3205        map as usual, then invert it at the end. However, we use a different opcode
3206        so that data characters > 255 can be handled correctly.
3207    
3208        If the class contains characters outside the 0-255 range, a different
3209        opcode is compiled. It may optionally have a bit map for characters < 256,
3210        but those above are are explicitly listed afterwards. A flag byte tells
3211        whether the bitmap is present, and whether this is a negated class or not.
3212    
3213        In JavaScript compatibility mode, an isolated ']' causes an error. In
3214        default (Perl) mode, it is treated as a data character. */
3215    
3216        case CHAR_RIGHT_SQUARE_BRACKET:
3217        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3218          {
3219          *errorcodeptr = ERR64;
3220          goto FAILED;
3221          }
3222        goto NORMAL_CHAR;
3223    
3224        case CHAR_LEFT_SQUARE_BRACKET:
3225        previous = code;
3226    
3227        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3228        they are encountered at the top level, so we'll do that too. */
3229    
3230        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3231             ptr[1] == CHAR_EQUALS_SIGN) &&
3232            check_posix_syntax(ptr, &tempptr))
3233          {
3234          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3235          goto FAILED;
3236          }
3237    
3238        /* If the first character is '^', set the negation flag and skip it. Also,
3239        if the first few characters (either before or after ^) are \Q\E or \E we
3240        skip them too. This makes for compatibility with Perl. */
3241    
3242        negate_class = FALSE;
3243        for (;;)
3244          {
3245          c = *(++ptr);
3246          if (c == CHAR_BACKSLASH)
3247            {
3248            if (ptr[1] == CHAR_E)
3249              ptr++;
3250            else if (strncmp((const char *)ptr+1,
3251                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3252              ptr += 3;
3253            else
3254              break;
3255            }
3256          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3257            negate_class = TRUE;
3258          else break;
3259          }
3260    
3261        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3262        an initial ']' is taken as a data character -- the code below handles
3263        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3264        [^] must match any character, so generate OP_ALLANY. */
3265    
3266        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3267            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3268          {
3269          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3270          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3271          zerofirstbyte = firstbyte;
3272          break;
3273          }
3274    
3275        /* If a class contains a negative special such as \S, we need to flip the
3276        negation flag at the end, so that support for characters > 255 works
3277        correctly (they are all included in the class). */
3278    
3279        should_flip_negation = FALSE;
3280    
3281        /* Keep a count of chars with values < 256 so that we can optimize the case
3282        of just a single character (as long as it's < 256). However, For higher
3283        valued UTF-8 characters, we don't yet do any optimization. */
3284    
3285        class_charcount = 0;
3286        class_lastchar = -1;
3287    
3288        /* Initialize the 32-char bit map to all zeros. We build the map in a
3289        temporary bit of memory, in case the class contains only 1 character (less
3290        than 256), because in that case the compiled code doesn't use the bit map.
3291        */
3292    
3293        memset(classbits, 0, 32 * sizeof(uschar));
3294    
3295    #ifdef SUPPORT_UTF8
3296        class_utf8 = FALSE;                       /* No chars >= 256 */
3297        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3298        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3299    #endif
3300    
3301        /* Process characters until ] is reached. By writing this as a "do" it
3302        means that an initial ] is taken as a data character. At the start of the
3303        loop, c contains the first byte of the character. */
3304    
3305        if (c != 0) do
3306          {
3307          const uschar *oldptr;
3308    
3309    #ifdef SUPPORT_UTF8
3310          if (utf8 && c > 127)
3311            {                           /* Braces are required because the */
3312            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3313            }
3314    
3315          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3316          data and reset the pointer. This is so that very large classes that
3317          contain a zillion UTF-8 characters no longer overwrite the work space
3318          (which is on the stack). */
3319    
3320          if (lengthptr != NULL)
3321            {
3322            *lengthptr += class_utf8data - class_utf8data_base;
3323            class_utf8data = class_utf8data_base;
3324            }
3325    
3326    #endif
3327    
3328          /* Inside \Q...\E everything is literal except \E */
3329    
3330          if (inescq)
3331            {
3332            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3333              {
3334              inescq = FALSE;                   /* Reset literal state */
3335              ptr++;                            /* Skip the 'E' */
3336              continue;                         /* Carry on with next */
3337              }
3338            goto CHECK_RANGE;                   /* Could be range if \E follows */
3339            }
3340    
3341          /* Handle POSIX class names. Perl allows a negation extension of the
3342          form [:^name:]. A square bracket that doesn't match the syntax is
3343          treated as a literal. We also recognize the POSIX constructions
3344          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3345          5.6 and 5.8 do. */
3346    
3347          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3348              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3349               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3350            {
3351            BOOL local_negate = FALSE;
3352            int posix_class, taboffset, tabopt;
3353            register const uschar *cbits = cd->cbits;
3354            uschar pbits[32];
3355    
3356            if (ptr[1] != CHAR_COLON)
3357              {
3358              *errorcodeptr = ERR31;
3359              goto FAILED;
3360              }
3361    
3362            ptr += 2;
3363            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3364              {
3365              local_negate = TRUE;
3366              should_flip_negation = TRUE;  /* Note negative special */
3367              ptr++;
3368              }
3369    
3370            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3371            if (posix_class < 0)
3372              {
3373              *errorcodeptr = ERR30;
3374              goto FAILED;
3375              }
3376    
3377            /* If matching is caseless, upper and lower are converted to
3378            alpha. This relies on the fact that the class table starts with
3379            alpha, lower, upper as the first 3 entries. */
3380    
3381            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3382              posix_class = 0;
3383    
3384            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3385            different escape sequences that use Unicode properties. */
3386    
3387    #ifdef SUPPORT_UCP
3388            if ((options & PCRE_UCP) != 0)
3389              {
3390              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3391              if (posix_substitutes[pc] != NULL)
3392                {
3393                nestptr = tempptr + 1;
3394                ptr = posix_substitutes[pc] - 1;
3395                continue;
3396                }
3397              }
3398    #endif
3399            /* In the non-UCP case, we build the bit map for the POSIX class in a
3400            chunk of local store because we may be adding and subtracting from it,
3401            and we don't want to subtract bits that may be in the main map already.
3402            At the end we or the result into the bit map that is being built. */
3403    
3404            posix_class *= 3;
3405    
3406            /* Copy in the first table (always present) */
3407    
3408            memcpy(pbits, cbits + posix_class_maps[posix_class],
3409              32 * sizeof(uschar));
3410    
3411            /* If there is a second table, add or remove it as required. */
3412    
3413            taboffset = posix_class_maps[posix_class + 1];
3414            tabopt = posix_class_maps[posix_class + 2];
3415    
3416            if (taboffset >= 0)
3417              {
3418              if (tabopt >= 0)
3419                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3420              else
3421                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3422              }
3423    
3424            /* Not see if we need to remove any special characters. An option
3425            value of 1 removes vertical space and 2 removes underscore. */
3426    
3427            if (tabopt < 0) tabopt = -tabopt;
3428            if (tabopt == 1) pbits[1] &= ~0x3c;
3429              else if (tabopt == 2) pbits[11] &= 0x7f;
3430    
3431            /* Add the POSIX table or its complement into the main table that is
3432            being built and we are done. */
3433    
3434            if (local_negate)
3435              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3436            else
3437              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3438    
3439            ptr = tempptr + 1;
3440            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3441            continue;    /* End of POSIX syntax handling */
3442            }
3443    
3444          /* Backslash may introduce a single character, or it may introduce one
3445          of the specials, which just set a flag. The sequence \b is a special
3446          case. Inside a class (and only there) it is treated as backspace. We
3447          assume that other escapes have more than one character in them, so set
3448          class_charcount bigger than one. Unrecognized escapes fall through and
3449          are either treated as literal characters (by default), or are faulted if
3450          PCRE_EXTRA is set. */
3451    
3452          if (c == CHAR_BACKSLASH)
3453            {
3454            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3455            if (*errorcodeptr != 0) goto FAILED;
3456    
3457            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3458            else if (-c == ESC_Q)            /* Handle start of quoted string */
3459              {
3460              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3461                {
3462                ptr += 2; /* avoid empty string */
3463                }
3464              else inescq = TRUE;
3465              continue;
3466              }
3467            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3468    
3469            if (c < 0)
3470              {
3471              register const uschar *cbits = cd->cbits;
3472              class_charcount += 2;     /* Greater than 1 is what matters */
3473    
3474              switch (-c)
3475                {
3476    #ifdef SUPPORT_UCP
3477                case ESC_du:     /* These are the values given for \d etc */
3478                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3479                case ESC_wu:     /* escape sequence with an appropriate \p */
3480                case ESC_WU:     /* or \P to test Unicode properties instead */
3481                case ESC_su:     /* of the default ASCII testing. */
3482                case ESC_SU:
3483                nestptr = ptr;
3484                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3485                class_charcount -= 2;                /* Undo! */
3486                continue;
3487    #endif
3488                case ESC_d:
3489                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3490                continue;
3491    
3492                case ESC_D:
3493                should_flip_negation = TRUE;
3494                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3495                continue;
3496    
3497                case ESC_w:
3498                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
3499                continue;
3500    
3501              case ESC_W:              case ESC_W:
3502                should_flip_negation = TRUE;
3503              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3504              continue;              continue;
3505    
3506                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3507                if it was previously set by something earlier in the character
3508                class. */
3509    
3510              case ESC_s:              case ESC_s:
3511              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3512              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3513                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3514              continue;              continue;
3515    
3516              case ESC_S:              case ESC_S:
3517                should_flip_negation = TRUE;
3518              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3519              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3520              continue;              continue;
3521    
3522                case ESC_h:
3523                SETBIT(classbits, 0x09); /* VT */
3524                SETBIT(classbits, 0x20); /* SPACE */
3525                SETBIT(classbits, 0xa0); /* NSBP */
3526    #ifdef SUPPORT_UTF8
3527                if (utf8)
3528                  {
3529                  class_utf8 = TRUE;
3530                  *class_utf8data++ = XCL_SINGLE;
3531                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3532                  *class_utf8data++ = XCL_SINGLE;
3533                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3534                  *class_utf8data++ = XCL_RANGE;
3535                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3536                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3537                  *class_utf8data++ = XCL_SINGLE;
3538                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3539                  *class_utf8data++ = XCL_SINGLE;
3540                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3541                  *class_utf8data++ = XCL_SINGLE;
3542                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3543                  }
3544    #endif
3545                continue;
3546    
3547                case ESC_H:
3548                for (c = 0; c < 32; c++)
3549                  {
3550                  int x = 0xff;
3551                  switch (c)
3552                    {
3553                    case 0x09/8: x ^= 1 << (0x09%8); break;
3554                    case 0x20/8: x ^= 1 << (0x20%8); break;
3555                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3556                    default: break;
3557                    }
3558                  classbits[c] |= x;
3559                  }
3560    
3561    #ifdef SUPPORT_UTF8
3562                if (utf8)
3563                  {
3564                  class_utf8 = TRUE;
3565                  *class_utf8data++ = XCL_RANGE;
3566                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3567                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3568                  *class_utf8data++ = XCL_RANGE;
3569                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3570                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3571                  *class_utf8data++ = XCL_RANGE;
3572                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3573                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3574                  *class_utf8data++ = XCL_RANGE;
3575                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3576                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3577                  *class_utf8data++ = XCL_RANGE;
3578                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3579                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3580                  *class_utf8data++ = XCL_RANGE;
3581                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3582                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3583                  *class_utf8data++ = XCL_RANGE;
3584                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3585                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3586                  }
3587    #endif
3588                continue;
3589    
3590                case ESC_v:
3591                SETBIT(classbits, 0x0a); /* LF */
3592                SETBIT(classbits, 0x0b); /* VT */
3593                SETBIT(classbits, 0x0c); /* FF */
3594                SETBIT(classbits, 0x0d); /* CR */
3595                SETBIT(classbits, 0x85); /* NEL */
3596    #ifdef SUPPORT_UTF8
3597                if (utf8)
3598                  {
3599                  class_utf8 = TRUE;
3600                  *class_utf8data++ = XCL_RANGE;
3601                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3602                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3603                  }
3604    #endif
3605                continue;
3606    
3607                case ESC_V:
3608                for (c = 0; c < 32; c++)
3609                  {
3610                  int x = 0xff;
3611                  switch (c)
3612                    {
3613                    case 0x0a/8: x ^= 1 << (0x0a%8);
3614                                 x ^= 1 << (0x0b%8);
3615                                 x ^= 1 << (0x0c%8);
3616                                 x ^= 1 << (0x0d%8);
3617                                 break;
3618                    case 0x85/8: x ^= 1 << (0x85%8); break;
3619                    default: break;
3620                    }
3621                  classbits[c] |= x;
3622                  }
3623    
3624    #ifdef SUPPORT_UTF8
3625                if (utf8)
3626                  {
3627                  class_utf8 = TRUE;
3628                  *class_utf8data++ = XCL_RANGE;
3629                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3630                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3631                  *class_utf8data++ = XCL_RANGE;
3632                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3633                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3634                  }
3635    #endif
3636                continue;
3637    
3638  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3639              case ESC_p:              case ESC_p:
3640              case ESC_P:              case ESC_P:
# Line 1980  for (;; ptr++) Line 3649  for (;; ptr++)
3649                *class_utf8data++ = ptype;                *class_utf8data++ = ptype;
3650                *class_utf8data++ = pdata;                *class_utf8data++ = pdata;
3651                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
3652                  continue;
3653                }                }
             continue;  
3654  #endif  #endif
   
3655              /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3656              strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3657              treated as literals. */              treated as literals. */
# Line 1994  for (;; ptr++) Line 3662  for (;; ptr++)
3662                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
3663                goto FAILED;                goto FAILED;
3664                }                }
             c = *ptr;              /* The final character */  
3665              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3666                c = *ptr;              /* Get the final character and fall through */
3667                break;
3668              }              }
3669            }            }
3670    
3671          /* 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
3672          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3673    
3674          }   /* End of backslash handling */          }   /* End of backslash handling */
3675    
3676        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3677        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
3678        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3679          entirely. The code for handling \Q and \E is messy. */
3680    
3681        if (ptr[1] == '-' && ptr[2] != ']')        CHECK_RANGE:
3682          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3683            {
3684            inescq = FALSE;
3685            ptr += 2;
3686            }
3687    
3688          oldptr = ptr;
3689    
3690          /* Remember \r or \n */
3691    
3692          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3693    
3694          /* Check for range */
3695    
3696          if (!inescq && ptr[1] == CHAR_MINUS)
3697          {          {
3698          int d;          int d;
3699          ptr += 2;          ptr += 2;
3700            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3701    
3702            /* If we hit \Q (not followed by \E) at this point, go into escaped
3703            mode. */
3704    
3705            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3706              {
3707              ptr += 2;
3708              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3709                { ptr += 2; continue; }
3710              inescq = TRUE;
3711              break;
3712              }
3713    
3714            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3715              {
3716              ptr = oldptr;
3717              goto LONE_SINGLE_CHARACTER;
3718              }
3719    
3720  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3721          if (utf8)          if (utf8)
# Line 2026  for (;; ptr++) Line 3730  for (;; ptr++)
3730          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3731          in such circumstances. */          in such circumstances. */
3732    
3733          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3734            {            {
3735            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3736            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3737    
3738            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; any other special means the '-' was literal */
           was literal */  
3739    
3740            if (d < 0)            if (d < 0)
3741              {              {
3742              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X'; else  
3743                {                {
3744                ptr = oldptr - 2;                ptr = oldptr;
3745                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3746                }                }
3747              }              }
3748            }            }
3749    
3750          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3751          the pre-pass. Optimize one-character ranges */          one-character ranges */
3752    
3753            if (d < c)
3754              {
3755              *errorcodeptr = ERR8;
3756              goto FAILED;
3757              }
3758    
3759          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3760    
3761            /* Remember \r or \n */
3762    
3763            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3764    
3765          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3766          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3767          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2067  for (;; ptr++) Line 3779  for (;; ptr++)
3779  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3780            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3781              {              {
3782              int occ, ocd;              unsigned int occ, ocd;
3783              int cc = c;              unsigned int cc = c;
3784              int origd = d;              unsigned int origd = d;
3785              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3786                {                {
3787                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3788                      ocd <= (unsigned int)d)
3789                    continue;                          /* Skip embedded ranges */
3790    
3791                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3792                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3793                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3794                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3795                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3796                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3797                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3798                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3799                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3800                  d = ocd;                  d = ocd;
3801                  continue;                  continue;
# Line 2127  for (;; ptr++) Line 3843  for (;; ptr++)
3843          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
3844          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3845    
3846          for (; c <= d; c++)          class_charcount += d - c + 1;
3847            class_lastchar = d;
3848    
3849            /* We can save a bit of time by skipping this in the pre-compile. */
3850    
3851            if (lengthptr == NULL) for (; c <= d; c++)
3852            {            {
3853            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3854            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2135  for (;; ptr++) Line 3856  for (;; ptr++)
3856              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3857              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3858              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3859            }            }
3860    
3861          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2160  for (;; ptr++) Line 3879  for (;; ptr++)
3879  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3880          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3881            {            {
3882            int othercase;            unsigned int othercase;
3883            if ((othercase = _pcre_ucp_othercase(c)) >= 0)            if ((othercase = UCD_OTHERCASE(c)) != c)
3884              {              {
3885              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3886              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2186  for (;; ptr++) Line 3905  for (;; ptr++)
3905          }          }
3906        }        }
3907    
3908      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" far above.
3909      loop. This "while" is the end of the "do" above. */      If we are at the end of an internal nested string, revert to the outer
3910        string. */
3911    
3912        while (((c = *(++ptr)) != 0 ||
3913               (nestptr != NULL &&
3914                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3915               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3916    
3917      while ((c = *(++ptr)) != ']' || inescq);      /* Check for missing terminating ']' */
3918    
3919        if (c == 0)
3920          {
3921          *errorcodeptr = ERR6;
3922          goto FAILED;
3923          }
3924    
3925      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3926      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3927      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3928      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3929      single-bytes only. This is an historical hangover. Maybe one day we can  
3930      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3931        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3932        operate on single-bytes only. This is an historical hangover. Maybe one day
3933        we can tidy these opcodes to handle multi-byte characters.
3934    
3935      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3936      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2206  for (;; ptr++) Line 3940  for (;; ptr++)
3940      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3941    
3942  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3943      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3944            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3945  #else  #else
3946      if (class_charcount == 1)      if (class_charcount == 1)
3947  #endif  #endif
# Line 2252  for (;; ptr++) Line 3984  for (;; ptr++)
3984      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3985    
3986      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3987      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3988      we can omit the bitmap. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
3989        characters > 255 are in the class, so any that were explicitly given as
3990        well can be ignored. If (when there are explicit characters > 255 that must
3991        be listed) there are no characters < 256, we can omit the bitmap in the
3992        actual compiled code. */
3993    
3994  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3995      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3996        {        {
3997        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3998        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3999        code += LINK_SIZE;        code += LINK_SIZE;
4000        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
4001    
4002        /* If the map is required, install it, and move on to the end of        /* If the map is required, move up the extra data to make room for it;
4003        the extra data */        otherwise just move the code pointer to the end of the extra data. */
4004    
4005        if (class_charcount > 0)        if (class_charcount > 0)
4006          {          {
4007          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4008            memmove(code + 32, code, class_utf8data - code);
4009          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4010          code = class_utf8data;          code = class_utf8data + 32;
         }  
   
       /* If the map is not required, slide down the extra data. */  
   
       else  
         {  
         int len = class_utf8data - (code + 33);  
         memmove(code + 1, code + 33, len);  
         code += len + 1;  
4011          }          }
4012          else code = class_utf8data;
4013    
4014        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4015    
# Line 2289  for (;; ptr++) Line 4018  for (;; ptr++)
4018        }        }
4019  #endif  #endif
4020    
4021      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
4022      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4023      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
4024      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4025        negating it if necessary. */
4026    
4027        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4028      if (negate_class)      if (negate_class)
4029        {        {
4030        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4031        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4032        }        }
4033      else      else
4034        {        {
       *code++ = OP_CLASS;  
4035        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4036        }        }
4037      code += 32;      code += 32;
4038      break;      break;
4039    
4040    
4041        /* ===================================================================*/
4042      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4043      has been tested above. */      has been tested above. */
4044    
4045      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4046      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4047      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4048      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4049      goto REPEAT;      goto REPEAT;
4050    
4051      case '*':      case CHAR_ASTERISK:
4052      repeat_min = 0;      repeat_min = 0;
4053      repeat_max = -1;      repeat_max = -1;
4054      goto REPEAT;      goto REPEAT;
4055    
4056      case '+':      case CHAR_PLUS:
4057      repeat_min = 1;      repeat_min = 1;
4058      repeat_max = -1;      repeat_max = -1;
4059      goto REPEAT;      goto REPEAT;
4060    
4061      case '?':      case CHAR_QUESTION_MARK:
4062      repeat_min = 0;      repeat_min = 0;
4063      repeat_max = 1;      repeat_max = 1;
4064    
# Line 2361  for (;; ptr++) Line 4093  for (;; ptr++)
4093      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
4094      repeat type to the non-default. */      repeat type to the non-default. */
4095    
4096      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4097        {        {
4098        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4099        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4100        ptr++;        ptr++;
4101        }        }
4102      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4103        {        {
4104        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4105        ptr++;        ptr++;
4106        }        }
4107      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4108    
     /* If previous was a recursion, we need to wrap it inside brackets so that  
     it can be replicated if necessary. */  
   
     if (*previous == OP_RECURSE)  
       {  
       memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);  
       code += 1 + LINK_SIZE;  
       *previous = OP_BRA;  
       PUT(previous, 1, code - previous);  
       *code = OP_KET;  
       PUT(code, 1, code - previous);  
       code += 1 + LINK_SIZE;  
       }  
   
4109      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4110      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4111      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2421  for (;; ptr++) Line 4139  for (;; ptr++)
4139          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
4140          }          }
4141    
4142          /* If the repetition is unlimited, it pays to see if the next thing on
4143          the line is something that cannot possibly match this character. If so,
4144          automatically possessifying this item gains some performance in the case
4145          where the match fails. */
4146    
4147          if (!possessive_quantifier &&
4148              repeat_max < 0 &&
4149              check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4150            {
4151            repeat_type = 0;    /* Force greedy */
4152            possessive_quantifier = TRUE;
4153            }
4154    
4155        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
4156        }        }
4157    
4158      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4159      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4160      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4161      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
4162        currently used only for single-byte chars. */
4163    
4164      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
4165        {        {
4166        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
4167        c = previous[1];        c = previous[1];
4168          if (!possessive_quantifier &&
4169              repeat_max < 0 &&
4170              check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4171            {
4172            repeat_type = 0;    /* Force greedy */
4173            possessive_quantifier = TRUE;
4174            }
4175        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
4176        }        }
4177    
# Line 2450  for (;; ptr++) Line 4189  for (;; ptr++)
4189        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4190        c = *previous;        c = *previous;
4191    
4192          if (!possessive_quantifier &&
4193              repeat_max < 0 &&
4194              check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4195            {
4196            repeat_type = 0;    /* Force greedy */
4197            possessive_quantifier = TRUE;
4198            }
4199    
4200        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
4201        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
4202          {          {
# Line 2466  for (;; ptr++) Line 4213  for (;; ptr++)
4213