/[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 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 504 by ph10, Mon Mar 8 08:57:04 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) 1997-2005 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 PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #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 BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    
96  /* 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 63  are simple data values; negative values Line 98  are simple data values; negative values
98  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
99  is invalid. */  is invalid. */
100    
101  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
102    
103    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104    in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
108       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
109     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
110       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
111  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
112  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
113     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
116       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
117         -ESC_D,                  -ESC_E,
118         0,                       -ESC_G,
119         -ESC_H,                  0,
120         0,                       -ESC_K,
121         0,                       0,
122         0,                       0,
123         -ESC_P,                  -ESC_Q,
124         -ESC_R,                  -ESC_S,
125         0,                       0,
126         -ESC_V,                  -ESC_W,
127         -ESC_X,                  0,
128         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131         CHAR_GRAVE_ACCENT,       7,
132         -ESC_b,                  0,
133         -ESC_d,                  ESC_e,
134         ESC_f,                   0,
135         -ESC_h,                  0,
136         0,                       -ESC_k,
137         0,                       0,
138         ESC_n,                   0,
139         -ESC_p,                  0,
140         ESC_r,                   -ESC_s,
141         ESC_tee,                 0,
142         -ESC_v,                  -ESC_w,
143         0,                       0,
144         -ESC_z
145  };  };
146    
147  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
148    
149    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
151  static const short int escapes[] = {  static const short int escapes[] = {
152  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
153  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 157  static const short int escapes[] = {
157  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
158  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
159  /*  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,
160  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
161  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
162  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
163  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
164  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
165  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
166  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
167  /*  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,
168  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
169  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
170  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
171  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
172  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
173  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
174  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 176  static const short int escapes[] = {
176  #endif  #endif
177    
178    
179  /* 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
180  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
181  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
182    string is built from string macros so that it works in UTF-8 mode on EBCDIC
183  static const char *const posix_names[] = {  platforms. */
184    "alpha", "lower", "upper",  
185    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
186    "print", "punct", "space", "word",  "xdigit" };    int   len;
187      int   op;
188    } verbitem;
189    
190    static const char verbnames[] =
191      STRING_ACCEPT0
192      STRING_COMMIT0
193      STRING_F0
194      STRING_FAIL0
195      STRING_PRUNE0
196      STRING_SKIP0
197      STRING_THEN;
198    
199    static const verbitem verbs[] = {
200      { 6, OP_ACCEPT },
201      { 6, OP_COMMIT },
202      { 1, OP_FAIL },
203      { 4, OP_FAIL },
204      { 5, OP_PRUNE },
205      { 4, OP_SKIP  },
206      { 4, OP_THEN  }
207    };
208    
209    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
210    
211    
212    /* Tables of names of POSIX character classes and their lengths. The names are
213    now all in a single string, to reduce the number of relocations when a shared
214    library is dynamically loaded. The list of lengths is terminated by a zero
215    length entry. The first three must be alpha, lower, upper, as this is assumed
216    for handling case independence. */
217    
218    static const char posix_names[] =
219      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222      STRING_word0  STRING_xdigit;
223    
224  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
225    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 };
226    
227  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
228  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
229  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
230    characters are removed, and for [:alpha:] and [:alnum:] the underscore
231    character is removed. The triples in the table consist of the base map offset,
232    second map offset or -1 if no second map, and a non-negative value for map
233    addition or a negative value for map subtraction (if there are two maps). The
234    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
235    remove vertical space characters, 2 => remove underscore. */
236    
237  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
238    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
239    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
240    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
241    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
242    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
243    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
244    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
245    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
246    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
247    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
248    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
249    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
250    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
251    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
252  };  };
253    
254    
255  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
256  are passed to the outside world. */  #define XSTRING(s) STRING(s)
257    
258  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
259    "no error",  are passed to the outside world. Do not ever re-use any error number, because
260    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
261    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
262    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
263    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
264    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
265    simply count through to the one we want - this isn't a performance issue
266    because these strings are used only when there is a compilation error.
267    
268    Each substring ends with \0 to insert a null character. This includes the final
269    substring, so that the whole string ends with \0\0, which can be detected when
270    counting through. */
271    
272    static const char error_texts[] =
273      "no error\0"
274      "\\ at end of pattern\0"
275      "\\c at end of pattern\0"
276      "unrecognized character follows \\\0"
277      "numbers out of order in {} quantifier\0"
278    /* 5 */    /* 5 */
279    "number too big in {} quantifier",    "number too big in {} quantifier\0"
280    "missing terminating ] for character class",    "missing terminating ] for character class\0"
281    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
282    "range out of order in character class",    "range out of order in character class\0"
283    "nothing to repeat",    "nothing to repeat\0"
284    /* 10 */    /* 10 */
285    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
286    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
287    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
288    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
289    "missing )",    "missing )\0"
290    /* 15 */    /* 15 */
291    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
292    "erroffset passed as NULL",    "erroffset passed as NULL\0"
293    "unknown option bit(s) set",    "unknown option bit(s) set\0"
294    "missing ) after comment",    "missing ) after comment\0"
295    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
296    /* 20 */    /* 20 */
297    "regular expression too large",    "regular expression is too large\0"
298    "failed to get memory",    "failed to get memory\0"
299    "unmatched parentheses",    "unmatched parentheses\0"
300    "internal error: code overflow",    "internal error: code overflow\0"
301    "unrecognized character after (?<",    "unrecognized character after (?<\0"
302    /* 25 */    /* 25 */
303    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
304    "malformed number after (?(",    "malformed number or name after (?(\0"
305    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
306    "assertion expected after (?(",    "assertion expected after (?(\0"
307    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
308    /* 30 */    /* 30 */
309    "unknown POSIX class name",    "unknown POSIX class name\0"
310    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
311    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
312    "spare error",    "spare error\0"  /** DEAD **/
313    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
314    /* 35 */    /* 35 */
315    "invalid condition (?(0)",    "invalid condition (?(0)\0"
316    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
317    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
318    "number after (?C is > 255",    "number after (?C is > 255\0"
319    "closing ) for (?C expected",    "closing ) for (?C expected\0"
320    /* 40 */    /* 40 */
321    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
322    "unrecognized character after (?P",    "unrecognized character after (?P\0"
323    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
324    "two named groups have the same name",    "two named subpatterns have the same name\0"
325    "invalid UTF-8 string",    "invalid UTF-8 string\0"
326    /* 45 */    /* 45 */
327    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
328    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
329    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
330  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
331      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
332      /* 50 */
333      "repeated subpattern is too long\0"    /** DEAD **/
334      "octal value is greater than \\377 (not in UTF-8 mode)\0"
335      "internal error: overran compiling workspace\0"
336      "internal error: previously-checked referenced subpattern not found\0"
337      "DEFINE group contains more than one branch\0"
338      /* 55 */
339      "repeating a DEFINE group is not allowed\0"
340      "inconsistent NEWLINE options\0"
341      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
342      "a numbered reference must not be zero\0"
343      "(*VERB) with an argument is not supported\0"
344      /* 60 */
345      "(*VERB) not recognized\0"
346      "number is too big\0"
347      "subpattern name expected\0"
348      "digit expected after (?+\0"
349      "] is an invalid data character in JavaScript compatibility mode\0"
350      /* 65 */
351      "different names for subpatterns of the same number are not allowed\0";
352    
353  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
354  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 220  For convenience, we use the same bit def Line 366  For convenience, we use the same bit def
366    
367  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
368    
369  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
370    
371    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
372    UTF-8 mode. */
373    
374  static const unsigned char digitab[] =  static const unsigned char digitab[] =
375    {    {
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 406  static const unsigned char digitab[] =
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
408    
409  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
410    
411    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
412    
413  static const unsigned char digitab[] =  static const unsigned char digitab[] =
414    {    {
415    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 423  static const unsigned char digitab[] =
423    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
424    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
425    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
426    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
427    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
428    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
429    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 457  static const unsigned char ebcdic_charta
457    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
458    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
459    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
460    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
461    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
462    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
463    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 484  static const unsigned char ebcdic_charta
484  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
485    
486  static BOOL  static BOOL
487    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
488      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
489    
490    
491    
492    /*************************************************
493    *            Find an error text                  *
494    *************************************************/
495    
496    /* The error texts are now all in one long string, to save on relocations. As
497    some of the text is of unknown length, we can't use a table of offsets.
498    Instead, just count through the strings. This is not a performance issue
499    because it happens only when there has been a compilation error.
500    
501    Argument:   the error number
502    Returns:    pointer to the error string
503    */
504    
505    static const char *
506    find_error_text(int n)
507    {
508    const char *s = error_texts;
509    for (; n > 0; n--)
510      {
511      while (*s++ != 0) {};
512      if (*s == 0) return "Error text not found (please report)";
513      }
514    return s;
515    }
516    
517    
518  /*************************************************  /*************************************************
# Line 342  static BOOL Line 521  static BOOL
521    
522  /* 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
523  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
524  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
525  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
526  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,
527    ptr is pointing at the \. On exit, it is on the final character of the escape
528    sequence.
529    
530  Arguments:  Arguments:
531    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 536  Arguments:
536    
537  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
538                   negative => a special escape sequence                   negative => a special escape sequence
539                   on error, errorptr is set                   on error, errorcodeptr is set
540  */  */
541    
542  static int  static int
543  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
544    int options, BOOL isclass)    int options, BOOL isclass)
545  {  {
546  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
547    const uschar *ptr = *ptrptr + 1;
548  int c, i;  int c, i;
549    
550    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
551    ptr--;                            /* Set pointer back to the last byte */
552    
553  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
554    
 c = *(++ptr);  
555  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
556    
557  /* 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
558  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.
559  Otherwise further processing may be required. */  Otherwise further processing may be required. */
560    
561  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
562  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
563  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
564    
565  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
566  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
567  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
568  #endif  #endif
569    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 572  else if ((i = escapes[c - 0x48]) != 0)
572  else  else
573    {    {
574    const uschar *oldptr;    const uschar *oldptr;
575      BOOL braced, negated;
576    
577    switch (c)    switch (c)
578      {      {
579      /* 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
580      error. */      error. */
581    
582      case 'l':      case CHAR_l:
583      case 'L':      case CHAR_L:
584      case 'N':      case CHAR_N:
585      case 'u':      case CHAR_u:
586      case 'U':      case CHAR_U:
587      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
588      break;      break;
589    
590        /* \g must be followed by one of a number of specific things:
591    
592        (1) A number, either plain or braced. If positive, it is an absolute
593        backreference. If negative, it is a relative backreference. This is a Perl
594        5.10 feature.
595    
596        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
597        is part of Perl's movement towards a unified syntax for back references. As
598        this is synonymous with \k{name}, we fudge it up by pretending it really
599        was \k.
600    
601        (3) For Oniguruma compatibility we also support \g followed by a name or a
602        number either in angle brackets or in single quotes. However, these are
603        (possibly recursive) subroutine calls, _not_ backreferences. Just return
604        the -ESC_g code (cf \k). */
605    
606        case CHAR_g:
607        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
608          {
609          c = -ESC_g;
610          break;
611          }
612    
613        /* Handle the Perl-compatible cases */
614    
615        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
616          {
617          const uschar *p;
618          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
619            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
620          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
621            {
622            c = -ESC_k;
623            break;
624            }
625          braced = TRUE;
626          ptr++;
627          }
628        else braced = FALSE;
629    
630        if (ptr[1] == CHAR_MINUS)
631          {
632          negated = TRUE;
633          ptr++;
634          }
635        else negated = FALSE;
636    
637        c = 0;
638        while ((digitab[ptr[1]] & ctype_digit) != 0)
639          c = c * 10 + *(++ptr) - CHAR_0;
640    
641        if (c < 0)   /* Integer overflow */
642          {
643          *errorcodeptr = ERR61;
644          break;
645          }
646    
647        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
648          {
649          *errorcodeptr = ERR57;
650          break;
651          }
652    
653        if (c == 0)
654          {
655          *errorcodeptr = ERR58;
656          break;
657          }
658    
659        if (negated)
660          {
661          if (c > bracount)
662            {
663            *errorcodeptr = ERR15;
664            break;
665            }
666          c = bracount - (c - 1);
667          }
668    
669        c = -(ESC_REF + c);
670        break;
671    
672      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
673      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
674      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 681  else
681      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
682      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
683    
684      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:
685      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
686    
687      if (!isclass)      if (!isclass)
688        {        {
689        oldptr = ptr;        oldptr = ptr;
690        c -= '0';        c -= CHAR_0;
691        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
692          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
693          if (c < 0)    /* Integer overflow */
694            {
695            *errorcodeptr = ERR61;
696            break;
697            }
698        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
699          {          {
700          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 707  else
707      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.
708      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
709    
710      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
711        {        {
712        ptr--;        ptr--;
713        c = 0;        c = 0;
# Line 442  else Line 715  else
715        }        }
716    
717      /* \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
718      larger first octal digit. */      larger first octal digit. The original code used just to take the least
719        significant 8 bits of octal numbers (I think this is what early Perls used
720      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
721      c -= '0';      than 3 octal digits. */
722      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
723          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
724      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
725        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
726            c = c * 8 + *(++ptr) - CHAR_0;
727        if (!utf8 && c > 255) *errorcodeptr = ERR51;
728      break;      break;
729    
730      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
731      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
732        treated as a data character. */
733    
734      case 'x':      case CHAR_x:
735  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
736        {        {
737        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
738        register int count = 0;        int count = 0;
739    
740        c = 0;        c = 0;
741        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
742          {          {
743          int cc = *pt++;          register int cc = *pt++;
744            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
745          count++;          count++;
746  #if !EBCDIC    /* ASCII coding */  
747          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
748          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
749  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
750          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
751          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
752            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
753  #endif  #endif
754          }          }
755        if (*pt == '}')  
756          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
757          {          {
758          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
759          ptr = pt;          ptr = pt;
760          break;          break;
761          }          }
762    
763        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
764        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
765        }        }
 #endif  
766    
767      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
768    
769      c = 0;      c = 0;
770      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
771        {        {
772        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
773        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
774  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
775        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
776        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
777  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
778        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
779        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
780  #endif  #endif
781        }        }
782      break;      break;
783    
784      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
785        This coding is ASCII-specific, but then the whole concept of \cx is
786        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
787    
788      case 'c':      case CHAR_c:
789      c = *(++ptr);      c = *(++ptr);
790      if (c == 0)      if (c == 0)
791        {        {
792        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
793        return 0;        break;
794        }        }
795    
796      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
797      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;  
798      c ^= 0x40;      c ^= 0x40;
799  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
800      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
801      c ^= 0xC0;      c ^= 0xC0;
802  #endif  #endif
803      break;      break;
804    
805      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
806      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
807      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
808      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
809      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
810    
811      default:      default:
812      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 838  escape sequence.
838  Argument:  Argument:
839    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
840    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
841      dptr           points to an int that is set to the detailed property value
842    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
843    
844  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
845  */  */
846    
847  static int  static int
848  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
849  {  {
850  int c, i, bot, top;  int c, i, bot, top;
851  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
852  char name[4];  char name[32];
853    
854  c = *(++ptr);  c = *(++ptr);
855  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
856    
857  *negptr = FALSE;  *negptr = FALSE;
858    
859  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
860  preceded by ^ for negation. */  negation. */
861    
862  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
863    {    {
864    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
865      {      {
866      *negptr = TRUE;      *negptr = TRUE;
867      ptr++;      ptr++;
868      }      }
869    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
870      {      {
871      c = *(++ptr);      c = *(++ptr);
872      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
873      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
874      name[i] = c;      name[i] = c;
875      }      }
876    if (c !='}')   /* Try to distinguish error cases */    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
877    name[i] = 0;    name[i] = 0;
878    }    }
879    
# Line 619  top = _pcre_utt_size; Line 894  top = _pcre_utt_size;
894    
895  while (bot < top)  while (bot < top)
896    {    {
897    i = (bot + top)/2;    i = (bot + top) >> 1;
898    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
899    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
900        {
901        *dptr = _pcre_utt[i].value;
902        return _pcre_utt[i].type;
903        }
904    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
905    }    }
906    
 UNKNOWN_RETURN:  
907  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
908  *ptrptr = ptr;  *ptrptr = ptr;
909  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 938  is_counted_repeat(const uschar *p)
938  {  {
939  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
940  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
941  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
944  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
945    
946  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
947  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
948    
949  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
950  }  }
951    
952    
# Line 698  read_repeat_counts(const uschar *p, int Line 976  read_repeat_counts(const uschar *p, int
976  int min = 0;  int min = 0;
977  int max = -1;  int max = -1;
978    
979  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
980    an integer overflow. */
981    
982    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
983    if (min < 0 || min > 65535)
984      {
985      *errorcodeptr = ERR5;
986      return p;
987      }
988    
989    /* Read the maximum value if there is one, and again do a paranoid on its size.
990    Also, max must not be less than min. */
991    
992  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
993    {    {
994    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
995      {      {
996      max = 0;      max = 0;
997      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
998        if (max < 0 || max > 65535)
999          {
1000          *errorcodeptr = ERR5;
1001          return p;
1002          }
1003      if (max < min)      if (max < min)
1004        {        {
1005        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1008  if (*p == '}') max = min; else
1008      }      }
1009    }    }
1010    
1011  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1012  pointer to the terminating '}'. */  '}'. */
1013    
1014  if (min > 65535 || max > 65535)  *minp = min;
1015    *errorcodeptr = ERR5;  *maxp = max;
1016  else  return p;
1017    }
1018    
1019    
1020    
1021    /*************************************************
1022    *  Subroutine for finding forward reference      *
1023    *************************************************/
1024    
1025    /* This recursive function is called only from find_parens() below. The
1026    top-level call starts at the beginning of the pattern. All other calls must
1027    start at a parenthesis. It scans along a pattern's text looking for capturing
1028    subpatterns, and counting them. If it finds a named pattern that matches the
1029    name it is given, it returns its number. Alternatively, if the name is NULL, it
1030    returns when it reaches a given numbered subpattern. We know that if (?P< is
1031    encountered, the name will be terminated by '>' because that is checked in the
1032    first pass. Recursion is used to keep track of subpatterns that reset the
1033    capturing group numbers - the (?| feature.
1034    
1035    Arguments:
1036      ptrptr       address of the current character pointer (updated)
1037      cd           compile background data
1038      name         name to seek, or NULL if seeking a numbered subpattern
1039      lorn         name length, or subpattern number if name is NULL
1040      xmode        TRUE if we are in /x mode
1041      count        pointer to the current capturing subpattern number (updated)
1042    
1043    Returns:       the number of the named subpattern, or -1 if not found
1044    */
1045    
1046    static int
1047    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1048      BOOL xmode, int *count)
1049    {
1050    uschar *ptr = *ptrptr;
1051    int start_count = *count;
1052    int hwm_count = start_count;
1053    BOOL dup_parens = FALSE;
1054    
1055    /* If the first character is a parenthesis, check on the type of group we are
1056    dealing with. The very first call may not start with a parenthesis. */
1057    
1058    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1059    {    {
1060    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1061    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1062        {
1063        ptr += 3;
1064        dup_parens = TRUE;
1065        }
1066    
1067      /* Handle a normal, unnamed capturing parenthesis */
1068    
1069      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1070        {
1071        *count += 1;
1072        if (name == NULL && *count == lorn) return *count;
1073        ptr++;
1074        }
1075    
1076      /* Handle a condition. If it is an assertion, just carry on so that it
1077      is processed as normal. If not, skip to the closing parenthesis of the
1078      condition (there can't be any nested parens. */
1079    
1080      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1081        {
1082        ptr += 2;
1083        if (ptr[1] != CHAR_QUESTION_MARK)
1084          {
1085          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1086          if (*ptr != 0) ptr++;
1087          }
1088        }
1089    
1090      /* We have either (? or (* and not a condition */
1091    
1092      else
1093        {
1094        ptr += 2;
1095        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1096    
1097        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1098    
1099        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1100            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1101          {
1102          int term;
1103          const uschar *thisname;
1104          *count += 1;
1105          if (name == NULL && *count == lorn) return *count;
1106          term = *ptr++;
1107          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1108          thisname = ptr;
1109          while (*ptr != term) ptr++;
1110          if (name != NULL && lorn == ptr - thisname &&
1111              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1112            return *count;
1113          term++;
1114          }
1115        }
1116    }    }
1117  return p;  
1118    /* Past any initial parenthesis handling, scan for parentheses or vertical
1119    bars. */
1120    
1121    for (; *ptr != 0; ptr++)
1122      {
1123      /* Skip over backslashed characters and also entire \Q...\E */
1124    
1125      if (*ptr == CHAR_BACKSLASH)
1126        {
1127        if (*(++ptr) == 0) goto FAIL_EXIT;
1128        if (*ptr == CHAR_Q) for (;;)
1129          {
1130          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1131          if (*ptr == 0) goto FAIL_EXIT;
1132          if (*(++ptr) == CHAR_E) break;
1133          }
1134        continue;
1135        }
1136    
1137      /* Skip over character classes; this logic must be similar to the way they
1138      are handled for real. If the first character is '^', skip it. Also, if the
1139      first few characters (either before or after ^) are \Q\E or \E we skip them
1140      too. This makes for compatibility with Perl. Note the use of STR macros to
1141      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1142    
1143      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1144        {
1145        BOOL negate_class = FALSE;
1146        for (;;)
1147          {
1148          if (ptr[1] == CHAR_BACKSLASH)
1149            {
1150            if (ptr[2] == CHAR_E)
1151              ptr+= 2;
1152            else if (strncmp((const char *)ptr+2,
1153                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1154              ptr += 4;
1155            else
1156              break;
1157            }
1158          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1159            {
1160            negate_class = TRUE;
1161            ptr++;
1162            }
1163          else break;
1164          }
1165    
1166        /* If the next character is ']', it is a data character that must be
1167        skipped, except in JavaScript compatibility mode. */
1168    
1169        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171          ptr++;
1172    
1173        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1174          {
1175          if (*ptr == 0) return -1;
1176          if (*ptr == CHAR_BACKSLASH)
1177            {
1178            if (*(++ptr) == 0) goto FAIL_EXIT;
1179            if (*ptr == CHAR_Q) for (;;)
1180              {
1181              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1182              if (*ptr == 0) goto FAIL_EXIT;
1183              if (*(++ptr) == CHAR_E) break;
1184              }
1185            continue;
1186            }
1187          }
1188        continue;
1189        }
1190    
1191      /* Skip comments in /x mode */
1192    
1193      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1194        {
1195        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1196        if (*ptr == 0) goto FAIL_EXIT;
1197        continue;
1198        }
1199    
1200      /* Check for the special metacharacters */
1201    
1202      if (*ptr == CHAR_LEFT_PARENTHESIS)
1203        {
1204        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1205        if (rc > 0) return rc;
1206        if (*ptr == 0) goto FAIL_EXIT;
1207        }
1208    
1209      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210        {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214        }
1215    
1216      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217        {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    FAIL_EXIT:
1224    *ptrptr = ptr;
1225    return -1;
1226    }
1227    
1228    
1229    
1230    
1231    /*************************************************
1232    *       Find forward referenced subpattern       *
1233    *************************************************/
1234    
1235    /* This function scans along a pattern's text looking for capturing
1236    subpatterns, and counting them. If it finds a named pattern that matches the
1237    name it is given, it returns its number. Alternatively, if the name is NULL, it
1238    returns when it reaches a given numbered subpattern. This is used for forward
1239    references to subpatterns. We used to be able to start this scan from the
1240    current compiling point, using the current count value from cd->bracount, and
1241    do it all in a single loop, but the addition of the possibility of duplicate
1242    subpattern numbers means that we have to scan from the very start, in order to
1243    take account of such duplicates, and to use a recursive function to keep track
1244    of the different types of group.
1245    
1246    Arguments:
1247      cd           compile background data
1248      name         name to seek, or NULL if seeking a numbered subpattern
1249      lorn         name length, or subpattern number if name is NULL
1250      xmode        TRUE if we are in /x mode
1251    
1252    Returns:       the number of the found subpattern, or -1 if not found
1253    */
1254    
1255    static int
1256    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    {
1258    uschar *ptr = (uschar *)cd->start_pattern;
1259    int count = 0;
1260    int rc;
1261    
1262    /* If the pattern does not start with an opening parenthesis, the first call
1263    to find_parens_sub() will scan right to the end (if necessary). However, if it
1264    does start with a parenthesis, find_parens_sub() will return when it hits the
1265    matching closing parens. That is why we have to have a loop. */
1266    
1267    for (;;)
1268      {
1269      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270      if (rc > 0 || *ptr++ == 0) break;
1271      }
1272    
1273    return rc;
1274  }  }
1275    
1276    
1277    
1278    
1279  /*************************************************  /*************************************************
1280  *      Find first significant op code            *  *      Find first significant op code            *
1281  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1325  for (;;)
1325    
1326      case OP_CALLOUT:      case OP_CALLOUT:
1327      case OP_CREF:      case OP_CREF:
1328      case OP_BRANUMBER:      case OP_NCREF:
1329        case OP_RREF:
1330        case OP_NRREF:
1331        case OP_DEF:
1332      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1333      break;      break;
1334    
# Line 793  for (;;) Line 1343  for (;;)
1343    
1344    
1345  /*************************************************  /*************************************************
1346  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1347  *************************************************/  *************************************************/
1348    
1349  /* 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,
1350  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.
1351  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
1352    temporarily terminated with OP_END when this function is called.
1353    
1354    This function is called when a backward assertion is encountered, so that if it
1355    fails, the error message can point to the correct place in the pattern.
1356    However, we cannot do this when the assertion contains subroutine calls,
1357    because they can be forward references. We solve this by remembering this case
1358    and doing the check at the end; a flag specifies which mode we are running in.
1359    
1360  Arguments:  Arguments:
1361    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1362    options  the compiling options    options  the compiling options
1363      atend    TRUE if called when the pattern is complete
1364      cd       the "compile data" structure
1365    
1366  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1367                 or -1 if there is no fixed length,
1368               or -2 if \C was encountered               or -2 if \C was encountered
1369                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1370  */  */
1371    
1372  static int  static int
1373  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1374  {  {
1375  int length = -1;  int length = -1;
1376    
# Line 822  branch, check the length against that of Line 1383  branch, check the length against that of
1383  for (;;)  for (;;)
1384    {    {
1385    int d;    int d;
1386      uschar *ce, *cs;
1387    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1388    switch (op)    switch (op)
1389      {      {
1390        case OP_CBRA:
1391      case OP_BRA:      case OP_BRA:
1392      case OP_ONCE:      case OP_ONCE:
1393      case OP_COND:      case OP_COND:
1394      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1395      if (d < 0) return d;      if (d < 0) return d;
1396      branchlength += d;      branchlength += d;
1397      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 853  for (;;) Line 1414  for (;;)
1414      branchlength = 0;      branchlength = 0;
1415      break;      break;
1416    
1417        /* A true recursion implies not fixed length, but a subroutine call may
1418        be OK. If the subroutine is a forward reference, we can't deal with
1419        it until the end of the pattern, so return -3. */
1420    
1421        case OP_RECURSE:
1422        if (!atend) return -3;
1423        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1424        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1425        if (cc > cs && cc < ce) return -1;                /* Recursion */
1426        d = find_fixedlength(cs + 2, options, atend, cd);
1427        if (d < 0) return d;
1428        branchlength += d;
1429        cc += 1 + LINK_SIZE;
1430        break;
1431    
1432      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 865  for (;;) Line 1441  for (;;)
1441      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1442    
1443      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1444      case OP_CREF:      case OP_CREF:
1445        case OP_NCREF:
1446        case OP_RREF:
1447        case OP_NRREF:
1448        case OP_DEF:
1449      case OP_OPT:      case OP_OPT:
1450      case OP_CALLOUT:      case OP_CALLOUT:
1451      case OP_SOD:      case OP_SOD:
1452      case OP_SOM:      case OP_SOM:
1453        case OP_SET_SOM:
1454      case OP_EOD:      case OP_EOD:
1455      case OP_EODN:      case OP_EODN:
1456      case OP_CIRC:      case OP_CIRC:
# Line 884  for (;;) Line 1464  for (;;)
1464    
1465      case OP_CHAR:      case OP_CHAR:
1466      case OP_CHARNC:      case OP_CHARNC:
1467        case OP_NOT:
1468      branchlength++;      branchlength++;
1469      cc += 2;      cc += 2;
1470  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1471      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1472        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1473  #endif  #endif
1474      break;      break;
1475    
# Line 901  for (;;) Line 1480  for (;;)
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481      cc += 4;      cc += 4;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
1488      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1489      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1490        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1491      cc += 4;      cc += 4;
1492      break;      break;
1493    
# Line 917  for (;;) Line 1495  for (;;)
1495    
1496      case OP_PROP:      case OP_PROP:
1497      case OP_NOTPROP:      case OP_NOTPROP:
1498      cc++;      cc += 2;
1499      /* Fall through */      /* Fall through */
1500    
1501      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1505  for (;;)
1505      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1506      case OP_WORDCHAR:      case OP_WORDCHAR:
1507      case OP_ANY:      case OP_ANY:
1508        case OP_ALLANY:
1509      branchlength++;      branchlength++;
1510      cc++;      cc++;
1511      break;      break;
# Line 981  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1564  *************************************************/  *************************************************/
1565    
1566  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1567  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1568    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1569    so that it can be called from pcre_study() when finding the minimum matching
1570    length.
1571    
1572  Arguments:  Arguments:
1573    code        points to start of expression    code        points to start of expression
1574    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1575    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1576    
1577  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
1578  */  */
1579    
1580  static const uschar *  const uschar *
1581  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1582  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1583  for (;;)  for (;;)
1584    {    {
1585    register int c = *code;    register int c = *code;
1586    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
   else if (c > OP_BRA)  
     {  
     int n = c - OP_BRA;  
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
     if (n == number) return (uschar *)code;  
     code += _pcre_OP_lengths[OP_BRA];  
     }  
   else  
     {  
     code += _pcre_OP_lengths[c];  
1587    
1588  #ifdef SUPPORT_UTF8    /* XCLASS is used for classes that cannot be represented just by a bit
1589      map. This includes negated single high-valued characters. The length in
1590      the table is zero; the actual length is stored in the compiled code. */
1591    
1592      /* In UTF-8 mode, opcodes that are followed by a character may be followed    if (c == OP_XCLASS) code += GET(code, 1);
     by a multi-byte character. The length in the table is a minimum, so we have  
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
1593    
1594      /* Handle recursion */
1595    
1596      else if (c == OP_REVERSE)
1597        {
1598        if (number < 0) return (uschar *)code;
1599        code += _pcre_OP_lengths[c];
1600        }
1601    
1602      /* Handle capturing bracket */
1603    
1604      else if (c == OP_CBRA)
1605        {
1606        int n = GET2(code, 1+LINK_SIZE);
1607        if (n == number) return (uschar *)code;
1608        code += _pcre_OP_lengths[c];
1609        }
1610    
1611      /* Otherwise, we can get the item's length from the table, except that for
1612      repeated character types, we have to test for \p and \P, which have an extra
1613      two bytes of parameters. */
1614    
1615      else
1616        {
1617        switch(c)
1618          {
1619          case OP_TYPESTAR:
1620          case OP_TYPEMINSTAR:
1621          case OP_TYPEPLUS:
1622          case OP_TYPEMINPLUS:
1623          case OP_TYPEQUERY:
1624          case OP_TYPEMINQUERY:
1625          case OP_TYPEPOSSTAR:
1626          case OP_TYPEPOSPLUS:
1627          case OP_TYPEPOSQUERY:
1628          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1629          break;
1630    
1631          case OP_TYPEUPTO:
1632          case OP_TYPEMINUPTO:
1633          case OP_TYPEEXACT:
1634          case OP_TYPEPOSUPTO:
1635          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1636          break;
1637          }
1638    
1639        /* Add in the fixed length from the table */
1640    
1641        code += _pcre_OP_lengths[c];
1642    
1643      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1644      a multi-byte character. The length in the table is a minimum, so we have to
1645      arrange to skip the extra bytes. */
1646    
1647    #ifdef SUPPORT_UTF8
1648      if (utf8) switch(c)      if (utf8) switch(c)
1649        {        {
1650        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1652  for (;;)
1652        case OP_EXACT:        case OP_EXACT:
1653        case OP_UPTO:        case OP_UPTO:
1654        case OP_MINUPTO:        case OP_MINUPTO:
1655          case OP_POSUPTO:
1656        case OP_STAR:        case OP_STAR:
1657        case OP_MINSTAR:        case OP_MINSTAR:
1658          case OP_POSSTAR:
1659        case OP_PLUS:        case OP_PLUS:
1660        case OP_MINPLUS:        case OP_MINPLUS:
1661          case OP_POSPLUS:
1662        case OP_QUERY:        case OP_QUERY:
1663        case OP_MINQUERY:        case OP_MINQUERY:
1664        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1665        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;  
1666        break;        break;
1667        }        }
1668    #else
1669        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1670  #endif  #endif
1671      }      }
1672    }    }
# Line 1072  Returns:      pointer to the opcode for Line 1691  Returns:      pointer to the opcode for
1691  static const uschar *  static const uschar *
1692  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1693  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1694  for (;;)  for (;;)
1695    {    {
1696    register int c = *code;    register int c = *code;
1697    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1698    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1699    else if (c > OP_BRA)  
1700      {    /* XCLASS is used for classes that cannot be represented just by a bit
1701      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1702      }    the table is zero; the actual length is stored in the compiled code. */
1703    
1704      if (c == OP_XCLASS) code += GET(code, 1);
1705    
1706      /* Otherwise, we can get the item's length from the table, except that for
1707      repeated character types, we have to test for \p and \P, which have an extra
1708      two bytes of parameters. */
1709    
1710    else    else
1711      {      {
1712      code += _pcre_OP_lengths[c];      switch(c)
1713          {
1714          case OP_TYPESTAR:
1715          case OP_TYPEMINSTAR:
1716          case OP_TYPEPLUS:
1717          case OP_TYPEMINPLUS:
1718          case OP_TYPEQUERY:
1719          case OP_TYPEMINQUERY:
1720          case OP_TYPEPOSSTAR:
1721          case OP_TYPEPOSPLUS:
1722          case OP_TYPEPOSQUERY:
1723          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1724          break;
1725    
1726  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1727          case OP_TYPEUPTO:
1728          case OP_TYPEMINUPTO:
1729          case OP_TYPEEXACT:
1730          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1731          break;
1732          }
1733    
1734        /* Add in the fixed length from the table */
1735    
1736        code += _pcre_OP_lengths[c];
1737    
1738      /* 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
1739      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
1740      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. */  
1741    
1742    #ifdef SUPPORT_UTF8
1743      if (utf8) switch(c)      if (utf8) switch(c)
1744        {        {
1745        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1747  for (;;)
1747        case OP_EXACT:        case OP_EXACT:
1748        case OP_UPTO:        case OP_UPTO:
1749        case OP_MINUPTO:        case OP_MINUPTO:
1750          case OP_POSUPTO:
1751        case OP_STAR:        case OP_STAR:
1752        case OP_MINSTAR:        case OP_MINSTAR:
1753          case OP_POSSTAR:
1754        case OP_PLUS:        case OP_PLUS:
1755        case OP_MINPLUS:        case OP_MINPLUS:
1756          case OP_POSPLUS:
1757        case OP_QUERY:        case OP_QUERY:
1758        case OP_MINQUERY:        case OP_MINQUERY:
1759        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1760        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;  
1761        break;        break;
1762        }        }
1763    #else
1764        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1765  #endif  #endif
1766      }      }
1767    }    }
# Line 1132  for (;;) Line 1774  for (;;)
1774  *************************************************/  *************************************************/
1775    
1776  /* 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
1777  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()
1778  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
1779  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
1780  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1781    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1782    bracket whose current branch will already have been scanned.
1783    
1784  Arguments:  Arguments:
1785    code        points to start of search    code        points to start of search
1786    endcode     points to where to stop    endcode     points to where to stop
1787    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1788      cd          contains pointers to tables etc.
1789    
1790  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1791  */  */
1792    
1793  static BOOL  static BOOL
1794  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1795      compile_data *cd)
1796  {  {
1797  register int c;  register int c;
1798  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);
1799       code < endcode;       code < endcode;
1800       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1801    {    {
1802    const uschar *ccode;    const uschar *ccode;
1803    
1804    c = *code;    c = *code;
1805    
1806      /* Skip over forward assertions; the other assertions are skipped by
1807      first_significant_code() with a TRUE final argument. */
1808    
1809    if (c >= OP_BRA)    if (c == OP_ASSERT)
1810      {      {
1811      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1812      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1813        continue;
1814        }
1815    
1816      /* Scan a closed bracket */    /* Groups with zero repeats can of course be empty; skip them. */
1817    
1818      empty_branch = FALSE;    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1819        {
1820        code += _pcre_OP_lengths[c];
1821        do code += GET(code, 1); while (*code == OP_ALT);
1822        c = *code;
1823        continue;
1824        }
1825    
1826      /* For a recursion/subroutine call, if its end has been reached, which
1827      implies a subroutine call, we can scan it. */
1828    
1829      if (c == OP_RECURSE)
1830        {
1831        BOOL empty_branch = FALSE;
1832        const uschar *scode = cd->start_code + GET(code, 1);
1833        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1834      do      do
1835        {        {
1836        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1837            {
1838          empty_branch = TRUE;          empty_branch = TRUE;
1839            break;
1840            }
1841          scode += GET(scode, 1);
1842          }
1843        while (*scode == OP_ALT);
1844        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1845        continue;
1846        }
1847    
1848      /* For other groups, scan the branches. */
1849    
1850      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1851        {
1852        BOOL empty_branch;
1853        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1854    
1855        /* If a conditional group has only one branch, there is a second, implied,
1856        empty branch, so just skip over the conditional, because it could be empty.
1857        Otherwise, scan the individual branches of the group. */
1858    
1859        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1860        code += GET(code, 1);        code += GET(code, 1);
1861        else
1862          {
1863          empty_branch = FALSE;
1864          do
1865            {
1866            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1867              empty_branch = TRUE;
1868            code += GET(code, 1);
1869            }
1870          while (*code == OP_ALT);
1871          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1872        }        }
1873      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1874      c = *code;      c = *code;
1875        continue;
1876      }      }
1877    
1878    else switch (c)    /* Handle the other opcodes */
1879    
1880      switch (c)
1881      {      {
1882      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1883        cannot be represented just by a bit map. This includes negated single
1884        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1885        actual length is stored in the compiled code, so we must update "code"
1886        here. */
1887    
1888  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1889      case OP_XCLASS:      case OP_XCLASS:
1890      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1891      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1892  #endif  #endif
1893    
# Line 1227  for (code = first_significant_code(code Line 1931  for (code = first_significant_code(code
1931      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1932      case OP_WORDCHAR:      case OP_WORDCHAR:
1933      case OP_ANY:      case OP_ANY:
1934        case OP_ALLANY:
1935      case OP_ANYBYTE:      case OP_ANYBYTE:
1936      case OP_CHAR:      case OP_CHAR:
1937      case OP_CHARNC:      case OP_CHARNC:
1938      case OP_NOT:      case OP_NOT:
1939      case OP_PLUS:      case OP_PLUS:
1940      case OP_MINPLUS:      case OP_MINPLUS:
1941        case OP_POSPLUS:
1942      case OP_EXACT:      case OP_EXACT:
1943      case OP_NOTPLUS:      case OP_NOTPLUS:
1944      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1945        case OP_NOTPOSPLUS:
1946      case OP_NOTEXACT:      case OP_NOTEXACT:
1947      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1948      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1949        case OP_TYPEPOSPLUS:
1950      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1951      return FALSE;      return FALSE;
1952    
1953        /* These are going to continue, as they may be empty, but we have to
1954        fudge the length for the \p and \P cases. */
1955    
1956        case OP_TYPESTAR:
1957        case OP_TYPEMINSTAR:
1958        case OP_TYPEPOSSTAR:
1959        case OP_TYPEQUERY:
1960        case OP_TYPEMINQUERY:
1961        case OP_TYPEPOSQUERY:
1962        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1963        break;
1964    
1965        /* Same for these */
1966    
1967        case OP_TYPEUPTO:
1968        case OP_TYPEMINUPTO:
1969        case OP_TYPEPOSUPTO:
1970        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1971        break;
1972    
1973      /* End of branch */      /* End of branch */
1974    
1975      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1978  for (code = first_significant_code(code
1978      case OP_ALT:      case OP_ALT:
1979      return TRUE;      return TRUE;
1980    
1981      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1982      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1983    
1984  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1985      case OP_STAR:      case OP_STAR:
1986      case OP_MINSTAR:      case OP_MINSTAR:
1987        case OP_POSSTAR:
1988      case OP_QUERY:      case OP_QUERY:
1989      case OP_MINQUERY:      case OP_MINQUERY:
1990        case OP_POSQUERY:
1991        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1992        break;
1993    
1994      case OP_UPTO:      case OP_UPTO:
1995      case OP_MINUPTO:      case OP_MINUPTO:
1996      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
1997        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1998      break;      break;
1999  #endif  #endif
2000    
2001        /* None of the remaining opcodes are required to match a character. */
2002    
2003        default:
2004        break;
2005      }      }
2006    }    }
2007    
# Line 1285  Arguments: Line 2024  Arguments:
2024    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2025    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2026    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2027      cd          pointers to tables etc
2028    
2029  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2030  */  */
2031    
2032  static BOOL  static BOOL
2033  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2034    BOOL utf8)    BOOL utf8, compile_data *cd)
2035  {  {
2036  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2037    {    {
2038    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2039        return FALSE;
2040    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2041    }    }
2042  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2049  return TRUE;
2049  *************************************************/  *************************************************/
2050    
2051  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2052  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
2053  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2054  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2055    
2056    Originally, this function only recognized a sequence of letters between the
2057    terminators, but it seems that Perl recognizes any sequence of characters,
2058    though of course unknown POSIX names are subsequently rejected. Perl gives an
2059    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2060    didn't consider this to be a POSIX class. Likewise for [:1234:].
2061    
2062    The problem in trying to be exactly like Perl is in the handling of escapes. We
2063    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2064    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2065    below handles the special case of \], but does not try to do any other escape
2066    processing. This makes it different from Perl for cases such as [:l\ower:]
2067    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2068    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2069    I think.
2070    
2071  Argument:  Arguments:
2072    ptr      pointer to the initial [    ptr      pointer to the initial [
2073    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2074    
2075  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2076  */  */
2077    
2078  static BOOL  static BOOL
2079  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2080  {  {
2081  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2082  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2083  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2084    {    {
2085    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2086    return TRUE;      {
2087        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2088        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2089          {
2090          *endptr = ptr;
2091          return TRUE;
2092          }
2093        }
2094    }    }
2095  return FALSE;  return FALSE;
2096  }  }
# Line 1355  Returns:     a value representing the na Line 2115  Returns:     a value representing the na
2115  static int  static int
2116  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2117  {  {
2118    const char *pn = posix_names;
2119  register int yield = 0;  register int yield = 0;
2120  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2121    {    {
2122    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2123      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2124      pn += posix_name_lengths[yield] + 1;
2125    yield++;    yield++;
2126    }    }
2127  return -1;  return -1;
# Line 1374  return -1; Line 2136  return -1;
2136  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2137  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2138  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
2139  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
2140  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
2141  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
2142  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
2143  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2144    OP_END.
2145    
2146    This function has been extended with the possibility of forward references for
2147    recursions and subroutine calls. It must also check the list of such references
2148    for the group we are dealing with. If it finds that one of the recursions in
2149    the current group is on this list, it adjusts the offset in the list, not the
2150    value in the reference (which is a group number).
2151    
2152  Arguments:  Arguments:
2153    group      points to the start of the group    group      points to the start of the group
2154    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2155    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2156    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2157      save_hwm   the hwm forward reference pointer at the start of the group
2158    
2159  Returns:     nothing  Returns:     nothing
2160  */  */
2161    
2162  static void  static void
2163  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2164      uschar *save_hwm)
2165  {  {
2166  uschar *ptr = group;  uschar *ptr = group;
2167    
2168  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2169    {    {
2170    int offset = GET(ptr, 1);    int offset;
2171    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2172    
2173      /* See if this recursion is on the forward reference list. If so, adjust the
2174      reference. */
2175    
2176      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2177        {
2178        offset = GET(hc, 0);
2179        if (cd->start_code + offset == ptr + 1)
2180          {
2181          PUT(hc, 0, offset + adjust);
2182          break;
2183          }
2184        }
2185    
2186      /* Otherwise, adjust the recursion offset if it's after the start of this
2187      group. */
2188    
2189      if (hc >= cd->hwm)
2190        {
2191        offset = GET(ptr, 1);
2192        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2193        }
2194    
2195    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2196    }    }
2197  }  }
# Line 1475  Yield:        TRUE when range returned; Line 2270  Yield:        TRUE when range returned;
2270  */  */
2271    
2272  static BOOL  static BOOL
2273  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2274      unsigned int *odptr)
2275  {  {
2276  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2277    
2278  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2279    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2280    
2281  if (c > d) return FALSE;  if (c > d) return FALSE;
2282    
# Line 1492  next = othercase + 1; Line 2285  next = othercase + 1;
2285    
2286  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2287    {    {
2288    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2289    next++;    next++;
2290    }    }
2291    
# Line 1506  return TRUE; Line 2297  return TRUE;
2297  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2298    
2299    
2300    
2301    /*************************************************
2302    *     Check if auto-possessifying is possible    *
2303    *************************************************/
2304    
2305    /* This function is called for unlimited repeats of certain items, to see
2306    whether the next thing could possibly match the repeated item. If not, it makes
2307    sense to automatically possessify the repeated item.
2308    
2309    Arguments:
2310      op_code       the repeated op code
2311      this          data for this item, depends on the opcode
2312      utf8          TRUE in UTF-8 mode
2313      utf8_char     used for utf8 character bytes, NULL if not relevant
2314      ptr           next character in pattern
2315      options       options bits
2316      cd            contains pointers to tables etc.
2317    
2318    Returns:        TRUE if possessifying is wanted
2319    */
2320    
2321    static BOOL
2322    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2323      const uschar *ptr, int options, compile_data *cd)
2324    {
2325    int next;
2326    
2327    /* Skip whitespace and comments in extended mode */
2328    
2329    if ((options & PCRE_EXTENDED) != 0)
2330      {
2331      for (;;)
2332        {
2333        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2334        if (*ptr == CHAR_NUMBER_SIGN)
2335          {
2336          while (*(++ptr) != 0)
2337            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2338          }
2339        else break;
2340        }
2341      }
2342    
2343    /* If the next item is one that we can handle, get its value. A non-negative
2344    value is a character, a negative value is an escape value. */
2345    
2346    if (*ptr == CHAR_BACKSLASH)
2347      {
2348      int temperrorcode = 0;
2349      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2350      if (temperrorcode != 0) return FALSE;
2351      ptr++;    /* Point after the escape sequence */
2352      }
2353    
2354    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2355      {
2356    #ifdef SUPPORT_UTF8
2357      if (utf8) { GETCHARINC(next, ptr); } else
2358    #endif
2359      next = *ptr++;
2360      }
2361    
2362    else return FALSE;
2363    
2364    /* Skip whitespace and comments in extended mode */
2365    
2366    if ((options & PCRE_EXTENDED) != 0)
2367      {
2368      for (;;)
2369        {
2370        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2371        if (*ptr == CHAR_NUMBER_SIGN)
2372          {
2373          while (*(++ptr) != 0)
2374            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2375          }
2376        else break;
2377        }
2378      }
2379    
2380    /* If the next thing is itself optional, we have to give up. */
2381    
2382    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2383      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2384        return FALSE;
2385    
2386    /* Now compare the next item with the previous opcode. If the previous is a
2387    positive single character match, "item" either contains the character or, if
2388    "item" is greater than 127 in utf8 mode, the character's bytes are in
2389    utf8_char. */
2390    
2391    
2392    /* Handle cases when the next item is a character. */
2393    
2394    if (next >= 0) switch(op_code)
2395      {
2396      case OP_CHAR:
2397    #ifdef SUPPORT_UTF8
2398      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2399    #else
2400      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2401    #endif
2402      return item != next;
2403    
2404      /* For CHARNC (caseless character) we must check the other case. If we have
2405      Unicode property support, we can use it to test the other case of
2406      high-valued characters. */
2407    
2408      case OP_CHARNC:
2409    #ifdef SUPPORT_UTF8
2410      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2411    #endif
2412      if (item == next) return FALSE;
2413    #ifdef SUPPORT_UTF8
2414      if (utf8)
2415        {
2416        unsigned int othercase;
2417        if (next < 128) othercase = cd->fcc[next]; else
2418    #ifdef SUPPORT_UCP
2419        othercase = UCD_OTHERCASE((unsigned int)next);
2420    #else
2421        othercase = NOTACHAR;
2422    #endif
2423        return (unsigned int)item != othercase;
2424        }
2425      else
2426    #endif  /* SUPPORT_UTF8 */
2427      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2428    
2429      /* For OP_NOT, "item" must be a single-byte character. */
2430    
2431      case OP_NOT:
2432      if (item == next) return TRUE;
2433      if ((options & PCRE_CASELESS) == 0) return FALSE;
2434    #ifdef SUPPORT_UTF8
2435      if (utf8)
2436        {
2437        unsigned int othercase;
2438        if (next < 128) othercase = cd->fcc[next]; else
2439    #ifdef SUPPORT_UCP
2440        othercase = UCD_OTHERCASE(next);
2441    #else
2442        othercase = NOTACHAR;
2443    #endif
2444        return (unsigned int)item == othercase;
2445        }
2446      else
2447    #endif  /* SUPPORT_UTF8 */
2448      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2449    
2450      case OP_DIGIT:
2451      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2452    
2453      case OP_NOT_DIGIT:
2454      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2455    
2456      case OP_WHITESPACE:
2457      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2458    
2459      case OP_NOT_WHITESPACE:
2460      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2461    
2462      case OP_WORDCHAR:
2463      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2464    
2465      case OP_NOT_WORDCHAR:
2466      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2467    
2468      case OP_HSPACE:
2469      case OP_NOT_HSPACE:
2470      switch(next)
2471        {
2472        case 0x09:
2473        case 0x20:
2474        case 0xa0:
2475        case 0x1680:
2476        case 0x180e:
2477        case 0x2000:
2478        case 0x2001:
2479        case 0x2002:
2480        case 0x2003:
2481        case 0x2004:
2482        case 0x2005:
2483        case 0x2006:
2484        case 0x2007:
2485        case 0x2008:
2486        case 0x2009:
2487        case 0x200A:
2488        case 0x202f:
2489        case 0x205f:
2490        case 0x3000:
2491        return op_code != OP_HSPACE;
2492        default:
2493        return op_code == OP_HSPACE;
2494        }
2495    
2496      case OP_VSPACE:
2497      case OP_NOT_VSPACE:
2498      switch(next)
2499        {
2500        case 0x0a:
2501        case 0x0b:
2502        case 0x0c:
2503        case 0x0d:
2504        case 0x85:
2505        case 0x2028:
2506        case 0x2029:
2507        return op_code != OP_VSPACE;
2508        default:
2509        return op_code == OP_VSPACE;
2510        }
2511    
2512      default:
2513      return FALSE;
2514      }
2515    
2516    
2517    /* Handle the case when the next item is \d, \s, etc. */
2518    
2519    switch(op_code)
2520      {
2521      case OP_CHAR:
2522      case OP_CHARNC:
2523    #ifdef SUPPORT_UTF8
2524      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2525    #endif
2526      switch(-next)
2527        {
2528        case ESC_d:
2529        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2530    
2531        case ESC_D:
2532        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2533    
2534        case ESC_s:
2535        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2536    
2537        case ESC_S:
2538        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2539    
2540        case ESC_w:
2541        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2542    
2543        case ESC_W:
2544        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2545    
2546        case ESC_h:
2547        case ESC_H:
2548        switch(item)
2549          {
2550          case 0x09:
2551          case 0x20:
2552          case 0xa0:
2553          case 0x1680:
2554          case 0x180e:
2555          case 0x2000:
2556          case 0x2001:
2557          case 0x2002:
2558          case 0x2003:
2559          case 0x2004:
2560          case 0x2005:
2561          case 0x2006:
2562          case 0x2007:
2563          case 0x2008:
2564          case 0x2009:
2565          case 0x200A:
2566          case 0x202f:
2567          case 0x205f:
2568          case 0x3000:
2569          return -next != ESC_h;
2570          default:
2571          return -next == ESC_h;
2572          }
2573    
2574        case ESC_v:
2575        case ESC_V:
2576        switch(item)
2577          {
2578          case 0x0a:
2579          case 0x0b:
2580          case 0x0c:
2581          case 0x0d:
2582          case 0x85:
2583          case 0x2028:
2584          case 0x2029:
2585          return -next != ESC_v;
2586          default:
2587          return -next == ESC_v;
2588          }
2589    
2590        default:
2591        return FALSE;
2592        }
2593    
2594      case OP_DIGIT:
2595      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2596             next == -ESC_h || next == -ESC_v;
2597    
2598      case OP_NOT_DIGIT:
2599      return next == -ESC_d;
2600    
2601      case OP_WHITESPACE:
2602      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2603    
2604      case OP_NOT_WHITESPACE:
2605      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2606    
2607      case OP_HSPACE:
2608      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2609    
2610      case OP_NOT_HSPACE:
2611      return next == -ESC_h;
2612    
2613      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2614      case OP_VSPACE:
2615      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2616    
2617      case OP_NOT_VSPACE:
2618      return next == -ESC_v;
2619    
2620      case OP_WORDCHAR:
2621      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2622    
2623      case OP_NOT_WORDCHAR:
2624      return next == -ESC_w || next == -ESC_d;
2625    
2626      default:
2627      return FALSE;
2628      }
2629    
2630    /* Control does not reach here */
2631    }
2632    
2633    
2634    
2635  /*************************************************  /*************************************************
2636  *           Compile one branch                   *  *           Compile one branch                   *
2637  *************************************************/  *************************************************/
2638    
2639  /* Scan the pattern, compiling it into the code vector. If the options are  /* Scan the pattern, compiling it into the a vector. If the options are
2640  changed during the branch, the pointer is used to change the external options  changed during the branch, the pointer is used to change the external options
2641  bits.  bits. This function is used during the pre-compile phase when we are trying
2642    to find out the amount of memory needed, as well as during the real compile
2643    phase. The value of lengthptr distinguishes the two phases.
2644    
2645  Arguments:  Arguments:
2646    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2647    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2648    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2649    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2651  Arguments:
2651    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2652    bcptr          points to current branch chain    bcptr          points to current branch chain
2653    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2654      lengthptr      NULL during the real compile phase
2655                     points to length accumulator during pre-compile phase
2656    
2657  Returns:         TRUE on success  Returns:         TRUE on success
2658                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2659  */  */
2660    
2661  static BOOL  static BOOL
2662  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2663    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2664    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2665  {  {
2666  int repeat_type, op_type;  int repeat_type, op_type;
2667  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1541  int greedy_default, greedy_non_default; Line 2670  int greedy_default, greedy_non_default;
2670  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2671  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2672  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2673  int options = *optionsptr;  int options = *optionsptr;
2674  int after_manual_callout = 0;  int after_manual_callout = 0;
2675    int length_prevgroup = 0;
2676  register int c;  register int c;
2677  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2678    uschar *last_code = code;
2679    uschar *orig_code = code;
2680  uschar *tempcode;  uschar *tempcode;
2681  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2682  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2684  const uschar *ptr = *ptrptr;
2684  const uschar *tempptr;  const uschar *tempptr;
2685  uschar *previous = NULL;  uschar *previous = NULL;
2686  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2687    uschar *save_hwm = NULL;
2688  uschar classbits[32];  uschar classbits[32];
2689    
2690  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2691  BOOL class_utf8;  BOOL class_utf8;
2692  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2693  uschar *class_utf8data;  uschar *class_utf8data;
2694    uschar *class_utf8data_base;
2695  uschar utf8_char[6];  uschar utf8_char[6];
2696  #else  #else
2697  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2698    uschar *utf8_char = NULL;
2699    #endif
2700    
2701    #ifdef PCRE_DEBUG
2702    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2703  #endif  #endif
2704    
2705  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1593  req_caseopt = ((options & PCRE_CASELESS) Line 2731  req_caseopt = ((options & PCRE_CASELESS)
2731  for (;; ptr++)  for (;; ptr++)
2732    {    {
2733    BOOL negate_class;    BOOL negate_class;
2734      BOOL should_flip_negation;
2735    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2736    BOOL is_quantifier;    BOOL is_quantifier;
2737      BOOL is_recurse;
2738      BOOL reset_bracount;
2739    int class_charcount;    int class_charcount;
2740    int class_lastchar;    int class_lastchar;
2741    int newoptions;    int newoptions;
2742    int recno;    int recno;
2743      int refsign;
2744    int skipbytes;    int skipbytes;
2745    int subreqbyte;    int subreqbyte;
2746    int subfirstbyte;    int subfirstbyte;
2747      int terminator;
2748    int mclength;    int mclength;
2749    uschar mcbuffer[8];    uschar mcbuffer[8];
2750    
2751    /* Next byte in the pattern */    /* Get next byte in the pattern */
2752    
2753    c = *ptr;    c = *ptr;
2754    
2755    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If we are in the pre-compile phase, accumulate the length used for the
2756      previous cycle of this loop. */
2757    
2758    if (inescq && c != 0)    if (lengthptr != NULL)
2759      {      {
2760      if (c == '\\' && ptr[1] == 'E')  #ifdef PCRE_DEBUG
2761        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2762    #endif
2763        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2764        {        {
2765        inescq = FALSE;        *errorcodeptr = ERR52;
2766        ptr++;        goto FAILED;
       continue;  
2767        }        }
2768      else  
2769        /* There is at least one situation where code goes backwards: this is the
2770        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2771        the class is simply eliminated. However, it is created first, so we have to
2772        allow memory for it. Therefore, don't ever reduce the length at this point.
2773        */
2774    
2775        if (code < last_code) code = last_code;
2776    
2777        /* Paranoid check for integer overflow */
2778    
2779        if (OFLOW_MAX - *lengthptr < code - last_code)
2780        {        {
2781        if (previous_callout != NULL)        *errorcodeptr = ERR20;
2782          {        goto FAILED;
2783          complete_callout(previous_callout, ptr, cd);        }
2784    
2785        *lengthptr += code - last_code;
2786        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2787    
2788        /* If "previous" is set and it is not at the start of the work space, move
2789        it back to there, in order to avoid filling up the work space. Otherwise,
2790        if "previous" is NULL, reset the current code pointer to the start. */
2791    
2792        if (previous != NULL)
2793          {
2794          if (previous > orig_code)
2795            {
2796            memmove(orig_code, previous, code - previous);
2797            code -= previous - orig_code;
2798            previous = orig_code;
2799            }
2800          }
2801        else code = orig_code;
2802    
2803        /* Remember where this code item starts so we can pick up the length
2804        next time round. */
2805    
2806        last_code = code;
2807        }
2808    
2809      /* In the real compile phase, just check the workspace used by the forward
2810      reference list. */
2811    
2812      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2813        {
2814        *errorcodeptr = ERR52;
2815        goto FAILED;
2816        }
2817    
2818      /* If in \Q...\E, check for the end; if not, we have a literal */
2819    
2820      if (inescq && c != 0)
2821        {
2822        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2823          {
2824          inescq = FALSE;
2825          ptr++;
2826          continue;
2827          }
2828        else
2829          {
2830          if (previous_callout != NULL)
2831            {
2832            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2833              complete_callout(previous_callout, ptr, cd);
2834          previous_callout = NULL;          previous_callout = NULL;
2835          }          }
2836        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2845  for (;; ptr++)
2845    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
2846    a quantifier. */    a quantifier. */
2847    
2848    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2849      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2850        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2851    
2852    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2853         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2854      {      {
2855      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2856          complete_callout(previous_callout, ptr, cd);
2857      previous_callout = NULL;      previous_callout = NULL;
2858      }      }
2859    
# Line 1653  for (;; ptr++) Line 2862  for (;; ptr++)
2862    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2863      {      {
2864      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2865      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2866        {        {
2867        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2868        on the Macintosh. */          {
2869        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2870        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2871          if (*ptr != 0) continue;
2872    
2873          /* Else fall through to handle end of string */
2874          c = 0;
2875        }        }
2876      }      }
2877    
# Line 1672  for (;; ptr++) Line 2885  for (;; ptr++)
2885    
2886    switch(c)    switch(c)
2887      {      {
2888      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2889        case 0:                        /* The branch terminates at string end */
2890      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2891      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2892      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2893      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2894      *codeptr = code;      *codeptr = code;
2895      *ptrptr = ptr;      *ptrptr = ptr;
2896        if (lengthptr != NULL)
2897          {
2898          if (OFLOW_MAX - *lengthptr < code - last_code)
2899            {
2900            *errorcodeptr = ERR20;
2901            goto FAILED;
2902            }
2903          *lengthptr += code - last_code;   /* To include callout length */
2904          DPRINTF((">> end branch\n"));
2905          }
2906      return TRUE;      return TRUE;
2907    
2908    
2909        /* ===================================================================*/
2910      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2911      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2912    
2913      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2914      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2915        {        {
2916        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2919  for (;; ptr++)
2919      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2920      break;      break;
2921    
2922      case '$':      case CHAR_DOLLAR_SIGN:
2923      previous = NULL;      previous = NULL;
2924      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2925      break;      break;
# Line 1703  for (;; ptr++) Line 2927  for (;; ptr++)
2927      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
2928      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2929    
2930      case '.':      case CHAR_DOT:
2931      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2932      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2933      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2934      previous = code;      previous = code;
2935      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2936      break;      break;
2937    
2938      /* Character classes. If the included characters are all < 255 in value, we  
2939      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2940      case where there is only one such character. For negated classes, we build      /* Character classes. If the included characters are all < 256, we build a
2941      the map as usual, then invert it at the end. However, we use a different      32-byte bitmap of the permitted characters, except in the special case
2942      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2943        map as usual, then invert it at the end. However, we use a different opcode
2944        so that data characters > 255 can be handled correctly.
2945    
2946      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2947      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
2948      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2949      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
2950    
2951      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2952        default (Perl) mode, it is treated as a data character. */
2953    
2954        case CHAR_RIGHT_SQUARE_BRACKET:
2955        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2956          {
2957          *errorcodeptr = ERR64;
2958          goto FAILED;
2959          }
2960        goto NORMAL_CHAR;
2961    
2962        case CHAR_LEFT_SQUARE_BRACKET:
2963      previous = code;      previous = code;
2964    
2965      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
2966      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2967    
2968      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2969          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2970            check_posix_syntax(ptr, &tempptr))
2971        {        {
2972        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2973        goto FAILED;        goto FAILED;
2974        }        }
2975    
2976      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2977        if the first few characters (either before or after ^) are \Q\E or \E we
2978        skip them too. This makes for compatibility with Perl. */
2979    
2980      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2981        for (;;)
2982        {        {
       negate_class = TRUE;  
2983        c = *(++ptr);        c = *(++ptr);
2984          if (c == CHAR_BACKSLASH)
2985            {
2986            if (ptr[1] == CHAR_E)
2987              ptr++;
2988            else if (strncmp((const char *)ptr+1,
2989                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2990              ptr += 3;
2991            else
2992              break;
2993            }
2994          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2995            negate_class = TRUE;
2996          else break;
2997        }        }
2998      else  
2999        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3000        an initial ']' is taken as a data character -- the code below handles
3001        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3002        [^] must match any character, so generate OP_ALLANY. */
3003    
3004        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3005            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3006        {        {
3007        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3008          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3009          zerofirstbyte = firstbyte;
3010          break;
3011        }        }
3012    
3013        /* If a class contains a negative special such as \S, we need to flip the
3014        negation flag at the end, so that support for characters > 255 works
3015        correctly (they are all included in the class). */
3016    
3017        should_flip_negation = FALSE;
3018    
3019      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
3020      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
3021      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3022    
3023      class_charcount = 0;      class_charcount = 0;
3024      class_lastchar = -1;      class_lastchar = -1;
3025    
3026        /* Initialize the 32-char bit map to all zeros. We build the map in a
3027        temporary bit of memory, in case the class contains only 1 character (less
3028        than 256), because in that case the compiled code doesn't use the bit map.
3029        */
3030    
3031        memset(classbits, 0, 32 * sizeof(uschar));
3032    
3033  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3034      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3035      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3036        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3037  #endif  #endif
3038    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
3039      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3040      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
3041      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3042    
3043      do      if (c != 0) do
3044        {        {
3045          const uschar *oldptr;
3046    
3047  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3048        if (utf8 && c > 127)        if (utf8 && c > 127)
3049          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3050          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3051          }          }
3052    
3053          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3054          data and reset the pointer. This is so that very large classes that
3055          contain a zillion UTF-8 characters no longer overwrite the work space
3056          (which is on the stack). */
3057    
3058          if (lengthptr != NULL)
3059            {
3060            *lengthptr += class_utf8data - class_utf8data_base;
3061            class_utf8data = class_utf8data_base;
3062            }
3063    
3064  #endif  #endif
3065    
3066        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3067    
3068        if (inescq)        if (inescq)
3069          {          {
3070          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3071            {            {
3072            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3073            ptr++;            ptr++;                            /* Skip the 'E' */
3074            continue;            continue;                         /* Carry on with next */
3075            }            }
3076          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3077          }          }
3078    
3079        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1801  for (;; ptr++) Line 3082  for (;; ptr++)
3082        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3083        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3084    
3085        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3086            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3087            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3088          {          {
3089          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3090          int posix_class, i;          int posix_class, taboffset, tabopt;
3091          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3092            uschar pbits[32];
3093    
3094          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3095            {            {
3096            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3097            goto FAILED;            goto FAILED;
3098            }            }
3099    
3100          ptr += 2;          ptr += 2;
3101          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3102            {            {
3103            local_negate = TRUE;            local_negate = TRUE;
3104              should_flip_negation = TRUE;  /* Note negative special */
3105            ptr++;            ptr++;
3106            }            }
3107    
# Line 1836  for (;; ptr++) Line 3119  for (;; ptr++)
3119          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3120            posix_class = 0;            posix_class = 0;
3121    
3122          /* Or into the map we are building up to 3 of the static class          /* We build the bit map for the POSIX class in a chunk of local store
3123          tables, or their negations. The [:blank:] class sets up the same          because we may be adding and subtracting from it, and we don't want to
3124          chars as the [:space:] class (all white space). We remove the vertical          subtract bits that may be in the main map already. At the end we or the
3125          white space chars afterwards. */          result into the bit map that is being built. */
3126    
3127          posix_class *= 3;          posix_class *= 3;
3128          for (i = 0; i < 3; i++)  
3129            /* Copy in the first table (always present) */
3130    
3131            memcpy(pbits, cbits + posix_class_maps[posix_class],
3132              32 * sizeof(uschar));
3133    
3134            /* If there is a second table, add or remove it as required. */
3135    
3136            taboffset = posix_class_maps[posix_class + 1];
3137            tabopt = posix_class_maps[posix_class + 2];
3138    
3139            if (taboffset >= 0)
3140            {            {
3141            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3142            int taboffset = posix_class_maps[posix_class + i];              for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
           if (taboffset < 0) break;  
           if (local_negate)  
             {  
             if (i == 0)  
               for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];  
             else  
               for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];  
             if (blankclass) classbits[1] |= 0x3c;  
             }  
3143            else            else
3144              {              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
3145            }            }
3146    
3147            /* Not see if we need to remove any special characters. An option
3148            value of 1 removes vertical space and 2 removes underscore. */
3149    
3150            if (tabopt < 0) tabopt = -tabopt;
3151            if (tabopt == 1) pbits[1] &= ~0x3c;
3152              else if (tabopt == 2) pbits[11] &= 0x7f;
3153    
3154            /* Add the POSIX table or its complement into the main table that is
3155            being built and we are done. */
3156    
3157            if (local_negate)
3158              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3159            else
3160              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3161    
3162          ptr = tempptr + 1;          ptr = tempptr + 1;
3163          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3164          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3165          }          }
3166    
3167        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3168        of the specials, which just set a flag. Escaped items are checked for        of the specials, which just set a flag. The sequence \b is a special
3169        validity in the pre-compiling pass. The sequence \b is a special case.        case. Inside a class (and only there) it is treated as backspace.
3170        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3171        it marks a word boundary. Other escapes have preset maps ready to        to 'or' into the one we are building. We assume they have more than one
       or into the one we are building. We assume they have more than one  
3172        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3173    
3174        if (c == '\\')        if (c == CHAR_BACKSLASH)
3175          {          {
3176          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3177            if (*errorcodeptr != 0) goto FAILED;
3178    
3179          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */
3180          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3181            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3182          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3183            {            {
3184            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3185              {              {
3186              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3187              }              }
3188            else inescq = TRUE;            else inescq = TRUE;
3189            continue;            continue;
3190            }            }
3191            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3192    
3193          if (c < 0)          if (c < 0)
3194            {            {
3195            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3196            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3197            switch (-c)  
3198              /* Save time by not doing this in the pre-compile phase. */
3199    
3200              if (lengthptr == NULL) switch (-c)
3201              {              {
3202              case ESC_d:              case ESC_d:
3203              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3204              continue;              continue;
3205    
3206              case ESC_D:              case ESC_D:
3207                should_flip_negation = TRUE;
3208              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3209              continue;              continue;
3210    
# Line 1910  for (;; ptr++) Line 3213  for (;; ptr++)
3213              continue;              continue;
3214    
3215              case ESC_W:              case ESC_W:
3216                should_flip_negation = TRUE;
3217              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3218              continue;              continue;
3219    
# Line 1919  for (;; ptr++) Line 3223  for (;; ptr++)
3223              continue;              continue;
3224    
3225              case ESC_S:              case ESC_S:
3226                should_flip_negation = TRUE;
3227              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3228              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3229              continue;              continue;
3230    
3231  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3232              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3233              case ESC_P:              }
3234    
3235              /* In the pre-compile phase, just do the recognition. */
3236    
3237              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3238                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3239    
3240              /* We need to deal with \H, \h, \V, and \v in both phases because
3241              they use extra memory. */
3242    
3243              if (-c == ESC_h)
3244                {
3245                SETBIT(classbits, 0x09); /* VT */
3246                SETBIT(classbits, 0x20); /* SPACE */
3247                SETBIT(classbits, 0xa0); /* NSBP */
3248    #ifdef SUPPORT_UTF8
3249                if (utf8)
3250                  {
3251                  class_utf8 = TRUE;
3252                  *class_utf8data++ = XCL_SINGLE;
3253                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3254                  *class_utf8data++ = XCL_SINGLE;
3255                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3256                  *class_utf8data++ = XCL_RANGE;
3257                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3258                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3259                  *class_utf8data++ = XCL_SINGLE;
3260                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3261                  *class_utf8data++ = XCL_SINGLE;
3262                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3263                  *class_utf8data++ = XCL_SINGLE;
3264                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3265                  }
3266    #endif
3267                continue;
3268                }
3269    
3270              if (-c == ESC_H)
3271                {
3272                for (c = 0; c < 32; c++)
3273                  {
3274                  int x = 0xff;
3275                  switch (c)
3276                    {
3277                    case 0x09/8: x ^= 1 << (0x09%8); break;
3278                    case 0x20/8: x ^= 1 << (0x20%8); break;
3279                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3280                    default: break;
3281                    }
3282                  classbits[c] |= x;
3283                  }
3284    
3285    #ifdef SUPPORT_UTF8
3286                if (utf8)
3287                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3288                class_utf8 = TRUE;                class_utf8 = TRUE;
3289                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3290                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3291                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3292                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3293                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3294                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3295                  *class_utf8data++ = XCL_RANGE;
3296                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3297                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3298                  *class_utf8data++ = XCL_RANGE;
3299                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3300                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3301                  *class_utf8data++ = XCL_RANGE;
3302                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3303                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3304                  *class_utf8data++ = XCL_RANGE;
3305                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3306                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3307                  *class_utf8data++ = XCL_RANGE;
3308                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3309                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3310                }                }
3311    #endif
3312              continue;              continue;
3313                }
3314    
3315              if (-c == ESC_v)
3316                {
3317                SETBIT(classbits, 0x0a); /* LF */
3318                SETBIT(classbits, 0x0b); /* VT */
3319                SETBIT(classbits, 0x0c); /* FF */
3320                SETBIT(classbits, 0x0d); /* CR */
3321                SETBIT(classbits, 0x85); /* NEL */
3322    #ifdef SUPPORT_UTF8
3323                if (utf8)
3324                  {
3325                  class_utf8 = TRUE;
3326                  *class_utf8data++ = XCL_RANGE;
3327                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3328                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3329                  }
3330  #endif  #endif
3331                continue;
3332                }
3333    
3334              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3335              strict mode. By default, for compatibility with Perl, they are              {
3336              treated as literals. */              for (c = 0; c < 32; c++)
3337                  {
3338                  int x = 0xff;
3339                  switch (c)
3340                    {
3341                    case 0x0a/8: x ^= 1 << (0x0a%8);
3342                                 x ^= 1 << (0x0b%8);
3343                                 x ^= 1 << (0x0c%8);
3344                                 x ^= 1 << (0x0d%8);
3345                                 break;
3346                    case 0x85/8: x ^= 1 << (0x85%8); break;
3347                    default: break;
3348                    }
3349                  classbits[c] |= x;
3350                  }
3351    
3352              default:  #ifdef SUPPORT_UTF8
3353              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3354                {                {
3355                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3356                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3357                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3358                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3359                  *class_utf8data++ = XCL_RANGE;
3360                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3361                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3362                }                }
3363              c = *ptr;              /* The final character */  #endif
3364              class_charcount -= 2;  /* Undo the default count from above */              continue;
3365                }
3366    
3367              /* We need to deal with \P and \p in both phases. */
3368    
3369    #ifdef SUPPORT_UCP
3370              if (-c == ESC_p || -c == ESC_P)
3371                {
3372                BOOL negated;
3373                int pdata;
3374                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3375                if (ptype < 0) goto FAILED;
3376                class_utf8 = TRUE;
3377                *class_utf8data++ = ((-c == ESC_p) != negated)?
3378                  XCL_PROP : XCL_NOTPROP;
3379                *class_utf8data++ = ptype;
3380                *class_utf8data++ = pdata;
3381                class_charcount -= 2;   /* Not a < 256 character */
3382                continue;
3383                }
3384    #endif
3385              /* Unrecognized escapes are faulted if PCRE is running in its
3386              strict mode. By default, for compatibility with Perl, they are
3387              treated as literals. */
3388    
3389              if ((options & PCRE_EXTRA) != 0)
3390                {
3391                *errorcodeptr = ERR7;
3392                goto FAILED;
3393              }              }
3394    
3395              class_charcount -= 2;  /* Undo the default count from above */
3396              c = *ptr;              /* Get the final character and fall through */
3397            }            }
3398    
3399          /* 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
3400          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3401    
3402          }   /* End of backslash handling */          }   /* End of backslash handling */
3403    
3404        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3405        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
3406        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3407          entirely. The code for handling \Q and \E is messy. */
3408    
3409          CHECK_RANGE:
3410          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3411            {
3412            inescq = FALSE;
3413            ptr += 2;
3414            }
3415    
3416          oldptr = ptr;
3417    
3418        if (ptr[1] == '-' && ptr[2] != ']')        /* Remember \r or \n */
3419    
3420          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3421    
3422          /* Check for range */
3423    
3424          if (!inescq && ptr[1] == CHAR_MINUS)
3425          {          {
3426          int d;          int d;
3427          ptr += 2;          ptr += 2;
3428            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3429    
3430            /* If we hit \Q (not followed by \E) at this point, go into escaped
3431            mode. */
3432    
3433            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3434              {
3435              ptr += 2;
3436              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3437                { ptr += 2; continue; }
3438              inescq = TRUE;
3439              break;
3440              }
3441    
3442            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3443              {
3444              ptr = oldptr;
3445              goto LONE_SINGLE_CHARACTER;
3446              }
3447    
3448  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3449          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3458  for (;; ptr++)
3458          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
3459          in such circumstances. */          in such circumstances. */
3460    
3461          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3462            {            {
3463            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3464            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3465    
3466            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; \X is literal X; \R is literal R; any other
3467            was literal */            special means the '-' was literal */
3468    
3469            if (d < 0)            if (d < 0)
3470              {              {
3471              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3472              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3473                else if (d == -ESC_R) d = CHAR_R; else
3474                {                {
3475                ptr = oldptr - 2;                ptr = oldptr;
3476                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3477                }                }
3478              }              }
3479            }            }
3480    
3481          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3482          the pre-pass. Optimize one-character ranges */          one-character ranges */
3483    
3484            if (d < c)
3485              {
3486              *errorcodeptr = ERR8;
3487              goto FAILED;
3488              }
3489    
3490          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3491    
3492            /* Remember \r or \n */
3493    
3494            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3495    
3496          /* 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
3497          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3498          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2022  for (;; ptr++) Line 3510  for (;; ptr++)
3510  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3511            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3512              {              {
3513              int occ, ocd;              unsigned int occ, ocd;
3514              int cc = c;              unsigned int cc = c;
3515              int origd = d;              unsigned int origd = d;
3516              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3517                {                {
3518                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3519                      ocd <= (unsigned int)d)
3520                    continue;                          /* Skip embedded ranges */
3521    
3522                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3523                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3524                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3525                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3526                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3527                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3528                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3529                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3530                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3531                  d = ocd;                  d = ocd;
3532                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3574  for (;; ptr++)
3574          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
3575          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3576    
3577          for (; c <= d; c++)          class_charcount += d - c + 1;
3578            class_lastchar = d;
3579    
3580            /* We can save a bit of time by skipping this in the pre-compile. */
3581    
3582            if (lengthptr == NULL) for (; c <= d; c++)
3583            {            {
3584            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3585            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3587  for (;; ptr++)
3587              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3588              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3589              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3590            }            }
3591    
3592          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3610  for (;; ptr++)
3610  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3611          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3612            {            {
3613            int chartype;            unsigned int othercase;
3614            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3615              {              {
3616              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3617              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3636  for (;; ptr++)
3636          }          }
3637        }        }
3638    
3639      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3640      loop. This "while" is the end of the "do" above. */  
3641        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3642    
3643        if (c == 0)                          /* Missing terminating ']' */
3644          {
3645          *errorcodeptr = ERR6;
3646          goto FAILED;
3647          }
3648    
3649    
3650    /* This code has been disabled because it would mean that \s counts as
3651    an explicit \r or \n reference, and that's not really what is wanted. Now
3652    we set the flag only if there is a literal "\r" or "\n" in the class. */
3653    
3654    #if 0
3655        /* Remember whether \r or \n are in this class */
3656    
3657        if (negate_class)
3658          {
3659          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3660          }
3661        else
3662          {
3663          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3664          }
3665    #endif
3666    
     while ((c = *(++ptr)) != ']' || inescq);  
3667    
3668      /* 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
3669      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
3670      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
3671      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3672      single-bytes only. This is an historical hangover. Maybe one day we can  
3673      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3674        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3675        operate on single-bytes only. This is an historical hangover. Maybe one day
3676        we can tidy these opcodes to handle multi-byte characters.
3677    
3678      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
3679      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 2163  for (;; ptr++) Line 3683  for (;; ptr++)
3683      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3684    
3685  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3686      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3687            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3688  #else  #else
3689      if (class_charcount == 1)      if (class_charcount == 1)
3690  #endif  #endif
# Line 2209  for (;; ptr++) Line 3727  for (;; ptr++)
3727      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3728    
3729      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3730      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3731      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3732        the class, so any that were explicitly given as well can be ignored. If
3733        (when there are explicit characters > 255 that must be listed) there are no
3734        characters < 256, we can omit the bitmap in the actual compiled code. */
3735    
3736  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3737      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3738        {        {
3739        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3740        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3741        code += LINK_SIZE;        code += LINK_SIZE;
3742        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3743    
3744        /* 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;
3745        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3746    
3747        if (class_charcount > 0)        if (class_charcount > 0)
3748          {          {
3749          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3750            memmove(code + 32, code, class_utf8data - code);
3751          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3752          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;  
3753          }          }
3754          else code = class_utf8data;
3755    
3756        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3757    
# Line 2246  for (;; ptr++) Line 3760  for (;; ptr++)
3760        }        }
3761  #endif  #endif
3762    
3763      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3764      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3765      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3766      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3767    
3768        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3769      if (negate_class)      if (negate_class)
3770        {        {
3771        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3772        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3773        }        }
3774      else      else
3775        {        {
       *code++ = OP_CLASS;  
3776        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3777        }        }
3778      code += 32;      code += 32;
3779      break;      break;
3780    
3781    
3782        /* ===================================================================*/
3783      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3784      has been tested above. */      has been tested above. */
3785    
3786      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3787      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3788      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3789      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3790      goto REPEAT;      goto REPEAT;
3791    
3792      case '*':      case CHAR_ASTERISK:
3793      repeat_min = 0;      repeat_min = 0;
3794      repeat_max = -1;      repeat_max = -1;
3795      goto REPEAT;      goto REPEAT;
3796    
3797      case '+':      case CHAR_PLUS:
3798      repeat_min = 1;      repeat_min = 1;
3799      repeat_max = -1;      repeat_max = -1;
3800      goto REPEAT;      goto REPEAT;
3801    
3802      case '?':      case CHAR_QUESTION_MARK:
3803      repeat_min = 0;      repeat_min = 0;
3804      repeat_max = 1;      repeat_max = 1;
3805    
# Line 2318  for (;; ptr++) Line 3834  for (;; ptr++)
3834      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
3835      repeat type to the non-default. */      repeat type to the non-default. */
3836    
3837      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3838        {        {
3839        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3840        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3841        ptr++;        ptr++;
3842        }        }
3843      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3844        {        {
3845        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3846        ptr++;        ptr++;
3847        }        }
3848      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3849    
     /* 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;  
       }  
   
3850      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3851      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
3852      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 2378  for (;; ptr++) Line 3880  for (;; ptr++)
3880          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3881          }          }
3882    
3883          /* If the repetition is unlimited, it pays to see if the next thing on
3884          the line is something that cannot possibly match this character. If so,
3885          automatically possessifying this item gains some performance in the case
3886          where the match fails. */
3887    
3888          if (!possessive_quantifier &&
3889              repeat_max < 0 &&
3890              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3891                options, cd))
3892            {
3893            repeat_type = 0;    /* Force greedy */
3894            possessive_quantifier = TRUE;
3895            }
3896    
3897        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3898        }        }
3899    
3900      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3901      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-
3902      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3903      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3904        currently used only for single-byte chars. */
3905    
3906      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3907        {        {
3908        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3909        c = previous[1];        c = previous[1];
3910          if (!possessive_quantifier &&
3911              repeat_max < 0 &&
3912              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3913            {
3914            repeat_type = 0;    /* Force greedy */
3915            possessive_quantifier = TRUE;
3916            }
3917        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3918        }        }
3919    
# Line 2403  for (;; ptr++) Line 3927  for (;; ptr++)
3927      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3928        {        {
3929        uschar *oldcode;        uschar *oldcode;
3930        int prop_type;        int prop_type, prop_value;
3931        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3932        c = *previous;        c = *previous;
3933    
3934          if (!possessive_quantifier &&
3935              repeat_max < 0 &&
3936              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3937            {
3938            repeat_type = 0;    /* Force greedy */
3939            possessive_quantifier = TRUE;
3940            }
3941    
3942        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3943        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3944          previous[1] : -1;          {
3945            prop_type = previous[1];
3946            prop_value = previous[2];
3947            }
3948          else prop_type = prop_value = -1;
3949    
3950        oldcode = code;        oldcode = code;
3951        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2419  for (;; ptr++) Line 3955  for (;; ptr++)
3955    
3956        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3957    
3958          /*--------------------------------------------------------------------*/
3959          /* This code is obsolete from release 8.00; the restriction was finally
3960          removed: */
3961    
3962        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3963        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3964    
3965        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3966          /*--------------------------------------------------------------------*/
3967    
3968        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3969    
# Line 2443  for (;; ptr++) Line 3984  for (;; ptr++)
3984          }          }
3985    
3986        /* A repeat minimum of 1 is optimized into some special cases. If the        /* A repeat minimum of 1 is optimized into some special cases. If the
3987        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3988        left in place and, if the maximum is greater than 1, we use OP_UPTO with        left in place and, if the maximum is greater than 1, we use OP_UPTO with
3989        one less than the maximum. */        one less than the maximum. */
3990    
# Line 2470  for (;; ptr++) Line 4011  for (;; ptr++)
4011    
4012          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
4013          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
4014          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
4015          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4016          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
4017    
# Line 2486  for (;; ptr++) Line 4027  for (;; ptr++)
4027  #endif  #endif
4028              {              {
4029              *code++ = c;              *code++ = c;
4030              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
4031                  {
4032                  *code++ = prop_type;
4033                  *code++ = prop_value;
4034                  }
4035              }              }
4036            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
4037            }            }
4038    
4039          /* Else insert an UPTO if the max is greater than the min, again          /* Else insert an UPTO if the max is greater than the min, again
4040          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
4041            UPTO is just for 1 instance, we can use QUERY instead. */
4042    
4043          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4044            {            {
#