/[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 454 by ph10, Tue Sep 22 09:42:11 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When DEBUG is defined, we need the pcre_printint() function, which is also
57    used by pcretest. DEBUG is not defined when building a production library. */
58    
59    #ifdef DEBUG
60    #include "pcre_printint.src"
61    #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
179  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
225    
226  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
227  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
228  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
229    characters are removed, and for [:alpha:] and [:alnum:] the underscore
230    character is removed. The triples in the table consist of the base map offset,
231    second map offset or -1 if no second map, and a non-negative value for map
232    addition or a negative value for map subtraction (if there are two maps). The
233    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
234    remove vertical space characters, 2 => remove underscore. */
235    
236  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
237    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
238    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
239    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
240    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
241    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
242    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
243    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
244    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
245    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
246    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
247    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
248    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
249    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
250    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
251  };  };
252    
253    
254  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
255  are passed to the outside world. */  #define XSTRING(s) STRING(s)
256    
257  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
258    "no error",  are passed to the outside world. Do not ever re-use any error number, because
259    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
260    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
261    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
262    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
263    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    simply count through to the one we want - this isn't a performance issue
265    because these strings are used only when there is a compilation error. */
266    
267    static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
319    "two named groups have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
325  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327      /* 50 */
328      "repeated subpattern is too long\0"    /** DEAD **/
329      "octal value is greater than \\377 (not in UTF-8 mode)\0"
330      "internal error: overran compiling workspace\0"
331      "internal error: previously-checked referenced subpattern not found\0"
332      "DEFINE group contains more than one branch\0"
333      /* 55 */
334      "repeating a DEFINE group is not allowed\0"
335      "inconsistent NEWLINE options\0"
336      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
337      "a numbered reference must not be zero\0"
338      "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 220  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    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 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    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 417  static const unsigned char digitab[] =
417    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
418    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
420    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
421    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
422    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
423    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 451  static const unsigned char ebcdic_charta
451    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
452    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
453    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
454    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
455    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
456    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
457    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 478  static const unsigned char ebcdic_charta
478  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
479    
480  static BOOL  static BOOL
481    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
482      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
483    
484    
485    
486    /*************************************************
487    *            Find an error text                  *
488    *************************************************/
489    
490    /* The error texts are now all in one long string, to save on relocations. As
491    some of the text is of unknown length, we can't use a table of offsets.
492    Instead, just count through the strings. This is not a performance issue
493    because it happens only when there has been a compilation error.
494    
495    Argument:   the error number
496    Returns:    pointer to the error string
497    */
498    
499    static const char *
500    find_error_text(int n)
501    {
502    const char *s = error_texts;
503    for (; n > 0; n--) while (*s++ != 0) {};
504    return s;
505    }
506    
507    
508  /*************************************************  /*************************************************
# Line 342  static BOOL Line 511  static BOOL
511    
512  /* 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
513  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
514  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
515  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
516  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,
517    ptr is pointing at the \. On exit, it is on the final character of the escape
518    sequence.
519    
520  Arguments:  Arguments:
521    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 526  Arguments:
526    
527  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
528                   negative => a special escape sequence                   negative => a special escape sequence
529                   on error, errorptr is set                   on error, errorcodeptr is set
530  */  */
531    
532  static int  static int
533  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
534    int options, BOOL isclass)    int options, BOOL isclass)
535  {  {
536  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
537    const uschar *ptr = *ptrptr + 1;
538  int c, i;  int c, i;
539    
540    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
541    ptr--;                            /* Set pointer back to the last byte */
542    
543  /* 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. */
544    
 c = *(++ptr);  
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* 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
548  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.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 562  else if ((i = escapes[c - 0x48]) != 0)
562  else  else
563    {    {
564    const uschar *oldptr;    const uschar *oldptr;
565      BOOL braced, negated;
566    
567    switch (c)    switch (c)
568      {      {
569      /* 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
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580        /* \g must be followed by one of a number of specific things:
581    
582        (1) A number, either plain or braced. If positive, it is an absolute
583        backreference. If negative, it is a relative backreference. This is a Perl
584        5.10 feature.
585    
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596        case CHAR_g:
597        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606          {
607          const uschar *p;
608          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611            {
612            c = -ESC_k;
613            break;
614            }
615          braced = TRUE;
616          ptr++;
617          }
618        else braced = FALSE;
619    
620        if (ptr[1] == CHAR_MINUS)
621          {
622          negated = TRUE;
623          ptr++;
624          }
625        else negated = FALSE;
626    
627        c = 0;
628        while ((digitab[ptr[1]] & ctype_digit) != 0)
629          c = c * 10 + *(++ptr) - CHAR_0;
630    
631        if (c < 0)   /* Integer overflow */
632          {
633          *errorcodeptr = ERR61;
634          break;
635          }
636    
637        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638          {
639          *errorcodeptr = ERR57;
640          break;
641          }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647          }
648    
649        if (negated)
650          {
651          if (c > bracount)
652            {
653            *errorcodeptr = ERR15;
654            break;
655            }
656          c = bracount - (c - 1);
657          }
658    
659        c = -(ESC_REF + c);
660        break;
661    
662      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
663      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
664      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 671  else
671      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
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      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:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683          if (c < 0)    /* Integer overflow */
684            {
685            *errorcodeptr = ERR61;
686            break;
687            }
688        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
689          {          {
690          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 697  else
697      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.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 442  else Line 705  else
705        }        }
706    
707      /* \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
708      larger first octal digit. */      larger first octal digit. The original code used just to take the least
709        significant 8 bits of octal numbers (I think this is what early Perls used
710      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
711      c -= '0';      than 3 octal digits. */
712      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
713          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
714      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
715        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716            c = c * 8 + *(++ptr) - CHAR_0;
717        if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
720      /* \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
721      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
722        treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        register int count = 0;        int count = 0;
729    
730        c = 0;        c = 0;
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          int cc = *pt++;          register int cc = *pt++;
734            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736  #if !EBCDIC    /* ASCII coding */  
737          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
741          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745        if (*pt == '}')  
746          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
750          break;          break;
751          }          }
752    
753        /* 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
754        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
755        }        }
 #endif  
756    
757      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
758    
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
773    
774      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
775        This coding is ASCII-specific, but then the whole concept of \cx is
776        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
782        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
783        return 0;        break;
784        }        }
785    
786      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      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;  
788      c ^= 0x40;      c ^= 0x40;
789  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      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
798      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
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 828  escape sequence.
828  Argument:  Argument:
829    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
830    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
831      dptr           points to an int that is set to the detailed property value
832    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
833    
834  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
835  */  */
836    
837  static int  static int
838  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
839  {  {
840  int c, i, bot, top;  int c, i, bot, top;
841  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
842  char name[4];  char name[32];
843    
844  c = *(++ptr);  c = *(++ptr);
845  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
846    
847  *negptr = FALSE;  *negptr = FALSE;
848    
849  /* \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
850  preceded by ^ for negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
858      }      }
859    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    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;  
     }  
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 619  top = _pcre_utt_size; Line 884  top = _pcre_utt_size;
884    
885  while (bot < top)  while (bot < top)
886    {    {
887    i = (bot + top)/2;    i = (bot + top) >> 1;
888    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
889    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
890        {
891        *dptr = _pcre_utt[i].value;
892        return _pcre_utt[i].type;
893        }
894    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
895    }    }
896    
 UNKNOWN_RETURN:  
897  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
898  *ptrptr = ptr;  *ptrptr = ptr;
899  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 698  read_repeat_counts(const uschar *p, int Line 966  read_repeat_counts(const uschar *p, int
966  int min = 0;  int min = 0;
967  int max = -1;  int max = -1;
968    
969  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
970    an integer overflow. */
971    
972    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973    if (min < 0 || min > 65535)
974      {
975      *errorcodeptr = ERR5;
976      return p;
977      }
978    
979    /* Read the maximum value if there is one, and again do a paranoid on its size.
980    Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988        if (max < 0 || max > 65535)
989          {
990          *errorcodeptr = ERR5;
991          return p;
992          }
993      if (max < min)      if (max < min)
994        {        {
995        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 998  if (*p == '}') max = min; else
998      }      }
999    }    }
1000    
1001  /* 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
1002  pointer to the terminating '}'. */  '}'. */
1003    
1004  if (min > 65535 || max > 65535)  *minp = min;
1005    *errorcodeptr = ERR5;  *maxp = max;
1006  else  return p;
1007    }
1008    
1009    
1010    
1011    /*************************************************
1012    *  Subroutine for finding forward reference      *
1013    *************************************************/
1014    
1015    /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018    subpatterns, and counting them. If it finds a named pattern that matches the
1019    name it is given, it returns its number. Alternatively, if the name is NULL, it
1020    returns when it reaches a given numbered subpattern. We know that if (?P< is
1021    encountered, the name will be terminated by '>' because that is checked in the
1022    first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025    Arguments:
1026      ptrptr       address of the current character pointer (updated)
1027      cd           compile background data
1028      name         name to seek, or NULL if seeking a numbered subpattern
1029      lorn         name length, or subpattern number if name is NULL
1030      xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033    Returns:       the number of the named subpattern, or -1 if not found
1034    */
1035    
1036    static int
1037    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1038      BOOL xmode, int *count)
1039    {
1040    uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045    /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          term++;
1104          }
1105        }
1106    }    }
1107  return p;  
1108    /* Past any initial parenthesis handling, scan for parentheses or vertical
1109    bars. */
1110    
1111    for (; *ptr != 0; ptr++)
1112      {
1113      /* Skip over backslashed characters and also entire \Q...\E */
1114    
1115      if (*ptr == CHAR_BACKSLASH)
1116        {
1117        if (*(++ptr) == 0) goto FAIL_EXIT;
1118        if (*ptr == CHAR_Q) for (;;)
1119          {
1120          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1121          if (*ptr == 0) goto FAIL_EXIT;
1122          if (*(++ptr) == CHAR_E) break;
1123          }
1124        continue;
1125        }
1126    
1127      /* Skip over character classes; this logic must be similar to the way they
1128      are handled for real. If the first character is '^', skip it. Also, if the
1129      first few characters (either before or after ^) are \Q\E or \E we skip them
1130      too. This makes for compatibility with Perl. Note the use of STR macros to
1131      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1132    
1133      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134        {
1135        BOOL negate_class = FALSE;
1136        for (;;)
1137          {
1138          if (ptr[1] == CHAR_BACKSLASH)
1139            {
1140            if (ptr[2] == CHAR_E)
1141              ptr+= 2;
1142            else if (strncmp((const char *)ptr+2,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 4;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1149            {
1150            negate_class = TRUE;
1151            ptr++;
1152            }
1153          else break;
1154          }
1155    
1156        /* If the next character is ']', it is a data character that must be
1157        skipped, except in JavaScript compatibility mode. */
1158    
1159        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1160            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1161          ptr++;
1162    
1163        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1164          {
1165          if (*ptr == 0) return -1;
1166          if (*ptr == CHAR_BACKSLASH)
1167            {
1168            if (*(++ptr) == 0) goto FAIL_EXIT;
1169            if (*ptr == CHAR_Q) for (;;)
1170              {
1171              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1172              if (*ptr == 0) goto FAIL_EXIT;
1173              if (*(++ptr) == CHAR_E) break;
1174              }
1175            continue;
1176            }
1177          }
1178        continue;
1179        }
1180    
1181      /* Skip comments in /x mode */
1182    
1183      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1184        {
1185        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1186        if (*ptr == 0) goto FAIL_EXIT;
1187        continue;
1188        }
1189    
1190      /* Check for the special metacharacters */
1191    
1192      if (*ptr == CHAR_LEFT_PARENTHESIS)
1193        {
1194        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1195        if (rc > 0) return rc;
1196        if (*ptr == 0) goto FAIL_EXIT;
1197        }
1198    
1199      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200        {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204        }
1205    
1206      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1207        {
1208        if (*count > hwm_count) hwm_count = *count;
1209        *count = start_count;
1210        }
1211      }
1212    
1213    FAIL_EXIT:
1214    *ptrptr = ptr;
1215    return -1;
1216    }
1217    
1218    
1219    
1220    
1221    /*************************************************
1222    *       Find forward referenced subpattern       *
1223    *************************************************/
1224    
1225    /* This function scans along a pattern's text looking for capturing
1226    subpatterns, and counting them. If it finds a named pattern that matches the
1227    name it is given, it returns its number. Alternatively, if the name is NULL, it
1228    returns when it reaches a given numbered subpattern. This is used for forward
1229    references to subpatterns. We used to be able to start this scan from the
1230    current compiling point, using the current count value from cd->bracount, and
1231    do it all in a single loop, but the addition of the possibility of duplicate
1232    subpattern numbers means that we have to scan from the very start, in order to
1233    take account of such duplicates, and to use a recursive function to keep track
1234    of the different types of group.
1235    
1236    Arguments:
1237      cd           compile background data
1238      name         name to seek, or NULL if seeking a numbered subpattern
1239      lorn         name length, or subpattern number if name is NULL
1240      xmode        TRUE if we are in /x mode
1241    
1242    Returns:       the number of the found subpattern, or -1 if not found
1243    */
1244    
1245    static int
1246    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1247    {
1248    uschar *ptr = (uschar *)cd->start_pattern;
1249    int count = 0;
1250    int rc;
1251    
1252    /* If the pattern does not start with an opening parenthesis, the first call
1253    to find_parens_sub() will scan right to the end (if necessary). However, if it
1254    does start with a parenthesis, find_parens_sub() will return when it hits the
1255    matching closing parens. That is why we have to have a loop. */
1256    
1257    for (;;)
1258      {
1259      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1260      if (rc > 0 || *ptr++ == 0) break;
1261      }
1262    
1263    return rc;
1264  }  }
1265    
1266    
1267    
1268    
1269  /*************************************************  /*************************************************
1270  *      Find first significant op code            *  *      Find first significant op code            *
1271  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1315  for (;;)
1315    
1316      case OP_CALLOUT:      case OP_CALLOUT:
1317      case OP_CREF:      case OP_CREF:
1318      case OP_BRANUMBER:      case OP_RREF:
1319        case OP_DEF:
1320      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1321      break;      break;
1322    
# Line 793  for (;;) Line 1331  for (;;)
1331    
1332    
1333  /*************************************************  /*************************************************
1334  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1335  *************************************************/  *************************************************/
1336    
1337  /* 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,
1338  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.
1339  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
1340    temporarily terminated with OP_END when this function is called.
1341    
1342    This function is called when a backward assertion is encountered, so that if it
1343    fails, the error message can point to the correct place in the pattern.
1344    However, we cannot do this when the assertion contains subroutine calls,
1345    because they can be forward references. We solve this by remembering this case
1346    and doing the check at the end; a flag specifies which mode we are running in.
1347    
1348  Arguments:  Arguments:
1349    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1350    options  the compiling options    options  the compiling options
1351      atend    TRUE if called when the pattern is complete
1352      cd       the "compile data" structure
1353    
1354  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1355                 or -1 if there is no fixed length,
1356               or -2 if \C was encountered               or -2 if \C was encountered
1357                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1358  */  */
1359    
1360  static int  static int
1361  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1362  {  {
1363  int length = -1;  int length = -1;
1364    
# Line 822  branch, check the length against that of Line 1371  branch, check the length against that of
1371  for (;;)  for (;;)
1372    {    {
1373    int d;    int d;
1374      uschar *ce, *cs;
1375    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1376    switch (op)    switch (op)
1377      {      {
1378        case OP_CBRA:
1379      case OP_BRA:      case OP_BRA:
1380      case OP_ONCE:      case OP_ONCE:
1381      case OP_COND:      case OP_COND:
1382      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1383      if (d < 0) return d;      if (d < 0) return d;
1384      branchlength += d;      branchlength += d;
1385      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 852  for (;;) Line 1401  for (;;)
1401      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1402      branchlength = 0;      branchlength = 0;
1403      break;      break;
1404    
1405        /* A true recursion implies not fixed length, but a subroutine call may
1406        be OK. If the subroutine is a forward reference, we can't deal with
1407        it until the end of the pattern, so return -3. */
1408    
1409        case OP_RECURSE:
1410        if (!atend) return -3;
1411        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1412        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1413        if (cc > cs && cc < ce) return -1;                /* Recursion */
1414        d = find_fixedlength(cs + 2, options, atend, cd);
1415        if (d < 0) return d;
1416        branchlength += d;
1417        cc += 1 + LINK_SIZE;
1418        break;
1419    
1420      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1421    
# Line 865  for (;;) Line 1429  for (;;)
1429      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1430    
1431      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1432      case OP_CREF:      case OP_CREF:
1433        case OP_RREF:
1434        case OP_DEF:
1435      case OP_OPT:      case OP_OPT:
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1449  for (;;)
1449    
1450      case OP_CHAR:      case OP_CHAR:
1451      case OP_CHARNC:      case OP_CHARNC:
1452        case OP_NOT:
1453      branchlength++;      branchlength++;
1454      cc += 2;      cc += 2;
1455  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1456      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1457        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1458  #endif  #endif
1459      break;      break;
1460    
# Line 901  for (;;) Line 1465  for (;;)
1465      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1466      cc += 4;      cc += 4;
1467  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1468      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1469        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1470  #endif  #endif
1471      break;      break;
1472    
1473      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1474      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1475        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1476      cc += 4;      cc += 4;
1477      break;      break;
1478    
# Line 917  for (;;) Line 1480  for (;;)
1480    
1481      case OP_PROP:      case OP_PROP:
1482      case OP_NOTPROP:      case OP_NOTPROP:
1483      cc++;      cc += 2;
1484      /* Fall through */      /* Fall through */
1485    
1486      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1490  for (;;)
1490      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1491      case OP_WORDCHAR:      case OP_WORDCHAR:
1492      case OP_ANY:      case OP_ANY:
1493        case OP_ALLANY:
1494      branchlength++;      branchlength++;
1495      cc++;      cc++;
1496      break;      break;
# Line 981  for (;;) Line 1545  for (;;)
1545    
1546    
1547  /*************************************************  /*************************************************
1548  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1549  *************************************************/  *************************************************/
1550    
1551  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1552  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1553    instance of OP_REVERSE for a lookbehind.
1554    
1555  Arguments:  Arguments:
1556    code        points to start of expression    code        points to start of expression
1557    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1558    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1559    
1560  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
1561  */  */
# Line 998  Returns: pointer to the opcode for Line 1563  Returns: pointer to the opcode for
1563  static const uschar *  static const uschar *
1564  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1565  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1566  for (;;)  for (;;)
1567    {    {
1568    register int c = *code;    register int c = *code;
1569    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1570    else if (c > OP_BRA)  
1571      /* XCLASS is used for classes that cannot be represented just by a bit
1572      map. This includes negated single high-valued characters. The length in
1573      the table is zero; the actual length is stored in the compiled code. */
1574    
1575      if (c == OP_XCLASS) code += GET(code, 1);
1576    
1577      /* Handle recursion */
1578    
1579      else if (c == OP_REVERSE)
1580      {      {
1581      int n = c - OP_BRA;      if (number < 0) return (uschar *)code;
1582      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];  
1583      }      }
1584    else  
1585      /* Handle capturing bracket */
1586    
1587      else if (c == OP_CBRA)
1588      {      {
1589        int n = GET2(code, 1+LINK_SIZE);
1590        if (n == number) return (uschar *)code;
1591      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1592        }
1593    
1594  #ifdef SUPPORT_UTF8    /* Otherwise, we can get the item's length from the table, except that for
1595      repeated character types, we have to test for \p and \P, which have an extra
1596      /* In UTF-8 mode, opcodes that are followed by a character may be followed    two bytes of parameters. */
     by a multi-byte character. The length in the table is a minimum, so we have  
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
1597    
1598      if (utf8) switch(c)    else
1599        {
1600        switch(c)
1601        {        {
1602        case OP_CHAR:        case OP_TYPESTAR:
1603          case OP_TYPEMINSTAR:
1604          case OP_TYPEPLUS:
1605          case OP_TYPEMINPLUS:
1606          case OP_TYPEQUERY:
1607          case OP_TYPEMINQUERY:
1608          case OP_TYPEPOSSTAR:
1609          case OP_TYPEPOSPLUS:
1610          case OP_TYPEPOSQUERY:
1611          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1612          break;
1613    
1614          case OP_TYPEUPTO:
1615          case OP_TYPEMINUPTO:
1616          case OP_TYPEEXACT:
1617          case OP_TYPEPOSUPTO:
1618          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1619          break;
1620          }
1621    
1622        /* Add in the fixed length from the table */
1623    
1624        code += _pcre_OP_lengths[c];
1625    
1626      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1627      a multi-byte character. The length in the table is a minimum, so we have to
1628      arrange to skip the extra bytes. */
1629    
1630    #ifdef SUPPORT_UTF8
1631        if (utf8) switch(c)
1632          {
1633          case OP_CHAR:
1634        case OP_CHARNC:        case OP_CHARNC:
1635        case OP_EXACT:        case OP_EXACT:
1636        case OP_UPTO:        case OP_UPTO:
1637        case OP_MINUPTO:        case OP_MINUPTO:
1638          case OP_POSUPTO:
1639        case OP_STAR:        case OP_STAR:
1640        case OP_MINSTAR:        case OP_MINSTAR:
1641          case OP_POSSTAR:
1642        case OP_PLUS:        case OP_PLUS:
1643        case OP_MINPLUS:        case OP_MINPLUS:
1644          case OP_POSPLUS:
1645        case OP_QUERY:        case OP_QUERY:
1646        case OP_MINQUERY:        case OP_MINQUERY:
1647        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1648        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;  
1649        break;        break;
1650        }        }
1651    #else
1652        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1653  #endif  #endif
1654      }      }
1655    }    }
# Line 1072  Returns: pointer to the opcode for Line 1674  Returns: pointer to the opcode for
1674  static const uschar *  static const uschar *
1675  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1676  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1677  for (;;)  for (;;)
1678    {    {
1679    register int c = *code;    register int c = *code;
1680    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1681    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1682    else if (c > OP_BRA)  
1683      {    /* XCLASS is used for classes that cannot be represented just by a bit
1684      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1685      }    the table is zero; the actual length is stored in the compiled code. */
1686    
1687      if (c == OP_XCLASS) code += GET(code, 1);
1688    
1689      /* Otherwise, we can get the item's length from the table, except that for
1690      repeated character types, we have to test for \p and \P, which have an extra
1691      two bytes of parameters. */
1692    
1693    else    else
1694      {      {
1695      code += _pcre_OP_lengths[c];      switch(c)
1696          {
1697          case OP_TYPESTAR:
1698          case OP_TYPEMINSTAR:
1699          case OP_TYPEPLUS:
1700          case OP_TYPEMINPLUS:
1701          case OP_TYPEQUERY:
1702          case OP_TYPEMINQUERY:
1703          case OP_TYPEPOSSTAR:
1704          case OP_TYPEPOSPLUS:
1705          case OP_TYPEPOSQUERY:
1706          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1707          break;
1708    
1709  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1710          case OP_TYPEUPTO:
1711          case OP_TYPEMINUPTO:
1712          case OP_TYPEEXACT:
1713          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1714          break;
1715          }
1716    
1717        /* Add in the fixed length from the table */
1718    
1719        code += _pcre_OP_lengths[c];
1720    
1721      /* 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
1722      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
1723      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. */  
1724    
1725    #ifdef SUPPORT_UTF8
1726      if (utf8) switch(c)      if (utf8) switch(c)
1727        {        {
1728        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1730  for (;;)
1730        case OP_EXACT:        case OP_EXACT:
1731        case OP_UPTO:        case OP_UPTO:
1732        case OP_MINUPTO:        case OP_MINUPTO:
1733          case OP_POSUPTO:
1734        case OP_STAR:        case OP_STAR:
1735        case OP_MINSTAR:        case OP_MINSTAR:
1736          case OP_POSSTAR:
1737        case OP_PLUS:        case OP_PLUS:
1738        case OP_MINPLUS:        case OP_MINPLUS:
1739          case OP_POSPLUS:
1740        case OP_QUERY:        case OP_QUERY:
1741        case OP_MINQUERY:        case OP_MINQUERY:
1742        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1743        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;  
1744        break;        break;
1745        }        }
1746    #else
1747        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1748  #endif  #endif
1749      }      }
1750    }    }
# Line 1132  for (;;) Line 1757  for (;;)
1757  *************************************************/  *************************************************/
1758    
1759  /* 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
1760  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()
1761  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
1762  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
1763  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1764    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1765    bracket whose current branch will already have been scanned.
1766    
1767  Arguments:  Arguments:
1768    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1776  static BOOL
1776  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1777  {  {
1778  register int c;  register int c;
1779  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);
1780       code < endcode;       code < endcode;
1781       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1782    {    {
# Line 1157  for (code = first_significant_code(code Line 1784  for (code = first_significant_code(code
1784    
1785    c = *code;    c = *code;
1786    
1787    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1788      first_significant_code() with a TRUE final argument. */
1789    
1790      if (c == OP_ASSERT)
1791        {
1792        do code += GET(code, 1); while (*code == OP_ALT);
1793        c = *code;
1794        continue;
1795        }
1796    
1797      /* Groups with zero repeats can of course be empty; skip them. */
1798    
1799      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1800        {
1801        code += _pcre_OP_lengths[c];
1802        do code += GET(code, 1); while (*code == OP_ALT);
1803        c = *code;
1804        continue;
1805        }
1806    
1807      /* For other groups, scan the branches. */
1808    
1809      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1810      {      {
1811      BOOL empty_branch;      BOOL empty_branch;
1812      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1813    
1814      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1815        empty branch, so just skip over the conditional, because it could be empty.
1816        Otherwise, scan the individual branches of the group. */
1817    
1818      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1819        code += GET(code, 1);        code += GET(code, 1);
1820        else
1821          {
1822          empty_branch = FALSE;
1823          do
1824            {
1825            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1826              empty_branch = TRUE;
1827            code += GET(code, 1);
1828            }
1829          while (*code == OP_ALT);
1830          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1831        }        }
1832      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1833      c = *code;      c = *code;
1834        continue;
1835      }      }
1836    
1837    else switch (c)    /* Handle the other opcodes */
1838    
1839      switch (c)
1840      {      {
1841      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1842        cannot be represented just by a bit map. This includes negated single
1843        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1844        actual length is stored in the compiled code, so we must update "code"
1845        here. */
1846    
1847  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1848      case OP_XCLASS:      case OP_XCLASS:
1849      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1850      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1851  #endif  #endif
1852    
# Line 1227  for (code = first_significant_code(code Line 1890  for (code = first_significant_code(code
1890      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1891      case OP_WORDCHAR:      case OP_WORDCHAR:
1892      case OP_ANY:      case OP_ANY:
1893        case OP_ALLANY:
1894      case OP_ANYBYTE:      case OP_ANYBYTE:
1895      case OP_CHAR:      case OP_CHAR:
1896      case OP_CHARNC:      case OP_CHARNC:
1897      case OP_NOT:      case OP_NOT:
1898      case OP_PLUS:      case OP_PLUS:
1899      case OP_MINPLUS:      case OP_MINPLUS:
1900        case OP_POSPLUS:
1901      case OP_EXACT:      case OP_EXACT:
1902      case OP_NOTPLUS:      case OP_NOTPLUS:
1903      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1904        case OP_NOTPOSPLUS:
1905      case OP_NOTEXACT:      case OP_NOTEXACT:
1906      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1907      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1908        case OP_TYPEPOSPLUS:
1909      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1910      return FALSE;      return FALSE;
1911    
1912        /* These are going to continue, as they may be empty, but we have to
1913        fudge the length for the \p and \P cases. */
1914    
1915        case OP_TYPESTAR:
1916        case OP_TYPEMINSTAR:
1917        case OP_TYPEPOSSTAR:
1918        case OP_TYPEQUERY:
1919        case OP_TYPEMINQUERY:
1920        case OP_TYPEPOSQUERY:
1921        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1922        break;
1923    
1924        /* Same for these */
1925    
1926        case OP_TYPEUPTO:
1927        case OP_TYPEMINUPTO:
1928        case OP_TYPEPOSUPTO:
1929        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1930        break;
1931    
1932      /* End of branch */      /* End of branch */
1933    
1934      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1937  for (code = first_significant_code(code
1937      case OP_ALT:      case OP_ALT:
1938      return TRUE;      return TRUE;
1939    
1940      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1941      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1942    
1943  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1944      case OP_STAR:      case OP_STAR:
1945      case OP_MINSTAR:      case OP_MINSTAR:
1946        case OP_POSSTAR:
1947      case OP_QUERY:      case OP_QUERY:
1948      case OP_MINQUERY:      case OP_MINQUERY:
1949        case OP_POSQUERY:
1950        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1951        break;
1952    
1953      case OP_UPTO:      case OP_UPTO:
1954      case OP_MINUPTO:      case OP_MINUPTO:
1955      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
1956        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1957      break;      break;
1958  #endif  #endif
1959      }      }
# Line 1308  return TRUE; Line 2001  return TRUE;
2001  *************************************************/  *************************************************/
2002    
2003  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2004  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
2005  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2006  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2007    
2008    Originally, this function only recognized a sequence of letters between the
2009    terminators, but it seems that Perl recognizes any sequence of characters,
2010    though of course unknown POSIX names are subsequently rejected. Perl gives an
2011    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2012    didn't consider this to be a POSIX class. Likewise for [:1234:].
2013    
2014    The problem in trying to be exactly like Perl is in the handling of escapes. We
2015    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2016    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2017    below handles the special case of \], but does not try to do any other escape
2018    processing. This makes it different from Perl for cases such as [:l\ower:]
2019    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2020    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2021    I think.
2022    
2023  Argument:  Arguments:
2024    ptr      pointer to the initial [    ptr      pointer to the initial [
2025    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2026    
2027  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2028  */  */
2029    
2030  static BOOL  static BOOL
2031  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2032  {  {
2033  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2034  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2035  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2036    {    {
2037    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2038    return TRUE;      {
2039        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2040        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2041          {
2042          *endptr = ptr;
2043          return TRUE;
2044          }
2045        }
2046    }    }
2047  return FALSE;  return FALSE;
2048  }  }
# Line 1355  Returns: a value representing the na Line 2067  Returns: a value representing the na
2067  static int  static int
2068  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2069  {  {
2070    const char *pn = posix_names;
2071  register int yield = 0;  register int yield = 0;
2072  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2073    {    {
2074    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2075      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2076      pn += posix_name_lengths[yield] + 1;
2077    yield++;    yield++;
2078    }    }
2079  return -1;  return -1;
# Line 1374  return -1; Line 2088  return -1;
2088  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2089  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2090  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
2091  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
2092  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
2093  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
2094  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
2095  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2096    OP_END.
2097    
2098    This function has been extended with the possibility of forward references for
2099    recursions and subroutine calls. It must also check the list of such references
2100    for the group we are dealing with. If it finds that one of the recursions in
2101    the current group is on this list, it adjusts the offset in the list, not the
2102    value in the reference (which is a group number).
2103    
2104  Arguments:  Arguments:
2105    group      points to the start of the group    group      points to the start of the group
2106    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2107    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2108    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2109      save_hwm   the hwm forward reference pointer at the start of the group
2110    
2111  Returns:     nothing  Returns:     nothing
2112  */  */
2113    
2114  static void  static void
2115  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2116      uschar *save_hwm)
2117  {  {
2118  uschar *ptr = group;  uschar *ptr = group;
2119    
2120  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2121    {    {
2122    int offset = GET(ptr, 1);    int offset;
2123    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2124    
2125      /* See if this recursion is on the forward reference list. If so, adjust the
2126      reference. */
2127    
2128      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2129        {
2130        offset = GET(hc, 0);
2131        if (cd->start_code + offset == ptr + 1)
2132          {
2133          PUT(hc, 0, offset + adjust);
2134          break;
2135          }
2136        }
2137    
2138      /* Otherwise, adjust the recursion offset if it's after the start of this
2139      group. */
2140    
2141      if (hc >= cd->hwm)
2142        {
2143        offset = GET(ptr, 1);
2144        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2145        }
2146    
2147    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2148    }    }
2149  }  }
# Line 1475  Yield: TRUE when range returned; Line 2222  Yield: TRUE when range returned;
2222  */  */
2223    
2224  static BOOL  static BOOL
2225  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2226      unsigned int *odptr)
2227  {  {
2228  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2229    
2230  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2231    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2232    
2233  if (c > d) return FALSE;  if (c > d) return FALSE;
2234    
# Line 1492  next = othercase + 1; Line 2237  next = othercase + 1;
2237    
2238  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2239    {    {
2240    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2241    next++;    next++;
2242    }    }
2243    
# Line 1506  return TRUE; Line 2249  return TRUE;
2249  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2250    
2251    
2252    
2253    /*************************************************
2254    *     Check if auto-possessifying is possible    *
2255    *************************************************/
2256    
2257    /* This function is called for unlimited repeats of certain items, to see
2258    whether the next thing could possibly match the repeated item. If not, it makes
2259    sense to automatically possessify the repeated item.
2260    
2261    Arguments:
2262      op_code       the repeated op code
2263      this          data for this item, depends on the opcode
2264      utf8          TRUE in UTF-8 mode
2265      utf8_char     used for utf8 character bytes, NULL if not relevant
2266      ptr           next character in pattern
2267      options       options bits
2268      cd            contains pointers to tables etc.
2269    
2270    Returns:        TRUE if possessifying is wanted
2271    */
2272    
2273    static BOOL
2274    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2275      const uschar *ptr, int options, compile_data *cd)
2276    {
2277    int next;
2278    
2279    /* Skip whitespace and comments in extended mode */
2280    
2281    if ((options & PCRE_EXTENDED) != 0)
2282      {
2283      for (;;)
2284        {
2285        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2286        if (*ptr == CHAR_NUMBER_SIGN)
2287          {
2288          while (*(++ptr) != 0)
2289            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2290          }
2291        else break;
2292        }
2293      }
2294    
2295    /* If the next item is one that we can handle, get its value. A non-negative
2296    value is a character, a negative value is an escape value. */
2297    
2298    if (*ptr == CHAR_BACKSLASH)
2299      {
2300      int temperrorcode = 0;
2301      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2302      if (temperrorcode != 0) return FALSE;
2303      ptr++;    /* Point after the escape sequence */
2304      }
2305    
2306    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2307      {
2308    #ifdef SUPPORT_UTF8
2309      if (utf8) { GETCHARINC(next, ptr); } else
2310    #endif
2311      next = *ptr++;
2312      }
2313    
2314    else return FALSE;
2315    
2316    /* Skip whitespace and comments in extended mode */
2317    
2318    if ((options & PCRE_EXTENDED) != 0)
2319      {
2320      for (;;)
2321        {
2322        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2323        if (*ptr == CHAR_NUMBER_SIGN)
2324          {
2325          while (*(++ptr) != 0)
2326            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2327          }
2328        else break;
2329        }
2330      }
2331    
2332    /* If the next thing is itself optional, we have to give up. */
2333    
2334    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2335      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2336        return FALSE;
2337    
2338    /* Now compare the next item with the previous opcode. If the previous is a
2339    positive single character match, "item" either contains the character or, if
2340    "item" is greater than 127 in utf8 mode, the character's bytes are in
2341    utf8_char. */
2342    
2343    
2344    /* Handle cases when the next item is a character. */
2345    
2346    if (next >= 0) switch(op_code)
2347      {
2348      case OP_CHAR:
2349    #ifdef SUPPORT_UTF8
2350      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2351    #else
2352      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2353    #endif
2354      return item != next;
2355    
2356      /* For CHARNC (caseless character) we must check the other case. If we have
2357      Unicode property support, we can use it to test the other case of
2358      high-valued characters. */
2359    
2360      case OP_CHARNC:
2361    #ifdef SUPPORT_UTF8
2362      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2363    #endif
2364      if (item == next) return FALSE;
2365    #ifdef SUPPORT_UTF8
2366      if (utf8)
2367        {
2368        unsigned int othercase;
2369        if (next < 128) othercase = cd->fcc[next]; else
2370    #ifdef SUPPORT_UCP
2371        othercase = UCD_OTHERCASE((unsigned int)next);
2372    #else
2373        othercase = NOTACHAR;
2374    #endif
2375        return (unsigned int)item != othercase;
2376        }
2377      else
2378    #endif  /* SUPPORT_UTF8 */
2379      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2380    
2381      /* For OP_NOT, "item" must be a single-byte character. */
2382    
2383      case OP_NOT:
2384      if (item == next) return TRUE;
2385      if ((options & PCRE_CASELESS) == 0) return FALSE;
2386    #ifdef SUPPORT_UTF8
2387      if (utf8)
2388        {
2389        unsigned int othercase;
2390        if (next < 128) othercase = cd->fcc[next]; else
2391    #ifdef SUPPORT_UCP
2392        othercase = UCD_OTHERCASE(next);
2393    #else
2394        othercase = NOTACHAR;
2395    #endif
2396        return (unsigned int)item == othercase;
2397        }
2398      else
2399    #endif  /* SUPPORT_UTF8 */
2400      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2401    
2402      case OP_DIGIT:
2403      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2404    
2405      case OP_NOT_DIGIT:
2406      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2407    
2408      case OP_WHITESPACE:
2409      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2410    
2411      case OP_NOT_WHITESPACE:
2412      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2413    
2414      case OP_WORDCHAR:
2415      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2416    
2417      case OP_NOT_WORDCHAR:
2418      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2419    
2420      case OP_HSPACE:
2421      case OP_NOT_HSPACE:
2422      switch(next)
2423        {
2424        case 0x09:
2425        case 0x20:
2426        case 0xa0:
2427        case 0x1680:
2428        case 0x180e:
2429        case 0x2000:
2430        case 0x2001:
2431        case 0x2002:
2432        case 0x2003:
2433        case 0x2004:
2434        case 0x2005:
2435        case 0x2006:
2436        case 0x2007:
2437        case 0x2008:
2438        case 0x2009:
2439        case 0x200A:
2440        case 0x202f:
2441        case 0x205f:
2442        case 0x3000:
2443        return op_code != OP_HSPACE;
2444        default:
2445        return op_code == OP_HSPACE;
2446        }
2447    
2448      case OP_VSPACE:
2449      case OP_NOT_VSPACE:
2450      switch(next)
2451        {
2452        case 0x0a:
2453        case 0x0b:
2454        case 0x0c:
2455        case 0x0d:
2456        case 0x85:
2457        case 0x2028:
2458        case 0x2029:
2459        return op_code != OP_VSPACE;
2460        default:
2461        return op_code == OP_VSPACE;
2462        }
2463    
2464      default:
2465      return FALSE;
2466      }
2467    
2468    
2469    /* Handle the case when the next item is \d, \s, etc. */
2470    
2471    switch(op_code)
2472      {
2473      case OP_CHAR:
2474      case OP_CHARNC:
2475    #ifdef SUPPORT_UTF8
2476      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2477    #endif
2478      switch(-next)
2479        {
2480        case ESC_d:
2481        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2482    
2483        case ESC_D:
2484        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2485    
2486        case ESC_s:
2487        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2488    
2489        case ESC_S:
2490        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2491    
2492        case ESC_w:
2493        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2494    
2495        case ESC_W:
2496        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2497    
2498        case ESC_h:
2499        case ESC_H:
2500        switch(item)
2501          {
2502          case 0x09:
2503          case 0x20:
2504          case 0xa0:
2505          case 0x1680:
2506          case 0x180e:
2507          case 0x2000:
2508          case 0x2001:
2509          case 0x2002:
2510          case 0x2003:
2511          case 0x2004:
2512          case 0x2005:
2513          case 0x2006:
2514          case 0x2007:
2515          case 0x2008:
2516          case 0x2009:
2517          case 0x200A:
2518          case 0x202f:
2519          case 0x205f:
2520          case 0x3000:
2521          return -next != ESC_h;
2522          default:
2523          return -next == ESC_h;
2524          }
2525    
2526        case ESC_v:
2527        case ESC_V:
2528        switch(item)
2529          {
2530          case 0x0a:
2531          case 0x0b:
2532          case 0x0c:
2533          case 0x0d:
2534          case 0x85:
2535          case 0x2028:
2536          case 0x2029:
2537          return -next != ESC_v;
2538          default:
2539          return -next == ESC_v;
2540          }
2541    
2542        default:
2543        return FALSE;
2544        }
2545    
2546      case OP_DIGIT:
2547      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2548             next == -ESC_h || next == -ESC_v;
2549    
2550      case OP_NOT_DIGIT:
2551      return next == -ESC_d;
2552    
2553      case OP_WHITESPACE:
2554      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2555    
2556      case OP_NOT_WHITESPACE:
2557      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2558    
2559      case OP_HSPACE:
2560      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2561    
2562      case OP_NOT_HSPACE:
2563      return next == -ESC_h;
2564    
2565      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2566      case OP_VSPACE:
2567      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2568    
2569      case OP_NOT_VSPACE:
2570      return next == -ESC_v;
2571    
2572      case OP_WORDCHAR:
2573      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2574    
2575      case OP_NOT_WORDCHAR:
2576      return next == -ESC_w || next == -ESC_d;
2577    
2578      default:
2579      return FALSE;
2580      }
2581    
2582    /* Control does not reach here */
2583    }
2584    
2585    
2586    
2587  /*************************************************  /*************************************************
2588  *           Compile one branch                   *  *           Compile one branch                   *
2589  *************************************************/  *************************************************/
2590    
2591  /* 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
2592  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
2593  bits.  bits. This function is used during the pre-compile phase when we are trying
2594    to find out the amount of memory needed, as well as during the real compile
2595    phase. The value of lengthptr distinguishes the two phases.
2596    
2597  Arguments:  Arguments:
2598    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2599    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2600    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2601    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2603  Arguments:
2603    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2604    bcptr          points to current branch chain    bcptr          points to current branch chain
2605    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2606      lengthptr      NULL during the real compile phase
2607                     points to length accumulator during pre-compile phase
2608    
2609  Returns:         TRUE on success  Returns:         TRUE on success
2610                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2611  */  */
2612    
2613  static BOOL  static BOOL
2614  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2615    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2616    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2617  {  {
2618  int repeat_type, op_type;  int repeat_type, op_type;
2619  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 2622  int greedy_default, greedy_non_default;
2622  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2623  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2624  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2625  int options = *optionsptr;  int options = *optionsptr;
2626  int after_manual_callout = 0;  int after_manual_callout = 0;
2627    int length_prevgroup = 0;
2628  register int c;  register int c;
2629  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2630    uschar *last_code = code;
2631    uschar *orig_code = code;
2632  uschar *tempcode;  uschar *tempcode;
2633  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2634  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2636  const uschar *ptr = *ptrptr;
2636  const uschar *tempptr;  const uschar *tempptr;
2637  uschar *previous = NULL;  uschar *previous = NULL;
2638  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2639    uschar *save_hwm = NULL;
2640  uschar classbits[32];  uschar classbits[32];
2641    
2642  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2643  BOOL class_utf8;  BOOL class_utf8;
2644  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2645  uschar *class_utf8data;  uschar *class_utf8data;
2646    uschar *class_utf8data_base;
2647  uschar utf8_char[6];  uschar utf8_char[6];
2648  #else  #else
2649  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2650    uschar *utf8_char = NULL;
2651    #endif
2652    
2653    #ifdef DEBUG
2654    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2655  #endif  #endif
2656    
2657  /* 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 2683  req_caseopt = ((options & PCRE_CASELESS)
2683  for (;; ptr++)  for (;; ptr++)
2684    {    {
2685    BOOL negate_class;    BOOL negate_class;
2686      BOOL should_flip_negation;
2687    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2688    BOOL is_quantifier;    BOOL is_quantifier;
2689      BOOL is_recurse;
2690      BOOL reset_bracount;
2691    int class_charcount;    int class_charcount;
2692    int class_lastchar;    int class_lastchar;
2693    int newoptions;    int newoptions;
2694    int recno;    int recno;
2695      int refsign;
2696    int skipbytes;    int skipbytes;
2697    int subreqbyte;    int subreqbyte;
2698    int subfirstbyte;    int subfirstbyte;
2699      int terminator;
2700    int mclength;    int mclength;
2701    uschar mcbuffer[8];    uschar mcbuffer[8];
2702    
2703    /* Next byte in the pattern */    /* Get next byte in the pattern */
2704    
2705    c = *ptr;    c = *ptr;
2706    
2707      /* If we are in the pre-compile phase, accumulate the length used for the
2708      previous cycle of this loop. */
2709    
2710      if (lengthptr != NULL)
2711        {
2712    #ifdef DEBUG
2713        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2714    #endif
2715        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2716          {
2717          *errorcodeptr = ERR52;
2718          goto FAILED;
2719          }
2720    
2721        /* There is at least one situation where code goes backwards: this is the
2722        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2723        the class is simply eliminated. However, it is created first, so we have to
2724        allow memory for it. Therefore, don't ever reduce the length at this point.
2725        */
2726    
2727        if (code < last_code) code = last_code;
2728    
2729        /* Paranoid check for integer overflow */
2730    
2731        if (OFLOW_MAX - *lengthptr < code - last_code)
2732          {
2733          *errorcodeptr = ERR20;
2734          goto FAILED;
2735          }
2736    
2737        *lengthptr += code - last_code;
2738        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2739    
2740        /* If "previous" is set and it is not at the start of the work space, move
2741        it back to there, in order to avoid filling up the work space. Otherwise,
2742        if "previous" is NULL, reset the current code pointer to the start. */
2743    
2744        if (previous != NULL)
2745          {
2746          if (previous > orig_code)
2747            {
2748            memmove(orig_code, previous, code - previous);
2749            code -= previous - orig_code;
2750            previous = orig_code;
2751            }
2752          }
2753        else code = orig_code;
2754    
2755        /* Remember where this code item starts so we can pick up the length
2756        next time round. */
2757    
2758        last_code = code;
2759        }
2760    
2761      /* In the real compile phase, just check the workspace used by the forward
2762      reference list. */
2763    
2764      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2765        {
2766        *errorcodeptr = ERR52;
2767        goto FAILED;
2768        }
2769    
2770    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
2771    
2772    if (inescq && c != 0)    if (inescq && c != 0)
2773      {      {
2774      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2775        {        {
2776        inescq = FALSE;        inescq = FALSE;
2777        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 2781  for (;; ptr++)
2781        {        {
2782        if (previous_callout != NULL)        if (previous_callout != NULL)
2783          {          {
2784          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2785              complete_callout(previous_callout, ptr, cd);
2786          previous_callout = NULL;          previous_callout = NULL;
2787          }          }
2788        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2797  for (;; ptr++)
2797    /* 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
2798    a quantifier. */    a quantifier. */
2799    
2800    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2801      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2802        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2803    
2804    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2805         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2806      {      {
2807      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2808          complete_callout(previous_callout, ptr, cd);
2809      previous_callout = NULL;      previous_callout = NULL;
2810      }      }
2811    
# Line 1653  for (;; ptr++) Line 2814  for (;; ptr++)
2814    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2815      {      {
2816      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2817      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2818        {        {
2819        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2820        on the Macintosh. */          {
2821        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2822        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2823          if (*ptr != 0) continue;
2824    
2825          /* Else fall through to handle end of string */
2826          c = 0;
2827        }        }
2828      }      }
2829    
# Line 1672  for (;; ptr++) Line 2837  for (;; ptr++)
2837    
2838    switch(c)    switch(c)
2839      {      {
2840      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2841        case 0:                        /* The branch terminates at string end */
2842      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2843      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2844      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2845      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2846      *codeptr = code;      *codeptr = code;
2847      *ptrptr = ptr;      *ptrptr = ptr;
2848      return TRUE;      if (lengthptr != NULL)
2849          {
2850          if (OFLOW_MAX - *lengthptr < code - last_code)
2851            {
2852            *errorcodeptr = ERR20;
2853            goto FAILED;
2854            }
2855          *lengthptr += code - last_code;   /* To include callout length */
2856          DPRINTF((">> end branch\n"));
2857          }
2858        return TRUE;
2859    
2860    
2861        /* ===================================================================*/
2862      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2863      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2864    
2865      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2866      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2867        {        {
2868        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2871  for (;; ptr++)
2871      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2872      break;      break;
2873    
2874      case '$':      case CHAR_DOLLAR_SIGN:
2875      previous = NULL;      previous = NULL;
2876      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2877      break;      break;
# Line 1703  for (;; ptr++) Line 2879  for (;; ptr++)
2879      /* 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
2880      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2881    
2882      case '.':      case CHAR_DOT:
2883      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2884      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2885      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2886      previous = code;      previous = code;
2887      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2888      break;      break;
2889    
2890      /* Character classes. If the included characters are all < 255 in value, we  
2891      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2892      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
2893      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
2894      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2895        map as usual, then invert it at the end. However, we use a different opcode
2896        so that data characters > 255 can be handled correctly.
2897    
2898      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2899      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,
2900      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2901      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.
     */  
2902    
2903      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2904        default (Perl) mode, it is treated as a data character. */
2905    
2906        case CHAR_RIGHT_SQUARE_BRACKET:
2907        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2908          {
2909          *errorcodeptr = ERR64;
2910          goto FAILED;
2911          }
2912        goto NORMAL_CHAR;
2913    
2914        case CHAR_LEFT_SQUARE_BRACKET:
2915      previous = code;      previous = code;
2916    
2917      /* 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
2918      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. */
2919    
2920      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2921          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2922            check_posix_syntax(ptr, &tempptr))
2923        {        {
2924        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2925        goto FAILED;        goto FAILED;
2926        }        }
2927    
2928      /* 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,
2929        if the first few characters (either before or after ^) are \Q\E or \E we
2930        skip them too. This makes for compatibility with Perl. */
2931    
2932      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2933        for (;;)
2934        {        {
       negate_class = TRUE;  
2935        c = *(++ptr);        c = *(++ptr);
2936          if (c == CHAR_BACKSLASH)
2937            {
2938            if (ptr[1] == CHAR_E)
2939              ptr++;
2940            else if (strncmp((const char *)ptr+1,
2941                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2942              ptr += 3;
2943            else
2944              break;
2945            }
2946          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2947            negate_class = TRUE;
2948          else break;
2949        }        }
2950      else  
2951        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2952        an initial ']' is taken as a data character -- the code below handles
2953        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2954        [^] must match any character, so generate OP_ALLANY. */
2955    
2956        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2957            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2958        {        {
2959        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2960          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2961          zerofirstbyte = firstbyte;
2962          break;
2963        }        }
2964    
2965        /* If a class contains a negative special such as \S, we need to flip the
2966        negation flag at the end, so that support for characters > 255 works
2967        correctly (they are all included in the class). */
2968    
2969        should_flip_negation = FALSE;
2970    
2971      /* 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
2972      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
2973      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2974    
2975      class_charcount = 0;      class_charcount = 0;
2976      class_lastchar = -1;      class_lastchar = -1;
2977    
2978        /* Initialize the 32-char bit map to all zeros. We build the map in a
2979        temporary bit of memory, in case the class contains only 1 character (less
2980        than 256), because in that case the compiled code doesn't use the bit map.
2981        */
2982    
2983        memset(classbits, 0, 32 * sizeof(uschar));
2984    
2985  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2986      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2987      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2988        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2989  #endif  #endif
2990    
     /* 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));  
   
2991      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2992      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
2993      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. */  
2994    
2995      do      if (c != 0) do
2996        {        {
2997          const uschar *oldptr;
2998    
2999  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3000        if (utf8 && c > 127)        if (utf8 && c > 127)
3001          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3002          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3003          }          }
3004    
3005          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3006          data and reset the pointer. This is so that very large classes that
3007          contain a zillion UTF-8 characters no longer overwrite the work space
3008          (which is on the stack). */
3009    
3010          if (lengthptr != NULL)
3011            {
3012            *lengthptr += class_utf8data - class_utf8data_base;
3013            class_utf8data = class_utf8data_base;
3014            }
3015    
3016  #endif  #endif
3017    
3018        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3019    
3020        if (inescq)        if (inescq)
3021          {          {
3022          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3023            {            {
3024            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3025            ptr++;            ptr++;                            /* Skip the 'E' */
3026            continue;            continue;                         /* Carry on with next */
3027            }            }
3028          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3029          }          }
3030    
3031        /* 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 3034  for (;; ptr++)
3034        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3035        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3036    
3037        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3038            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3039            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3040          {          {
3041          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3042          int posix_class, i;          int posix_class, taboffset, tabopt;
3043          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3044            uschar pbits[32];
3045    
3046          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3047            {            {
3048            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3049            goto FAILED;            goto FAILED;
3050            }            }
3051    
3052          ptr += 2;          ptr += 2;
3053          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3054            {            {
3055            local_negate = TRUE;            local_negate = TRUE;
3056              should_flip_negation = TRUE;  /* Note negative special */
3057            ptr++;            ptr++;
3058            }            }
3059    
# Line 1836  for (;; ptr++) Line 3071  for (;; ptr++)
3071          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3072            posix_class = 0;            posix_class = 0;
3073    
3074          /* 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
3075          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
3076          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
3077          white space chars afterwards. */          result into the bit map that is being built. */
3078    
3079          posix_class *= 3;          posix_class *= 3;
3080          for (i = 0; i < 3; i++)  
3081            /* Copy in the first table (always present) */
3082    
3083            memcpy(pbits, cbits + posix_class_maps[posix_class],
3084              32 * sizeof(uschar));
3085    
3086            /* If there is a second table, add or remove it as required. */
3087    
3088            taboffset = posix_class_maps[posix_class + 1];
3089            tabopt = posix_class_maps[posix_class + 2];
3090    
3091            if (taboffset >= 0)
3092            {            {
3093            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3094            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;  
             }  
3095            else            else
3096              {              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;  
             }  
3097            }            }
3098    
3099            /* Not see if we need to remove any special characters. An option
3100            value of 1 removes vertical space and 2 removes underscore. */
3101    
3102            if (tabopt < 0) tabopt = -tabopt;
3103            if (tabopt == 1) pbits[1] &= ~0x3c;
3104              else if (tabopt == 2) pbits[11] &= 0x7f;
3105    
3106            /* Add the POSIX table or its complement into the main table that is
3107            being built and we are done. */
3108    
3109            if (local_negate)
3110              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3111            else
3112              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3113    
3114          ptr = tempptr + 1;          ptr = tempptr + 1;
3115          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3116          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3117          }          }
3118    
3119        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3120        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
3121        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.
3122        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3123        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  
3124        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3125    
3126        if (c == '\\')        if (c == CHAR_BACKSLASH)
3127          {          {
3128          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3129            if (*errorcodeptr != 0) goto FAILED;
3130    
3131          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 */
3132          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 */
3133            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3134          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3135            {            {
3136            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3137              {              {
3138              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3139              }              }
3140            else inescq = TRUE;            else inescq = TRUE;
3141            continue;            continue;
3142            }            }
3143            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3144    
3145          if (c < 0)          if (c < 0)
3146            {            {
3147            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3148            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3149            switch (-c)  
3150              /* Save time by not doing this in the pre-compile phase. */
3151    
3152              if (lengthptr == NULL) switch (-c)
3153              {              {
3154              case ESC_d:              case ESC_d:
3155              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3156              continue;              continue;
3157    
3158              case ESC_D:              case ESC_D:
3159                should_flip_negation = TRUE;
3160              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3161              continue;              continue;
3162    
# Line 1910  for (;; ptr++) Line 3165  for (;; ptr++)
3165              continue;              continue;
3166    
3167              case ESC_W:              case ESC_W:
3168                should_flip_negation = TRUE;
3169              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3170              continue;              continue;
3171    
# Line 1919  for (;; ptr++) Line 3175  for (;; ptr++)
3175              continue;              continue;
3176    
3177              case ESC_S:              case ESC_S:
3178                should_flip_negation = TRUE;
3179              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3180              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3181              continue;              continue;
3182    
3183  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3184              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3185              case ESC_P:              }
3186    
3187              /* In the pre-compile phase, just do the recognition. */
3188    
3189              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3190                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3191    
3192              /* We need to deal with \H, \h, \V, and \v in both phases because
3193              they use extra memory. */
3194    
3195              if (-c == ESC_h)
3196                {
3197                SETBIT(classbits, 0x09); /* VT */
3198                SETBIT(classbits, 0x20); /* SPACE */
3199                SETBIT(classbits, 0xa0); /* NSBP */
3200    #ifdef SUPPORT_UTF8
3201                if (utf8)
3202                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3203                class_utf8 = TRUE;                class_utf8 = TRUE;
3204                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
3205                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3206                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
3207                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3208                  *class_utf8data++ = XCL_RANGE;
3209                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3210                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3211                  *class_utf8data++ = XCL_SINGLE;
3212                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3213                  *class_utf8data++ = XCL_SINGLE;
3214                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3215                  *class_utf8data++ = XCL_SINGLE;
3216                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3217                }                }
3218    #endif
3219              continue;              continue;
3220                }
3221    
3222              if (-c == ESC_H)
3223                {
3224                for (c = 0; c < 32; c++)
3225                  {
3226                  int x = 0xff;
3227                  switch (c)
3228                    {
3229                    case 0x09/8: x ^= 1 << (0x09%8); break;
3230                    case 0x20/8: x ^= 1 << (0x20%8); break;
3231                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3232                    default: break;
3233                    }
3234                  classbits[c] |= x;
3235                  }
3236    
3237    #ifdef SUPPORT_UTF8
3238                if (utf8)
3239                  {
3240                  class_utf8 = TRUE;
3241                  *class_utf8data++ = XCL_RANGE;
3242                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3243                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3244                  *class_utf8data++ = XCL_RANGE;
3245                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3246                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3247                  *class_utf8data++ = XCL_RANGE;
3248                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3249                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3250                  *class_utf8data++ = XCL_RANGE;
3251                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3252                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3253                  *class_utf8data++ = XCL_RANGE;
3254                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3255                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3256                  *class_utf8data++ = XCL_RANGE;
3257                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3258                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3259                  *class_utf8data++ = XCL_RANGE;
3260                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3261                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3262                  }
3263  #endif  #endif
3264                continue;
3265                }
3266    
3267              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_v)
3268              strict mode. By default, for compatibility with Perl, they are              {
3269              treated as literals. */              SETBIT(classbits, 0x0a); /* LF */
3270                SETBIT(classbits, 0x0b); /* VT */
3271                SETBIT(classbits, 0x0c); /* FF */
3272                SETBIT(classbits, 0x0d); /* CR */
3273                SETBIT(classbits, 0x85); /* NEL */
3274    #ifdef SUPPORT_UTF8
3275                if (utf8)
3276                  {
3277                  class_utf8 = TRUE;
3278                  *class_utf8data++ = XCL_RANGE;
3279                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3280                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3281                  }
3282    #endif
3283                continue;
3284                }
3285    
3286              default:            if (-c == ESC_V)
3287              if ((options & PCRE_EXTRA) != 0)              {
3288                for (c = 0; c < 32; c++)
3289                {                {
3290                *errorcodeptr = ERR7;                int x = 0xff;
3291                goto FAILED;                switch (c)
3292                    {
3293                    case 0x0a/8: x ^= 1 << (0x0a%8);
3294                                 x ^= 1 << (0x0b%8);
3295                                 x ^= 1 << (0x0c%8);
3296                                 x ^= 1 << (0x0d%8);
3297                                 break;
3298                    case 0x85/8: x ^= 1 << (0x85%8); break;
3299                    default: break;
3300                    }
3301                  classbits[c] |= x;
3302                  }
3303    
3304    #ifdef SUPPORT_UTF8
3305                if (utf8)
3306                  {
3307                  class_utf8 = TRUE;
3308                  *class_utf8data++ = XCL_RANGE;
3309                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3310                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3311                  *class_utf8data++ = XCL_RANGE;
3312                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3313                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3314                }                }
3315              c = *ptr;              /* The final character */  #endif
3316              class_charcount -= 2;  /* Undo the default count from above */              continue;
3317                }
3318    
3319              /* We need to deal with \P and \p in both phases. */
3320    
3321    #ifdef SUPPORT_UCP
3322              if (-c == ESC_p || -c == ESC_P)
3323                {
3324                BOOL negated;
3325                int pdata;
3326                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3327                if (ptype < 0) goto FAILED;
3328                class_utf8 = TRUE;
3329                *class_utf8data++ = ((-c == ESC_p) != negated)?
3330                  XCL_PROP : XCL_NOTPROP;
3331                *class_utf8data++ = ptype;
3332                *class_utf8data++ = pdata;
3333                class_charcount -= 2;   /* Not a < 256 character */
3334                continue;
3335              }              }
3336    #endif
3337              /* Unrecognized escapes are faulted if PCRE is running in its
3338              strict mode. By default, for compatibility with Perl, they are
3339              treated as literals. */
3340    
3341              if ((options & PCRE_EXTRA) != 0)
3342                {
3343                *errorcodeptr = ERR7;
3344                goto FAILED;
3345                }
3346    
3347              class_charcount -= 2;  /* Undo the default count from above */
3348              c = *ptr;              /* Get the final character and fall through */
3349            }            }
3350    
3351          /* 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
3352          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3353    
3354          }   /* End of backslash handling */          }   /* End of backslash handling */
3355    
3356        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3357        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
3358        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3359          entirely. The code for handling \Q and \E is messy. */
3360    
3361          CHECK_RANGE:
3362          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3363            {
3364            inescq = FALSE;
3365            ptr += 2;
3366            }
3367    
3368          oldptr = ptr;
3369    
3370          /* Remember \r or \n */
3371    
3372          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3373    
3374          /* Check for range */
3375    
3376        if (ptr[1] == '-' && ptr[2] != ']')        if (!inescq && ptr[1] == CHAR_MINUS)
3377          {          {
3378          int d;          int d;
3379          ptr += 2;          ptr += 2;
3380            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3381    
3382            /* If we hit \Q (not followed by \E) at this point, go into escaped
3383            mode. */
3384    
3385            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3386              {
3387              ptr += 2;
3388              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3389                { ptr += 2; continue; }
3390              inescq = TRUE;
3391              break;
3392              }
3393    
3394            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3395              {
3396              ptr = oldptr;
3397              goto LONE_SINGLE_CHARACTER;
3398              }
3399    
3400  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3401          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3410  for (;; ptr++)
3410          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
3411          in such circumstances. */          in such circumstances. */
3412    
3413          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3414            {            {
3415            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3416            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3417    
3418            /* \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
3419            was literal */            special means the '-' was literal */
3420    
3421            if (d < 0)            if (d < 0)
3422              {              {
3423              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3424              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3425                else if (d == -ESC_R) d = CHAR_R; else
3426                {                {
3427                ptr = oldptr - 2;                ptr = oldptr;
3428                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3429                }                }
3430              }              }
3431            }            }
3432    
3433          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3434          the pre-pass. Optimize one-character ranges */          one-character ranges */
3435    
3436            if (d < c)
3437              {
3438              *errorcodeptr = ERR8;
3439              goto FAILED;
3440              }
3441    
3442          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3443    
3444            /* Remember \r or \n */
3445    
3446            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3447    
3448          /* 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
3449          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3450          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 3462  for (;; ptr++)
3462  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3463            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3464              {              {
3465              int occ, ocd;              unsigned int occ, ocd;
3466              int cc = c;              unsigned int cc = c;
3467              int origd = d;              unsigned int origd = d;
3468              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3469                {                {
3470                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3471                      ocd <= (unsigned int)d)
3472                    continue;                          /* Skip embedded ranges */
3473    
3474                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3475                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3476                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3477                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3478                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3479                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3480                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3481                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3482                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3483                  d = ocd;                  d = ocd;
3484                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3526  for (;; ptr++)
3526          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
3527          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3528    
3529          for (; c <= d; c++)          class_charcount += d - c + 1;
3530            class_lastchar = d;
3531    
3532            /* We can save a bit of time by skipping this in the pre-compile. */
3533    
3534            if (lengthptr == NULL) for (; c <= d; c++)
3535            {            {
3536            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3537            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3539  for (;; ptr++)
3539              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3540              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3541              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3542            }            }
3543    
3544          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3562  for (;; ptr++)
3562  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3563          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3564            {            {
3565            int chartype;            unsigned int othercase;
3566            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3567              {              {
3568              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3569              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3588  for (;; ptr++)
3588          }          }
3589        }        }
3590    
3591      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3592      loop. This "while" is the end of the "do" above. */  
3593        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3594    
3595        if (c == 0)                          /* Missing terminating ']' */
3596          {
3597          *errorcodeptr = ERR6;
3598          goto FAILED;
3599          }
3600    
3601    
3602    /* This code has been disabled because it would mean that \s counts as
3603    an explicit \r or \n reference, and that's not really what is wanted. Now
3604    we set the flag only if there is a literal "\r" or "\n" in the class. */
3605    
3606    #if 0
3607        /* Remember whether \r or \n are in this class */
3608    
3609        if (negate_class)
3610          {
3611          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3612          }
3613        else
3614          {
3615          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3616          }
3617    #endif
3618    
     while ((c = *(++ptr)) != ']' || inescq);  
3619    
3620      /* 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
3621      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
3622      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
3623      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3624      single-bytes only. This is an historical hangover. Maybe one day we can  
3625      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3626        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3627        operate on single-bytes only. This is an historical hangover. Maybe one day
3628        we can tidy these opcodes to handle multi-byte characters.
3629    
3630      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
3631      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 3635  for (;; ptr++)
3635      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3636    
3637  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3638      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3639            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3640  #else  #else
3641      if (class_charcount == 1)      if (class_charcount == 1)
3642  #endif  #endif
# Line 2209  for (;; ptr++) Line 3679  for (;; ptr++)
3679      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3680    
3681      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3682      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3683      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3684        the class, so any that were explicitly given as well can be ignored. If
3685        (when there are explicit characters > 255 that must be listed) there are no
3686        characters < 256, we can omit the bitmap in the actual compiled code. */
3687    
3688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3689      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3690        {        {
3691        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3692        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3693        code += LINK_SIZE;        code += LINK_SIZE;
3694        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3695    
3696        /* 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;
3697        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3698    
3699        if (class_charcount > 0)        if (class_charcount > 0)
3700          {          {
3701          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3702            memmove(code + 32, code, class_utf8data - code);
3703          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3704          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;  
3705          }          }
3706          else code = class_utf8data;
3707    
3708        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3709    
# Line 2246  for (;; ptr++) Line 3712  for (;; ptr++)
3712        }        }
3713  #endif  #endif
3714    
3715      /* 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
3716      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
3717      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
3718      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3719    
3720        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3721      if (negate_class)      if (negate_class)
3722        {        {
3723        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3724        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3725        }        }
3726      else      else
3727        {        {
       *code++ = OP_CLASS;  
3728        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3729        }        }
3730      code += 32;      code += 32;
3731      break;      break;
3732    
3733    
3734        /* ===================================================================*/
3735      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3736      has been tested above. */      has been tested above. */
3737    
3738      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3739      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3740      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3741      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3742      goto REPEAT;      goto REPEAT;
3743    
3744      case '*':      case CHAR_ASTERISK:
3745      repeat_min = 0;      repeat_min = 0;
3746      repeat_max = -1;      repeat_max = -1;
3747      goto REPEAT;      goto REPEAT;
3748    
3749      case '+':      case CHAR_PLUS:
3750      repeat_min = 1;      repeat_min = 1;
3751      repeat_max = -1;      repeat_max = -1;
3752      goto REPEAT;      goto REPEAT;
3753    
3754      case '?':      case CHAR_QUESTION_MARK:
3755      repeat_min = 0;      repeat_min = 0;
3756      repeat_max = 1;      repeat_max = 1;
3757    
# Line 2318  for (;; ptr++) Line 3786  for (;; ptr++)
3786      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
3787      repeat type to the non-default. */      repeat type to the non-default. */
3788    
3789      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3790        {        {
3791        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3792        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3793        ptr++;        ptr++;
3794        }        }
3795      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3796        {        {
3797        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3798        ptr++;        ptr++;
3799        }        }
3800      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3801    
     /* 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;  
       }  
   
3802      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3803      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
3804      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 3832  for (;; ptr++)
3832          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3833          }          }
3834    
3835          /* If the repetition is unlimited, it pays to see if the next thing on
3836          the line is something that cannot possibly match this character. If so,
3837          automatically possessifying this item gains some performance in the case
3838          where the match fails. */
3839    
3840          if (!possessive_quantifier &&
3841              repeat_max < 0 &&
3842              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3843                options, cd))
3844            {
3845            repeat_type = 0;    /* Force greedy */
3846            possessive_quantifier = TRUE;
3847            }
3848    
3849        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3850        }        }
3851    
3852      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3853      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-
3854      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3855      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3856        currently used only for single-byte chars. */
3857    
3858      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3859        {        {
3860        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3861        c = previous[1];        c = previous[1];
3862          if (!possessive_quantifier &&
3863              repeat_max < 0 &&
3864              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3865            {
3866            repeat_type = 0;    /* Force greedy */
3867            possessive_quantifier = TRUE;
3868            }
3869        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3870        }        }
3871    
# Line 2403  for (;; ptr++) Line 3879  for (;; ptr++)
3879      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3880        {        {
3881        uschar *oldcode;        uschar *oldcode;
3882        int prop_type;        int prop_type, prop_value;
3883        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3884        c = *previous;        c = *previous;
3885    
3886          if (!possessive_quantifier &&
3887              repeat_max < 0 &&
3888              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3889            {
3890            repeat_type = 0;    /* Force greedy */
3891            possessive_quantifier = TRUE;
3892            }
3893    
3894        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3895        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3896          previous[1] : -1;          {
3897            prop_type = previous[1];
3898            prop_value = previous[2];
3899            }
3900          else prop_type = prop_value = -1;
3901    
3902        oldcode = code;        oldcode = code;
3903        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2419  for (;; ptr++) Line 3907  for (;; ptr++)
3907    
3908        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3909    
3910          /*--------------------------------------------------------------------*/
3911          /* This code is obsolete from release 8.00; the restriction was finally
3912          removed: */
3913    
3914        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3915        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3916    
3917        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3918          /*--------------------------------------------------------------------*/
3919    
3920        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3921    
# Line 2443  for (;; ptr++) Line 3936  for (;; ptr++)
3936          }          }
3937    
3938        /* A repeat minimum of 1 is optimized into some special cases. If the        /* A repeat minimum of 1 is optimized into some special cases. If the
3939        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3940        left in place and, if the maximum is greater than 1, we use OP_UPTO with        left in place and, if the maximum is greater than 1, we use OP_UPTO with
3941        one less than the maximum. */        one less than the maximum. */
3942    
# Line 2470  for (;; ptr++) Line 3963  for (;; ptr++)
3963    
3964          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3965          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
3966          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3967          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3968          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3969    
# Line 2486  for (;; ptr++) Line 3979  for (;; ptr++)
3979  #endif  #endif
3980              {              {
3981              *code++ = c;              *code++ = c;
3982              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3983                  {
3984                  *code++ = prop_type;
3985                  *code++ = prop_value;
3986                  }
3987              }              }
3988            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3989            }            }
3990    
3991          /* Else insert an UPTO if the max is greater than the min, again          /* Else insert an UPTO if the max is greater than the min, again
3992          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3993            UPTO is just for 1 instance, we can use QUERY instead. */
3994    
3995          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3996            {            {
# Line 2505  for (;; ptr++) Line 4003  for (;; ptr++)
4003            else            else
4004  #endif  #endif
4005            *code++ = c;            *code++ = c;
4006            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
4007                {
4008                *code++ = prop_type;
4009                *code++ = prop_value;
4010                }
4011            repeat_max -= repeat_min;            repeat_max -= repeat_min;
4012            *code++ = OP_UPTO + repeat_type;  
4013            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
4014                {
4015                *code++ = OP_QUERY + repeat_type;
4016                }
4017              else
4018                {
4019                *code++ = OP_UPTO + repeat_type;
4020                PUT2INC(code, 0, repeat_max);
4021                }
4022            }            }
4023          }          }
4024    
# Line 2524  for (;; ptr++) Line 4034  for (;; ptr++)
4034  #endif  #endif
4035        *code++ = c;        *code++ = c;
4036    
4037        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
4038        defines the required property. */        define the required property. */
4039    
4040  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4041        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
4042            {
4043            *code++ = prop_type;
4044            *code++ = prop_value;
4045            }
4046  #endif  #endif
4047        }        }
4048    
#