/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 512 by ph10, Tue Mar 30 11:11:52 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         0,                       0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
236    
237  /* 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
238  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
239  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
240    characters are removed, and for [:alpha:] and [:alnum:] the underscore
241    character is removed. The triples in the table consist of the base map offset,
242    second map offset or -1 if no second map, and a non-negative value for map
243    addition or a negative value for map subtraction (if there are two maps). The
244    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
245    remove vertical space characters, 2 => remove underscore. */
246    
247  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
248    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
249    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
250    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
251    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
252    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
253    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
254    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
255    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
256    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
257    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
258    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
259    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
260    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
261    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    
265  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
266  are passed to the outside world. */  #define XSTRING(s) STRING(s)
267    
268  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
269    "no error",  are passed to the outside world. Do not ever re-use any error number, because
270    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
271    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
272    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
273    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
274    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
275    simply count through to the one we want - this isn't a performance issue
276    because these strings are used only when there is a compilation error.
277    
278    Each substring ends with \0 to insert a null character. This includes the final
279    substring, so that the whole string ends with \0\0, which can be detected when
280    counting through. */
281    
282    static const char error_texts[] =
283      "no error\0"
284      "\\ at end of pattern\0"
285      "\\c at end of pattern\0"
286      "unrecognized character follows \\\0"
287      "numbers out of order in {} quantifier\0"
288    /* 5 */    /* 5 */
289    "number too big in {} quantifier",    "number too big in {} quantifier\0"
290    "missing terminating ] for character class",    "missing terminating ] for character class\0"
291    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
292    "range out of order in character class",    "range out of order in character class\0"
293    "nothing to repeat",    "nothing to repeat\0"
294    /* 10 */    /* 10 */
295    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
296    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
297    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
298    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
299    "missing )",    "missing )\0"
300    /* 15 */    /* 15 */
301    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
302    "erroffset passed as NULL",    "erroffset passed as NULL\0"
303    "unknown option bit(s) set",    "unknown option bit(s) set\0"
304    "missing ) after comment",    "missing ) after comment\0"
305    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
306    /* 20 */    /* 20 */
307    "regular expression too large",    "regular expression is too large\0"
308    "failed to get memory",    "failed to get memory\0"
309    "unmatched parentheses",    "unmatched parentheses\0"
310    "internal error: code overflow",    "internal error: code overflow\0"
311    "unrecognized character after (?<",    "unrecognized character after (?<\0"
312    /* 25 */    /* 25 */
313    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
314    "malformed number after (?(",    "malformed number or name after (?(\0"
315    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
316    "assertion expected after (?(",    "assertion expected after (?(\0"
317    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
318    /* 30 */    /* 30 */
319    "unknown POSIX class name",    "unknown POSIX class name\0"
320    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
321    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
322    "spare error",    "spare error\0"  /** DEAD **/
323    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
324    /* 35 */    /* 35 */
325    "invalid condition (?(0)",    "invalid condition (?(0)\0"
326    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
327    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
328    "number after (?C is > 255",    "number after (?C is > 255\0"
329    "closing ) for (?C expected",    "closing ) for (?C expected\0"
330    /* 40 */    /* 40 */
331    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
332    "unrecognized character after (?P",    "unrecognized character after (?P\0"
333    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
334    "two named groups have the same name",    "two named subpatterns have the same name\0"
335    "invalid UTF-8 string",    "invalid UTF-8 string\0"
336    /* 45 */    /* 45 */
337    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
338    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
339    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
340  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
341      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
342      /* 50 */
343      "repeated subpattern is too long\0"    /** DEAD **/
344      "octal value is greater than \\377 (not in UTF-8 mode)\0"
345      "internal error: overran compiling workspace\0"
346      "internal error: previously-checked referenced subpattern not found\0"
347      "DEFINE group contains more than one branch\0"
348      /* 55 */
349      "repeating a DEFINE group is not allowed\0"
350      "inconsistent NEWLINE options\0"
351      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
352      "a numbered reference must not be zero\0"
353      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
354      /* 60 */
355      "(*VERB) not recognized\0"
356      "number is too big\0"
357      "subpattern name expected\0"
358      "digit expected after (?+\0"
359      "] is an invalid data character in JavaScript compatibility mode\0"
360      /* 65 */
361      "different names for subpatterns of the same number are not allowed\0"
362      "(*MARK) must have an argument\0"
363      ;
364    
365  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
366  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 378  For convenience, we use the same bit def
378    
379  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
380    
381  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
382    
383    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
384    UTF-8 mode. */
385    
386  static const unsigned char digitab[] =  static const unsigned char digitab[] =
387    {    {
388    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 418  static const unsigned char digitab[] =
418    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
420    
421  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
422    
423    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
424    
425  static const unsigned char digitab[] =  static const unsigned char digitab[] =
426    {    {
427    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 435  static const unsigned char digitab[] =
435    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
439    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
440    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
441    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 469  static const unsigned char ebcdic_charta
469    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
470    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
471    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
472    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
473    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
474    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 496  static const unsigned char ebcdic_charta
496  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
497    
498  static BOOL  static BOOL
499    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
500      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
501    
502    
503    
504  /*************************************************  /*************************************************
505    *            Find an error text                  *
506    *************************************************/
507    
508    /* The error texts are now all in one long string, to save on relocations. As
509    some of the text is of unknown length, we can't use a table of offsets.
510    Instead, just count through the strings. This is not a performance issue
511    because it happens only when there has been a compilation error.
512    
513    Argument:   the error number
514    Returns:    pointer to the error string
515    */
516    
517    static const char *
518    find_error_text(int n)
519    {
520    const char *s = error_texts;
521    for (; n > 0; n--)
522      {
523      while (*s++ != 0) {};
524      if (*s == 0) return "Error text not found (please report)";
525      }
526    return s;
527    }
528    
529    
530    /*************************************************
531  *            Handle escapes                      *  *            Handle escapes                      *
532  *************************************************/  *************************************************/
533    
534  /* 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
535  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
536  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
537  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
538  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,
539    ptr is pointing at the \. On exit, it is on the final character of the escape
540    sequence.
541    
542  Arguments:  Arguments:
543    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 548  Arguments:
548    
549  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
550                   negative => a special escape sequence                   negative => a special escape sequence
551                   on error, errorptr is set                   on error, errorcodeptr is set
552  */  */
553    
554  static int  static int
555  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
556    int options, BOOL isclass)    int options, BOOL isclass)
557  {  {
558  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
559    const uschar *ptr = *ptrptr + 1;
560  int c, i;  int c, i;
561    
562    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
563    ptr--;                            /* Set pointer back to the last byte */
564    
565  /* 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. */
566    
 c = *(++ptr);  
567  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
568    
569  /* 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
570  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.
571  Otherwise further processing may be required. */  Otherwise further processing may be required. */
572    
573  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
574  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
575  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
576    
577  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
578  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
579  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
580  #endif  #endif
581    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 584  else if ((i = escapes[c - 0x48]) != 0)
584  else  else
585    {    {
586    const uschar *oldptr;    const uschar *oldptr;
587      BOOL braced, negated;
588    
589    switch (c)    switch (c)
590      {      {
591      /* 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
592      error. */      error. */
593    
594      case 'l':      case CHAR_l:
595      case 'L':      case CHAR_L:
596      case 'N':      case CHAR_N:
597      case 'u':      case CHAR_u:
598      case 'U':      case CHAR_U:
599      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
600      break;      break;
601    
602        /* \g must be followed by one of a number of specific things:
603    
604        (1) A number, either plain or braced. If positive, it is an absolute
605        backreference. If negative, it is a relative backreference. This is a Perl
606        5.10 feature.
607    
608        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
609        is part of Perl's movement towards a unified syntax for back references. As
610        this is synonymous with \k{name}, we fudge it up by pretending it really
611        was \k.
612    
613        (3) For Oniguruma compatibility we also support \g followed by a name or a
614        number either in angle brackets or in single quotes. However, these are
615        (possibly recursive) subroutine calls, _not_ backreferences. Just return
616        the -ESC_g code (cf \k). */
617    
618        case CHAR_g:
619        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
620          {
621          c = -ESC_g;
622          break;
623          }
624    
625        /* Handle the Perl-compatible cases */
626    
627        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
628          {
629          const uschar *p;
630          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
631            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
632          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
633            {
634            c = -ESC_k;
635            break;
636            }
637          braced = TRUE;
638          ptr++;
639          }
640        else braced = FALSE;
641    
642        if (ptr[1] == CHAR_MINUS)
643          {
644          negated = TRUE;
645          ptr++;
646          }
647        else negated = FALSE;
648    
649        c = 0;
650        while ((digitab[ptr[1]] & ctype_digit) != 0)
651          c = c * 10 + *(++ptr) - CHAR_0;
652    
653        if (c < 0)   /* Integer overflow */
654          {
655          *errorcodeptr = ERR61;
656          break;
657          }
658    
659        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
660          {
661          *errorcodeptr = ERR57;
662          break;
663          }
664    
665        if (c == 0)
666          {
667          *errorcodeptr = ERR58;
668          break;
669          }
670    
671        if (negated)
672          {
673          if (c > bracount)
674            {
675            *errorcodeptr = ERR15;
676            break;
677            }
678          c = bracount - (c - 1);
679          }
680    
681        c = -(ESC_REF + c);
682        break;
683    
684      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
685      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
686      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 693  else
693      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
694      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
695    
696      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:
697      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
698    
699      if (!isclass)      if (!isclass)
700        {        {
701        oldptr = ptr;        oldptr = ptr;
702        c -= '0';        c -= CHAR_0;
703        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
704          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
705          if (c < 0)    /* Integer overflow */
706            {
707            *errorcodeptr = ERR61;
708            break;
709            }
710        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
711          {          {
712          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 719  else
719      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.
720      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
721    
722      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
723        {        {
724        ptr--;        ptr--;
725        c = 0;        c = 0;
# Line 442  else Line 727  else
727        }        }
728    
729      /* \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
730      larger first octal digit. */      larger first octal digit. The original code used just to take the least
731        significant 8 bits of octal numbers (I think this is what early Perls used
732      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
733      c -= '0';      than 3 octal digits. */
734      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
735          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
736      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
737        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
738            c = c * 8 + *(++ptr) - CHAR_0;
739        if (!utf8 && c > 255) *errorcodeptr = ERR51;
740      break;      break;
741    
742      /* \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
743      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
744        treated as a data character. */
745    
746      case 'x':      case CHAR_x:
747  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
748        {        {
749        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
750        register int count = 0;        int count = 0;
751    
752        c = 0;        c = 0;
753        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
754          {          {
755          int cc = *pt++;          register int cc = *pt++;
756            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
757          count++;          count++;
758  #if !EBCDIC    /* ASCII coding */  
759          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
760          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
761  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
762          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
763          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
764            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
765  #endif  #endif
766          }          }
767        if (*pt == '}')  
768          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
769          {          {
770          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
771          ptr = pt;          ptr = pt;
772          break;          break;
773          }          }
774    
775        /* 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
776        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
777        }        }
 #endif  
778    
779      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
780    
781      c = 0;      c = 0;
782      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
783        {        {
784        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
785        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
786  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
788        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
789  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
790        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
791        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
792  #endif  #endif
793        }        }
794      break;      break;
795    
796      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
797        This coding is ASCII-specific, but then the whole concept of \cx is
798        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
799    
800      case 'c':      case CHAR_c:
801      c = *(++ptr);      c = *(++ptr);
802      if (c == 0)      if (c == 0)
803        {        {
804        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
805        return 0;        break;
806        }        }
807    
808      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
809      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;  
810      c ^= 0x40;      c ^= 0x40;
811  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
812      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
813      c ^= 0xC0;      c ^= 0xC0;
814  #endif  #endif
815      break;      break;
816    
817      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
818      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
819      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
820      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
821      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
822    
823      default:      default:
824      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 850  escape sequence.
850  Argument:  Argument:
851    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
852    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
853      dptr           points to an int that is set to the detailed property value
854    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
855    
856  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
857  */  */
858    
859  static int  static int
860  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
861  {  {
862  int c, i, bot, top;  int c, i, bot, top;
863  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
864  char name[4];  char name[32];
865    
866  c = *(++ptr);  c = *(++ptr);
867  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
868    
869  *negptr = FALSE;  *negptr = FALSE;
870    
871  /* \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
872  preceded by ^ for negation. */  negation. */
873    
874  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
875    {    {
876    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
877      {      {
878      *negptr = TRUE;      *negptr = TRUE;
879      ptr++;      ptr++;
880      }      }
881    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
882      {      {
883      c = *(++ptr);      c = *(++ptr);
884      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
885      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
886      name[i] = c;      name[i] = c;
887      }      }
888    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;  
     }  
889    name[i] = 0;    name[i] = 0;
890    }    }
891    
# Line 619  top = _pcre_utt_size; Line 906  top = _pcre_utt_size;
906    
907  while (bot < top)  while (bot < top)
908    {    {
909    i = (bot + top)/2;    i = (bot + top) >> 1;
910    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
911    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
912        {
913        *dptr = _pcre_utt[i].value;
914        return _pcre_utt[i].type;
915        }
916    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
917    }    }
918    
 UNKNOWN_RETURN:  
919  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
920  *ptrptr = ptr;  *ptrptr = ptr;
921  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 950  is_counted_repeat(const uschar *p)
950  {  {
951  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
952  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
953  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
954    
955  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
956  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
957    
958  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
959  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
960    
961  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
962  }  }
963    
964    
# Line 698  read_repeat_counts(const uschar *p, int Line 988  read_repeat_counts(const uschar *p, int
988  int min = 0;  int min = 0;
989  int max = -1;  int max = -1;
990    
991  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
992    an integer overflow. */
993    
994    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
995    if (min < 0 || min > 65535)
996      {
997      *errorcodeptr = ERR5;
998      return p;
999      }
1000    
1001    /* Read the maximum value if there is one, and again do a paranoid on its size.
1002    Also, max must not be less than min. */
1003    
1004  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1005    {    {
1006    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1007      {      {
1008      max = 0;      max = 0;
1009      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1010        if (max < 0 || max > 65535)
1011          {
1012          *errorcodeptr = ERR5;
1013          return p;
1014          }
1015      if (max < min)      if (max < min)
1016        {        {
1017        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1020  if (*p == '}') max = min; else
1020      }      }
1021    }    }
1022    
1023  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1024  pointer to the terminating '}'. */  '}'. */
1025    
1026  if (min > 65535 || max > 65535)  *minp = min;
1027    *errorcodeptr = ERR5;  *maxp = max;
1028  else  return p;
1029    }
1030    
1031    
1032    
1033    /*************************************************
1034    *  Subroutine for finding forward reference      *
1035    *************************************************/
1036    
1037    /* This recursive function is called only from find_parens() below. The
1038    top-level call starts at the beginning of the pattern. All other calls must
1039    start at a parenthesis. It scans along a pattern's text looking for capturing
1040    subpatterns, and counting them. If it finds a named pattern that matches the
1041    name it is given, it returns its number. Alternatively, if the name is NULL, it
1042    returns when it reaches a given numbered subpattern. We know that if (?P< is
1043    encountered, the name will be terminated by '>' because that is checked in the
1044    first pass. Recursion is used to keep track of subpatterns that reset the
1045    capturing group numbers - the (?| feature.
1046    
1047    Arguments:
1048      ptrptr       address of the current character pointer (updated)
1049      cd           compile background data
1050      name         name to seek, or NULL if seeking a numbered subpattern
1051      lorn         name length, or subpattern number if name is NULL
1052      xmode        TRUE if we are in /x mode
1053      count        pointer to the current capturing subpattern number (updated)
1054    
1055    Returns:       the number of the named subpattern, or -1 if not found
1056    */
1057    
1058    static int
1059    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1060      BOOL xmode, int *count)
1061    {
1062    uschar *ptr = *ptrptr;
1063    int start_count = *count;
1064    int hwm_count = start_count;
1065    BOOL dup_parens = FALSE;
1066    
1067    /* If the first character is a parenthesis, check on the type of group we are
1068    dealing with. The very first call may not start with a parenthesis. */
1069    
1070    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1071    {    {
1072    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1073    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1074        {
1075        ptr += 3;
1076        dup_parens = TRUE;
1077        }
1078    
1079      /* Handle a normal, unnamed capturing parenthesis */
1080    
1081      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1082        {
1083        *count += 1;
1084        if (name == NULL && *count == lorn) return *count;
1085        ptr++;
1086        }
1087    
1088      /* Handle a condition. If it is an assertion, just carry on so that it
1089      is processed as normal. If not, skip to the closing parenthesis of the
1090      condition (there can't be any nested parens. */
1091    
1092      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1093        {
1094        ptr += 2;
1095        if (ptr[1] != CHAR_QUESTION_MARK)
1096          {
1097          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1098          if (*ptr != 0) ptr++;
1099          }
1100        }
1101    
1102      /* We have either (? or (* and not a condition */
1103    
1104      else
1105        {
1106        ptr += 2;
1107        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1108    
1109        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1110    
1111        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1112            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1113          {
1114          int term;
1115          const uschar *thisname;
1116          *count += 1;
1117          if (name == NULL && *count == lorn) return *count;
1118          term = *ptr++;
1119          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1120          thisname = ptr;
1121          while (*ptr != term) ptr++;
1122          if (name != NULL && lorn == ptr - thisname &&
1123              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1124            return *count;
1125          term++;
1126          }
1127        }
1128    }    }
1129  return p;  
1130    /* Past any initial parenthesis handling, scan for parentheses or vertical
1131    bars. */
1132    
1133    for (; *ptr != 0; ptr++)
1134      {
1135      /* Skip over backslashed characters and also entire \Q...\E */
1136    
1137      if (*ptr == CHAR_BACKSLASH)
1138        {
1139        if (*(++ptr) == 0) goto FAIL_EXIT;
1140        if (*ptr == CHAR_Q) for (;;)
1141          {
1142          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1143          if (*ptr == 0) goto FAIL_EXIT;
1144          if (*(++ptr) == CHAR_E) break;
1145          }
1146        continue;
1147        }
1148    
1149      /* Skip over character classes; this logic must be similar to the way they
1150      are handled for real. If the first character is '^', skip it. Also, if the
1151      first few characters (either before or after ^) are \Q\E or \E we skip them
1152      too. This makes for compatibility with Perl. Note the use of STR macros to
1153      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1154    
1155      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1156        {
1157        BOOL negate_class = FALSE;
1158        for (;;)
1159          {
1160          if (ptr[1] == CHAR_BACKSLASH)
1161            {
1162            if (ptr[2] == CHAR_E)
1163              ptr+= 2;
1164            else if (strncmp((const char *)ptr+2,
1165                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1166              ptr += 4;
1167            else
1168              break;
1169            }
1170          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1171            {
1172            negate_class = TRUE;
1173            ptr++;
1174            }
1175          else break;
1176          }
1177    
1178        /* If the next character is ']', it is a data character that must be
1179        skipped, except in JavaScript compatibility mode. */
1180    
1181        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1182            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1183          ptr++;
1184    
1185        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1186          {
1187          if (*ptr == 0) return -1;
1188          if (*ptr == CHAR_BACKSLASH)
1189            {
1190            if (*(++ptr) == 0) goto FAIL_EXIT;
1191            if (*ptr == CHAR_Q) for (;;)
1192              {
1193              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1194              if (*ptr == 0) goto FAIL_EXIT;
1195              if (*(++ptr) == CHAR_E) break;
1196              }
1197            continue;
1198            }
1199          }
1200        continue;
1201        }
1202    
1203      /* Skip comments in /x mode */
1204    
1205      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1206        {
1207        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1208        if (*ptr == 0) goto FAIL_EXIT;
1209        continue;
1210        }
1211    
1212      /* Check for the special metacharacters */
1213    
1214      if (*ptr == CHAR_LEFT_PARENTHESIS)
1215        {
1216        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1217        if (rc > 0) return rc;
1218        if (*ptr == 0) goto FAIL_EXIT;
1219        }
1220    
1221      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1222        {
1223        if (dup_parens && *count < hwm_count) *count = hwm_count;
1224        *ptrptr = ptr;
1225        return -1;
1226        }
1227    
1228      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1229        {
1230        if (*count > hwm_count) hwm_count = *count;
1231        *count = start_count;
1232        }
1233      }
1234    
1235    FAIL_EXIT:
1236    *ptrptr = ptr;
1237    return -1;
1238    }
1239    
1240    
1241    
1242    
1243    /*************************************************
1244    *       Find forward referenced subpattern       *
1245    *************************************************/
1246    
1247    /* This function scans along a pattern's text looking for capturing
1248    subpatterns, and counting them. If it finds a named pattern that matches the
1249    name it is given, it returns its number. Alternatively, if the name is NULL, it
1250    returns when it reaches a given numbered subpattern. This is used for forward
1251    references to subpatterns. We used to be able to start this scan from the
1252    current compiling point, using the current count value from cd->bracount, and
1253    do it all in a single loop, but the addition of the possibility of duplicate
1254    subpattern numbers means that we have to scan from the very start, in order to
1255    take account of such duplicates, and to use a recursive function to keep track
1256    of the different types of group.
1257    
1258    Arguments:
1259      cd           compile background data
1260      name         name to seek, or NULL if seeking a numbered subpattern
1261      lorn         name length, or subpattern number if name is NULL
1262      xmode        TRUE if we are in /x mode
1263    
1264    Returns:       the number of the found subpattern, or -1 if not found
1265    */
1266    
1267    static int
1268    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1269    {
1270    uschar *ptr = (uschar *)cd->start_pattern;
1271    int count = 0;
1272    int rc;
1273    
1274    /* If the pattern does not start with an opening parenthesis, the first call
1275    to find_parens_sub() will scan right to the end (if necessary). However, if it
1276    does start with a parenthesis, find_parens_sub() will return when it hits the
1277    matching closing parens. That is why we have to have a loop. */
1278    
1279    for (;;)
1280      {
1281      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1282      if (rc > 0 || *ptr++ == 0) break;
1283      }
1284    
1285    return rc;
1286  }  }
1287    
1288    
1289    
1290    
1291  /*************************************************  /*************************************************
1292  *      Find first significant op code            *  *      Find first significant op code            *
1293  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1337  for (;;)
1337    
1338      case OP_CALLOUT:      case OP_CALLOUT:
1339      case OP_CREF:      case OP_CREF:
1340      case OP_BRANUMBER:      case OP_NCREF:
1341        case OP_RREF:
1342        case OP_NRREF:
1343        case OP_DEF:
1344      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1345      break;      break;
1346    
# Line 793  for (;;) Line 1355  for (;;)
1355    
1356    
1357  /*************************************************  /*************************************************
1358  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1359  *************************************************/  *************************************************/
1360    
1361  /* 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,
1362  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.
1363  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
1364    temporarily terminated with OP_END when this function is called.
1365    
1366    This function is called when a backward assertion is encountered, so that if it
1367    fails, the error message can point to the correct place in the pattern.
1368    However, we cannot do this when the assertion contains subroutine calls,
1369    because they can be forward references. We solve this by remembering this case
1370    and doing the check at the end; a flag specifies which mode we are running in.
1371    
1372  Arguments:  Arguments:
1373    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1374    options  the compiling options    options  the compiling options
1375      atend    TRUE if called when the pattern is complete
1376      cd       the "compile data" structure
1377    
1378  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1379                 or -1 if there is no fixed length,
1380               or -2 if \C was encountered               or -2 if \C was encountered
1381                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1382  */  */
1383    
1384  static int  static int
1385  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1386  {  {
1387  int length = -1;  int length = -1;
1388    
# Line 822  branch, check the length against that of Line 1395  branch, check the length against that of
1395  for (;;)  for (;;)
1396    {    {
1397    int d;    int d;
1398      uschar *ce, *cs;
1399    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1400    switch (op)    switch (op)
1401      {      {
1402        case OP_CBRA:
1403      case OP_BRA:      case OP_BRA:
1404      case OP_ONCE:      case OP_ONCE:
1405      case OP_COND:      case OP_COND:
1406      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1407      if (d < 0) return d;      if (d < 0) return d;
1408      branchlength += d;      branchlength += d;
1409      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 853  for (;;) Line 1426  for (;;)
1426      branchlength = 0;      branchlength = 0;
1427      break;      break;
1428    
1429        /* A true recursion implies not fixed length, but a subroutine call may
1430        be OK. If the subroutine is a forward reference, we can't deal with
1431        it until the end of the pattern, so return -3. */
1432    
1433        case OP_RECURSE:
1434        if (!atend) return -3;
1435        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1436        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1437        if (cc > cs && cc < ce) return -1;                /* Recursion */
1438        d = find_fixedlength(cs + 2, options, atend, cd);
1439        if (d < 0) return d;
1440        branchlength += d;
1441        cc += 1 + LINK_SIZE;
1442        break;
1443    
1444      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1445    
1446      case OP_ASSERT:      case OP_ASSERT:
# Line 865  for (;;) Line 1453  for (;;)
1453      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1454    
1455      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1456      case OP_CREF:      case OP_CREF:
1457        case OP_NCREF:
1458        case OP_RREF:
1459        case OP_NRREF:
1460        case OP_DEF:
1461      case OP_OPT:      case OP_OPT:
1462      case OP_CALLOUT:      case OP_CALLOUT:
1463      case OP_SOD:      case OP_SOD:
1464      case OP_SOM:      case OP_SOM:
1465        case OP_SET_SOM:
1466      case OP_EOD:      case OP_EOD:
1467      case OP_EODN:      case OP_EODN:
1468      case OP_CIRC:      case OP_CIRC:
# Line 884  for (;;) Line 1476  for (;;)
1476    
1477      case OP_CHAR:      case OP_CHAR:
1478      case OP_CHARNC:      case OP_CHARNC:
1479        case OP_NOT:
1480      branchlength++;      branchlength++;
1481      cc += 2;      cc += 2;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 901  for (;;) Line 1492  for (;;)
1492      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1493      cc += 4;      cc += 4;
1494  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1495      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1496        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1497  #endif  #endif
1498      break;      break;
1499    
1500      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1501      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1502        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1503      cc += 4;      cc += 4;
1504      break;      break;
1505    
# Line 917  for (;;) Line 1507  for (;;)
1507    
1508      case OP_PROP:      case OP_PROP:
1509      case OP_NOTPROP:      case OP_NOTPROP:
1510      cc++;      cc += 2;
1511      /* Fall through */      /* Fall through */
1512    
1513      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1517  for (;;)
1517      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1518      case OP_WORDCHAR:      case OP_WORDCHAR:
1519      case OP_ANY:      case OP_ANY:
1520        case OP_ALLANY:
1521      branchlength++;      branchlength++;
1522      cc++;      cc++;
1523      break;      break;
# Line 981  for (;;) Line 1572  for (;;)
1572    
1573    
1574  /*************************************************  /*************************************************
1575  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1576  *************************************************/  *************************************************/
1577    
1578  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1579  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1580    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1581    so that it can be called from pcre_study() when finding the minimum matching
1582    length.
1583    
1584  Arguments:  Arguments:
1585    code        points to start of expression    code        points to start of expression
1586    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1587    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1588    
1589  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
1590  */  */
1591    
1592  static const uschar *  const uschar *
1593  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1594  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1595  for (;;)  for (;;)
1596    {    {
1597    register int c = *code;    register int c = *code;
1598    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1599    else if (c > OP_BRA)  
1600      /* XCLASS is used for classes that cannot be represented just by a bit
1601      map. This includes negated single high-valued characters. The length in
1602      the table is zero; the actual length is stored in the compiled code. */
1603    
1604      if (c == OP_XCLASS) code += GET(code, 1);
1605    
1606      /* Handle recursion */
1607    
1608      else if (c == OP_REVERSE)
1609      {      {
1610      int n = c - OP_BRA;      if (number < 0) return (uschar *)code;
1611      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      code += _pcre_OP_lengths[c];
     if (n == number) return (uschar *)code;  
     code += _pcre_OP_lengths[OP_BRA];  
1612      }      }
1613    else  
1614      /* Handle capturing bracket */
1615    
1616      else if (c == OP_CBRA)
1617      {      {
1618        int n = GET2(code, 1+LINK_SIZE);
1619        if (n == number) return (uschar *)code;
1620      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1621        }
1622    
1623  #ifdef SUPPORT_UTF8    /* Otherwise, we can get the item's length from the table, except that for
1624      repeated character types, we have to test for \p and \P, which have an extra
1625      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1626      must add in its length. */
1627    
1628      /* In UTF-8 mode, opcodes that are followed by a character may be followed    else
1629      by a multi-byte character. The length in the table is a minimum, so we have      {
1630      to scan along to skip the extra bytes. All opcodes are less than 128, so we      switch(c)
1631      can use relatively efficient code. */        {
1632          case OP_TYPESTAR:
1633          case OP_TYPEMINSTAR:
1634          case OP_TYPEPLUS:
1635          case OP_TYPEMINPLUS:
1636          case OP_TYPEQUERY:
1637          case OP_TYPEMINQUERY:
1638          case OP_TYPEPOSSTAR:
1639          case OP_TYPEPOSPLUS:
1640          case OP_TYPEPOSQUERY:
1641          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1642          break;
1643    
1644          case OP_TYPEUPTO:
1645          case OP_TYPEMINUPTO:
1646          case OP_TYPEEXACT:
1647          case OP_TYPEPOSUPTO:
1648          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1649          break;
1650    
1651          case OP_MARK:
1652          case OP_PRUNE_ARG:
1653          case OP_SKIP_ARG:
1654          case OP_THEN_ARG:
1655          code += code[1];
1656          break;
1657          }
1658    
1659        /* Add in the fixed length from the table */
1660    
1661        code += _pcre_OP_lengths[c];
1662    
1663      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1664      a multi-byte character. The length in the table is a minimum, so we have to
1665      arrange to skip the extra bytes. */
1666    
1667    #ifdef SUPPORT_UTF8
1668      if (utf8) switch(c)      if (utf8) switch(c)
1669        {        {
1670        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1672  for (;;)
1672        case OP_EXACT:        case OP_EXACT:
1673        case OP_UPTO:        case OP_UPTO:
1674        case OP_MINUPTO:        case OP_MINUPTO:
1675          case OP_POSUPTO:
1676        case OP_STAR:        case OP_STAR:
1677        case OP_MINSTAR:        case OP_MINSTAR:
1678          case OP_POSSTAR:
1679        case OP_PLUS:        case OP_PLUS:
1680        case OP_MINPLUS:        case OP_MINPLUS:
1681          case OP_POSPLUS:
1682        case OP_QUERY:        case OP_QUERY:
1683        case OP_MINQUERY:        case OP_MINQUERY:
1684        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1685        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;  
1686        break;        break;
1687        }        }
1688    #else
1689        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1690  #endif  #endif
1691      }      }
1692    }    }
# Line 1072  Returns: pointer to the opcode for Line 1711  Returns: pointer to the opcode for
1711  static const uschar *  static const uschar *
1712  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1713  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1714  for (;;)  for (;;)
1715    {    {
1716    register int c = *code;    register int c = *code;
1717    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1718    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1719    else if (c > OP_BRA)  
1720      {    /* XCLASS is used for classes that cannot be represented just by a bit
1721      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1722      }    the table is zero; the actual length is stored in the compiled code. */
1723    
1724      if (c == OP_XCLASS) code += GET(code, 1);
1725    
1726      /* Otherwise, we can get the item's length from the table, except that for
1727      repeated character types, we have to test for \p and \P, which have an extra
1728      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731    else    else
1732      {      {
1733      code += _pcre_OP_lengths[c];      switch(c)
1734          {
1735          case OP_TYPESTAR:
1736          case OP_TYPEMINSTAR:
1737          case OP_TYPEPLUS:
1738          case OP_TYPEMINPLUS:
1739          case OP_TYPEQUERY:
1740          case OP_TYPEMINQUERY:
1741          case OP_TYPEPOSSTAR:
1742          case OP_TYPEPOSPLUS:
1743          case OP_TYPEPOSQUERY:
1744          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1745          break;
1746    
1747  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1748          case OP_TYPEUPTO:
1749          case OP_TYPEMINUPTO:
1750          case OP_TYPEEXACT:
1751          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752          break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          case OP_THEN_ARG:
1758          code += code[1];
1759          break;
1760          }
1761    
1762        /* Add in the fixed length from the table */
1763    
1764        code += _pcre_OP_lengths[c];
1765    
1766      /* 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
1767      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
1768      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. */  
1769    
1770    #ifdef SUPPORT_UTF8
1771      if (utf8) switch(c)      if (utf8) switch(c)
1772        {        {
1773        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1775  for (;;)
1775        case OP_EXACT:        case OP_EXACT:
1776        case OP_UPTO:        case OP_UPTO:
1777        case OP_MINUPTO:        case OP_MINUPTO:
1778          case OP_POSUPTO:
1779        case OP_STAR:        case OP_STAR:
1780        case OP_MINSTAR:        case OP_MINSTAR:
1781          case OP_POSSTAR:
1782        case OP_PLUS:        case OP_PLUS:
1783        case OP_MINPLUS:        case OP_MINPLUS:
1784          case OP_POSPLUS:
1785        case OP_QUERY:        case OP_QUERY:
1786        case OP_MINQUERY:        case OP_MINQUERY:
1787        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1788        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;  
1789        break;        break;
1790        }        }
1791    #else
1792        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1793  #endif  #endif
1794      }      }
1795    }    }
# Line 1132  for (;;) Line 1802  for (;;)
1802  *************************************************/  *************************************************/
1803    
1804  /* 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
1805  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()
1806  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
1807  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
1808  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1809    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1810    bracket whose current branch will already have been scanned.
1811    
1812  Arguments:  Arguments:
1813    code        points to start of search    code        points to start of search
1814    endcode     points to where to stop    endcode     points to where to stop
1815    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1816      cd          contains pointers to tables etc.
1817    
1818  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1819  */  */
1820    
1821  static BOOL  static BOOL
1822  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1823      compile_data *cd)
1824  {  {
1825  register int c;  register int c;
1826  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);
1827       code < endcode;       code < endcode;
1828       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1829    {    {
# Line 1157  for (code = first_significant_code(code Line 1831  for (code = first_significant_code(code
1831    
1832    c = *code;    c = *code;
1833    
1834    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1835      first_significant_code() with a TRUE final argument. */
1836    
1837      if (c == OP_ASSERT)
1838      {      {
1839      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1840      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1841        continue;
1842        }
1843    
1844      /* Groups with zero repeats can of course be empty; skip them. */
1845    
1846      /* Scan a closed bracket */    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1847        {
1848        code += _pcre_OP_lengths[c];
1849        do code += GET(code, 1); while (*code == OP_ALT);
1850        c = *code;
1851        continue;
1852        }
1853    
1854      empty_branch = FALSE;    /* For a recursion/subroutine call, if its end has been reached, which
1855      implies a subroutine call, we can scan it. */
1856    
1857      if (c == OP_RECURSE)
1858        {
1859        BOOL empty_branch = FALSE;
1860        const uschar *scode = cd->start_code + GET(code, 1);
1861        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1862      do      do
1863        {        {
1864        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1865            {
1866          empty_branch = TRUE;          empty_branch = TRUE;
1867            break;
1868            }
1869          scode += GET(scode, 1);
1870          }
1871        while (*scode == OP_ALT);
1872        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1873        continue;
1874        }
1875    
1876      /* For other groups, scan the branches. */
1877    
1878      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1879        {
1880        BOOL empty_branch;
1881        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1882    
1883        /* If a conditional group has only one branch, there is a second, implied,
1884        empty branch, so just skip over the conditional, because it could be empty.
1885        Otherwise, scan the individual branches of the group. */
1886    
1887        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1888        code += GET(code, 1);        code += GET(code, 1);
1889        else
1890          {
1891          empty_branch = FALSE;
1892          do
1893            {
1894            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1895              empty_branch = TRUE;
1896            code += GET(code, 1);
1897            }
1898          while (*code == OP_ALT);
1899          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1900        }        }
1901      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1902      c = *code;      c = *code;
1903        continue;
1904      }      }
1905    
1906    else switch (c)    /* Handle the other opcodes */
1907    
1908      switch (c)
1909      {      {
1910      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1911        cannot be represented just by a bit map. This includes negated single
1912        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1913        actual length is stored in the compiled code, so we must update "code"
1914        here. */
1915    
1916  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1917      case OP_XCLASS:      case OP_XCLASS:
1918      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1919      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1920  #endif  #endif
1921    
# Line 1227  for (code = first_significant_code(code Line 1959  for (code = first_significant_code(code
1959      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1960      case OP_WORDCHAR:      case OP_WORDCHAR:
1961      case OP_ANY:      case OP_ANY:
1962        case OP_ALLANY:
1963      case OP_ANYBYTE:      case OP_ANYBYTE:
1964      case OP_CHAR:      case OP_CHAR:
1965      case OP_CHARNC:      case OP_CHARNC:
1966      case OP_NOT:      case OP_NOT:
1967      case OP_PLUS:      case OP_PLUS:
1968      case OP_MINPLUS:      case OP_MINPLUS:
1969        case OP_POSPLUS:
1970      case OP_EXACT:      case OP_EXACT:
1971      case OP_NOTPLUS:      case OP_NOTPLUS:
1972      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1973        case OP_NOTPOSPLUS:
1974      case OP_NOTEXACT:      case OP_NOTEXACT:
1975      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1976      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1977        case OP_TYPEPOSPLUS:
1978      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1979      return FALSE;      return FALSE;
1980    
1981        /* These are going to continue, as they may be empty, but we have to
1982        fudge the length for the \p and \P cases. */
1983    
1984        case OP_TYPESTAR:
1985        case OP_TYPEMINSTAR:
1986        case OP_TYPEPOSSTAR:
1987        case OP_TYPEQUERY:
1988        case OP_TYPEMINQUERY:
1989        case OP_TYPEPOSQUERY:
1990        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1991        break;
1992    
1993        /* Same for these */
1994    
1995        case OP_TYPEUPTO:
1996        case OP_TYPEMINUPTO:
1997        case OP_TYPEPOSUPTO:
1998        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1999        break;
2000    
2001      /* End of branch */      /* End of branch */
2002    
2003      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 2006  for (code = first_significant_code(code
2006      case OP_ALT:      case OP_ALT:
2007      return TRUE;      return TRUE;
2008    
2009      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2010      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2011    
2012  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2013      case OP_STAR:      case OP_STAR:
2014      case OP_MINSTAR:      case OP_MINSTAR:
2015        case OP_POSSTAR:
2016      case OP_QUERY:      case OP_QUERY:
2017      case OP_MINQUERY:      case OP_MINQUERY:
2018        case OP_POSQUERY:
2019        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2020        break;
2021    
2022      case OP_UPTO:      case OP_UPTO:
2023      case OP_MINUPTO:      case OP_MINUPTO:
2024      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2025        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2026      break;      break;
2027  #endif  #endif
2028    
2029        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2030        string. */
2031    
2032        case OP_MARK:
2033        case OP_PRUNE_ARG:
2034        case OP_SKIP_ARG:
2035        case OP_THEN_ARG:
2036        code += code[1];
2037        break;
2038    
2039        /* None of the remaining opcodes are required to match a character. */
2040    
2041        default:
2042        break;
2043      }      }
2044    }    }
2045    
# Line 1285  Arguments: Line 2062  Arguments:
2062    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2063    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2064    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2065      cd          pointers to tables etc
2066    
2067  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2068  */  */
2069    
2070  static BOOL  static BOOL
2071  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2072    BOOL utf8)    BOOL utf8, compile_data *cd)
2073  {  {
2074  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2075    {    {
2076    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2077        return FALSE;
2078    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2079    }    }
2080  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2087  return TRUE;
2087  *************************************************/  *************************************************/
2088    
2089  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2090  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
2091  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2092  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2093    
2094    Originally, this function only recognized a sequence of letters between the
2095    terminators, but it seems that Perl recognizes any sequence of characters,
2096    though of course unknown POSIX names are subsequently rejected. Perl gives an
2097    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2098    didn't consider this to be a POSIX class. Likewise for [:1234:].
2099    
2100    The problem in trying to be exactly like Perl is in the handling of escapes. We
2101    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2102    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2103    below handles the special case of \], but does not try to do any other escape
2104    processing. This makes it different from Perl for cases such as [:l\ower:]
2105    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2106    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2107    I think.
2108    
2109  Argument:  Arguments:
2110    ptr      pointer to the initial [    ptr      pointer to the initial [
2111    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2112    
2113  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2114  */  */
2115    
2116  static BOOL  static BOOL
2117  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2118  {  {
2119  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2120  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2121  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2122    {    {
2123    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2124    return TRUE;      {
2125        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2126        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2127          {
2128          *endptr = ptr;
2129          return TRUE;
2130          }
2131        }
2132    }    }
2133  return FALSE;  return FALSE;
2134  }  }
# Line 1355  Returns: a value representing the na Line 2153  Returns: a value representing the na
2153  static int  static int
2154  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2155  {  {
2156    const char *pn = posix_names;
2157  register int yield = 0;  register int yield = 0;
2158  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2159    {    {
2160    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2161      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2162      pn += posix_name_lengths[yield] + 1;
2163    yield++;    yield++;
2164    }    }
2165  return -1;  return -1;
# Line 1374  return -1; Line 2174  return -1;
2174  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2175  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2176  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
2177  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
2178  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
2179  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
2180  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
2181  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2182    OP_END.
2183    
2184    This function has been extended with the possibility of forward references for
2185    recursions and subroutine calls. It must also check the list of such references
2186    for the group we are dealing with. If it finds that one of the recursions in
2187    the current group is on this list, it adjusts the offset in the list, not the
2188    value in the reference (which is a group number).
2189    
2190  Arguments:  Arguments:
2191    group      points to the start of the group    group      points to the start of the group
2192    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2193    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2194    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2195      save_hwm   the hwm forward reference pointer at the start of the group
2196    
2197  Returns:     nothing  Returns:     nothing
2198  */  */
2199    
2200  static void  static void
2201  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2202      uschar *save_hwm)
2203  {  {
2204  uschar *ptr = group;  uschar *ptr = group;
2205    
2206  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2207    {    {
2208    int offset = GET(ptr, 1);    int offset;
2209    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2210    
2211      /* See if this recursion is on the forward reference list. If so, adjust the
2212      reference. */
2213    
2214      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2215        {
2216        offset = GET(hc, 0);
2217        if (cd->start_code + offset == ptr + 1)
2218          {
2219          PUT(hc, 0, offset + adjust);
2220          break;
2221          }
2222        }
2223    
2224      /* Otherwise, adjust the recursion offset if it's after the start of this
2225      group. */
2226    
2227      if (hc >= cd->hwm)
2228        {
2229        offset = GET(ptr, 1);
2230        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2231        }
2232    
2233    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2234    }    }
2235  }  }
# Line 1475  Yield: TRUE when range returned; Line 2308  Yield: TRUE when range returned;
2308  */  */
2309    
2310  static BOOL  static BOOL
2311  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2312      unsigned int *odptr)
2313  {  {
2314  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2315    
2316  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2317    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2318    
2319  if (c > d) return FALSE;  if (c > d) return FALSE;
2320    
# Line 1492  next = othercase + 1; Line 2323  next = othercase + 1;
2323    
2324  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2325    {    {
2326    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2327    next++;    next++;
2328    }    }
2329    
# Line 1506  return TRUE; Line 2335  return TRUE;
2335  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2336    
2337    
2338    
2339    /*************************************************
2340    *     Check if auto-possessifying is possible    *
2341    *************************************************/
2342    
2343    /* This function is called for unlimited repeats of certain items, to see
2344    whether the next thing could possibly match the repeated item. If not, it makes
2345    sense to automatically possessify the repeated item.
2346    
2347    Arguments:
2348      op_code       the repeated op code
2349      this          data for this item, depends on the opcode
2350      utf8          TRUE in UTF-8 mode
2351      utf8_char     used for utf8 character bytes, NULL if not relevant
2352      ptr           next character in pattern
2353      options       options bits
2354      cd            contains pointers to tables etc.
2355    
2356    Returns:        TRUE if possessifying is wanted
2357    */
2358    
2359    static BOOL
2360    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2361      const uschar *ptr, int options, compile_data *cd)
2362    {
2363    int next;
2364    
2365    /* Skip whitespace and comments in extended mode */
2366    
2367    if ((options & PCRE_EXTENDED) != 0)
2368      {
2369      for (;;)
2370        {
2371        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2372        if (*ptr == CHAR_NUMBER_SIGN)
2373          {
2374          while (*(++ptr) != 0)
2375            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2376          }
2377        else break;
2378        }
2379      }
2380    
2381    /* If the next item is one that we can handle, get its value. A non-negative
2382    value is a character, a negative value is an escape value. */
2383    
2384    if (*ptr == CHAR_BACKSLASH)
2385      {
2386      int temperrorcode = 0;
2387      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2388      if (temperrorcode != 0) return FALSE;
2389      ptr++;    /* Point after the escape sequence */
2390      }
2391    
2392    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2393      {
2394    #ifdef SUPPORT_UTF8
2395      if (utf8) { GETCHARINC(next, ptr); } else
2396    #endif
2397      next = *ptr++;
2398      }
2399    
2400    else return FALSE;
2401    
2402    /* Skip whitespace and comments in extended mode */
2403    
2404    if ((options & PCRE_EXTENDED) != 0)
2405      {
2406      for (;;)
2407        {
2408        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2409        if (*ptr == CHAR_NUMBER_SIGN)
2410          {
2411          while (*(++ptr) != 0)
2412            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2413          }
2414        else break;
2415        }
2416      }
2417    
2418    /* If the next thing is itself optional, we have to give up. */
2419    
2420    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2421      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2422        return FALSE;
2423    
2424    /* Now compare the next item with the previous opcode. If the previous is a
2425    positive single character match, "item" either contains the character or, if
2426    "item" is greater than 127 in utf8 mode, the character's bytes are in
2427    utf8_char. */
2428    
2429    
2430    /* Handle cases when the next item is a character. */
2431    
2432    if (next >= 0) switch(op_code)
2433      {
2434      case OP_CHAR:
2435    #ifdef SUPPORT_UTF8
2436      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2437    #else
2438      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2439    #endif
2440      return item != next;
2441    
2442      /* For CHARNC (caseless character) we must check the other case. If we have
2443      Unicode property support, we can use it to test the other case of
2444      high-valued characters. */
2445    
2446      case OP_CHARNC:
2447    #ifdef SUPPORT_UTF8
2448      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2449    #endif
2450      if (item == next) return FALSE;
2451    #ifdef SUPPORT_UTF8
2452      if (utf8)
2453        {
2454        unsigned int othercase;
2455        if (next < 128) othercase = cd->fcc[next]; else
2456    #ifdef SUPPORT_UCP
2457        othercase = UCD_OTHERCASE((unsigned int)next);
2458    #else
2459        othercase = NOTACHAR;
2460    #endif
2461        return (unsigned int)item != othercase;
2462        }
2463      else
2464    #endif  /* SUPPORT_UTF8 */
2465      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2466    
2467      /* For OP_NOT, "item" must be a single-byte character. */
2468    
2469      case OP_NOT:
2470      if (item == next) return TRUE;
2471      if ((options & PCRE_CASELESS) == 0) return FALSE;
2472    #ifdef SUPPORT_UTF8
2473      if (utf8)
2474        {
2475        unsigned int othercase;
2476        if (next < 128) othercase = cd->fcc[next]; else
2477    #ifdef SUPPORT_UCP
2478        othercase = UCD_OTHERCASE(next);
2479    #else
2480        othercase = NOTACHAR;
2481    #endif
2482        return (unsigned int)item == othercase;
2483        }
2484      else
2485    #endif  /* SUPPORT_UTF8 */
2486      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2487    
2488      case OP_DIGIT:
2489      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2490    
2491      case OP_NOT_DIGIT:
2492      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2493    
2494      case OP_WHITESPACE:
2495      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2496    
2497      case OP_NOT_WHITESPACE:
2498      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2499    
2500      case OP_WORDCHAR:
2501      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2502    
2503      case OP_NOT_WORDCHAR:
2504      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2505    
2506      case OP_HSPACE:
2507      case OP_NOT_HSPACE:
2508      switch(next)
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 op_code != OP_HSPACE;
2530        default:
2531        return op_code == OP_HSPACE;
2532        }
2533    
2534      case OP_VSPACE:
2535      case OP_NOT_VSPACE:
2536      switch(next)
2537        {
2538        case 0x0a:
2539        case 0x0b:
2540        case 0x0c:
2541        case 0x0d:
2542        case 0x85:
2543        case 0x2028:
2544        case 0x2029:
2545        return op_code != OP_VSPACE;
2546        default:
2547        return op_code == OP_VSPACE;
2548        }
2549    
2550      default:
2551      return FALSE;
2552      }
2553    
2554    
2555    /* Handle the case when the next item is \d, \s, etc. */
2556    
2557    switch(op_code)
2558      {
2559      case OP_CHAR:
2560      case OP_CHARNC:
2561    #ifdef SUPPORT_UTF8
2562      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2563    #endif
2564      switch(-next)
2565        {
2566        case ESC_d:
2567        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2568    
2569        case ESC_D:
2570        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2571    
2572        case ESC_s:
2573        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2574    
2575        case ESC_S:
2576        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2577    
2578        case ESC_w:
2579        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2580    
2581        case ESC_W:
2582        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2583    
2584        case ESC_h:
2585        case ESC_H:
2586        switch(item)
2587          {
2588          case 0x09:
2589          case 0x20:
2590          case 0xa0:
2591          case 0x1680:
2592          case 0x180e:
2593          case 0x2000:
2594          case 0x2001:
2595          case 0x2002:
2596          case 0x2003:
2597          case 0x2004:
2598          case 0x2005:
2599          case 0x2006:
2600          case 0x2007:
2601          case 0x2008:
2602          case 0x2009:
2603          case 0x200A:
2604          case 0x202f:
2605          case 0x205f:
2606          case 0x3000:
2607          return -next != ESC_h;
2608          default:
2609          return -next == ESC_h;
2610          }
2611    
2612        case ESC_v:
2613        case ESC_V:
2614        switch(item)
2615          {
2616          case 0x0a:
2617          case 0x0b:
2618          case 0x0c:
2619          case 0x0d:
2620          case 0x85:
2621          case 0x2028:
2622          case 0x2029:
2623          return -next != ESC_v;
2624          default:
2625          return -next == ESC_v;
2626          }
2627    
2628        default:
2629        return FALSE;
2630        }
2631    
2632      case OP_DIGIT:
2633      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2634             next == -ESC_h || next == -ESC_v;
2635    
2636      case OP_NOT_DIGIT:
2637      return next == -ESC_d;
2638    
2639      case OP_WHITESPACE:
2640      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2641    
2642      case OP_NOT_WHITESPACE:
2643      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2644    
2645      case OP_HSPACE:
2646      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2647    
2648      case OP_NOT_HSPACE:
2649      return next == -ESC_h;
2650    
2651      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2652      case OP_VSPACE:
2653      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2654    
2655      case OP_NOT_VSPACE:
2656      return next == -ESC_v;
2657    
2658      case OP_WORDCHAR:
2659      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2660    
2661      case OP_NOT_WORDCHAR:
2662      return next == -ESC_w || next == -ESC_d;
2663    
2664      default:
2665      return FALSE;
2666      }
2667    
2668    /* Control does not reach here */
2669    }
2670    
2671    
2672    
2673  /*************************************************  /*************************************************
2674  *           Compile one branch                   *  *           Compile one branch                   *
2675  *************************************************/  *************************************************/
2676    
2677  /* 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
2678  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
2679  bits.  bits. This function is used during the pre-compile phase when we are trying
2680    to find out the amount of memory needed, as well as during the real compile
2681    phase. The value of lengthptr distinguishes the two phases.
2682    
2683  Arguments:  Arguments:
2684    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2685    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2686    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2687    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2689  Arguments:
2689    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2690    bcptr          points to current branch chain    bcptr          points to current branch chain
2691    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2692      lengthptr      NULL during the real compile phase
2693                     points to length accumulator during pre-compile phase
2694    
2695  Returns:         TRUE on success  Returns:         TRUE on success
2696                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2697  */  */
2698    
2699  static BOOL  static BOOL
2700  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2701    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2702    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2703  {  {
2704  int repeat_type, op_type;  int repeat_type, op_type;
2705  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1541  int greedy_default, greedy_non_default; Line 2708  int greedy_default, greedy_non_default;
2708  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2709  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2710  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2711  int options = *optionsptr;  int options = *optionsptr;
2712  int after_manual_callout = 0;  int after_manual_callout = 0;
2713    int length_prevgroup = 0;
2714  register int c;  register int c;
2715  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2716    uschar *last_code = code;
2717    uschar *orig_code = code;
2718  uschar *tempcode;  uschar *tempcode;
2719  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2720  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2722  const uschar *ptr = *ptrptr;
2722  const uschar *tempptr;  const uschar *tempptr;
2723  uschar *previous = NULL;  uschar *previous = NULL;
2724  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2725    uschar *save_hwm = NULL;
2726  uschar classbits[32];  uschar classbits[32];
2727    
2728  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2729  BOOL class_utf8;  BOOL class_utf8;
2730  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2731  uschar *class_utf8data;  uschar *class_utf8data;
2732    uschar *class_utf8data_base;
2733  uschar utf8_char[6];  uschar utf8_char[6];
2734  #else  #else
2735  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2736    uschar *utf8_char = NULL;
2737    #endif
2738    
2739    #ifdef PCRE_DEBUG
2740    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2741  #endif  #endif
2742    
2743  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1593  req_caseopt = ((options & PCRE_CASELESS) Line 2769  req_caseopt = ((options & PCRE_CASELESS)
2769  for (;; ptr++)  for (;; ptr++)
2770    {    {
2771    BOOL negate_class;    BOOL negate_class;
2772      BOOL should_flip_negation;
2773    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2774    BOOL is_quantifier;    BOOL is_quantifier;
2775      BOOL is_recurse;
2776      BOOL reset_bracount;
2777    int class_charcount;    int class_charcount;
2778    int class_lastchar;    int class_lastchar;
2779    int newoptions;    int newoptions;
2780    int recno;    int recno;
2781      int refsign;
2782    int skipbytes;    int skipbytes;
2783    int subreqbyte;    int subreqbyte;
2784    int subfirstbyte;    int subfirstbyte;
2785      int terminator;
2786    int mclength;    int mclength;
2787    uschar mcbuffer[8];    uschar mcbuffer[8];
2788    
2789    /* Next byte in the pattern */    /* Get next byte in the pattern */
2790    
2791    c = *ptr;    c = *ptr;
2792    
2793    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If we are in the pre-compile phase, accumulate the length used for the
2794      previous cycle of this loop. */
2795    
2796    if (inescq && c != 0)    if (lengthptr != NULL)
2797      {      {
2798      if (c == '\\' && ptr[1] == 'E')  #ifdef PCRE_DEBUG
2799        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2800    #endif
2801        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2802        {        {
2803        inescq = FALSE;        *errorcodeptr = ERR52;
2804        ptr++;        goto FAILED;
2805        continue;        }
2806    
2807        /* There is at least one situation where code goes backwards: this is the
2808        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2809        the class is simply eliminated. However, it is created first, so we have to
2810        allow memory for it. Therefore, don't ever reduce the length at this point.
2811        */
2812    
2813        if (code < last_code) code = last_code;
2814    
2815        /* Paranoid check for integer overflow */
2816    
2817        if (OFLOW_MAX - *lengthptr < code - last_code)
2818          {
2819          *errorcodeptr = ERR20;
2820          goto FAILED;
2821          }
2822    
2823        *lengthptr += code - last_code;
2824        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2825    
2826        /* If "previous" is set and it is not at the start of the work space, move
2827        it back to there, in order to avoid filling up the work space. Otherwise,
2828        if "previous" is NULL, reset the current code pointer to the start. */
2829    
2830        if (previous != NULL)
2831          {
2832          if (previous > orig_code)
2833            {
2834            memmove(orig_code, previous, code - previous);
2835            code -= previous - orig_code;
2836            previous = orig_code;
2837            }
2838          }
2839        else code = orig_code;
2840    
2841        /* Remember where this code item starts so we can pick up the length
2842        next time round. */
2843    
2844        last_code = code;
2845        }
2846    
2847      /* In the real compile phase, just check the workspace used by the forward
2848      reference list. */
2849    
2850      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2851        {
2852        *errorcodeptr = ERR52;
2853        goto FAILED;
2854        }
2855    
2856      /* If in \Q...\E, check for the end; if not, we have a literal */
2857    
2858      if (inescq && c != 0)
2859        {
2860        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2861          {
2862          inescq = FALSE;
2863          ptr++;
2864          continue;
2865        }        }
2866      else      else
2867        {        {
2868        if (previous_callout != NULL)        if (previous_callout != NULL)
2869          {          {
2870          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2871              complete_callout(previous_callout, ptr, cd);
2872          previous_callout = NULL;          previous_callout = NULL;
2873          }          }
2874        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2883  for (;; ptr++)
2883    /* 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
2884    a quantifier. */    a quantifier. */
2885    
2886    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2887      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2888        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2889    
2890    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2891         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2892      {      {
2893      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2894          complete_callout(previous_callout, ptr, cd);
2895      previous_callout = NULL;      previous_callout = NULL;
2896      }      }
2897    
# Line 1653  for (;; ptr++) Line 2900  for (;; ptr++)
2900    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2901      {      {
2902      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2903      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2904        {        {
2905        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2906        on the Macintosh. */          {
2907        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2908        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2909          if (*ptr != 0) continue;
2910    
2911          /* Else fall through to handle end of string */
2912          c = 0;
2913        }        }
2914      }      }
2915    
# Line 1672  for (;; ptr++) Line 2923  for (;; ptr++)
2923    
2924    switch(c)    switch(c)
2925      {      {
2926      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2927        case 0:                        /* The branch terminates at string end */
2928      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2929      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2930      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2931      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2932      *codeptr = code;      *codeptr = code;
2933      *ptrptr = ptr;      *ptrptr = ptr;
2934        if (lengthptr != NULL)
2935          {
2936          if (OFLOW_MAX - *lengthptr < code - last_code)
2937            {
2938            *errorcodeptr = ERR20;
2939            goto FAILED;
2940            }
2941          *lengthptr += code - last_code;   /* To include callout length */
2942          DPRINTF((">> end branch\n"));
2943          }
2944      return TRUE;      return TRUE;
2945    
2946    
2947        /* ===================================================================*/
2948      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2949      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2950    
2951      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2952      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2953        {        {
2954        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2957  for (;; ptr++)
2957      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2958      break;      break;
2959    
2960      case '$':      case CHAR_DOLLAR_SIGN:
2961      previous = NULL;      previous = NULL;
2962      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2963      break;      break;
# Line 1703  for (;; ptr++) Line 2965  for (;; ptr++)
2965      /* 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
2966      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2967    
2968      case '.':      case CHAR_DOT:
2969      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2970      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2971      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2972      previous = code;      previous = code;
2973      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2974      break;      break;
2975    
2976      /* Character classes. If the included characters are all < 255 in value, we  
2977      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2978      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
2979      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
2980      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2981        map as usual, then invert it at the end. However, we use a different opcode
2982        so that data characters > 255 can be handled correctly.
2983    
2984      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2985      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,
2986      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2987      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.
     */  
2988    
2989      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2990        default (Perl) mode, it is treated as a data character. */
2991    
2992        case CHAR_RIGHT_SQUARE_BRACKET:
2993        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2994          {
2995          *errorcodeptr = ERR64;
2996          goto FAILED;
2997          }
2998        goto NORMAL_CHAR;
2999    
3000        case CHAR_LEFT_SQUARE_BRACKET:
3001      previous = code;      previous = code;
3002    
3003      /* 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
3004      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. */
3005    
3006      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3007          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3008            check_posix_syntax(ptr, &tempptr))
3009        {        {
3010        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3011        goto FAILED;        goto FAILED;
3012        }        }
3013    
3014      /* 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,
3015        if the first few characters (either before or after ^) are \Q\E or \E we
3016        skip them too. This makes for compatibility with Perl. */
3017    
3018      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3019        for (;;)
3020        {        {
       negate_class = TRUE;  
3021        c = *(++ptr);        c = *(++ptr);
3022          if (c == CHAR_BACKSLASH)
3023            {
3024            if (ptr[1] == CHAR_E)
3025              ptr++;
3026            else if (strncmp((const char *)ptr+1,
3027                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3028              ptr += 3;
3029            else
3030              break;
3031            }
3032          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3033            negate_class = TRUE;
3034          else break;
3035        }        }
3036      else  
3037        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3038        an initial ']' is taken as a data character -- the code below handles
3039        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3040        [^] must match any character, so generate OP_ALLANY. */
3041    
3042        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3043            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3044        {        {
3045        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3046          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3047          zerofirstbyte = firstbyte;
3048          break;
3049        }        }
3050    
3051        /* If a class contains a negative special such as \S, we need to flip the
3052        negation flag at the end, so that support for characters > 255 works
3053        correctly (they are all included in the class). */
3054    
3055        should_flip_negation = FALSE;
3056    
3057      /* 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
3058      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
3059      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3060    
3061      class_charcount = 0;      class_charcount = 0;
3062      class_lastchar = -1;      class_lastchar = -1;
3063    
3064        /* Initialize the 32-char bit map to all zeros. We build the map in a
3065        temporary bit of memory, in case the class contains only 1 character (less
3066        than 256), because in that case the compiled code doesn't use the bit map.
3067        */
3068    
3069        memset(classbits, 0, 32 * sizeof(uschar));
3070    
3071  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3072      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3073      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3074        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3075  #endif  #endif
3076    
     /* 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));  
   
3077      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3078      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
3079      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. */  
3080    
3081      do      if (c != 0) do
3082        {        {
3083          const uschar *oldptr;
3084    
3085  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3086        if (utf8 && c > 127)        if (utf8 && c > 127)
3087          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3088          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3089          }          }
3090    
3091          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3092          data and reset the pointer. This is so that very large classes that
3093          contain a zillion UTF-8 characters no longer overwrite the work space
3094          (which is on the stack). */
3095    
3096          if (lengthptr != NULL)
3097            {
3098            *lengthptr += class_utf8data - class_utf8data_base;
3099            class_utf8data = class_utf8data_base;
3100            }
3101    
3102  #endif  #endif
3103    
3104        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3105    
3106        if (inescq)        if (inescq)
3107          {          {
3108          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3109            {            {
3110            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3111            ptr++;            ptr++;                            /* Skip the 'E' */
3112            continue;            continue;                         /* Carry on with next */
3113            }            }
3114          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3115          }          }
3116    
3117        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1801  for (;; ptr++) Line 3120  for (;; ptr++)
3120        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3121        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3122    
3123        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3124            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3125            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3126          {          {
3127          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3128          int posix_class, i;          int posix_class, taboffset, tabopt;
3129          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3130            uschar pbits[32];
3131    
3132          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3133            {            {
3134            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3135            goto FAILED;            goto FAILED;
3136            }            }
3137    
3138          ptr += 2;          ptr += 2;
3139          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3140            {            {
3141            local_negate = TRUE;            local_negate = TRUE;
3142              should_flip_negation = TRUE;  /* Note negative special */
3143            ptr++;            ptr++;
3144            }            }
3145    
# Line 1836  for (;; ptr++) Line 3157  for (;; ptr++)
3157          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3158            posix_class = 0;            posix_class = 0;
3159    
3160          /* 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
3161          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
3162          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
3163          white space chars afterwards. */          result into the bit map that is being built. */
3164    
3165          posix_class *= 3;          posix_class *= 3;
3166          for (i = 0; i < 3; i++)  
3167            /* Copy in the first table (always present) */
3168    
3169            memcpy(pbits, cbits + posix_class_maps[posix_class],
3170              32 * sizeof(uschar));
3171    
3172            /* If there is a second table, add or remove it as required. */
3173    
3174            taboffset = posix_class_maps[posix_class + 1];
3175            tabopt = posix_class_maps[posix_class + 2];
3176    
3177            if (taboffset >= 0)
3178            {            {
3179            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3180            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;  
             }  
3181            else            else
3182              {              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;  
             }  
3183            }            }
3184    
3185            /* Not see if we need to remove any special characters. An option
3186            value of 1 removes vertical space and 2 removes underscore. */
3187    
3188            if (tabopt < 0) tabopt = -tabopt;
3189            if (tabopt == 1) pbits[1] &= ~0x3c;
3190              else if (tabopt == 2) pbits[11] &= 0x7f;
3191    
3192            /* Add the POSIX table or its complement into the main table that is
3193            being built and we are done. */
3194    
3195            if (local_negate)
3196              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3197            else
3198              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3199    
3200          ptr = tempptr + 1;          ptr = tempptr + 1;
3201          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3202          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3203          }          }
3204    
3205        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3206        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
3207        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.
3208        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3209        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  
3210        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3211    
3212        if (c == '\\')        if (c == CHAR_BACKSLASH)
3213          {          {
3214          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3215            if (*errorcodeptr != 0) goto FAILED;
3216    
3217          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 */
3218          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 */
3219            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3220          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3221            {            {
3222            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3223              {              {
3224              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3225              }              }
3226            else inescq = TRUE;            else inescq = TRUE;
3227            continue;            continue;
3228            }            }
3229            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3230    
3231          if (c < 0)          if (c < 0)
3232            {            {
3233            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3234            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3235            switch (-c)  
3236              /* Save time by not doing this in the pre-compile phase. */
3237    
3238              if (lengthptr == NULL) switch (-c)
3239              {              {
3240              case ESC_d:              case ESC_d:
3241              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3242              continue;              continue;
3243    
3244              case ESC_D:              case ESC_D:
3245                should_flip_negation = TRUE;
3246              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3247              continue;              continue;
3248    
# Line 1910  for (;; ptr++) Line 3251  for (;; ptr++)
3251              continue;              continue;
3252    
3253              case ESC_W:              case ESC_W:
3254                should_flip_negation = TRUE;
3255              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3256              continue;              continue;
3257    
# Line 1919  for (;; ptr++) Line 3261  for (;; ptr++)
3261              continue;              continue;
3262    
3263              case ESC_S:              case ESC_S:
3264                should_flip_negation = TRUE;
3265              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3266              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3267              continue;              continue;
3268    
3269  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3270              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3271              case ESC_P:              }
3272    
3273              /* In the pre-compile phase, just do the recognition. */
3274    
3275              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3276                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3277    
3278              /* We need to deal with \H, \h, \V, and \v in both phases because
3279              they use extra memory. */
3280    
3281              if (-c == ESC_h)
3282                {
3283                SETBIT(classbits, 0x09); /* VT */
3284                SETBIT(classbits, 0x20); /* SPACE */
3285                SETBIT(classbits, 0xa0); /* NSBP */
3286    #ifdef SUPPORT_UTF8
3287                if (utf8)
3288                  {
3289                  class_utf8 = TRUE;
3290                  *class_utf8data++ = XCL_SINGLE;
3291                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3292                  *class_utf8data++ = XCL_SINGLE;
3293                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3294                  *class_utf8data++ = XCL_RANGE;
3295                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3296                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3297                  *class_utf8data++ = XCL_SINGLE;
3298                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3299                  *class_utf8data++ = XCL_SINGLE;
3300                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3301                  *class_utf8data++ = XCL_SINGLE;
3302                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3303                  }
3304    #endif
3305                continue;
3306                }
3307    
3308              if (-c == ESC_H)
3309                {
3310                for (c = 0; c < 32; c++)
3311                  {
3312                  int x = 0xff;
3313                  switch (c)
3314                    {
3315                    case 0x09/8: x ^= 1 << (0x09%8); break;
3316                    case 0x20/8: x ^= 1 << (0x20%8); break;
3317                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3318                    default: break;
3319                    }
3320                  classbits[c] |= x;
3321                  }
3322    
3323    #ifdef SUPPORT_UTF8
3324                if (utf8)
3325                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3326                class_utf8 = TRUE;                class_utf8 = TRUE;
3327                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3328                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3329                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3330                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3331                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3332                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3333                  *class_utf8data++ = XCL_RANGE;
3334                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3335                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3336                  *class_utf8data++ = XCL_RANGE;
3337                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3338                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3339                  *class_utf8data++ = XCL_RANGE;
3340                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3341                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3342                  *class_utf8data++ = XCL_RANGE;
3343                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3344                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3345                  *class_utf8data++ = XCL_RANGE;
3346                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3347                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3348                }                }
3349    #endif
3350              continue;              continue;
3351                }
3352    
3353              if (-c == ESC_v)
3354                {
3355                SETBIT(classbits, 0x0a); /* LF */
3356                SETBIT(classbits, 0x0b); /* VT */
3357                SETBIT(classbits, 0x0c); /* FF */
3358                SETBIT(classbits, 0x0d); /* CR */
3359                SETBIT(classbits, 0x85); /* NEL */
3360    #ifdef SUPPORT_UTF8
3361                if (utf8)
3362                  {
3363                  class_utf8 = TRUE;
3364                  *class_utf8data++ = XCL_RANGE;
3365                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3366                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3367                  }
3368  #endif  #endif
3369                continue;
3370                }
3371    
3372              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3373              strict mode. By default, for compatibility with Perl, they are              {
3374              treated as literals. */              for (c = 0; c < 32; c++)
3375                  {
3376                  int x = 0xff;
3377                  switch (c)
3378                    {
3379                    case 0x0a/8: x ^= 1 << (0x0a%8);
3380                                 x ^= 1 << (0x0b%8);
3381                                 x ^= 1 << (0x0c%8);
3382                                 x ^= 1 << (0x0d%8);
3383                                 break;
3384                    case 0x85/8: x ^= 1 << (0x85%8); break;
3385                    default: break;
3386                    }
3387                  classbits[c] |= x;
3388                  }
3389    
3390              default:  #ifdef SUPPORT_UTF8
3391              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3392                {                {
3393                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3394                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3395                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3396                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3397                  *class_utf8data++ = XCL_RANGE;
3398                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3399                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3400                }                }
3401              c = *ptr;              /* The final character */  #endif
3402              class_charcount -= 2;  /* Undo the default count from above */              continue;
3403                }
3404    
3405              /* We need to deal with \P and \p in both phases. */
3406    
3407    #ifdef SUPPORT_UCP
3408              if (-c == ESC_p || -c == ESC_P)
3409                {
3410                BOOL negated;
3411                int pdata;
3412                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3413                if (ptype < 0) goto FAILED;
3414                class_utf8 = TRUE;
3415                *class_utf8data++ = ((-c == ESC_p) != negated)?
3416                  XCL_PROP : XCL_NOTPROP;
3417                *class_utf8data++ = ptype;
3418                *class_utf8data++ = pdata;
3419                class_charcount -= 2;   /* Not a < 256 character */
3420                continue;
3421                }
3422    #endif
3423              /* Unrecognized escapes are faulted if PCRE is running in its
3424              strict mode. By default, for compatibility with Perl, they are
3425              treated as literals. */
3426    
3427              if ((options & PCRE_EXTRA) != 0)
3428                {
3429                *errorcodeptr = ERR7;
3430                goto FAILED;
3431              }              }
3432    
3433              class_charcount -= 2;  /* Undo the default count from above */
3434              c = *ptr;              /* Get the final character and fall through */
3435            }            }
3436    
3437          /* 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
3438          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3439    
3440          }   /* End of backslash handling */          }   /* End of backslash handling */
3441    
3442        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3443        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
3444        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3445          entirely. The code for handling \Q and \E is messy. */
3446    
3447          CHECK_RANGE:
3448          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3449            {
3450            inescq = FALSE;
3451            ptr += 2;
3452            }
3453    
3454        if (ptr[1] == '-' && ptr[2] != ']')        oldptr = ptr;
3455    
3456          /* Remember \r or \n */
3457    
3458          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3459    
3460          /* Check for range */
3461    
3462          if (!inescq && ptr[1] == CHAR_MINUS)
3463          {          {
3464          int d;          int d;
3465          ptr += 2;          ptr += 2;
3466            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3467    
3468            /* If we hit \Q (not followed by \E) at this point, go into escaped
3469            mode. */
3470    
3471            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3472              {
3473              ptr += 2;
3474              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3475                { ptr += 2; continue; }
3476              inescq = TRUE;
3477              break;
3478              }
3479    
3480            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3481              {
3482              ptr = oldptr;
3483              goto LONE_SINGLE_CHARACTER;
3484              }
3485    
3486  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3487          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3496  for (;; ptr++)
3496          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
3497          in such circumstances. */          in such circumstances. */
3498    
3499          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3500            {            {
3501            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3502            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3503    
3504            /* \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
3505            was literal */            special means the '-' was literal */
3506    
3507            if (d < 0)            if (d < 0)
3508              {              {
3509              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3510              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3511                else if (d == -ESC_R) d = CHAR_R; else
3512                {                {
3513                ptr = oldptr - 2;                ptr = oldptr;
3514                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3515                }                }
3516              }              }
3517            }            }
3518    
3519          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3520          the pre-pass. Optimize one-character ranges */          one-character ranges */
3521    
3522            if (d < c)
3523              {
3524              *errorcodeptr = ERR8;
3525              goto FAILED;
3526              }
3527    
3528          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3529    
3530            /* Remember \r or \n */
3531    
3532            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3533    
3534          /* 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
3535          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3536          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2022  for (;; ptr++) Line 3548  for (;; ptr++)
3548  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3549            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3550              {              {
3551              int occ, ocd;              unsigned int occ, ocd;
3552              int cc = c;              unsigned int cc = c;
3553              int origd = d;              unsigned int origd = d;
3554              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3555                {                {
3556                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3557                      ocd <= (unsigned int)d)
3558                    continue;                          /* Skip embedded ranges */
3559    
3560                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3561                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3562                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3563                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3564                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3565                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3566                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3567                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3568                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3569                  d = ocd;                  d = ocd;
3570                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3612  for (;; ptr++)
3612          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
3613          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3614    
3615          for (; c <= d; c++)          class_charcount += d - c + 1;
3616            class_lastchar = d;
3617    
3618            /* We can save a bit of time by skipping this in the pre-compile. */
3619    
3620            if (lengthptr == NULL) for (; c <= d; c++)
3621            {            {
3622            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3623            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3625  for (;; ptr++)
3625              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3626              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3627              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3628            }            }
3629    
3630          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3648  for (;; ptr++)
3648  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3649          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3650            {            {
3651            int chartype;            unsigned int othercase;
3652            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3653              {              {
3654              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3655              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3674  for (;; ptr++)
3674          }          }
3675        }        }
3676    
3677      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3678      loop. This "while" is the end of the "do" above. */  
3679        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3680    
3681        if (c == 0)                          /* Missing terminating ']' */
3682          {
3683          *errorcodeptr = ERR6;
3684          goto FAILED;
3685          }
3686    
3687    
3688    /* This code has been disabled because it would mean that \s counts as
3689    an explicit \r or \n reference, and that's not really what is wanted. Now
3690    we set the flag only if there is a literal "\r" or "\n" in the class. */
3691    
3692    #if 0
3693        /* Remember whether \r or \n are in this class */
3694    
3695        if (negate_class)
3696          {
3697          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3698          }
3699        else
3700          {
3701          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3702          }
3703    #endif
3704    
     while ((c = *(++ptr)) != ']' || inescq);  
3705    
3706      /* 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
3707      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
3708      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
3709      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3710      single-bytes only. This is an historical hangover. Maybe one day we can  
3711      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3712        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3713        operate on single-bytes only. This is an historical hangover. Maybe one day
3714        we can tidy these opcodes to handle multi-byte characters.
3715    
3716      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
3717      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2163  for (;; ptr++) Line 3721  for (;; ptr++)
3721      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3722    
3723  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3724      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3725            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3726  #else  #else
3727      if (class_charcount == 1)      if (class_charcount == 1)
3728  #endif  #endif
# Line 2209  for (;; ptr++) Line 3765  for (;; ptr++)
3765      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3766    
3767      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3768      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3769      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3770        the class, so any that were explicitly given as well can be ignored. If
3771        (when there are explicit characters > 255 that must be listed) there are no
3772        characters < 256, we can omit the bitmap in the actual compiled code. */
3773    
3774  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3775      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3776        {        {
3777        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3778        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3779        code += LINK_SIZE;        code += LINK_SIZE;
3780        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3781    
3782        /* 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;
3783        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3784    
3785        if (class_charcount > 0)        if (class_charcount > 0)
3786          {          {
3787          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3788            memmove(code + 32, code, class_utf8data - code);
3789          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3790          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;  
3791          }          }
3792          else code = class_utf8data;
3793    
3794        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3795    
# Line 2246  for (;; ptr++) Line 3798  for (;; ptr++)
3798        }        }
3799  #endif  #endif
3800    
3801      /* 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
3802      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
3803      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
3804      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3805    
3806        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3807      if (negate_class)      if (negate_class)
3808        {        {
3809        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3810        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3811        }        }
3812      else      else
3813        {        {
       *code++ = OP_CLASS;  
3814        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3815        }        }
3816      code += 32;      code += 32;
3817      break;      break;
3818    
3819    
3820        /* ===================================================================*/
3821      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3822      has been tested above. */      has been tested above. */
3823    
3824      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3825      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3826      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3827      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3828      goto REPEAT;      goto REPEAT;
3829    
3830      case '*':      case CHAR_ASTERISK:
3831      repeat_min = 0;      repeat_min = 0;
3832      repeat_max = -1;      repeat_max = -1;
3833      goto REPEAT;      goto REPEAT;
3834    
3835      case '+':      case CHAR_PLUS:
3836      repeat_min = 1;      repeat_min = 1;
3837      repeat_max = -1;      repeat_max = -1;
3838      goto REPEAT;      goto REPEAT;
3839    
3840      case '?':      case CHAR_QUESTION_MARK:
3841      repeat_min = 0;      repeat_min = 0;
3842      repeat_max = 1;      repeat_max = 1;
3843    
# Line 2318  for (;; ptr++) Line 3872  for (;; ptr++)
3872      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
3873      repeat type to the non-default. */      repeat type to the non-default. */
3874    
3875      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3876        {        {
3877        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3878        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3879        ptr++;        ptr++;
3880        }        }
3881      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3882        {        {
3883        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3884        ptr++;        ptr++;
3885        }        }
3886      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3887    
     /* 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;  
       }  
   
3888      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3889      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
3890      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2378  for (;; ptr++) Line 3918  for (;; ptr++)
3918          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3919          }          }
3920    
3921          /* If the repetition is unlimited, it pays to see if the next thing on
3922          the line is something that cannot possibly match this character. If so,
3923          automatically possessifying this item gains some performance in the case
3924          where the match fails. */
3925    
3926          if (!possessive_quantifier &&
3927              repeat_max < 0 &&
3928              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3929                options, cd))
3930            {
3931            repeat_type = 0;    /* Force greedy */
3932            possessive_quantifier = TRUE;
3933            }
3934    
3935        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3936        }        }
3937    
3938      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3939      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-
3940      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into