/[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 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 459 by ph10, Sun Oct 4 09:21:39 2009 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) 1997-2009 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    
# Line 53  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* 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
81  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
82  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
83  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
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 97  are simple data values; negative values
97  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
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  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,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  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,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* 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
179  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
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    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 };
225    
226  /* 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
227  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
228  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
229    characters are removed, and for [:alpha:] and [:alnum:] the underscore
230    character is removed. The triples in the table consist of the base map offset,
231    second map offset or -1 if no second map, and a non-negative value for map
232    addition or a negative value for map subtraction (if there are two maps). The
233    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
234    remove vertical space characters, 2 => remove underscore. */
235    
236  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
237    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
238    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
239    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
240    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
241    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
242    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
243    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
244    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
245    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
246    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
247    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
248    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
249    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
250    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
251  };  };
252    
253    
254  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
255  are passed to the outside world. */  #define XSTRING(s) STRING(s)
256    
257  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
258    "no error",  are passed to the outside world. Do not ever re-use any error number, because
259    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
260    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
261    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
262    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
263    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    simply count through to the one we want - this isn't a performance issue
265    because these strings are used only when there is a compilation error. */
266    
267    static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
319    "two named groups have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
325  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327      /* 50 */
328      "repeated subpattern is too long\0"    /** DEAD **/
329      "octal value is greater than \\377 (not in UTF-8 mode)\0"
330      "internal error: overran compiling workspace\0"
331      "internal error: previously-checked referenced subpattern not found\0"
332      "DEFINE group contains more than one branch\0"
333      /* 55 */
334      "repeating a DEFINE group is not allowed\0"
335      "inconsistent NEWLINE options\0"
336      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
337      "a numbered reference must not be zero\0"
338      "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode\0"
345      /* 65 */
346      "different names for subpatterns of the same number are not allowed";
347    
348    
349  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 229  For convenience, we use the same bit def Line 362  For convenience, we use the same bit def
362    
363  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
364    
365  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
366    
367    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
368    UTF-8 mode. */
369    
370  static const unsigned char digitab[] =  static const unsigned char digitab[] =
371    {    {
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 265  static const unsigned char digitab[] = Line 402  static const unsigned char digitab[] =
402    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
403    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
404    
405  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
406    
407    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
408    
409  static const unsigned char digitab[] =  static const unsigned char digitab[] =
410    {    {
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 279  static const unsigned char digitab[] = Line 419  static const unsigned char digitab[] =
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
420    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
421    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
422    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
423    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
424    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
425    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 313  static const unsigned char ebcdic_charta Line 453  static const unsigned char ebcdic_charta
453    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
454    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
455    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
456    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
457    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
458    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
459    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 340  static const unsigned char ebcdic_charta Line 480  static const unsigned char ebcdic_charta
480  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
481    
482  static BOOL  static BOOL
483    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
484      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
485    
486    
487    
488    /*************************************************
489    *            Find an error text                  *
490    *************************************************/
491    
492    /* The error texts are now all in one long string, to save on relocations. As
493    some of the text is of unknown length, we can't use a table of offsets.
494    Instead, just count through the strings. This is not a performance issue
495    because it happens only when there has been a compilation error.
496    
497    Argument:   the error number
498    Returns:    pointer to the error string
499    */
500    
501    static const char *
502    find_error_text(int n)
503    {
504    const char *s = error_texts;
505    for (; n > 0; n--) while (*s++ != 0) {};
506    return s;
507    }
508    
509    
510  /*************************************************  /*************************************************
# Line 351  static BOOL Line 513  static BOOL
513    
514  /* 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
515  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
516  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
517  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
518  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,
519    ptr is pointing at the \. On exit, it is on the final character of the escape
520    sequence.
521    
522  Arguments:  Arguments:
523    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 364  Arguments: Line 528  Arguments:
528    
529  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
530                   negative => a special escape sequence                   negative => a special escape sequence
531                   on error, errorptr is set                   on error, errorcodeptr is set
532  */  */
533    
534  static int  static int
535  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
536    int options, BOOL isclass)    int options, BOOL isclass)
537  {  {
538  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
539    const uschar *ptr = *ptrptr + 1;
540  int c, i;  int c, i;
541    
542    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
543    ptr--;                            /* Set pointer back to the last byte */
544    
545  /* 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. */
546    
 c = *(++ptr);  
547  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
548    
549  /* 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
550  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.
551  Otherwise further processing may be required. */  Otherwise further processing may be required. */
552    
553  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
554  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
555  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
556    
557  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
558  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
559  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
560  #endif  #endif
561    
# Line 397  else if ((i = escapes[c - 0x48]) != 0) Line 564  else if ((i = escapes[c - 0x48]) != 0)
564  else  else
565    {    {
566    const uschar *oldptr;    const uschar *oldptr;
567      BOOL braced, negated;
568    
569    switch (c)    switch (c)
570      {      {
571      /* 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
572      error. */      error. */
573    
574      case 'l':      case CHAR_l:
575      case 'L':      case CHAR_L:
576      case 'N':      case CHAR_N:
577      case 'u':      case CHAR_u:
578      case 'U':      case CHAR_U:
579      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
580      break;      break;
581    
582        /* \g must be followed by one of a number of specific things:
583    
584        (1) A number, either plain or braced. If positive, it is an absolute
585        backreference. If negative, it is a relative backreference. This is a Perl
586        5.10 feature.
587    
588        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
589        is part of Perl's movement towards a unified syntax for back references. As
590        this is synonymous with \k{name}, we fudge it up by pretending it really
591        was \k.
592    
593        (3) For Oniguruma compatibility we also support \g followed by a name or a
594        number either in angle brackets or in single quotes. However, these are
595        (possibly recursive) subroutine calls, _not_ backreferences. Just return
596        the -ESC_g code (cf \k). */
597    
598        case CHAR_g:
599        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
600          {
601          c = -ESC_g;
602          break;
603          }
604    
605        /* Handle the Perl-compatible cases */
606    
607        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
608          {
609          const uschar *p;
610          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
611            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
612          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
613            {
614            c = -ESC_k;
615            break;
616            }
617          braced = TRUE;
618          ptr++;
619          }
620        else braced = FALSE;
621    
622        if (ptr[1] == CHAR_MINUS)
623          {
624          negated = TRUE;
625          ptr++;
626          }
627        else negated = FALSE;
628    
629        c = 0;
630        while ((digitab[ptr[1]] & ctype_digit) != 0)
631          c = c * 10 + *(++ptr) - CHAR_0;
632    
633        if (c < 0)   /* Integer overflow */
634          {
635          *errorcodeptr = ERR61;
636          break;
637          }
638    
639        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
640          {
641          *errorcodeptr = ERR57;
642          break;
643          }
644    
645        if (c == 0)
646          {
647          *errorcodeptr = ERR58;
648          break;
649          }
650    
651        if (negated)
652          {
653          if (c > bracount)
654            {
655            *errorcodeptr = ERR15;
656            break;
657            }
658          c = bracount - (c - 1);
659          }
660    
661        c = -(ESC_REF + c);
662        break;
663    
664      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
665      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
666      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 673  else
673      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
674      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
675    
676      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:
677      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
678    
679      if (!isclass)      if (!isclass)
680        {        {
681        oldptr = ptr;        oldptr = ptr;
682        c -= '0';        c -= CHAR_0;
683        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
684          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
685          if (c < 0)    /* Integer overflow */
686            {
687            *errorcodeptr = ERR61;
688            break;
689            }
690        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
691          {          {
692          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 443  else Line 699  else
699      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.
700      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
701    
702      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
703        {        {
704        ptr--;        ptr--;
705        c = 0;        c = 0;
# Line 451  else Line 707  else
707        }        }
708    
709      /* \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
710      larger first octal digit. */      larger first octal digit. The original code used just to take the least
711        significant 8 bits of octal numbers (I think this is what early Perls used
712      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
713      c -= '0';      than 3 octal digits. */
714      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
715          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
716      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
717        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
718            c = c * 8 + *(++ptr) - CHAR_0;
719        if (!utf8 && c > 255) *errorcodeptr = ERR51;
720      break;      break;
721    
722      /* \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
723      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
724        treated as a data character. */
725    
726      case 'x':      case CHAR_x:
727  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
728        {        {
729        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
730        register int count = 0;        int count = 0;
731    
732        c = 0;        c = 0;
733        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
734          {          {
735          int cc = *pt++;          register int cc = *pt++;
736            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
737          count++;          count++;
738  #if !EBCDIC    /* ASCII coding */  
739          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
740          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
741  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
742          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
743          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
744            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
745  #endif  #endif
746          }          }
747        if (*pt == '}')  
748          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
749          {          {
750          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
751          ptr = pt;          ptr = pt;
752          break;          break;
753          }          }
754    
755        /* 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
756        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
757        }        }
 #endif  
758    
759      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
760    
761      c = 0;      c = 0;
762      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
763        {        {
764        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
765        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
766  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
767        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
768        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
769  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
770        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
771        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
772  #endif  #endif
773        }        }
774      break;      break;
775    
776      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
777        This coding is ASCII-specific, but then the whole concept of \cx is
778        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
779    
780      case 'c':      case CHAR_c:
781      c = *(++ptr);      c = *(++ptr);
782      if (c == 0)      if (c == 0)
783        {        {
784        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
785        return 0;        break;
786        }        }
787    
788      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
789      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;  
790      c ^= 0x40;      c ^= 0x40;
791  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
792      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
793      c ^= 0xC0;      c ^= 0xC0;
794  #endif  #endif
795      break;      break;
796    
797      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
798      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
799      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
800      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
801      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
802    
803      default:      default:
804      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 569  escape sequence. Line 830  escape sequence.
830  Argument:  Argument:
831    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
832    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
833      dptr           points to an int that is set to the detailed property value
834    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
835    
836  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
837  */  */
838    
839  static int  static int
840  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
841  {  {
842  int c, i, bot, top;  int c, i, bot, top;
843  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
844  char name[4];  char name[32];
845    
846  c = *(++ptr);  c = *(++ptr);
847  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
848    
849  *negptr = FALSE;  *negptr = FALSE;
850    
851  /* \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
852  preceded by ^ for negation. */  negation. */
853    
854  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
855    {    {
856    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
857      {      {
858      *negptr = TRUE;      *negptr = TRUE;
859      ptr++;      ptr++;
860      }      }
861    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
862      {      {
863      c = *(++ptr);      c = *(++ptr);
864      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
865      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
866      name[i] = c;      name[i] = c;
867      }      }
868    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;  
     }  
869    name[i] = 0;    name[i] = 0;
870    }    }
871    
# Line 628  top = _pcre_utt_size; Line 886  top = _pcre_utt_size;
886    
887  while (bot < top)  while (bot < top)
888    {    {
889    i = (bot + top)/2;    i = (bot + top) >> 1;
890    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
891    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
892        {
893        *dptr = _pcre_utt[i].value;
894        return _pcre_utt[i].type;
895        }
896    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
897    }    }
898    
 UNKNOWN_RETURN:  
899  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return -1;  return -1;
# Line 669  is_counted_repeat(const uschar *p) Line 930  is_counted_repeat(const uschar *p)
930  {  {
931  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
932  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
933  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
934    
935  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
936  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
937    
938  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
939  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
940    
941  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
942  }  }
943    
944    
# Line 710  int max = -1; Line 971  int max = -1;
971  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
972  an integer overflow. */  an integer overflow. */
973    
974  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
975  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
976    {    {
977    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 720  if (min < 0 || min > 65535) Line 981  if (min < 0 || min > 65535)
981  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
982  Also, max must not be less than min. */  Also, max must not be less than min. */
983    
984  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
985    {    {
986    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
987      {      {
988      max = 0;      max = 0;
989      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
990      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
991        {        {
992        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 750  return p; Line 1011  return p;
1011    
1012    
1013  /*************************************************  /*************************************************
1014    *  Subroutine for finding forward reference      *
1015    *************************************************/
1016    
1017    /* This recursive function is called only from find_parens() below. The
1018    top-level call starts at the beginning of the pattern. All other calls must
1019    start at a parenthesis. It scans along a pattern's text looking for capturing
1020    subpatterns, and counting them. If it finds a named pattern that matches the
1021    name it is given, it returns its number. Alternatively, if the name is NULL, it
1022    returns when it reaches a given numbered subpattern. We know that if (?P< is
1023    encountered, the name will be terminated by '>' because that is checked in the
1024    first pass. Recursion is used to keep track of subpatterns that reset the
1025    capturing group numbers - the (?| feature.
1026    
1027    Arguments:
1028      ptrptr       address of the current character pointer (updated)
1029      cd           compile background data
1030      name         name to seek, or NULL if seeking a numbered subpattern
1031      lorn         name length, or subpattern number if name is NULL
1032      xmode        TRUE if we are in /x mode
1033      count        pointer to the current capturing subpattern number (updated)
1034    
1035    Returns:       the number of the named subpattern, or -1 if not found
1036    */
1037    
1038    static int
1039    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1040      BOOL xmode, int *count)
1041    {
1042    uschar *ptr = *ptrptr;
1043    int start_count = *count;
1044    int hwm_count = start_count;
1045    BOOL dup_parens = FALSE;
1046    
1047    /* If the first character is a parenthesis, check on the type of group we are
1048    dealing with. The very first call may not start with a parenthesis. */
1049    
1050    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1051      {
1052      if (ptr[1] == CHAR_QUESTION_MARK &&
1053          ptr[2] == CHAR_VERTICAL_LINE)
1054        {
1055        ptr += 3;
1056        dup_parens = TRUE;
1057        }
1058    
1059      /* Handle a normal, unnamed capturing parenthesis */
1060    
1061      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1062        {
1063        *count += 1;
1064        if (name == NULL && *count == lorn) return *count;
1065        ptr++;
1066        }
1067    
1068      /* Handle a condition. If it is an assertion, just carry on so that it
1069      is processed as normal. If not, skip to the closing parenthesis of the
1070      condition (there can't be any nested parens. */
1071    
1072      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1073        {
1074        ptr += 2;
1075        if (ptr[1] != CHAR_QUESTION_MARK)
1076          {
1077          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1078          if (*ptr != 0) ptr++;
1079          }
1080        }
1081    
1082      /* We have either (? or (* and not a condition */
1083    
1084      else
1085        {
1086        ptr += 2;
1087        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1088    
1089        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1090    
1091        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1092            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1093          {
1094          int term;
1095          const uschar *thisname;
1096          *count += 1;
1097          if (name == NULL && *count == lorn) return *count;
1098          term = *ptr++;
1099          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1100          thisname = ptr;
1101          while (*ptr != term) ptr++;
1102          if (name != NULL && lorn == ptr - thisname &&
1103              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1104            return *count;
1105          term++;
1106          }
1107        }
1108      }
1109    
1110    /* Past any initial parenthesis handling, scan for parentheses or vertical
1111    bars. */
1112    
1113    for (; *ptr != 0; ptr++)
1114      {
1115      /* Skip over backslashed characters and also entire \Q...\E */
1116    
1117      if (*ptr == CHAR_BACKSLASH)
1118        {
1119        if (*(++ptr) == 0) goto FAIL_EXIT;
1120        if (*ptr == CHAR_Q) for (;;)
1121          {
1122          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1123          if (*ptr == 0) goto FAIL_EXIT;
1124          if (*(++ptr) == CHAR_E) break;
1125          }
1126        continue;
1127        }
1128    
1129      /* Skip over character classes; this logic must be similar to the way they
1130      are handled for real. If the first character is '^', skip it. Also, if the
1131      first few characters (either before or after ^) are \Q\E or \E we skip them
1132      too. This makes for compatibility with Perl. Note the use of STR macros to
1133      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1134    
1135      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1136        {
1137        BOOL negate_class = FALSE;
1138        for (;;)
1139          {
1140          if (ptr[1] == CHAR_BACKSLASH)
1141            {
1142            if (ptr[2] == CHAR_E)
1143              ptr+= 2;
1144            else if (strncmp((const char *)ptr+2,
1145                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1146              ptr += 4;
1147            else
1148              break;
1149            }
1150          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1151            {
1152            negate_class = TRUE;
1153            ptr++;
1154            }
1155          else break;
1156          }
1157    
1158        /* If the next character is ']', it is a data character that must be
1159        skipped, except in JavaScript compatibility mode. */
1160    
1161        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1162            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1163          ptr++;
1164    
1165        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1166          {
1167          if (*ptr == 0) return -1;
1168          if (*ptr == CHAR_BACKSLASH)
1169            {
1170            if (*(++ptr) == 0) goto FAIL_EXIT;
1171            if (*ptr == CHAR_Q) for (;;)
1172              {
1173              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1174              if (*ptr == 0) goto FAIL_EXIT;
1175              if (*(++ptr) == CHAR_E) break;
1176              }
1177            continue;
1178            }
1179          }
1180        continue;
1181        }
1182    
1183      /* Skip comments in /x mode */
1184    
1185      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1186        {
1187        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1188        if (*ptr == 0) goto FAIL_EXIT;
1189        continue;
1190        }
1191    
1192      /* Check for the special metacharacters */
1193    
1194      if (*ptr == CHAR_LEFT_PARENTHESIS)
1195        {
1196        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1197        if (rc > 0) return rc;
1198        if (*ptr == 0) goto FAIL_EXIT;
1199        }
1200    
1201      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1202        {
1203        if (dup_parens && *count < hwm_count) *count = hwm_count;
1204        *ptrptr = ptr;
1205        return -1;
1206        }
1207    
1208      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1209        {
1210        if (*count > hwm_count) hwm_count = *count;
1211        *count = start_count;
1212        }
1213      }
1214    
1215    FAIL_EXIT:
1216    *ptrptr = ptr;
1217    return -1;
1218    }
1219    
1220    
1221    
1222    
1223    /*************************************************
1224    *       Find forward referenced subpattern       *
1225    *************************************************/
1226    
1227    /* This function scans along a pattern's text looking for capturing
1228    subpatterns, and counting them. If it finds a named pattern that matches the
1229    name it is given, it returns its number. Alternatively, if the name is NULL, it
1230    returns when it reaches a given numbered subpattern. This is used for forward
1231    references to subpatterns. We used to be able to start this scan from the
1232    current compiling point, using the current count value from cd->bracount, and
1233    do it all in a single loop, but the addition of the possibility of duplicate
1234    subpattern numbers means that we have to scan from the very start, in order to
1235    take account of such duplicates, and to use a recursive function to keep track
1236    of the different types of group.
1237    
1238    Arguments:
1239      cd           compile background data
1240      name         name to seek, or NULL if seeking a numbered subpattern
1241      lorn         name length, or subpattern number if name is NULL
1242      xmode        TRUE if we are in /x mode
1243    
1244    Returns:       the number of the found subpattern, or -1 if not found
1245    */
1246    
1247    static int
1248    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1249    {
1250    uschar *ptr = (uschar *)cd->start_pattern;
1251    int count = 0;
1252    int rc;
1253    
1254    /* If the pattern does not start with an opening parenthesis, the first call
1255    to find_parens_sub() will scan right to the end (if necessary). However, if it
1256    does start with a parenthesis, find_parens_sub() will return when it hits the
1257    matching closing parens. That is why we have to have a loop. */
1258    
1259    for (;;)
1260      {
1261      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1262      if (rc > 0 || *ptr++ == 0) break;
1263      }
1264    
1265    return rc;
1266    }
1267    
1268    
1269    
1270    
1271    /*************************************************
1272  *      Find first significant op code            *  *      Find first significant op code            *
1273  *************************************************/  *************************************************/
1274    
# Line 798  for (;;) Line 1317  for (;;)
1317    
1318      case OP_CALLOUT:      case OP_CALLOUT:
1319      case OP_CREF:      case OP_CREF:
1320      case OP_BRANUMBER:      case OP_NCREF:
1321        case OP_RREF:
1322        case OP_NRREF:
1323        case OP_DEF:
1324      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1325      break;      break;
1326    
# Line 813  for (;;) Line 1335  for (;;)
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1339  *************************************************/  *************************************************/
1340    
1341  /* 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,
1342  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.
1343  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
1344    temporarily terminated with OP_END when this function is called.
1345    
1346    This function is called when a backward assertion is encountered, so that if it
1347    fails, the error message can point to the correct place in the pattern.
1348    However, we cannot do this when the assertion contains subroutine calls,
1349    because they can be forward references. We solve this by remembering this case
1350    and doing the check at the end; a flag specifies which mode we are running in.
1351    
1352  Arguments:  Arguments:
1353    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1354    options  the compiling options    options  the compiling options
1355      atend    TRUE if called when the pattern is complete
1356      cd       the "compile data" structure
1357    
1358  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1359                 or -1 if there is no fixed length,
1360               or -2 if \C was encountered               or -2 if \C was encountered
1361                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1362  */  */
1363    
1364  static int  static int
1365  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1366  {  {
1367  int length = -1;  int length = -1;
1368    
# Line 842  branch, check the length against that of Line 1375  branch, check the length against that of
1375  for (;;)  for (;;)
1376    {    {
1377    int d;    int d;
1378      uschar *ce, *cs;
1379    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1380    switch (op)    switch (op)
1381      {      {
1382        case OP_CBRA:
1383      case OP_BRA:      case OP_BRA:
1384      case OP_ONCE:      case OP_ONCE:
1385      case OP_COND:      case OP_COND:
1386      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1387      if (d < 0) return d;      if (d < 0) return d;
1388      branchlength += d;      branchlength += d;
1389      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 872  for (;;) Line 1405  for (;;)
1405      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1406      branchlength = 0;      branchlength = 0;
1407      break;      break;
1408    
1409        /* A true recursion implies not fixed length, but a subroutine call may
1410        be OK. If the subroutine is a forward reference, we can't deal with
1411        it until the end of the pattern, so return -3. */
1412    
1413        case OP_RECURSE:
1414        if (!atend) return -3;
1415        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1416        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1417        if (cc > cs && cc < ce) return -1;                /* Recursion */
1418        d = find_fixedlength(cs + 2, options, atend, cd);
1419        if (d < 0) return d;
1420        branchlength += d;
1421        cc += 1 + LINK_SIZE;
1422        break;
1423    
1424      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1425    
# Line 885  for (;;) Line 1433  for (;;)
1433      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1434    
1435      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438        case OP_RREF:
1439        case OP_NRREF:
1440        case OP_DEF:
1441      case OP_OPT:      case OP_OPT:
1442      case OP_CALLOUT:      case OP_CALLOUT:
1443      case OP_SOD:      case OP_SOD:
# Line 904  for (;;) Line 1455  for (;;)
1455    
1456      case OP_CHAR:      case OP_CHAR:
1457      case OP_CHARNC:      case OP_CHARNC:
1458        case OP_NOT:
1459      branchlength++;      branchlength++;
1460      cc += 2;      cc += 2;
1461  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1462      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1463        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1464  #endif  #endif
1465      break;      break;
1466    
# Line 921  for (;;) Line 1471  for (;;)
1471      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1472      cc += 4;      cc += 4;
1473  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1474      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1475        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1476  #endif  #endif
1477      break;      break;
1478    
1479      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1482      cc += 4;      cc += 4;
1483      break;      break;
1484    
# Line 937  for (;;) Line 1486  for (;;)
1486    
1487      case OP_PROP:      case OP_PROP:
1488      case OP_NOTPROP:      case OP_NOTPROP:
1489      cc++;      cc += 2;
1490      /* Fall through */      /* Fall through */
1491    
1492      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 947  for (;;) Line 1496  for (;;)
1496      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1497      case OP_WORDCHAR:      case OP_WORDCHAR:
1498      case OP_ANY:      case OP_ANY:
1499        case OP_ALLANY:
1500      branchlength++;      branchlength++;
1501      cc++;      cc++;
1502      break;      break;
# Line 1001  for (;;) Line 1551  for (;;)
1551    
1552    
1553  /*************************************************  /*************************************************
1554  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1555  *************************************************/  *************************************************/
1556    
1557  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1558  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1559    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1560    so that it can be called from pcre_study() when finding the minimum matching
1561    length.
1562    
1563  Arguments:  Arguments:
1564    code        points to start of expression    code        points to start of expression
1565    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1566    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1567    
1568  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
1569  */  */
1570    
1571  static const uschar *  const uschar *
1572  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1573  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1574  for (;;)  for (;;)
1575    {    {
1576    register int c = *code;    register int c = *code;
1577    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1578    else if (c > OP_BRA)  
1579      /* XCLASS is used for classes that cannot be represented just by a bit
1580      map. This includes negated single high-valued characters. The length in
1581      the table is zero; the actual length is stored in the compiled code. */
1582    
1583      if (c == OP_XCLASS) code += GET(code, 1);
1584    
1585      /* Handle recursion */
1586    
1587      else if (c == OP_REVERSE)
1588        {
1589        if (number < 0) return (uschar *)code;
1590        code += _pcre_OP_lengths[c];
1591        }
1592    
1593      /* Handle capturing bracket */
1594    
1595      else if (c == OP_CBRA)
1596      {      {
1597      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1598      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1599      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1600      }      }
1601    
1602      /* Otherwise, we can get the item's length from the table, except that for
1603      repeated character types, we have to test for \p and \P, which have an extra
1604      two bytes of parameters. */
1605    
1606    else    else
1607      {      {
1608      code += _pcre_OP_lengths[c];      switch(c)
1609          {
1610          case OP_TYPESTAR:
1611          case OP_TYPEMINSTAR:
1612          case OP_TYPEPLUS:
1613          case OP_TYPEMINPLUS:
1614          case OP_TYPEQUERY:
1615          case OP_TYPEMINQUERY:
1616          case OP_TYPEPOSSTAR:
1617          case OP_TYPEPOSPLUS:
1618          case OP_TYPEPOSQUERY:
1619          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1620          break;
1621    
1622  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1623          case OP_TYPEMINUPTO:
1624          case OP_TYPEEXACT:
1625          case OP_TYPEPOSUPTO:
1626          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1627          break;
1628          }
1629    
1630      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1631      by a multi-byte character. The length in the table is a minimum, so we have  
1632      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
     can use relatively efficient code. */  
1633    
1634      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1635      a multi-byte character. The length in the table is a minimum, so we have to
1636      arrange to skip the extra bytes. */
1637    
1638    #ifdef SUPPORT_UTF8
1639      if (utf8) switch(c)      if (utf8) switch(c)
1640        {        {
1641        case OP_CHAR:        case OP_CHAR:
# Line 1051  for (;;) Line 1643  for (;;)
1643        case OP_EXACT:        case OP_EXACT:
1644        case OP_UPTO:        case OP_UPTO:
1645        case OP_MINUPTO:        case OP_MINUPTO:
1646          case OP_POSUPTO:
1647        case OP_STAR:        case OP_STAR:
1648        case OP_MINSTAR:        case OP_MINSTAR:
1649          case OP_POSSTAR:
1650        case OP_PLUS:        case OP_PLUS:
1651        case OP_MINPLUS:        case OP_MINPLUS:
1652          case OP_POSPLUS:
1653        case OP_QUERY:        case OP_QUERY:
1654        case OP_MINQUERY:        case OP_MINQUERY:
1655        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1656        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;  
1657        break;        break;
1658        }        }
1659    #else
1660        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1661  #endif  #endif
1662      }      }
1663    }    }
# Line 1092  Returns:      pointer to the opcode for Line 1682  Returns:      pointer to the opcode for
1682  static const uschar *  static const uschar *
1683  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1684  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1685  for (;;)  for (;;)
1686    {    {
1687    register int c = *code;    register int c = *code;
1688    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1689    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
   else if (c > OP_BRA)  
     {  
     code += _pcre_OP_lengths[OP_BRA];  
     }  
   else  
     {  
     code += _pcre_OP_lengths[c];  
1690    
1691  #ifdef SUPPORT_UTF8    /* XCLASS is used for classes that cannot be represented just by a bit
1692      map. This includes negated single high-valued characters. The length in
1693      the table is zero; the actual length is stored in the compiled code. */
1694    
1695      if (c == OP_XCLASS) code += GET(code, 1);
1696    
1697      /* Otherwise, we can get the item's length from the table, except that for
1698      repeated character types, we have to test for \p and \P, which have an extra
1699      two bytes of parameters. */
1700    
1701      else
1702        {
1703        switch(c)
1704          {
1705          case OP_TYPESTAR:
1706          case OP_TYPEMINSTAR:
1707          case OP_TYPEPLUS:
1708          case OP_TYPEMINPLUS:
1709          case OP_TYPEQUERY:
1710          case OP_TYPEMINQUERY:
1711          case OP_TYPEPOSSTAR:
1712          case OP_TYPEPOSPLUS:
1713          case OP_TYPEPOSQUERY:
1714          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1715          break;
1716    
1717          case OP_TYPEPOSUPTO:
1718          case OP_TYPEUPTO:
1719          case OP_TYPEMINUPTO:
1720          case OP_TYPEEXACT:
1721          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722          break;
1723          }
1724    
1725        /* Add in the fixed length from the table */
1726    
1727        code += _pcre_OP_lengths[c];
1728    
1729      /* 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
1730      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
1731      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. */  
1732    
1733    #ifdef SUPPORT_UTF8
1734      if (utf8) switch(c)      if (utf8) switch(c)
1735        {        {
1736        case OP_CHAR:        case OP_CHAR:
# Line 1123  for (;;) Line 1738  for (;;)
1738        case OP_EXACT:        case OP_EXACT:
1739        case OP_UPTO:        case OP_UPTO:
1740        case OP_MINUPTO:        case OP_MINUPTO:
1741          case OP_POSUPTO:
1742        case OP_STAR:        case OP_STAR:
1743        case OP_MINSTAR:        case OP_MINSTAR:
1744          case OP_POSSTAR:
1745        case OP_PLUS:        case OP_PLUS:
1746        case OP_MINPLUS:        case OP_MINPLUS:
1747          case OP_POSPLUS:
1748        case OP_QUERY:        case OP_QUERY:
1749        case OP_MINQUERY:        case OP_MINQUERY:
1750        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1751        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;  
1752        break;        break;
1753        }        }
1754    #else
1755        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1756  #endif  #endif
1757      }      }
1758    }    }
# Line 1152  for (;;) Line 1765  for (;;)
1765  *************************************************/  *************************************************/
1766    
1767  /* 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
1768  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()
1769  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
1770  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
1771  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1772    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1773    bracket whose current branch will already have been scanned.
1774    
1775  Arguments:  Arguments:
1776    code        points to start of search    code        points to start of search
# Line 1169  static BOOL Line 1784  static BOOL
1784  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1785  {  {
1786  register int c;  register int c;
1787  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);
1788       code < endcode;       code < endcode;
1789       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1790    {    {
# Line 1177  for (code = first_significant_code(code Line 1792  for (code = first_significant_code(code
1792    
1793    c = *code;    c = *code;
1794    
1795    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1796      first_significant_code() with a TRUE final argument. */
1797    
1798      if (c == OP_ASSERT)
1799        {
1800        do code += GET(code, 1); while (*code == OP_ALT);
1801        c = *code;
1802        continue;
1803        }
1804    
1805      /* Groups with zero repeats can of course be empty; skip them. */
1806    
1807      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1808        {
1809        code += _pcre_OP_lengths[c];
1810        do code += GET(code, 1); while (*code == OP_ALT);
1811        c = *code;
1812        continue;
1813        }
1814    
1815      /* For other groups, scan the branches. */
1816    
1817      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1818      {      {
1819      BOOL empty_branch;      BOOL empty_branch;
1820      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1821    
1822      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1823        empty branch, so just skip over the conditional, because it could be empty.
1824        Otherwise, scan the individual branches of the group. */
1825    
1826      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1827        code += GET(code, 1);        code += GET(code, 1);
1828        else
1829          {
1830          empty_branch = FALSE;
1831          do
1832            {
1833            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1834              empty_branch = TRUE;
1835            code += GET(code, 1);
1836            }
1837          while (*code == OP_ALT);
1838          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1839        }        }
1840      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1841      c = *code;      c = *code;
1842        continue;
1843      }      }
1844    
1845    else switch (c)    /* Handle the other opcodes */
1846    
1847      switch (c)
1848      {      {
1849      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1850        cannot be represented just by a bit map. This includes negated single
1851        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1852        actual length is stored in the compiled code, so we must update "code"
1853        here. */
1854    
1855  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1856      case OP_XCLASS:      case OP_XCLASS:
1857      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1858      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1859  #endif  #endif
1860    
# Line 1247  for (code = first_significant_code(code Line 1898  for (code = first_significant_code(code
1898      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1899      case OP_WORDCHAR:      case OP_WORDCHAR:
1900      case OP_ANY:      case OP_ANY:
1901        case OP_ALLANY:
1902      case OP_ANYBYTE:      case OP_ANYBYTE:
1903      case OP_CHAR:      case OP_CHAR:
1904      case OP_CHARNC:      case OP_CHARNC:
1905      case OP_NOT:      case OP_NOT:
1906      case OP_PLUS:      case OP_PLUS:
1907      case OP_MINPLUS:      case OP_MINPLUS:
1908        case OP_POSPLUS:
1909      case OP_EXACT:      case OP_EXACT:
1910      case OP_NOTPLUS:      case OP_NOTPLUS:
1911      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1912        case OP_NOTPOSPLUS:
1913      case OP_NOTEXACT:      case OP_NOTEXACT:
1914      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1915      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1916        case OP_TYPEPOSPLUS:
1917      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1918      return FALSE;      return FALSE;
1919    
1920        /* These are going to continue, as they may be empty, but we have to
1921        fudge the length for the \p and \P cases. */
1922    
1923        case OP_TYPESTAR:
1924        case OP_TYPEMINSTAR:
1925        case OP_TYPEPOSSTAR:
1926        case OP_TYPEQUERY:
1927        case OP_TYPEMINQUERY:
1928        case OP_TYPEPOSQUERY:
1929        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1930        break;
1931    
1932        /* Same for these */
1933    
1934        case OP_TYPEUPTO:
1935        case OP_TYPEMINUPTO:
1936        case OP_TYPEPOSUPTO:
1937        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1938        break;
1939    
1940      /* End of branch */      /* End of branch */
1941    
1942      case OP_KET:      case OP_KET:
# Line 1270  for (code = first_significant_code(code Line 1945  for (code = first_significant_code(code
1945      case OP_ALT:      case OP_ALT:
1946      return TRUE;      return TRUE;
1947    
1948      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1949      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1950    
1951  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1952      case OP_STAR:      case OP_STAR:
1953      case OP_MINSTAR:      case OP_MINSTAR:
1954        case OP_POSSTAR:
1955      case OP_QUERY:      case OP_QUERY:
1956      case OP_MINQUERY:      case OP_MINQUERY:
1957        case OP_POSQUERY:
1958        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1959        break;
1960    
1961      case OP_UPTO:      case OP_UPTO:
1962      case OP_MINUPTO:      case OP_MINUPTO:
1963      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
1964        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1965      break;      break;
1966  #endif  #endif
1967      }      }
# Line 1328  return TRUE; Line 2009  return TRUE;
2009  *************************************************/  *************************************************/
2010    
2011  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2012  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
2013  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2014  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2015    
2016    Originally, this function only recognized a sequence of letters between the
2017    terminators, but it seems that Perl recognizes any sequence of characters,
2018    though of course unknown POSIX names are subsequently rejected. Perl gives an
2019    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2020    didn't consider this to be a POSIX class. Likewise for [:1234:].
2021    
2022    The problem in trying to be exactly like Perl is in the handling of escapes. We
2023    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2024    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2025    below handles the special case of \], but does not try to do any other escape
2026    processing. This makes it different from Perl for cases such as [:l\ower:]
2027    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2028    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2029    I think.
2030    
2031  Argument:  Arguments:
2032    ptr      pointer to the initial [    ptr      pointer to the initial [
2033    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2034    
2035  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2036  */  */
2037    
2038  static BOOL  static BOOL
2039  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2040  {  {
2041  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2042  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2043  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2044    {    {
2045    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2046    return TRUE;      {
2047        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2048        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2049          {
2050          *endptr = ptr;
2051          return TRUE;
2052          }
2053        }
2054    }    }
2055  return FALSE;  return FALSE;
2056  }  }
# Line 1375  Returns:     a value representing the na Line 2075  Returns:     a value representing the na
2075  static int  static int
2076  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2077  {  {
2078    const char *pn = posix_names;
2079  register int yield = 0;  register int yield = 0;
2080  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2081    {    {
2082    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2083      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2084      pn += posix_name_lengths[yield] + 1;
2085    yield++;    yield++;
2086    }    }
2087  return -1;  return -1;
# Line 1394  return -1; Line 2096  return -1;
2096  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2097  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2098  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
2099  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
2100  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
2101  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
2102  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
2103  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2104    OP_END.
2105    
2106    This function has been extended with the possibility of forward references for
2107    recursions and subroutine calls. It must also check the list of such references
2108    for the group we are dealing with. If it finds that one of the recursions in
2109    the current group is on this list, it adjusts the offset in the list, not the
2110    value in the reference (which is a group number).
2111    
2112  Arguments:  Arguments:
2113    group      points to the start of the group    group      points to the start of the group
2114    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2115    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2116    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2117      save_hwm   the hwm forward reference pointer at the start of the group
2118    
2119  Returns:     nothing  Returns:     nothing
2120  */  */
2121    
2122  static void  static void
2123  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2124      uschar *save_hwm)
2125  {  {
2126  uschar *ptr = group;  uschar *ptr = group;
2127    
2128  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2129    {    {
2130    int offset = GET(ptr, 1);    int offset;
2131    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2132    
2133      /* See if this recursion is on the forward reference list. If so, adjust the
2134      reference. */
2135    
2136      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2137        {
2138        offset = GET(hc, 0);
2139        if (cd->start_code + offset == ptr + 1)
2140          {
2141          PUT(hc, 0, offset + adjust);
2142          break;
2143          }
2144        }
2145    
2146      /* Otherwise, adjust the recursion offset if it's after the start of this
2147      group. */
2148    
2149      if (hc >= cd->hwm)
2150        {
2151        offset = GET(ptr, 1);
2152        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2153        }
2154    
2155    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2156    }    }
2157  }  }
# Line 1495  Yield:        TRUE when range returned; Line 2230  Yield:        TRUE when range returned;
2230  */  */
2231    
2232  static BOOL  static BOOL
2233  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2234      unsigned int *odptr)
2235  {  {
2236  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2237    
2238  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2239    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2240    
2241  if (c > d) return FALSE;  if (c > d) return FALSE;
2242    
# Line 1512  next = othercase + 1; Line 2245  next = othercase + 1;
2245    
2246  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2247    {    {
2248    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2249    next++;    next++;
2250    }    }
2251    
# Line 1526  return TRUE; Line 2257  return TRUE;
2257  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2258    
2259    
2260    
2261    /*************************************************
2262    *     Check if auto-possessifying is possible    *
2263    *************************************************/
2264    
2265    /* This function is called for unlimited repeats of certain items, to see
2266    whether the next thing could possibly match the repeated item. If not, it makes
2267    sense to automatically possessify the repeated item.
2268    
2269    Arguments:
2270      op_code       the repeated op code
2271      this          data for this item, depends on the opcode
2272      utf8          TRUE in UTF-8 mode
2273      utf8_char     used for utf8 character bytes, NULL if not relevant
2274      ptr           next character in pattern
2275      options       options bits
2276      cd            contains pointers to tables etc.
2277    
2278    Returns:        TRUE if possessifying is wanted
2279    */
2280    
2281    static BOOL
2282    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2283      const uschar *ptr, int options, compile_data *cd)
2284    {
2285    int next;
2286    
2287    /* Skip whitespace and comments in extended mode */
2288    
2289    if ((options & PCRE_EXTENDED) != 0)
2290      {
2291      for (;;)
2292        {
2293        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2294        if (*ptr == CHAR_NUMBER_SIGN)
2295          {
2296          while (*(++ptr) != 0)
2297            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2298          }
2299        else break;
2300        }
2301      }
2302    
2303    /* If the next item is one that we can handle, get its value. A non-negative
2304    value is a character, a negative value is an escape value. */
2305    
2306    if (*ptr == CHAR_BACKSLASH)
2307      {
2308      int temperrorcode = 0;
2309      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2310      if (temperrorcode != 0) return FALSE;
2311      ptr++;    /* Point after the escape sequence */
2312      }
2313    
2314    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2315      {
2316    #ifdef SUPPORT_UTF8
2317      if (utf8) { GETCHARINC(next, ptr); } else
2318    #endif
2319      next = *ptr++;
2320      }
2321    
2322    else return FALSE;
2323    
2324    /* Skip whitespace and comments in extended mode */
2325    
2326    if ((options & PCRE_EXTENDED) != 0)
2327      {
2328      for (;;)
2329        {
2330        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2331        if (*ptr == CHAR_NUMBER_SIGN)
2332          {
2333          while (*(++ptr) != 0)
2334            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2335          }
2336        else break;
2337        }
2338      }
2339    
2340    /* If the next thing is itself optional, we have to give up. */
2341    
2342    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2343      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2344        return FALSE;
2345    
2346    /* Now compare the next item with the previous opcode. If the previous is a
2347    positive single character match, "item" either contains the character or, if
2348    "item" is greater than 127 in utf8 mode, the character's bytes are in
2349    utf8_char. */
2350    
2351    
2352    /* Handle cases when the next item is a character. */
2353    
2354    if (next >= 0) switch(op_code)
2355      {
2356      case OP_CHAR:
2357    #ifdef SUPPORT_UTF8
2358      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2359    #else
2360      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2361    #endif
2362      return item != next;
2363    
2364      /* For CHARNC (caseless character) we must check the other case. If we have
2365      Unicode property support, we can use it to test the other case of
2366      high-valued characters. */
2367    
2368      case OP_CHARNC:
2369    #ifdef SUPPORT_UTF8
2370      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2371    #endif
2372      if (item == next) return FALSE;
2373    #ifdef SUPPORT_UTF8
2374      if (utf8)
2375        {
2376        unsigned int othercase;
2377        if (next < 128) othercase = cd->fcc[next]; else
2378    #ifdef SUPPORT_UCP
2379        othercase = UCD_OTHERCASE((unsigned int)next);
2380    #else
2381        othercase = NOTACHAR;
2382    #endif
2383        return (unsigned int)item != othercase;
2384        }
2385      else
2386    #endif  /* SUPPORT_UTF8 */
2387      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2388    
2389      /* For OP_NOT, "item" must be a single-byte character. */
2390    
2391      case OP_NOT:
2392      if (item == next) return TRUE;
2393      if ((options & PCRE_CASELESS) == 0) return FALSE;
2394    #ifdef SUPPORT_UTF8
2395      if (utf8)
2396        {
2397        unsigned int othercase;
2398        if (next < 128) othercase = cd->fcc[next]; else
2399    #ifdef SUPPORT_UCP
2400        othercase = UCD_OTHERCASE(next);
2401    #else
2402        othercase = NOTACHAR;
2403    #endif
2404        return (unsigned int)item == othercase;
2405        }
2406      else
2407    #endif  /* SUPPORT_UTF8 */
2408      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2409    
2410      case OP_DIGIT:
2411      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2412    
2413      case OP_NOT_DIGIT:
2414      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2415    
2416      case OP_WHITESPACE:
2417      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2418    
2419      case OP_NOT_WHITESPACE:
2420      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2421    
2422      case OP_WORDCHAR:
2423      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2424    
2425      case OP_NOT_WORDCHAR:
2426      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2427    
2428      case OP_HSPACE:
2429      case OP_NOT_HSPACE:
2430      switch(next)
2431        {
2432        case 0x09:
2433        case 0x20:
2434        case 0xa0:
2435        case 0x1680:
2436        case 0x180e:
2437        case 0x2000:
2438        case 0x2001:
2439        case 0x2002:
2440        case 0x2003:
2441        case 0x2004:
2442        case 0x2005:
2443        case 0x2006:
2444        case 0x2007:
2445        case 0x2008:
2446        case 0x2009:
2447        case 0x200A:
2448        case 0x202f:
2449        case 0x205f:
2450        case 0x3000:
2451        return op_code != OP_HSPACE;
2452        default:
2453        return op_code == OP_HSPACE;
2454        }
2455    
2456      case OP_VSPACE:
2457      case OP_NOT_VSPACE:
2458      switch(next)
2459        {
2460        case 0x0a:
2461        case 0x0b:
2462        case 0x0c:
2463        case 0x0d:
2464        case 0x85:
2465        case 0x2028:
2466        case 0x2029:
2467        return op_code != OP_VSPACE;
2468        default:
2469        return op_code == OP_VSPACE;
2470        }
2471    
2472      default:
2473      return FALSE;
2474      }
2475    
2476    
2477    /* Handle the case when the next item is \d, \s, etc. */
2478    
2479    switch(op_code)
2480      {
2481      case OP_CHAR:
2482      case OP_CHARNC:
2483    #ifdef SUPPORT_UTF8
2484      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2485    #endif
2486      switch(-next)
2487        {
2488        case ESC_d:
2489        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2490    
2491        case ESC_D:
2492        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2493    
2494        case ESC_s:
2495        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2496    
2497        case ESC_S:
2498        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2499    
2500        case ESC_w:
2501        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2502    
2503        case ESC_W:
2504        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2505    
2506        case ESC_h:
2507        case ESC_H:
2508        switch(item)
2509          {
2510          case 0x09:
2511          case 0x20:
2512          case 0xa0:
2513          case 0x1680:
2514          case 0x180e:
2515          case 0x2000:
2516          case 0x2001:
2517          case 0x2002:
2518          case 0x2003:
2519          case 0x2004:
2520          case 0x2005:
2521          case 0x2006:
2522          case 0x2007:
2523          case 0x2008:
2524          case 0x2009:
2525          case 0x200A:
2526          case 0x202f:
2527          case 0x205f:
2528          case 0x3000:
2529          return -next != ESC_h;
2530          default:
2531          return -next == ESC_h;
2532          }
2533    
2534        case ESC_v:
2535        case ESC_V:
2536        switch(item)
2537          {
2538          case 0x0a:
2539          case 0x0b:
2540          case 0x0c:
2541          case 0x0d:
2542          case 0x85:
2543          case 0x2028:
2544          case 0x2029:
2545          return -next != ESC_v;
2546          default:
2547          return -next == ESC_v;
2548          }
2549    
2550        default:
2551        return FALSE;
2552        }
2553    
2554      case OP_DIGIT:
2555      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2556             next == -ESC_h || next == -ESC_v;
2557    
2558      case OP_NOT_DIGIT:
2559      return next == -ESC_d;
2560    
2561      case OP_WHITESPACE:
2562      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2563    
2564      case OP_NOT_WHITESPACE:
2565      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2566    
2567      case OP_HSPACE:
2568      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2569    
2570      case OP_NOT_HSPACE:
2571      return next == -ESC_h;
2572    
2573      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2574      case OP_VSPACE:
2575      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2576    
2577      case OP_NOT_VSPACE:
2578      return next == -ESC_v;
2579    
2580      case OP_WORDCHAR:
2581      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2582    
2583      case OP_NOT_WORDCHAR:
2584      return next == -ESC_w || next == -ESC_d;
2585    
2586      default:
2587      return FALSE;
2588      }
2589    
2590    /* Control does not reach here */
2591    }
2592    
2593    
2594    
2595  /*************************************************  /*************************************************
2596  *           Compile one branch                   *  *           Compile one branch                   *
2597  *************************************************/  *************************************************/
2598    
2599  /* 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
2600  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
2601  bits.  bits. This function is used during the pre-compile phase when we are trying
2602    to find out the amount of memory needed, as well as during the real compile
2603    phase. The value of lengthptr distinguishes the two phases.
2604    
2605  Arguments:  Arguments:
2606    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2607    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2608    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2609    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1544  Arguments: Line 2611  Arguments:
2611    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2612    bcptr          points to current branch chain    bcptr          points to current branch chain
2613    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2614      lengthptr      NULL during the real compile phase
2615                     points to length accumulator during pre-compile phase
2616    
2617  Returns:         TRUE on success  Returns:         TRUE on success
2618                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2619  */  */
2620    
2621  static BOOL  static BOOL
2622  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2623    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2624    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2625  {  {
2626  int repeat_type, op_type;  int repeat_type, op_type;
2627  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1561  int greedy_default, greedy_non_default; Line 2630  int greedy_default, greedy_non_default;
2630  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2631  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2632  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2633  int options = *optionsptr;  int options = *optionsptr;
2634  int after_manual_callout = 0;  int after_manual_callout = 0;
2635    int length_prevgroup = 0;
2636  register int c;  register int c;
2637  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2638    uschar *last_code = code;
2639    uschar *orig_code = code;
2640  uschar *tempcode;  uschar *tempcode;
2641  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2642  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1573  const uschar *ptr = *ptrptr; Line 2644  const uschar *ptr = *ptrptr;
2644  const uschar *tempptr;  const uschar *tempptr;
2645  uschar *previous = NULL;  uschar *previous = NULL;
2646  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2647    uschar *save_hwm = NULL;
2648  uschar classbits[32];  uschar classbits[32];
2649    
2650  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2651  BOOL class_utf8;  BOOL class_utf8;
2652  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2653  uschar *class_utf8data;  uschar *class_utf8data;
2654    uschar *class_utf8data_base;
2655  uschar utf8_char[6];  uschar utf8_char[6];
2656  #else  #else
2657  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2658    uschar *utf8_char = NULL;
2659    #endif
2660    
2661    #ifdef DEBUG
2662    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2663  #endif  #endif
2664    
2665  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1613  req_caseopt = ((options & PCRE_CASELESS) Line 2691  req_caseopt = ((options & PCRE_CASELESS)
2691  for (;; ptr++)  for (;; ptr++)
2692    {    {
2693    BOOL negate_class;    BOOL negate_class;
2694      BOOL should_flip_negation;
2695    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2696    BOOL is_quantifier;    BOOL is_quantifier;
2697      BOOL is_recurse;
2698      BOOL reset_bracount;
2699    int class_charcount;    int class_charcount;
2700    int class_lastchar;    int class_lastchar;
2701    int newoptions;    int newoptions;
2702    int recno;    int recno;
2703      int refsign;
2704    int skipbytes;    int skipbytes;
2705    int subreqbyte;    int subreqbyte;
2706    int subfirstbyte;    int subfirstbyte;
2707      int terminator;
2708    int mclength;    int mclength;
2709    uschar mcbuffer[8];    uschar mcbuffer[8];
2710    
2711    /* Next byte in the pattern */    /* Get next byte in the pattern */
2712    
2713    c = *ptr;    c = *ptr;
2714    
2715      /* If we are in the pre-compile phase, accumulate the length used for the
2716      previous cycle of this loop. */
2717    
2718      if (lengthptr != NULL)
2719        {
2720    #ifdef DEBUG
2721        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2722    #endif
2723        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2724          {
2725          *errorcodeptr = ERR52;
2726          goto FAILED;
2727          }
2728    
2729        /* There is at least one situation where code goes backwards: this is the
2730        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2731        the class is simply eliminated. However, it is created first, so we have to
2732        allow memory for it. Therefore, don't ever reduce the length at this point.
2733        */
2734    
2735        if (code < last_code) code = last_code;
2736    
2737        /* Paranoid check for integer overflow */
2738    
2739        if (OFLOW_MAX - *lengthptr < code - last_code)
2740          {
2741          *errorcodeptr = ERR20;
2742          goto FAILED;
2743          }
2744    
2745        *lengthptr += code - last_code;
2746        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2747    
2748        /* If "previous" is set and it is not at the start of the work space, move
2749        it back to there, in order to avoid filling up the work space. Otherwise,
2750        if "previous" is NULL, reset the current code pointer to the start. */
2751    
2752        if (previous != NULL)
2753          {
2754          if (previous > orig_code)
2755            {
2756            memmove(orig_code, previous, code - previous);
2757            code -= previous - orig_code;
2758            previous = orig_code;
2759            }
2760          }
2761        else code = orig_code;
2762    
2763        /* Remember where this code item starts so we can pick up the length
2764        next time round. */
2765    
2766        last_code = code;
2767        }
2768    
2769      /* In the real compile phase, just check the workspace used by the forward
2770      reference list. */
2771    
2772      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2773        {
2774        *errorcodeptr = ERR52;
2775        goto FAILED;
2776        }
2777    
2778    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
2779    
2780    if (inescq && c != 0)    if (inescq && c != 0)
2781      {      {
2782      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2783        {        {
2784        inescq = FALSE;        inescq = FALSE;
2785        ptr++;        ptr++;
# Line 1643  for (;; ptr++) Line 2789  for (;; ptr++)
2789        {        {
2790        if (previous_callout != NULL)        if (previous_callout != NULL)
2791          {          {
2792          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2793              complete_callout(previous_callout, ptr, cd);
2794          previous_callout = NULL;          previous_callout = NULL;
2795          }          }
2796        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1658  for (;; ptr++) Line 2805  for (;; ptr++)
2805    /* 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
2806    a quantifier. */    a quantifier. */
2807    
2808    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2809      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2810        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2811    
2812    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2813         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2814      {      {
2815      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2816          complete_callout(previous_callout, ptr, cd);
2817      previous_callout = NULL;      previous_callout = NULL;
2818      }      }
2819    
# Line 1673  for (;; ptr++) Line 2822  for (;; ptr++)
2822    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2823      {      {
2824      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2825      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2826        {        {
2827        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2828        on the Macintosh. */          {
2829        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2830        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2831          if (*ptr != 0) continue;
2832    
2833          /* Else fall through to handle end of string */
2834          c = 0;
2835        }        }
2836      }      }
2837    
# Line 1692  for (;; ptr++) Line 2845  for (;; ptr++)
2845    
2846    switch(c)    switch(c)
2847      {      {
2848      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2849        case 0:                        /* The branch terminates at string end */
2850      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2851      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2852      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2853      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2854      *codeptr = code;      *codeptr = code;
2855      *ptrptr = ptr;      *ptrptr = ptr;
2856        if (lengthptr != NULL)
2857          {
2858          if (OFLOW_MAX - *lengthptr < code - last_code)
2859            {
2860            *errorcodeptr = ERR20;
2861            goto FAILED;
2862            }
2863          *lengthptr += code - last_code;   /* To include callout length */
2864          DPRINTF((">> end branch\n"));
2865          }
2866      return TRUE;      return TRUE;
2867    
2868    
2869        /* ===================================================================*/
2870      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2871      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2872    
2873      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2874      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2875        {        {
2876        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1715  for (;; ptr++) Line 2879  for (;; ptr++)
2879      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2880      break;      break;
2881    
2882      case '$':      case CHAR_DOLLAR_SIGN:
2883      previous = NULL;      previous = NULL;
2884      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2885      break;      break;
# Line 1723  for (;; ptr++) Line 2887  for (;; ptr++)
2887      /* 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
2888      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2889    
2890      case '.':      case CHAR_DOT:
2891      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2892      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2893      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2894      previous = code;      previous = code;
2895      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2896      break;      break;
2897    
2898      /* Character classes. If the included characters are all < 255 in value, we  
2899      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2900      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
2901      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
2902      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2903        map as usual, then invert it at the end. However, we use a different opcode
2904        so that data characters > 255 can be handled correctly.
2905    
2906      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2907      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,
2908      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2909      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.
     */  
2910    
2911      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2912        default (Perl) mode, it is treated as a data character. */
2913    
2914        case CHAR_RIGHT_SQUARE_BRACKET:
2915        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2916          {
2917          *errorcodeptr = ERR64;
2918          goto FAILED;
2919          }
2920        goto NORMAL_CHAR;
2921    
2922        case CHAR_LEFT_SQUARE_BRACKET:
2923      previous = code;      previous = code;
2924    
2925      /* 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
2926      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. */
2927    
2928      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2929          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2930            check_posix_syntax(ptr, &tempptr))
2931        {        {
2932        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2933        goto FAILED;        goto FAILED;
2934        }        }
2935    
2936      /* 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,
2937        if the first few characters (either before or after ^) are \Q\E or \E we
2938        skip them too. This makes for compatibility with Perl. */
2939    
2940      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2941        for (;;)
2942        {        {
       negate_class = TRUE;  
2943        c = *(++ptr);        c = *(++ptr);
2944          if (c == CHAR_BACKSLASH)
2945            {
2946            if (ptr[1] == CHAR_E)
2947              ptr++;
2948            else if (strncmp((const char *)ptr+1,
2949                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2950              ptr += 3;
2951            else
2952              break;
2953            }
2954          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2955            negate_class = TRUE;
2956          else break;
2957        }        }
2958      else  
2959        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2960        an initial ']' is taken as a data character -- the code below handles
2961        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2962        [^] must match any character, so generate OP_ALLANY. */
2963    
2964        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2965            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2966        {        {
2967        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2968          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2969          zerofirstbyte = firstbyte;
2970          break;
2971        }        }
2972    
2973        /* If a class contains a negative special such as \S, we need to flip the
2974        negation flag at the end, so that support for characters > 255 works
2975        correctly (they are all included in the class). */
2976    
2977        should_flip_negation = FALSE;
2978    
2979      /* 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
2980      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
2981      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2982    
2983      class_charcount = 0;      class_charcount = 0;
2984      class_lastchar = -1;      class_lastchar = -1;
2985    
2986        /* Initialize the 32-char bit map to all zeros. We build the map in a
2987        temporary bit of memory, in case the class contains only 1 character (less
2988        than 256), because in that case the compiled code doesn't use the bit map.
2989        */
2990    
2991        memset(classbits, 0, 32 * sizeof(uschar));
2992    
2993  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2994      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2995      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2996        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2997  #endif  #endif
2998    
     /* 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));  
   
2999      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3000      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
3001      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. */  
3002    
3003      do      if (c != 0) do
3004        {        {
3005          const uschar *oldptr;
3006    
3007  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3008        if (utf8 && c > 127)        if (utf8 && c > 127)
3009          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3010          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3011          }          }
3012    
3013          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3014          data and reset the pointer. This is so that very large classes that
3015          contain a zillion UTF-8 characters no longer overwrite the work space
3016          (which is on the stack). */
3017    
3018          if (lengthptr != NULL)
3019            {
3020            *lengthptr += class_utf8data - class_utf8data_base;
3021            class_utf8data = class_utf8data_base;
3022            }
3023    
3024  #endif  #endif
3025    
3026        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3027    
3028        if (inescq)        if (inescq)
3029          {          {
3030          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3031            {            {
3032            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3033            ptr++;            ptr++;                            /* Skip the 'E' */
3034            continue;            continue;                         /* Carry on with next */
3035            }            }
3036          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3037          }          }
3038    
3039        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1821  for (;; ptr++) Line 3042  for (;; ptr++)
3042        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3043        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3044    
3045        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3046            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3047            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3048          {          {
3049          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3050          int posix_class, i;          int posix_class, taboffset, tabopt;
3051          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3052            uschar pbits[32];
3053    
3054          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3055            {            {
3056            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3057            goto FAILED;            goto FAILED;
3058            }            }
3059    
3060          ptr += 2;          ptr += 2;
3061          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3062            {            {
3063            local_negate = TRUE;            local_negate = TRUE;
3064              should_flip_negation = TRUE;  /* Note negative special */
3065            ptr++;            ptr++;
3066            }            }
3067    
# Line 1856  for (;; ptr++) Line 3079  for (;; ptr++)
3079          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3080            posix_class = 0;            posix_class = 0;
3081    
3082          /* 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
3083          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
3084          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
3085          white space chars afterwards. */          result into the bit map that is being built. */
3086    
3087          posix_class *= 3;          posix_class *= 3;
3088          for (i = 0; i < 3; i++)  
3089            /* Copy in the first table (always present) */
3090    
3091            memcpy(pbits, cbits + posix_class_maps[posix_class],
3092              32 * sizeof(uschar));
3093    
3094            /* If there is a second table, add or remove it as required. */
3095    
3096            taboffset = posix_class_maps[posix_class + 1];
3097            tabopt = posix_class_maps[posix_class + 2];
3098    
3099            if (taboffset >= 0)
3100            {            {
3101            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3102            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;  
             }  
3103            else            else
3104              {              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;  
             }  
3105            }            }
3106    
3107            /* Not see if we need to remove any special characters. An option
3108            value of 1 removes vertical space and 2 removes underscore. */
3109    
3110            if (tabopt < 0) tabopt = -tabopt;
3111            if (tabopt == 1) pbits[1] &= ~0x3c;
3112              else if (tabopt == 2) pbits[11] &= 0x7f;
3113    
3114            /* Add the POSIX table or its complement into the main table that is
3115            being built and we are done. */
3116    
3117            if (local_negate)
3118              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3119            else
3120              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3121    
3122          ptr = tempptr + 1;          ptr = tempptr + 1;
3123          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3124          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3125          }          }
3126    
3127        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3128        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
3129        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.
3130        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3131        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  
3132        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3133    
3134        if (c == '\\')        if (c == CHAR_BACKSLASH)
3135          {          {
3136          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3137            if (*errorcodeptr != 0) goto FAILED;
3138    
3139          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 */
3140          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 */
3141            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3142          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3143            {            {
3144            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3145              {              {
3146              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3147              }              }
3148            else inescq = TRUE;            else inescq = TRUE;
3149            continue;            continue;
3150            }            }
3151            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3152    
3153          if (c < 0)          if (c < 0)
3154            {            {
3155            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3156            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3157            switch (-c)  
3158              /* Save time by not doing this in the pre-compile phase. */
3159    
3160              if (lengthptr == NULL) switch (-c)
3161              {              {
3162              case ESC_d:              case ESC_d:
3163              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3164              continue;              continue;
3165    
3166              case ESC_D:              case ESC_D:
3167                should_flip_negation = TRUE;
3168              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3169              continue;              continue;
3170    
# Line 1930  for (;; ptr++) Line 3173  for (;; ptr++)
3173              continue;              continue;
3174    
3175              case ESC_W:              case ESC_W:
3176                should_flip_negation = TRUE;
3177              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3178              continue;              continue;
3179    
# Line 1939  for (;; ptr++) Line 3183  for (;; ptr++)
3183              continue;              continue;
3184    
3185              case ESC_S:              case ESC_S:
3186                should_flip_negation = TRUE;
3187              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3188              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3189              continue;              continue;
3190    
3191  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3192              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3193              case ESC_P:              }
3194    
3195              /* In the pre-compile phase, just do the recognition. */
3196    
3197              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3198                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3199    
3200              /* We need to deal with \H, \h, \V, and \v in both phases because
3201              they use extra memory. */
3202    
3203              if (-c == ESC_h)
3204                {
3205                SETBIT(classbits, 0x09); /* VT */
3206                SETBIT(classbits, 0x20); /* SPACE */
3207                SETBIT(classbits, 0xa0); /* NSBP */
3208    #ifdef SUPPORT_UTF8
3209                if (utf8)
3210                  {
3211                  class_utf8 = TRUE;
3212                  *class_utf8data++ = XCL_SINGLE;
3213                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3214                  *class_utf8data++ = XCL_SINGLE;
3215                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3216                  *class_utf8data++ = XCL_RANGE;
3217                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3218                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3219                  *class_utf8data++ = XCL_SINGLE;
3220                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3221                  *class_utf8data++ = XCL_SINGLE;
3222                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3223                  *class_utf8data++ = XCL_SINGLE;
3224                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3225                  }
3226    #endif
3227                continue;
3228                }
3229    
3230              if (-c == ESC_H)
3231                {
3232                for (c = 0; c < 32; c++)
3233                  {
3234                  int x = 0xff;
3235                  switch (c)
3236                    {
3237                    case 0x09/8: x ^= 1 << (0x09%8); break;
3238                    case 0x20/8: x ^= 1 << (0x20%8); break;
3239                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3240                    default: break;
3241                    }
3242                  classbits[c] |= x;
3243                  }
3244    
3245    #ifdef SUPPORT_UTF8
3246                if (utf8)
3247                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3248                class_utf8 = TRUE;                class_utf8 = TRUE;
3249                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3250                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3251                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3252                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3253                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3254                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3255                  *class_utf8data++ = XCL_RANGE;
3256                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3257                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3258                  *class_utf8data++ = XCL_RANGE;
3259                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3260                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3261                  *class_utf8data++ = XCL_RANGE;
3262                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3263                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3264                  *class_utf8data++ = XCL_RANGE;
3265                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3266                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3267                  *class_utf8data++ = XCL_RANGE;
3268                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3269                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3270                }                }
3271    #endif
3272              continue;              continue;
3273                }
3274    
3275              if (-c == ESC_v)
3276                {
3277                SETBIT(classbits, 0x0a); /* LF */
3278                SETBIT(classbits, 0x0b); /* VT */
3279                SETBIT(classbits, 0x0c); /* FF */
3280                SETBIT(classbits, 0x0d); /* CR */
3281                SETBIT(classbits, 0x85); /* NEL */
3282    #ifdef SUPPORT_UTF8
3283                if (utf8)
3284                  {
3285                  class_utf8 = TRUE;
3286                  *class_utf8data++ = XCL_RANGE;
3287                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3288                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3289                  }
3290  #endif  #endif
3291                continue;
3292                }
3293    
3294              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3295              strict mode. By default, for compatibility with Perl, they are              {
3296              treated as literals. */              for (c = 0; c < 32; c++)
3297                  {
3298                  int x = 0xff;
3299                  switch (c)
3300                    {
3301                    case 0x0a/8: x ^= 1 << (0x0a%8);
3302                                 x ^= 1 << (0x0b%8);
3303                                 x ^= 1 << (0x0c%8);
3304                                 x ^= 1 << (0x0d%8);
3305                                 break;
3306                    case 0x85/8: x ^= 1 << (0x85%8); break;
3307                    default: break;
3308                    }
3309                  classbits[c] |= x;
3310                  }
3311    
3312              default:  #ifdef SUPPORT_UTF8
3313              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3314                {                {
3315                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3316                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3317                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3318                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3319                  *class_utf8data++ = XCL_RANGE;
3320                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3321                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3322                }                }
3323              c = *ptr;              /* The final character */  #endif
3324              class_charcount -= 2;  /* Undo the default count from above */              continue;
3325                }
3326    
3327              /* We need to deal with \P and \p in both phases. */
3328    
3329    #ifdef SUPPORT_UCP
3330              if (-c == ESC_p || -c == ESC_P)
3331                {
3332                BOOL negated;
3333                int pdata;
3334                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3335                if (ptype < 0) goto FAILED;
3336                class_utf8 = TRUE;
3337                *class_utf8data++ = ((-c == ESC_p) != negated)?
3338                  XCL_PROP : XCL_NOTPROP;
3339                *class_utf8data++ = ptype;
3340                *class_utf8data++ = pdata;
3341                class_charcount -= 2;   /* Not a < 256 character */
3342                continue;
3343                }
3344    #endif
3345              /* Unrecognized escapes are faulted if PCRE is running in its
3346              strict mode. By default, for compatibility with Perl, they are
3347              treated as literals. */
3348    
3349              if ((options & PCRE_EXTRA) != 0)
3350                {
3351                *errorcodeptr = ERR7;
3352                goto FAILED;
3353              }              }
3354    
3355              class_charcount -= 2;  /* Undo the default count from above */
3356              c = *ptr;              /* Get the final character and fall through */
3357            }            }
3358    
3359          /* 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
3360          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3361    
3362          }   /* End of backslash handling */          }   /* End of backslash handling */
3363    
3364        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3365        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
3366        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3367          entirely. The code for handling \Q and \E is messy. */
3368    
3369          CHECK_RANGE:
3370          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3371            {
3372            inescq = FALSE;
3373            ptr += 2;
3374            }
3375    
3376          oldptr = ptr;
3377    
3378        if (ptr[1] == '-' && ptr[2] != ']')        /* Remember \r or \n */
3379    
3380          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3381    
3382          /* Check for range */
3383    
3384          if (!inescq && ptr[1] == CHAR_MINUS)
3385          {          {
3386          int d;          int d;
3387          ptr += 2;          ptr += 2;
3388            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3389    
3390            /* If we hit \Q (not followed by \E) at this point, go into escaped
3391            mode. */
3392    
3393            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3394              {
3395              ptr += 2;
3396              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3397                { ptr += 2; continue; }
3398              inescq = TRUE;
3399              break;
3400              }
3401    
3402            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3403              {
3404              ptr = oldptr;
3405              goto LONE_SINGLE_CHARACTER;
3406              }
3407    
3408  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3409          if (utf8)          if (utf8)
# Line 2001  for (;; ptr++) Line 3418  for (;; ptr++)
3418          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
3419          in such circumstances. */          in such circumstances. */
3420    
3421          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3422            {            {
3423            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3424            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3425    
3426            /* \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
3427            was literal */            special means the '-' was literal */
3428    
3429            if (d < 0)            if (d < 0)
3430              {              {
3431              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3432              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3433                else if (d == -ESC_R) d = CHAR_R; else
3434                {                {
3435                ptr = oldptr - 2;                ptr = oldptr;
3436                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3437                }                }
3438              }              }
3439            }            }
3440    
3441          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3442          the pre-pass. Optimize one-character ranges */          one-character ranges */
3443    
3444            if (d < c)
3445              {
3446              *errorcodeptr = ERR8;
3447              goto FAILED;
3448              }
3449    
3450          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3451    
3452            /* Remember \r or \n */
3453    
3454            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3455    
3456          /* 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
3457          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3458          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2042  for (;; ptr++) Line 3470  for (;; ptr++)
3470  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3471            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3472              {              {
3473              int occ, ocd;              unsigned int occ, ocd;
3474              int cc = c;              unsigned int cc = c;
3475              int origd = d;              unsigned int origd = d;
3476              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3477                {                {
3478                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3479                      ocd <= (unsigned int)d)
3480                    continue;                          /* Skip embedded ranges */
3481    
3482                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3483                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3484                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3485                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3486                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3487                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3488                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3489                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3490                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3491                  d = ocd;                  d = ocd;
3492                  continue;                  continue;
# Line 2102  for (;; ptr++) Line 3534  for (;; ptr++)
3534          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
3535          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3536    
3537          for (; c <= d; c++)          class_charcount += d - c + 1;
3538            class_lastchar = d;
3539    
3540            /* We can save a bit of time by skipping this in the pre-compile. */
3541    
3542            if (lengthptr == NULL) for (; c <= d; c++)
3543            {            {
3544            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3545            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2110  for (;; ptr++) Line 3547  for (;; ptr++)
3547              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3548              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3549              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3550            }            }
3551    
3552          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2135  for (;; ptr++) Line 3570  for (;; ptr++)
3570  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3571          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3572            {            {
3573            int chartype;            unsigned int othercase;
3574            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3575              {              {
3576              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3577              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2163  for (;; ptr++) Line 3596  for (;; ptr++)
3596          }          }
3597        }        }
3598    
3599      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3600      loop. This "while" is the end of the "do" above. */  
3601        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3602    
3603        if (c == 0)                          /* Missing terminating ']' */
3604          {
3605          *errorcodeptr = ERR6;
3606          goto FAILED;
3607          }
3608    
3609    
3610    /* This code has been disabled because it would mean that \s counts as
3611    an explicit \r or \n reference, and that's not really what is wanted. Now
3612    we set the flag only if there is a literal "\r" or "\n" in the class. */
3613    
3614    #if 0
3615        /* Remember whether \r or \n are in this class */
3616    
3617        if (negate_class)
3618          {
3619          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3620          }
3621        else
3622          {
3623          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3624          }
3625    #endif
3626    
     while ((c = *(++ptr)) != ']' || inescq);  
3627    
3628      /* 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
3629      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
3630      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
3631      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3632      single-bytes only. This is an historical hangover. Maybe one day we can  
3633      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3634        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3635        operate on single-bytes only. This is an historical hangover. Maybe one day
3636        we can tidy these opcodes to handle multi-byte characters.
3637    
3638      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
3639      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 2183  for (;; ptr++) Line 3643  for (;; ptr++)
3643      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3644    
3645  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3646      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3647            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3648  #else  #else
3649      if (class_charcount == 1)      if (class_charcount == 1)
3650  #endif  #endif
# Line 2229  for (;; ptr++) Line 3687  for (;; ptr++)
3687      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3688    
3689      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3690      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3691      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3692        the class, so any that were explicitly given as well can be ignored. If
3693        (when there are explicit characters > 255 that must be listed) there are no
3694        characters < 256, we can omit the bitmap in the actual compiled code. */
3695    
3696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3697      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3698        {        {
3699        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3700        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3701        code += LINK_SIZE;        code += LINK_SIZE;
3702        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3703    
3704        /* 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;
3705        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3706    
3707        if (class_charcount > 0)        if (class_charcount > 0)
3708          {          {
3709          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3710            memmove(code + 32, code, class_utf8data - code);
3711          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3712          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;  
3713          }          }
3714          else code = class_utf8data;
3715    
3716        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3717    
# Line 2266  for (;; ptr++) Line 3720  for (;; ptr++)
3720        }        }
3721  #endif  #endif
3722    
3723      /* 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
3724      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
3725      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
3726      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3727    
3728        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3729      if (negate_class)      if (negate_class)
3730        {        {
3731        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3732        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3733        }        }
3734      else      else
3735        {        {
       *code++ = OP_CLASS;  
3736        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3737        }        }
3738      code += 32;      code += 32;
3739      break;      break;
3740    
3741    
3742        /* ===================================================================*/
3743      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3744      has been tested above. */      has been tested above. */
3745    
3746      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3747      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3748      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3749      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3750      goto REPEAT;      goto REPEAT;
3751    
3752      case '*':      case CHAR_ASTERISK:
3753      repeat_min = 0;      repeat_min = 0;
3754      repeat_max = -1;      repeat_max = -1;
3755      goto REPEAT;      goto REPEAT;
3756    
3757      case '+':      case CHAR_PLUS:
3758      repeat_min = 1;      repeat_min = 1;
3759      repeat_max = -1;      repeat_max = -1;
3760      goto REPEAT;      goto REPEAT;
3761    
3762      case '?':      case CHAR_QUESTION_MARK:
3763      repeat_min = 0;      repeat_min = 0;
3764      repeat_max = 1;      repeat_max = 1;
3765    
# Line 2338  for (;; ptr++) Line 3794  for (;; ptr++)
3794      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
3795      repeat type to the non-default. */      repeat type to the non-default. */
3796    
3797      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3798        {        {
3799        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3800        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3801        ptr++;        ptr++;
3802        }        }
3803      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3804        {        {
3805        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3806        ptr++;        ptr++;
3807        }        }
3808      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3809    
     /* 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;  
       }  
   
3810      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3811      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
3812      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 2398  for (;; ptr++) Line 3840  for (;; ptr++)
3840          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3841          }          }
3842    
3843          /* If the repetition is unlimited, it pays to see if the next thing on
3844          the line is something that cannot possibly match this character. If so,
3845          automatically possessifying this item gains some performance in the case
3846          where the match fails. */
3847    
3848          if (!possessive_quantifier &&
3849              repeat_max < 0 &&
3850              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3851                options, cd))
3852            {
3853            repeat_type = 0;    /* Force greedy */
3854            possessive_quantifier = TRUE;
3855            }
3856    
3857        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3858        }        }
3859    
3860      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3861      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-
3862      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3863      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3864        currently used only for single-byte chars. */
3865    
3866      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3867        {        {
3868        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3869        c = previous[1];        c = previous[1];
3870          if (!possessive_quantifier &&
3871              repeat_max < 0 &&
3872              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3873            {
3874            repeat_type = 0;    /* Force greedy */
3875            possessive_quantifier = TRUE;
3876            }
3877        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3878        }        }
3879    
# Line 2423  for (;; ptr++) Line 3887  for (;; ptr++)
3887      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3888        {        {
3889        uschar *oldcode;        uschar *oldcode;
3890        int prop_type;        int prop_type, prop_value;
3891        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3892        c = *previous;        c = *previous;
3893    
3894          if (!possessive_quantifier &&
3895              repeat_max < 0 &&
3896              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3897            {
3898            repeat_type = 0;    /* Force greedy */
3899            possessive_quantifier = TRUE;
3900            }
3901    
3902        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3903        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3904          previous[1] : -1;          {
3905            prop_type = previous[1];
3906            prop_value = previous[2];
3907            }
3908          else prop_type = prop_value = -1;
3909    
3910        oldcode = code;        oldcode = code;
3911        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2439  for (;; ptr++) Line 3915  for (;; ptr++)
3915    
3916        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3917    
3918          /*--------------------------------------------------------------------*/
3919          /* This code is obsolete from release 8.00; the restriction was finally
3920          removed: */
3921    
3922        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3923        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3924    
3925        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3926          /*--------------------------------------------------------------------*/
3927    
3928        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3929    
# Line 2463  for (;; ptr++) Line 3944  for (;; ptr++)
3944          }          }
3945    
3946        /* 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
3947        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3948        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
3949        one less than the maximum. */        one less than the maximum. */
3950    
# Line 2490  for (;; ptr++) Line 3971  for (;; ptr++)
3971    
3972          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3973          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
3974          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3975          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3976          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3977    
# Line 2506  for (;; ptr++) Line 3987  for (;; ptr++)
3987  #endif  #endif
3988              {              {
3989              *code++ = c;              *code++ = c;
3990              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3991                  {
3992                  *code++ = prop_type;
3993                  *code++ = prop_value;
3994                  }
3995              }              }
3996            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3997            }            }
3998    
3999          /* 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
4000          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
4001            UPTO is just for 1 instance, we can use QUERY instead. */
4002    
4003          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4004            {            {
# Line 2525  for (;; ptr++) Line 4011  for (;; ptr++)
4011            else            else
4012  #endif  #endif
4013            *code++ = c;            *code++ = c;
4014            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
4015                {
4016                *code++ = prop_type;
4017                *code++ = prop_value;
4018                }
4019            repeat_max -= repeat_min;            repeat_max -= repeat_min;
4020            *code++ = OP_UPTO + repeat_type;  
4021            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
4022                {
4023                *code++ = OP_QUERY + repeat_type;
4024                }
4025              else
4026                {
4027                *code++ = OP_UPTO + repeat_type;
4028                PUT2INC(code, 0, repeat_max);
4029                }
4030            }            }
4031          }          }
4032    
# Line 2544  for (;; ptr++) Line 4042  for (;; ptr++)
4042  #endif  #endif
4043        *code++ = c;        *code++ = c;
4044    
4045        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
4046        defines the required property. */        define the required property. */
4047    
4048  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4049        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
4050            {
4051            *code++ = prop_type;
4052            *code++ = prop_value;
4053            }
4054  #endif  #endif
4055        }        }
4056    
# Line 2568  for (;; ptr++) Line 4070  for (;; ptr++)
4070          goto END_REPEAT;          goto END_REPEAT;
4071          }          }
4072    
4073          /*--------------------------------------------------------------------*/
4074          /* This code is obsolete from release 8.00; the restriction was finally
4075          removed: */
4076    
4077        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4078        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4079    
4080        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4081          /*--------------------------------------------------------------------*/
4082    
4083        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4084          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2591  for (;; ptr++) Line 4098  for (;; ptr++)
4098      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4099      cases. */      cases. */
4100    
4101      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4102               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4103        {        {
4104        register int i;        register int i;
4105        int ketoffset = 0;        int ketoffset = 0;
4106        int len = code - previous;        int len = code - previous;
4107        uschar *bralink = NULL;        uschar *bralink = NULL;
4108    
4109          /* Repeating a DEFINE group is pointless */
4110    
4111          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4112            {
4113            *errorcodeptr = ERR55;
4114            goto FAILED;
4115            }
4116    
4117        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4118        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4119        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2621  for (;; ptr++) Line 4136  for (;; ptr++)
4136    
4137        if (repeat_min == 0)        if (repeat_min == 0)
4138          {          {
4139          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4140          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4141    
4142          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4143          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4144          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4145          any internal group, because the offset is from the start of the whole          **   goto END_REPEAT;
4146          regex. Temporarily terminate the pattern while doing this. */          **   }
4147    
4148            However, that fails when a group is referenced as a subroutine from
4149            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4150            so that it is skipped on execution. As we don't have a list of which
4151            groups are referenced, we cannot do this selectively.
4152    
4153            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4154            and do no more at this point. However, we do need to adjust any