/[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 391 by ph10, Tue Mar 17 21:16:01 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  static const short int escapes[] = {  
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */  /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */  in UTF-8 mode. */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */  
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  static const short int escapes[] = {
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       0,                       0,
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       0,                       0,
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       0,                       0,
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       0,                       0,
111       0,      0, -ESC_z                                            /* x - z */       CHAR_COLON,              CHAR_SEMICOLON,
112         CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113         CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114         CHAR_COMMERCIAL_AT,      -ESC_A,
115         -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    *       Find forward referenced subpattern       *
1013    *************************************************/
1014    
1015    /* This function scans along a pattern's text looking for capturing
1016    subpatterns, and counting them. If it finds a named pattern that matches the
1017    name it is given, it returns its number. Alternatively, if the name is NULL, it
1018    returns when it reaches a given numbered subpattern. This is used for forward
1019    references to subpatterns. We know that if (?P< is encountered, the name will
1020    be terminated by '>' because that is checked in the first pass.
1021    
1022    Arguments:
1023      ptr          current position in the pattern
1024      cd           compile background data
1025      name         name to seek, or NULL if seeking a numbered subpattern
1026      lorn         name length, or subpattern number if name is NULL
1027      xmode        TRUE if we are in /x mode
1028    
1029    Returns:       the number of the named subpattern, or -1 if not found
1030    */
1031    
1032    static int
1033    find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,
1034      BOOL xmode)
1035    {
1036    const uschar *thisname;
1037    int count = cd->bracount;
1038    
1039    for (; *ptr != 0; ptr++)
1040    {    {
1041    *minp = min;    int term;
1042    *maxp = max;  
1043      /* Skip over backslashed characters and also entire \Q...\E */
1044    
1045      if (*ptr == CHAR_BACKSLASH)
1046        {
1047        if (*(++ptr) == 0) return -1;
1048        if (*ptr == CHAR_Q) for (;;)
1049          {
1050          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1051          if (*ptr == 0) return -1;
1052          if (*(++ptr) == CHAR_E) break;
1053          }
1054        continue;
1055        }
1056    
1057      /* Skip over character classes; this logic must be similar to the way they
1058      are handled for real. If the first character is '^', skip it. Also, if the
1059      first few characters (either before or after ^) are \Q\E or \E we skip them
1060      too. This makes for compatibility with Perl. Note the use of STR macros to
1061      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1062    
1063      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1064        {
1065        BOOL negate_class = FALSE;
1066        for (;;)
1067          {
1068          int c = *(++ptr);
1069          if (c == CHAR_BACKSLASH)
1070            {
1071            if (ptr[1] == CHAR_E)
1072              ptr++;
1073            else if (strncmp((const char *)ptr+1,
1074                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1075              ptr += 3;
1076            else
1077              break;
1078            }
1079          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
1080            negate_class = TRUE;
1081          else break;
1082          }
1083    
1084        /* If the next character is ']', it is a data character that must be
1085        skipped, except in JavaScript compatibility mode. */
1086    
1087        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1088            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1089          ptr++;
1090    
1091        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1092          {
1093          if (*ptr == 0) return -1;
1094          if (*ptr == CHAR_BACKSLASH)
1095            {
1096            if (*(++ptr) == 0) return -1;
1097            if (*ptr == CHAR_Q) for (;;)
1098              {
1099              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1100              if (*ptr == 0) return -1;
1101              if (*(++ptr) == CHAR_E) break;
1102              }
1103            continue;
1104            }
1105          }
1106        continue;
1107        }
1108    
1109      /* Skip comments in /x mode */
1110    
1111      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1112        {
1113        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1114        if (*ptr == 0) return -1;
1115        continue;
1116        }
1117    
1118      /* An opening parens must now be a real metacharacter */
1119    
1120      if (*ptr != CHAR_LEFT_PARENTHESIS) continue;
1121      if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1122        {
1123        count++;
1124        if (name == NULL && count == lorn) return count;
1125        continue;
1126        }
1127    
1128      ptr += 2;
1129      if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1130    
1131      /* We have to disambiguate (?<! and (?<= from (?<name> */
1132    
1133      if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||
1134          ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)
1135        continue;
1136    
1137      count++;
1138    
1139      if (name == NULL && count == lorn) return count;
1140      term = *ptr++;
1141      if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1142      thisname = ptr;
1143      while (*ptr != term) ptr++;
1144      if (name != NULL && lorn == ptr - thisname &&
1145          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1146        return count;
1147    }    }
1148  return p;  
1149    return -1;
1150  }  }
1151    
1152    
# Line 778  for (;;) Line 1200  for (;;)
1200    
1201      case OP_CALLOUT:      case OP_CALLOUT:
1202      case OP_CREF:      case OP_CREF:
1203      case OP_BRANUMBER:      case OP_RREF:
1204        case OP_DEF:
1205      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1206      break;      break;
1207    
# Line 823  for (;;) Line 1246  for (;;)
1246    {    {
1247    int d;    int d;
1248    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1249    switch (op)    switch (op)
1250      {      {
1251        case OP_CBRA:
1252      case OP_BRA:      case OP_BRA:
1253      case OP_ONCE:      case OP_ONCE:
1254      case OP_COND:      case OP_COND:
1255      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1256      if (d < 0) return d;      if (d < 0) return d;
1257      branchlength += d;      branchlength += d;
1258      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1287  for (;;)
1287      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1288    
1289      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1290      case OP_CREF:      case OP_CREF:
1291        case OP_RREF:
1292        case OP_DEF:
1293      case OP_OPT:      case OP_OPT:
1294      case OP_CALLOUT:      case OP_CALLOUT:
1295      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1307  for (;;)
1307    
1308      case OP_CHAR:      case OP_CHAR:
1309      case OP_CHARNC:      case OP_CHARNC:
1310        case OP_NOT:
1311      branchlength++;      branchlength++;
1312      cc += 2;      cc += 2;
1313  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 910  for (;;) Line 1334  for (;;)
1334    
1335      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1336      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1337        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1338      cc += 4;      cc += 4;
1339      break;      break;
1340    
# Line 917  for (;;) Line 1342  for (;;)
1342    
1343      case OP_PROP:      case OP_PROP:
1344      case OP_NOTPROP:      case OP_NOTPROP:
1345      cc++;      cc += 2;
1346      /* Fall through */      /* Fall through */
1347    
1348      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1352  for (;;)
1352      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1353      case OP_WORDCHAR:      case OP_WORDCHAR:
1354      case OP_ANY:      case OP_ANY:
1355        case OP_ALLANY:
1356      branchlength++;      branchlength++;
1357      cc++;      cc++;
1358      break;      break;
# Line 998  Returns:      pointer to the opcode for Line 1424  Returns:      pointer to the opcode for
1424  static const uschar *  static const uschar *
1425  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1426  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1427  for (;;)  for (;;)
1428    {    {
1429    register int c = *code;    register int c = *code;
1430    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1431    else if (c > OP_BRA)  
1432      /* XCLASS is used for classes that cannot be represented just by a bit
1433      map. This includes negated single high-valued characters. The length in
1434      the table is zero; the actual length is stored in the compiled code. */
1435    
1436      if (c == OP_XCLASS) code += GET(code, 1);
1437    
1438      /* Handle capturing bracket */
1439    
1440      else if (c == OP_CBRA)
1441      {      {
1442      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1443      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1444      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1445      }      }
1446    
1447      /* Otherwise, we can get the item's length from the table, except that for
1448      repeated character types, we have to test for \p and \P, which have an extra
1449      two bytes of parameters. */
1450    
1451    else    else
1452      {      {
1453      code += _pcre_OP_lengths[c];      switch(c)
1454          {
1455          case OP_TYPESTAR:
1456          case OP_TYPEMINSTAR:
1457          case OP_TYPEPLUS:
1458          case OP_TYPEMINPLUS:
1459          case OP_TYPEQUERY:
1460          case OP_TYPEMINQUERY:
1461          case OP_TYPEPOSSTAR:
1462          case OP_TYPEPOSPLUS:
1463          case OP_TYPEPOSQUERY:
1464          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1465          break;
1466    
1467  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1468          case OP_TYPEMINUPTO:
1469          case OP_TYPEEXACT:
1470          case OP_TYPEPOSUPTO:
1471          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1472          break;
1473          }
1474    
1475      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1476      by a multi-byte character. The length in the table is a minimum, so we have  
1477      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1478      can use relatively efficient code. */  
1479      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1480      a multi-byte character. The length in the table is a minimum, so we have to
1481      arrange to skip the extra bytes. */
1482    
1483    #ifdef SUPPORT_UTF8
1484      if (utf8) switch(c)      if (utf8) switch(c)
1485        {        {
1486        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1488  for (;;)
1488        case OP_EXACT:        case OP_EXACT:
1489        case OP_UPTO:        case OP_UPTO:
1490        case OP_MINUPTO:        case OP_MINUPTO:
1491          case OP_POSUPTO:
1492        case OP_STAR:        case OP_STAR:
1493        case OP_MINSTAR:        case OP_MINSTAR:
1494          case OP_POSSTAR:
1495        case OP_PLUS:        case OP_PLUS:
1496        case OP_MINPLUS:        case OP_MINPLUS:
1497          case OP_POSPLUS:
1498        case OP_QUERY:        case OP_QUERY:
1499        case OP_MINQUERY:        case OP_MINQUERY:
1500        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1501        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;  
1502        break;        break;
1503        }        }
1504    #else
1505        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1506  #endif  #endif
1507      }      }
1508    }    }
# Line 1072  Returns:      pointer to the opcode for Line 1527  Returns:      pointer to the opcode for
1527  static const uschar *  static const uschar *
1528  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1529  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1530  for (;;)  for (;;)
1531    {    {
1532    register int c = *code;    register int c = *code;
1533    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1534    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1535    else if (c > OP_BRA)  
1536      {    /* XCLASS is used for classes that cannot be represented just by a bit
1537      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1538      }    the table is zero; the actual length is stored in the compiled code. */
1539    
1540      if (c == OP_XCLASS) code += GET(code, 1);
1541    
1542      /* Otherwise, we can get the item's length from the table, except that for
1543      repeated character types, we have to test for \p and \P, which have an extra
1544      two bytes of parameters. */
1545    
1546    else    else
1547      {      {
1548      code += _pcre_OP_lengths[c];      switch(c)
1549          {
1550          case OP_TYPESTAR:
1551          case OP_TYPEMINSTAR:
1552          case OP_TYPEPLUS:
1553          case OP_TYPEMINPLUS:
1554          case OP_TYPEQUERY:
1555          case OP_TYPEMINQUERY:
1556          case OP_TYPEPOSSTAR:
1557          case OP_TYPEPOSPLUS:
1558          case OP_TYPEPOSQUERY:
1559          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1560          break;
1561    
1562  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1563          case OP_TYPEUPTO:
1564          case OP_TYPEMINUPTO:
1565          case OP_TYPEEXACT:
1566          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1567          break;
1568          }
1569    
1570        /* Add in the fixed length from the table */
1571    
1572        code += _pcre_OP_lengths[c];
1573    
1574      /* 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
1575      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
1576      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. */  
1577    
1578    #ifdef SUPPORT_UTF8
1579      if (utf8) switch(c)      if (utf8) switch(c)
1580        {        {
1581        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1583  for (;;)
1583        case OP_EXACT:        case OP_EXACT:
1584        case OP_UPTO:        case OP_UPTO:
1585        case OP_MINUPTO:        case OP_MINUPTO:
1586          case OP_POSUPTO:
1587        case OP_STAR:        case OP_STAR:
1588        case OP_MINSTAR:        case OP_MINSTAR:
1589          case OP_POSSTAR:
1590        case OP_PLUS:        case OP_PLUS:
1591        case OP_MINPLUS:        case OP_MINPLUS:
1592          case OP_POSPLUS:
1593        case OP_QUERY:        case OP_QUERY:
1594        case OP_MINQUERY:        case OP_MINQUERY:
1595        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1596        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;  
1597        break;        break;
1598        }        }
1599    #else
1600        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1601  #endif  #endif
1602      }      }
1603    }    }
# Line 1132  for (;;) Line 1610  for (;;)
1610  *************************************************/  *************************************************/
1611    
1612  /* 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
1613  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()
1614  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
1615  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
1616  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1617    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1618    bracket whose current branch will already have been scanned.
1619    
1620  Arguments:  Arguments:
1621    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1629  static BOOL
1629  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1630  {  {
1631  register int c;  register int c;
1632  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);
1633       code < endcode;       code < endcode;
1634       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1635    {    {
# Line 1157  for (code = first_significant_code(code Line 1637  for (code = first_significant_code(code
1637    
1638    c = *code;    c = *code;
1639    
1640    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1641      first_significant_code() with a TRUE final argument. */
1642    
1643      if (c == OP_ASSERT)
1644        {
1645        do code += GET(code, 1); while (*code == OP_ALT);
1646        c = *code;
1647        continue;
1648        }
1649    
1650      /* Groups with zero repeats can of course be empty; skip them. */
1651    
1652      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1653        {
1654        code += _pcre_OP_lengths[c];
1655        do code += GET(code, 1); while (*code == OP_ALT);
1656        c = *code;
1657        continue;
1658        }
1659    
1660      /* For other groups, scan the branches. */
1661    
1662      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1663      {      {
1664      BOOL empty_branch;      BOOL empty_branch;
1665      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1173  for (code = first_significant_code(code Line 1675  for (code = first_significant_code(code
1675        }        }
1676      while (*code == OP_ALT);      while (*code == OP_ALT);
1677      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1678      c = *code;      c = *code;
1679        continue;
1680      }      }
1681    
1682    else switch (c)    /* Handle the other opcodes */
1683    
1684      switch (c)
1685      {      {
1686      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1687        cannot be represented just by a bit map. This includes negated single
1688        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1689        actual length is stored in the compiled code, so we must update "code"
1690        here. */
1691    
1692  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1693      case OP_XCLASS:      case OP_XCLASS:
1694      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1695      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1696  #endif  #endif
1697    
# Line 1227  for (code = first_significant_code(code Line 1735  for (code = first_significant_code(code
1735      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1736      case OP_WORDCHAR:      case OP_WORDCHAR:
1737      case OP_ANY:      case OP_ANY:
1738        case OP_ALLANY:
1739      case OP_ANYBYTE:      case OP_ANYBYTE:
1740      case OP_CHAR:      case OP_CHAR:
1741      case OP_CHARNC:      case OP_CHARNC:
1742      case OP_NOT:      case OP_NOT:
1743      case OP_PLUS:      case OP_PLUS:
1744      case OP_MINPLUS:      case OP_MINPLUS:
1745        case OP_POSPLUS:
1746      case OP_EXACT:      case OP_EXACT:
1747      case OP_NOTPLUS:      case OP_NOTPLUS:
1748      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1749        case OP_NOTPOSPLUS:
1750      case OP_NOTEXACT:      case OP_NOTEXACT:
1751      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1752      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1753        case OP_TYPEPOSPLUS:
1754      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1755      return FALSE;      return FALSE;
1756    
1757        /* These are going to continue, as they may be empty, but we have to
1758        fudge the length for the \p and \P cases. */
1759    
1760        case OP_TYPESTAR:
1761        case OP_TYPEMINSTAR:
1762        case OP_TYPEPOSSTAR:
1763        case OP_TYPEQUERY:
1764        case OP_TYPEMINQUERY:
1765        case OP_TYPEPOSQUERY:
1766        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1767        break;
1768    
1769        /* Same for these */
1770    
1771        case OP_TYPEUPTO:
1772        case OP_TYPEMINUPTO:
1773        case OP_TYPEPOSUPTO:
1774        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1775        break;
1776    
1777      /* End of branch */      /* End of branch */
1778    
1779      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1782  for (code = first_significant_code(code
1782      case OP_ALT:      case OP_ALT:
1783      return TRUE;      return TRUE;
1784    
1785      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1786      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1787    
1788  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1789      case OP_STAR:      case OP_STAR:
1790      case OP_MINSTAR:      case OP_MINSTAR:
1791        case OP_POSSTAR:
1792      case OP_QUERY:      case OP_QUERY:
1793      case OP_MINQUERY:      case OP_MINQUERY:
1794        case OP_POSQUERY:
1795      case OP_UPTO:      case OP_UPTO:
1796      case OP_MINUPTO:      case OP_MINUPTO:
1797        case OP_POSUPTO:
1798      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1799      break;      break;
1800  #endif  #endif
# Line 1308  return TRUE; Line 1843  return TRUE;
1843  *************************************************/  *************************************************/
1844    
1845  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1846  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
1847  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1848  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1849    
1850    Originally, this function only recognized a sequence of letters between the
1851    terminators, but it seems that Perl recognizes any sequence of characters,
1852    though of course unknown POSIX names are subsequently rejected. Perl gives an
1853    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1854    didn't consider this to be a POSIX class. Likewise for [:1234:].
1855    
1856    The problem in trying to be exactly like Perl is in the handling of escapes. We
1857    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1858    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1859    below handles the special case of \], but does not try to do any other escape
1860    processing. This makes it different from Perl for cases such as [:l\ower:]
1861    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1862    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1863    I think.
1864    
1865  Argument:  Arguments:
1866    ptr      pointer to the initial [    ptr      pointer to the initial [
1867    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1868    
1869  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1870  */  */
1871    
1872  static BOOL  static BOOL
1873  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1874  {  {
1875  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1876  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1877  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1878    {    {
1879    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
1880    return TRUE;      {
1881    }      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
1882        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
1883          {
1884          *endptr = ptr;
1885          return TRUE;
1886          }
1887        }
1888      }
1889  return FALSE;  return FALSE;
1890  }  }
1891    
# Line 1355  Returns:     a value representing the na Line 1909  Returns:     a value representing the na
1909  static int  static int
1910  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1911  {  {
1912    const char *pn = posix_names;
1913  register int yield = 0;  register int yield = 0;
1914  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1915    {    {
1916    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1917      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1918      pn += posix_name_lengths[yield] + 1;
1919    yield++;    yield++;
1920    }    }
1921  return -1;  return -1;
# Line 1374  return -1; Line 1930  return -1;
1930  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1931  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1932  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
1933  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
1934  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
1935  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
1936  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
1937  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1938    OP_END.
1939    
1940    This function has been extended with the possibility of forward references for
1941    recursions and subroutine calls. It must also check the list of such references
1942    for the group we are dealing with. If it finds that one of the recursions in
1943    the current group is on this list, it adjusts the offset in the list, not the
1944    value in the reference (which is a group number).
1945    
1946  Arguments:  Arguments:
1947    group      points to the start of the group    group      points to the start of the group
1948    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1949    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1950    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1951      save_hwm   the hwm forward reference pointer at the start of the group
1952    
1953  Returns:     nothing  Returns:     nothing
1954  */  */
1955    
1956  static void  static void
1957  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1958      uschar *save_hwm)
1959  {  {
1960  uschar *ptr = group;  uschar *ptr = group;
1961    
1962  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1963    {    {
1964    int offset = GET(ptr, 1);    int offset;
1965    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1966    
1967      /* See if this recursion is on the forward reference list. If so, adjust the
1968      reference. */
1969    
1970      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1971        {
1972        offset = GET(hc, 0);
1973        if (cd->start_code + offset == ptr + 1)
1974          {
1975          PUT(hc, 0, offset + adjust);
1976          break;
1977          }
1978        }
1979    
1980      /* Otherwise, adjust the recursion offset if it's after the start of this
1981      group. */
1982    
1983      if (hc >= cd->hwm)
1984        {
1985        offset = GET(ptr, 1);
1986        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1987        }
1988    
1989    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1990    }    }
1991  }  }
# Line 1475  Yield:        TRUE when range returned; Line 2064  Yield:        TRUE when range returned;
2064  */  */
2065    
2066  static BOOL  static BOOL
2067  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2068      unsigned int *odptr)
2069  {  {
2070  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2071    
2072  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2073    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2074    
2075  if (c > d) return FALSE;  if (c > d) return FALSE;
2076    
# Line 1492  next = othercase + 1; Line 2079  next = othercase + 1;
2079    
2080  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2081    {    {
2082    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2083    next++;    next++;
2084    }    }
2085    
# Line 1506  return TRUE; Line 2091  return TRUE;
2091  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2092    
2093    
2094    
2095    /*************************************************
2096    *     Check if auto-possessifying is possible    *
2097    *************************************************/
2098    
2099    /* This function is called for unlimited repeats of certain items, to see
2100    whether the next thing could possibly match the repeated item. If not, it makes
2101    sense to automatically possessify the repeated item.
2102    
2103    Arguments:
2104      op_code       the repeated op code
2105      this          data for this item, depends on the opcode
2106      utf8          TRUE in UTF-8 mode
2107      utf8_char     used for utf8 character bytes, NULL if not relevant
2108      ptr           next character in pattern
2109      options       options bits
2110      cd            contains pointers to tables etc.
2111    
2112    Returns:        TRUE if possessifying is wanted
2113    */
2114    
2115    static BOOL
2116    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2117      const uschar *ptr, int options, compile_data *cd)
2118    {
2119    int next;
2120    
2121    /* Skip whitespace and comments in extended mode */
2122    
2123    if ((options & PCRE_EXTENDED) != 0)
2124      {
2125      for (;;)
2126        {
2127        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2128        if (*ptr == CHAR_NUMBER_SIGN)
2129          {
2130          while (*(++ptr) != 0)
2131            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2132          }
2133        else break;
2134        }
2135      }
2136    
2137    /* If the next item is one that we can handle, get its value. A non-negative
2138    value is a character, a negative value is an escape value. */
2139    
2140    if (*ptr == CHAR_BACKSLASH)
2141      {
2142      int temperrorcode = 0;
2143      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2144      if (temperrorcode != 0) return FALSE;
2145      ptr++;    /* Point after the escape sequence */
2146      }
2147    
2148    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2149      {
2150    #ifdef SUPPORT_UTF8
2151      if (utf8) { GETCHARINC(next, ptr); } else
2152    #endif
2153      next = *ptr++;
2154      }
2155    
2156    else return FALSE;
2157    
2158    /* Skip whitespace and comments in extended mode */
2159    
2160    if ((options & PCRE_EXTENDED) != 0)
2161      {
2162      for (;;)
2163        {
2164        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2165        if (*ptr == CHAR_NUMBER_SIGN)
2166          {
2167          while (*(++ptr) != 0)
2168            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2169          }
2170        else break;
2171        }
2172      }
2173    
2174    /* If the next thing is itself optional, we have to give up. */
2175    
2176    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2177      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2178        return FALSE;
2179    
2180    /* Now compare the next item with the previous opcode. If the previous is a
2181    positive single character match, "item" either contains the character or, if
2182    "item" is greater than 127 in utf8 mode, the character's bytes are in
2183    utf8_char. */
2184    
2185    
2186    /* Handle cases when the next item is a character. */
2187    
2188    if (next >= 0) switch(op_code)
2189      {
2190      case OP_CHAR:
2191    #ifdef SUPPORT_UTF8
2192      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2193    #else
2194      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2195    #endif
2196      return item != next;
2197    
2198      /* For CHARNC (caseless character) we must check the other case. If we have
2199      Unicode property support, we can use it to test the other case of
2200      high-valued characters. */
2201    
2202      case OP_CHARNC:
2203    #ifdef SUPPORT_UTF8
2204      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2205    #endif
2206      if (item == next) return FALSE;
2207    #ifdef SUPPORT_UTF8
2208      if (utf8)
2209        {
2210        unsigned int othercase;
2211        if (next < 128) othercase = cd->fcc[next]; else
2212    #ifdef SUPPORT_UCP
2213        othercase = UCD_OTHERCASE((unsigned int)next);
2214    #else
2215        othercase = NOTACHAR;
2216    #endif
2217        return (unsigned int)item != othercase;
2218        }
2219      else
2220    #endif  /* SUPPORT_UTF8 */
2221      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2222    
2223      /* For OP_NOT, "item" must be a single-byte character. */
2224    
2225      case OP_NOT:
2226      if (item == next) return TRUE;
2227      if ((options & PCRE_CASELESS) == 0) return FALSE;
2228    #ifdef SUPPORT_UTF8
2229      if (utf8)
2230        {
2231        unsigned int othercase;
2232        if (next < 128) othercase = cd->fcc[next]; else
2233    #ifdef SUPPORT_UCP
2234        othercase = UCD_OTHERCASE(next);
2235    #else
2236        othercase = NOTACHAR;
2237    #endif
2238        return (unsigned int)item == othercase;
2239        }
2240      else
2241    #endif  /* SUPPORT_UTF8 */
2242      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2243    
2244      case OP_DIGIT:
2245      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2246    
2247      case OP_NOT_DIGIT:
2248      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2249    
2250      case OP_WHITESPACE:
2251      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2252    
2253      case OP_NOT_WHITESPACE:
2254      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2255    
2256      case OP_WORDCHAR:
2257      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2258    
2259      case OP_NOT_WORDCHAR:
2260      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2261    
2262      case OP_HSPACE:
2263      case OP_NOT_HSPACE:
2264      switch(next)
2265        {
2266        case 0x09:
2267        case 0x20:
2268        case 0xa0:
2269        case 0x1680:
2270        case 0x180e:
2271        case 0x2000:
2272        case 0x2001:
2273        case 0x2002:
2274        case 0x2003:
2275        case 0x2004:
2276        case 0x2005:
2277        case 0x2006:
2278        case 0x2007:
2279        case 0x2008:
2280        case 0x2009:
2281        case 0x200A:
2282        case 0x202f:
2283        case 0x205f:
2284        case 0x3000:
2285        return op_code != OP_HSPACE;
2286        default:
2287        return op_code == OP_HSPACE;
2288        }
2289    
2290      case OP_VSPACE:
2291      case OP_NOT_VSPACE:
2292      switch(next)
2293        {
2294        case 0x0a:
2295        case 0x0b:
2296        case 0x0c:
2297        case 0x0d:
2298        case 0x85:
2299        case 0x2028:
2300        case 0x2029:
2301        return op_code != OP_VSPACE;
2302        default:
2303        return op_code == OP_VSPACE;
2304        }
2305    
2306      default:
2307      return FALSE;
2308      }
2309    
2310    
2311    /* Handle the case when the next item is \d, \s, etc. */
2312    
2313    switch(op_code)
2314      {
2315      case OP_CHAR:
2316      case OP_CHARNC:
2317    #ifdef SUPPORT_UTF8
2318      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2319    #endif
2320      switch(-next)
2321        {
2322        case ESC_d:
2323        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2324    
2325        case ESC_D:
2326        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2327    
2328        case ESC_s:
2329        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2330    
2331        case ESC_S:
2332        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2333    
2334        case ESC_w:
2335        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2336    
2337        case ESC_W:
2338        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2339    
2340        case ESC_h:
2341        case ESC_H:
2342        switch(item)
2343          {
2344          case 0x09:
2345          case 0x20:
2346          case 0xa0:
2347          case 0x1680:
2348          case 0x180e:
2349          case 0x2000:
2350          case 0x2001:
2351          case 0x2002:
2352          case 0x2003:
2353          case 0x2004:
2354          case 0x2005:
2355          case 0x2006:
2356          case 0x2007:
2357          case 0x2008:
2358          case 0x2009:
2359          case 0x200A:
2360          case 0x202f:
2361          case 0x205f:
2362          case 0x3000:
2363          return -next != ESC_h;
2364          default:
2365          return -next == ESC_h;
2366          }
2367    
2368        case ESC_v:
2369        case ESC_V:
2370        switch(item)
2371          {
2372          case 0x0a:
2373          case 0x0b:
2374          case 0x0c:
2375          case 0x0d:
2376          case 0x85:
2377          case 0x2028:
2378          case 0x2029:
2379          return -next != ESC_v;
2380          default:
2381          return -next == ESC_v;
2382          }
2383    
2384        default:
2385        return FALSE;
2386        }
2387    
2388      case OP_DIGIT:
2389      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2390             next == -ESC_h || next == -ESC_v;
2391    
2392      case OP_NOT_DIGIT:
2393      return next == -ESC_d;
2394    
2395      case OP_WHITESPACE:
2396      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2397    
2398      case OP_NOT_WHITESPACE:
2399      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2400    
2401      case OP_HSPACE:
2402      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2403    
2404      case OP_NOT_HSPACE:
2405      return next == -ESC_h;
2406    
2407      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2408      case OP_VSPACE:
2409      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2410    
2411      case OP_NOT_VSPACE:
2412      return next == -ESC_v;
2413    
2414      case OP_WORDCHAR:
2415      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2416    
2417      case OP_NOT_WORDCHAR:
2418      return next == -ESC_w || next == -ESC_d;
2419    
2420      default:
2421      return FALSE;
2422      }
2423    
2424    /* Control does not reach here */
2425    }
2426    
2427    
2428    
2429  /*************************************************  /*************************************************
2430  *           Compile one branch                   *  *           Compile one branch                   *
2431  *************************************************/  *************************************************/
2432    
2433  /* 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
2434  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
2435  bits.  bits. This function is used during the pre-compile phase when we are trying
2436    to find out the amount of memory needed, as well as during the real compile
2437    phase. The value of lengthptr distinguishes the two phases.
2438    
2439  Arguments:  Arguments:
2440    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2441    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2442    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2443    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2445  Arguments:
2445    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2446    bcptr          points to current branch chain    bcptr          points to current branch chain
2447    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2448      lengthptr      NULL during the real compile phase
2449                     points to length accumulator during pre-compile phase
2450    
2451  Returns:         TRUE on success  Returns:         TRUE on success
2452                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2453  */  */
2454    
2455  static BOOL  static BOOL
2456  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2457    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2458    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2459  {  {
2460  int repeat_type, op_type;  int repeat_type, op_type;
2461  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 2464  int greedy_default, greedy_non_default;
2464  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2465  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2466  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2467  int options = *optionsptr;  int options = *optionsptr;
2468  int after_manual_callout = 0;  int after_manual_callout = 0;
2469    int length_prevgroup = 0;
2470  register int c;  register int c;
2471  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2472    uschar *last_code = code;
2473    uschar *orig_code = code;
2474  uschar *tempcode;  uschar *tempcode;
2475  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2476  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2478  const uschar *ptr = *ptrptr;
2478  const uschar *tempptr;  const uschar *tempptr;
2479  uschar *previous = NULL;  uschar *previous = NULL;
2480  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2481    uschar *save_hwm = NULL;
2482  uschar classbits[32];  uschar classbits[32];
2483    
2484  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2485  BOOL class_utf8;  BOOL class_utf8;
2486  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2487  uschar *class_utf8data;  uschar *class_utf8data;
2488    uschar *class_utf8data_base;
2489  uschar utf8_char[6];  uschar utf8_char[6];
2490  #else  #else
2491  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2492    uschar *utf8_char = NULL;
2493    #endif
2494    
2495    #ifdef DEBUG
2496    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2497  #endif  #endif
2498    
2499  /* 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 2525  req_caseopt = ((options & PCRE_CASELESS)
2525  for (;; ptr++)  for (;; ptr++)
2526    {    {
2527    BOOL negate_class;    BOOL negate_class;
2528      BOOL should_flip_negation;
2529    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2530    BOOL is_quantifier;    BOOL is_quantifier;
2531      BOOL is_recurse;
2532      BOOL reset_bracount;
2533    int class_charcount;    int class_charcount;
2534    int class_lastchar;    int class_lastchar;
2535    int newoptions;    int newoptions;
2536    int recno;    int recno;
2537      int refsign;
2538    int skipbytes;    int skipbytes;
2539    int subreqbyte;    int subreqbyte;
2540    int subfirstbyte;    int subfirstbyte;
2541      int terminator;
2542    int mclength;    int mclength;
2543    uschar mcbuffer[8];    uschar mcbuffer[8];
2544    
2545    /* Next byte in the pattern */    /* Get next byte in the pattern */
2546    
2547    c = *ptr;    c = *ptr;
2548    
2549      /* If we are in the pre-compile phase, accumulate the length used for the
2550      previous cycle of this loop. */
2551    
2552      if (lengthptr != NULL)
2553        {
2554    #ifdef DEBUG
2555        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2556    #endif
2557        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2558          {
2559          *errorcodeptr = ERR52;
2560          goto FAILED;
2561          }
2562    
2563        /* There is at least one situation where code goes backwards: this is the
2564        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2565        the class is simply eliminated. However, it is created first, so we have to
2566        allow memory for it. Therefore, don't ever reduce the length at this point.
2567        */
2568    
2569        if (code < last_code) code = last_code;
2570    
2571        /* Paranoid check for integer overflow */
2572    
2573        if (OFLOW_MAX - *lengthptr < code - last_code)
2574          {
2575          *errorcodeptr = ERR20;
2576          goto FAILED;
2577          }
2578    
2579        *lengthptr += code - last_code;
2580        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2581    
2582        /* If "previous" is set and it is not at the start of the work space, move
2583        it back to there, in order to avoid filling up the work space. Otherwise,
2584        if "previous" is NULL, reset the current code pointer to the start. */
2585    
2586        if (previous != NULL)
2587          {
2588          if (previous > orig_code)
2589            {
2590            memmove(orig_code, previous, code - previous);
2591            code -= previous - orig_code;
2592            previous = orig_code;
2593            }
2594          }
2595        else code = orig_code;
2596    
2597        /* Remember where this code item starts so we can pick up the length
2598        next time round. */
2599    
2600        last_code = code;
2601        }
2602    
2603      /* In the real compile phase, just check the workspace used by the forward
2604      reference list. */
2605    
2606      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2607        {
2608        *errorcodeptr = ERR52;
2609        goto FAILED;
2610        }
2611    
2612    /* 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 */
2613    
2614    if (inescq && c != 0)    if (inescq && c != 0)
2615      {      {
2616      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2617        {        {
2618        inescq = FALSE;        inescq = FALSE;
2619        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 2623  for (;; ptr++)
2623        {        {
2624        if (previous_callout != NULL)        if (previous_callout != NULL)
2625          {          {
2626          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2627              complete_callout(previous_callout, ptr, cd);
2628          previous_callout = NULL;          previous_callout = NULL;
2629          }          }
2630        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2639  for (;; ptr++)
2639    /* 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
2640    a quantifier. */    a quantifier. */
2641    
2642    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2643      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2644        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2645    
2646    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2647         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2648      {      {
2649      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2650          complete_callout(previous_callout, ptr, cd);
2651      previous_callout = NULL;      previous_callout = NULL;
2652      }      }
2653    
# Line 1653  for (;; ptr++) Line 2656  for (;; ptr++)
2656    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2657      {      {
2658      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2659      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2660        {        {
2661        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2662        on the Macintosh. */          {
2663        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2664        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2665          if (*ptr != 0) continue;
2666    
2667          /* Else fall through to handle end of string */
2668          c = 0;
2669        }        }
2670      }      }
2671    
# Line 1672  for (;; ptr++) Line 2679  for (;; ptr++)
2679    
2680    switch(c)    switch(c)
2681      {      {
2682      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2683        case 0:                        /* The branch terminates at string end */
2684      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2685      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2686      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2687      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2688      *codeptr = code;      *codeptr = code;
2689      *ptrptr = ptr;      *ptrptr = ptr;
2690        if (lengthptr != NULL)
2691          {
2692          if (OFLOW_MAX - *lengthptr < code - last_code)
2693            {
2694            *errorcodeptr = ERR20;
2695            goto FAILED;
2696            }
2697          *lengthptr += code - last_code;   /* To include callout length */
2698          DPRINTF((">> end branch\n"));
2699          }
2700      return TRUE;      return TRUE;
2701    
2702    
2703        /* ===================================================================*/
2704      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2705      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2706    
2707      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2708      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2709        {        {
2710        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2713  for (;; ptr++)
2713      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2714      break;      break;
2715    
2716      case '$':      case CHAR_DOLLAR_SIGN:
2717      previous = NULL;      previous = NULL;
2718      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2719      break;      break;
# Line 1703  for (;; ptr++) Line 2721  for (;; ptr++)
2721      /* 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
2722      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2723    
2724      case '.':      case CHAR_DOT:
2725      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2726      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2727      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2728      previous = code;      previous = code;
2729      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2730      break;      break;
2731    
2732      /* Character classes. If the included characters are all < 255 in value, we  
2733      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2734      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
2735      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
2736      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2737        map as usual, then invert it at the end. However, we use a different opcode
2738        so that data characters > 255 can be handled correctly.
2739    
2740      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2741      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,
2742      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2743      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.
     */  
2744    
2745      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2746        default (Perl) mode, it is treated as a data character. */
2747    
2748        case CHAR_RIGHT_SQUARE_BRACKET:
2749        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2750          {
2751          *errorcodeptr = ERR64;
2752          goto FAILED;
2753          }
2754        goto NORMAL_CHAR;
2755    
2756        case CHAR_LEFT_SQUARE_BRACKET:
2757      previous = code;      previous = code;
2758    
2759      /* 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
2760      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. */
2761    
2762      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2763          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2764            check_posix_syntax(ptr, &tempptr))
2765        {        {
2766        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2767        goto FAILED;        goto FAILED;
2768        }        }
2769    
2770      /* 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,
2771        if the first few characters (either before or after ^) are \Q\E or \E we
2772        skip them too. This makes for compatibility with Perl. */
2773    
2774      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2775        for (;;)
2776        {        {
       negate_class = TRUE;  
2777        c = *(++ptr);        c = *(++ptr);
2778          if (c == CHAR_BACKSLASH)
2779            {
2780            if (ptr[1] == CHAR_E)
2781              ptr++;
2782            else if (strncmp((const char *)ptr+1,
2783                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2784              ptr += 3;
2785            else
2786              break;
2787            }
2788          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2789            negate_class = TRUE;
2790          else break;
2791        }        }
2792      else  
2793        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2794        an initial ']' is taken as a data character -- the code below handles
2795        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2796        [^] must match any character, so generate OP_ALLANY. */
2797    
2798        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2799            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2800        {        {
2801        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2802          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2803          zerofirstbyte = firstbyte;
2804          break;
2805        }        }
2806    
2807        /* If a class contains a negative special such as \S, we need to flip the
2808        negation flag at the end, so that support for characters > 255 works
2809        correctly (they are all included in the class). */
2810    
2811        should_flip_negation = FALSE;
2812    
2813      /* 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
2814      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
2815      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2816    
2817      class_charcount = 0;      class_charcount = 0;
2818      class_lastchar = -1;      class_lastchar = -1;
2819    
2820        /* Initialize the 32-char bit map to all zeros. We build the map in a
2821        temporary bit of memory, in case the class contains only 1 character (less
2822        than 256), because in that case the compiled code doesn't use the bit map.
2823        */
2824    
2825        memset(classbits, 0, 32 * sizeof(uschar));
2826    
2827  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2828      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2829      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2830        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2831  #endif  #endif
2832    
     /* 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));  
   
2833      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2834      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
2835      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. */  
2836    
2837      do      if (c != 0) do
2838        {        {
2839          const uschar *oldptr;
2840    
2841  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2842        if (utf8 && c > 127)        if (utf8 && c > 127)
2843          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2844          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2845          }          }
2846    
2847          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2848          data and reset the pointer. This is so that very large classes that
2849          contain a zillion UTF-8 characters no longer overwrite the work space
2850          (which is on the stack). */
2851    
2852          if (lengthptr != NULL)
2853            {
2854            *lengthptr += class_utf8data - class_utf8data_base;
2855            class_utf8data = class_utf8data_base;
2856            }
2857    
2858  #endif  #endif
2859    
2860        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2861    
2862        if (inescq)        if (inescq)
2863          {          {
2864          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2865            {            {
2866            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2867            ptr++;            ptr++;                            /* Skip the 'E' */
2868            continue;            continue;                         /* Carry on with next */
2869            }            }
2870          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2871          }          }
2872    
2873        /* 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 2876  for (;; ptr++)
2876        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2877        5.6 and 5.8 do. */        5.6 and 5.8 do. */
2878    
2879        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
2880            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2881            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
2882          {          {
2883          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2884          int posix_class, i;          int posix_class, taboffset, tabopt;
2885          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2886            uschar pbits[32];
2887    
2888          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
2889            {            {
2890            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
2891            goto FAILED;            goto FAILED;
2892            }            }
2893    
2894          ptr += 2;          ptr += 2;
2895          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
2896            {            {
2897            local_negate = TRUE;            local_negate = TRUE;
2898              should_flip_negation = TRUE;  /* Note negative special */
2899            ptr++;            ptr++;
2900            }            }
2901    
# Line 1836  for (;; ptr++) Line 2913  for (;; ptr++)
2913          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2914            posix_class = 0;            posix_class = 0;
2915    
2916          /* 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
2917          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
2918          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
2919          white space chars afterwards. */          result into the bit map that is being built. */
2920    
2921          posix_class *= 3;          posix_class *= 3;
2922          for (i = 0; i < 3; i++)  
2923            /* Copy in the first table (always present) */
2924    
2925            memcpy(pbits, cbits + posix_class_maps[posix_class],
2926              32 * sizeof(uschar));
2927    
2928            /* If there is a second table, add or remove it as required. */
2929    
2930            taboffset = posix_class_maps[posix_class + 1];
2931            tabopt = posix_class_maps[posix_class + 2];
2932    
2933            if (taboffset >= 0)
2934            {            {
2935            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2936            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;  
             }  
2937            else            else
2938              {              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;  
             }  
2939            }            }
2940    
2941            /* Not see if we need to remove any special characters. An option
2942            value of 1 removes vertical space and 2 removes underscore. */
2943    
2944            if (tabopt < 0) tabopt = -tabopt;
2945            if (tabopt == 1) pbits[1] &= ~0x3c;
2946              else if (tabopt == 2) pbits[11] &= 0x7f;
2947    
2948            /* Add the POSIX table or its complement into the main table that is
2949            being built and we are done. */
2950    
2951            if (local_negate)
2952              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2953            else
2954              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2955    
2956          ptr = tempptr + 1;          ptr = tempptr + 1;
2957          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2958          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2959          }          }
2960    
2961        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2962        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
2963        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.
2964        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2965        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  
2966        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2967    
2968        if (c == '\\')        if (c == CHAR_BACKSLASH)
2969          {          {
2970          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2971            if (*errorcodeptr != 0) goto FAILED;
2972    
2973          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 */
2974          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 */
2975            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
2976          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2977            {            {
2978            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
2979              {              {
2980              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
2981              }              }
2982            else inescq = TRUE;            else inescq = TRUE;
2983            continue;            continue;
2984            }            }
2985            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2986    
2987          if (c < 0)          if (c < 0)
2988            {            {
2989            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2990            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2991            switch (-c)  
2992              /* Save time by not doing this in the pre-compile phase. */
2993    
2994              if (lengthptr == NULL) switch (-c)
2995              {              {
2996              case ESC_d:              case ESC_d:
2997              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2998              continue;              continue;
2999    
3000              case ESC_D:              case ESC_D:
3001                should_flip_negation = TRUE;
3002              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3003              continue;              continue;
3004    
# Line 1910  for (;; ptr++) Line 3007  for (;; ptr++)
3007              continue;              continue;
3008    
3009              case ESC_W:              case ESC_W:
3010                should_flip_negation = TRUE;
3011              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3012              continue;              continue;
3013    
# Line 1919  for (;; ptr++) Line 3017  for (;; ptr++)
3017              continue;              continue;
3018    
3019              case ESC_S:              case ESC_S:
3020                should_flip_negation = TRUE;
3021              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3022              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3023              continue;              continue;
3024    
3025  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3026              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3027              case ESC_P:              }
3028    
3029              /* In the pre-compile phase, just do the recognition. */
3030    
3031              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3032                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3033    
3034              /* We need to deal with \H, \h, \V, and \v in both phases because
3035              they use extra memory. */
3036    
3037              if (-c == ESC_h)
3038                {
3039                SETBIT(classbits, 0x09); /* VT */
3040                SETBIT(classbits, 0x20); /* SPACE */
3041                SETBIT(classbits, 0xa0); /* NSBP */
3042    #ifdef SUPPORT_UTF8
3043                if (utf8)
3044                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3045                class_utf8 = TRUE;                class_utf8 = TRUE;
3046                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
3047                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3048                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
3049                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3050                  *class_utf8data++ = XCL_RANGE;
3051                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3052                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3053                  *class_utf8data++ = XCL_SINGLE;
3054                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3055                  *class_utf8data++ = XCL_SINGLE;
3056                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3057                  *class_utf8data++ = XCL_SINGLE;
3058                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3059                }                }
             continue;  
3060  #endif  #endif
3061                continue;
3062                }
3063    
3064              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
3065              strict mode. By default, for compatibility with Perl, they are              {
3066              treated as literals. */              for (c = 0; c < 32; c++)
   
             default:  
             if ((options & PCRE_EXTRA) != 0)  
3067                {                {
3068                *errorcodeptr = ERR7;                int x = 0xff;
3069                goto FAILED;                switch (c)
3070                }                  {
3071              c = *ptr;              /* The final character */                  case 0x09/8: x ^= 1 << (0x09%8); break;
3072              class_charcount -= 2;  /* Undo the default count from above */                  case 0x20/8: x ^= 1 << (0x20%8); break;
3073                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3074                    default: break;
3075                    }
3076                  classbits[c] |= x;
3077                  }
3078    
3079    #ifdef SUPPORT_UTF8
3080                if (utf8)
3081                  {
3082                  class_utf8 = TRUE;
3083                  *class_utf8data++ = XCL_RANGE;
3084                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3085                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3086                  *class_utf8data++ = XCL_RANGE;
3087                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3088                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3089                  *class_utf8data++ = XCL_RANGE;
3090                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3091                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3092                  *class_utf8data++ = XCL_RANGE;
3093                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3094                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3095                  *class_utf8data++ = XCL_RANGE;
3096                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3097                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3098                  *class_utf8data++ = XCL_RANGE;
3099                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3100                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3101                  *class_utf8data++ = XCL_RANGE;
3102                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3103                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3104                  }
3105    #endif
3106                continue;
3107                }
3108    
3109              if (-c == ESC_v)
3110                {
3111                SETBIT(classbits, 0x0a); /* LF */
3112                SETBIT(classbits, 0x0b); /* VT */
3113                SETBIT(classbits, 0x0c); /* FF */
3114                SETBIT(classbits, 0x0d); /* CR */
3115                SETBIT(classbits, 0x85); /* NEL */
3116    #ifdef SUPPORT_UTF8
3117                if (utf8)
3118                  {
3119                  class_utf8 = TRUE;
3120                  *class_utf8data++ = XCL_RANGE;
3121                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3122                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3123                  }
3124    #endif
3125                continue;
3126                }
3127    
3128              if (-c == ESC_V)
3129                {
3130                for (c = 0; c < 32; c++)
3131                  {
3132                  int x = 0xff;
3133                  switch (c)
3134                    {
3135                    case 0x0a/8: x ^= 1 << (0x0a%8);
3136                                 x ^= 1 << (0x0b%8);
3137                                 x ^= 1 << (0x0c%8);
3138                                 x ^= 1 << (0x0d%8);
3139                                 break;
3140                    case 0x85/8: x ^= 1 << (0x85%8); break;
3141                    default: break;
3142                    }
3143                  classbits[c] |= x;
3144                  }
3145    
3146    #ifdef SUPPORT_UTF8
3147                if (utf8)
3148                  {
3149                  class_utf8 = TRUE;
3150                  *class_utf8data++ = XCL_RANGE;
3151                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3152                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3153                  *class_utf8data++ = XCL_RANGE;
3154                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3155                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3156                  }
3157    #endif
3158                continue;
3159                }
3160    
3161              /* We need to deal with \P and \p in both phases. */
3162    
3163    #ifdef SUPPORT_UCP
3164              if (-c == ESC_p || -c == ESC_P)
3165                {
3166                BOOL negated;
3167                int pdata;
3168                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3169                if (ptype < 0) goto FAILED;
3170                class_utf8 = TRUE;
3171                *class_utf8data++ = ((-c == ESC_p) != negated)?
3172                  XCL_PROP : XCL_NOTPROP;
3173                *class_utf8data++ = ptype;
3174                *class_utf8data++ = pdata;
3175                class_charcount -= 2;   /* Not a < 256 character */
3176                continue;
3177                }
3178    #endif
3179              /* Unrecognized escapes are faulted if PCRE is running in its
3180              strict mode. By default, for compatibility with Perl, they are
3181              treated as literals. */
3182    
3183              if ((options & PCRE_EXTRA) != 0)
3184                {
3185                *errorcodeptr = ERR7;
3186                goto FAILED;
3187              }              }
3188    
3189              class_charcount -= 2;  /* Undo the default count from above */
3190              c = *ptr;              /* Get the final character and fall through */
3191            }            }
3192    
3193          /* 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
3194          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3195    
3196          }   /* End of backslash handling */          }   /* End of backslash handling */
3197    
3198        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3199        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
3200        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3201          entirely. The code for handling \Q and \E is messy. */
3202    
3203          CHECK_RANGE:
3204          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3205            {
3206            inescq = FALSE;
3207            ptr += 2;
3208            }
3209    
3210          oldptr = ptr;
3211    
3212        if (ptr[1] == '-' && ptr[2] != ']')        /* Remember \r or \n */
3213    
3214          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3215    
3216          /* Check for range */
3217    
3218          if (!inescq && ptr[1] == CHAR_MINUS)
3219          {          {
3220          int d;          int d;
3221          ptr += 2;          ptr += 2;
3222            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3223    
3224            /* If we hit \Q (not followed by \E) at this point, go into escaped
3225            mode. */
3226    
3227            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3228              {
3229              ptr += 2;
3230              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3231                { ptr += 2; continue; }
3232              inescq = TRUE;
3233              break;
3234              }
3235    
3236            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3237              {
3238              ptr = oldptr;
3239              goto LONE_SINGLE_CHARACTER;
3240              }
3241    
3242  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3243          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3252  for (;; ptr++)
3252          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
3253          in such circumstances. */          in such circumstances. */
3254    
3255          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3256            {            {
3257            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3258            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3259    
3260            /* \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
3261            was literal */            special means the '-' was literal */
3262    
3263            if (d < 0)            if (d < 0)
3264              {              {
3265              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3266              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3267                else if (d == -ESC_R) d = CHAR_R; else
3268                {                {
3269                ptr = oldptr - 2;                ptr = oldptr;
3270                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3271                }                }
3272              }              }
3273            }            }
3274    
3275          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3276          the pre-pass. Optimize one-character ranges */          one-character ranges */
3277    
3278            if (d < c)
3279              {
3280              *errorcodeptr = ERR8;
3281              goto FAILED;
3282              }
3283    
3284          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3285    
3286            /* Remember \r or \n */
3287    
3288            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3289    
3290          /* 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
3291          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3292          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 3304  for (;; ptr++)
3304  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3305            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3306              {              {
3307              int occ, ocd;              unsigned int occ, ocd;
3308              int cc = c;              unsigned int cc = c;
3309              int origd = d;              unsigned int origd = d;
3310              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3311                {                {
3312                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3313                      ocd <= (unsigned int)d)
3314                    continue;                          /* Skip embedded ranges */
3315    
3316                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3317                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3318                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3319                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3320                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3321                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3322                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3323                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3324                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3325                  d = ocd;                  d = ocd;
3326                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3368  for (;; ptr++)
3368          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
3369          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3370    
3371          for (; c <= d; c++)          class_charcount += d - c + 1;
3372            class_lastchar = d;
3373    
3374            /* We can save a bit of time by skipping this in the pre-compile. */
3375    
3376            if (lengthptr == NULL) for (; c <= d; c++)
3377            {            {
3378            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3379            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3381  for (;; ptr++)
3381              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3382              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3383              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3384            }            }
3385    
3386          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3404  for (;; ptr++)
3404  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3405          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3406            {            {
3407            int chartype;            unsigned int othercase;
3408            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3409              {              {
3410              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3411              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3430  for (;; ptr++)
3430          }          }
3431        }        }
3432    
3433      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3434      loop. This "while" is the end of the "do" above. */  
3435        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3436    
3437        if (c == 0)                          /* Missing terminating ']' */
3438          {
3439          *errorcodeptr = ERR6;
3440          goto FAILED;
3441          }
3442    
3443    
3444    /* This code has been disabled because it would mean that \s counts as
3445    an explicit \r or \n reference, and that's not really what is wanted. Now
3446    we set the flag only if there is a literal "\r" or "\n" in the class. */
3447    
3448    #if 0
3449        /* Remember whether \r or \n are in this class */
3450    
3451        if (negate_class)
3452          {
3453          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3454          }
3455        else
3456          {
3457          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3458          }
3459    #endif
3460    
     while ((c = *(++ptr)) != ']' || inescq);  
3461    
3462      /* 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
3463      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
3464      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
3465      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3466      single-bytes only. This is an historical hangover. Maybe one day we can  
3467      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3468        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3469        operate on single-bytes only. This is an historical hangover. Maybe one day
3470        we can tidy these opcodes to handle multi-byte characters.
3471    
3472      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
3473      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 3477  for (;; ptr++)
3477      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3478    
3479  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3480      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3481            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3482  #else  #else
3483      if (class_charcount == 1)      if (class_charcount == 1)
3484  #endif  #endif
# Line 2209  for (;; ptr++) Line 3521  for (;; ptr++)
3521      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3522    
3523      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3524      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3525      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3526        the class, so any that were explicitly given as well can be ignored. If
3527        (when there are explicit characters > 255 that must be listed) there are no
3528        characters < 256, we can omit the bitmap in the actual compiled code. */
3529    
3530  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3531      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3532        {        {
3533        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3534        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3535        code += LINK_SIZE;        code += LINK_SIZE;
3536        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3537    
3538        /* 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;
3539        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3540    
3541        if (class_charcount > 0)        if (class_charcount > 0)
3542          {          {
3543          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3544            memmove(code + 32, code, class_utf8data - code);
3545          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3546          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;  
3547          }          }
3548          else code = class_utf8data;
3549    
3550        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3551    
# Line 2246  for (;; ptr++) Line 3554  for (;; ptr++)
3554        }        }
3555  #endif  #endif
3556    
3557      /* 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
3558      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
3559      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
3560      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3561    
3562        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3563      if (negate_class)      if (negate_class)
3564        {        {
3565        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3566        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3567        }        }
3568      else      else
3569        {        {
       *code++ = OP_CLASS;  
3570        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3571        }        }
3572      code += 32;      code += 32;
3573      break;      break;
3574    
3575    
3576        /* ===================================================================*/
3577      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3578      has been tested above. */      has been tested above. */
3579    
3580      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3581      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3582      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3583      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3584      goto REPEAT;      goto REPEAT;
3585    
3586      case '*':      case CHAR_ASTERISK:
3587      repeat_min = 0;      repeat_min = 0;
3588      repeat_max = -1;      repeat_max = -1;
3589      goto REPEAT;      goto REPEAT;
3590    
3591      case '+':      case CHAR_PLUS:
3592      repeat_min = 1;      repeat_min = 1;
3593      repeat_max = -1;      repeat_max = -1;
3594      goto REPEAT;      goto REPEAT;
3595    
3596      case '?':      case CHAR_QUESTION_MARK:
3597      repeat_min = 0;      repeat_min = 0;
3598      repeat_max = 1;      repeat_max = 1;
3599    
# Line 2318  for (;; ptr++) Line 3628  for (;; ptr++)
3628      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
3629      repeat type to the non-default. */      repeat type to the non-default. */
3630    
3631      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3632        {        {
3633        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3634        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3635        ptr++;        ptr++;
3636        }        }
3637      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3638        {        {
3639        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3640        ptr++;        ptr++;
3641        }        }
3642      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3643    
     /* 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;  
       }  
   
3644      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3645      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
3646      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 3674  for (;; ptr++)
3674          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3675          }          }
3676    
3677          /* If the repetition is unlimited, it pays to see if the next thing on
3678          the line is something that cannot possibly match this character. If so,
3679          automatically possessifying this item gains some performance in the case
3680          where the match fails. */
3681    
3682          if (!possessive_quantifier &&
3683              repeat_max < 0 &&
3684              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3685                options, cd))
3686            {
3687            repeat_type = 0;    /* Force greedy */
3688            possessive_quantifier = TRUE;
3689            }
3690    
3691        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3692        }        }
3693    
3694      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3695      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-
3696      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3697      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3698        currently used only for single-byte chars. */
3699    
3700      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3701        {        {
3702        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3703        c = previous[1];        c = previous[1];
3704          if (!possessive_quantifier &&
3705              repeat_max < 0 &&
3706              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3707            {
3708            repeat_type = 0;    /* Force greedy */
3709            possessive_quantifier = TRUE;
3710            }
3711        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3712        }        }
3713    
# Line 2403  for (;; ptr++) Line 3721  for (;; ptr++)
3721      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3722        {        {
3723        uschar *oldcode;        uschar *oldcode;
3724        int prop_type;        int prop_type, prop_value;
3725        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3726        c = *previous;        c = *previous;
3727    
3728          if (!possessive_quantifier &&
3729              repeat_max < 0 &&
3730              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3731            {
3732            repeat_type = 0;    /* Force greedy */
3733            possessive_quantifier = TRUE;
3734            }
3735    
3736        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3737        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3738          previous[1] : -1;          {
3739            prop_type = previous[1];
3740            prop_value = previous[2];
3741            }
3742          else prop_type = prop_value = -1;
3743    
3744        oldcode = code;        oldcode = code;
3745        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3752  for (;; ptr++)
3752        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3753        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3754    
3755        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3756    
3757        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3758    
# Line 2443  for (;; ptr++) Line 3773  for (;; ptr++)
3773          }          }
3774    
3775        /* 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
3776        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3777        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
3778        one less than the maximum. */        one less than the maximum. */
3779    
# Line 2470  for (;; ptr++) Line 3800  for (;; ptr++)
3800    
3801          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3802          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
3803          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3804          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3805          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3806    
# Line 2486  for (;; ptr++) Line 3816  for (;; ptr++)
3816  #endif  #endif
3817              {              {
3818              *code++ = c;              *code++ = c;
3819              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3820                  {
3821                  *code++ = prop_type;
3822                  *code++ = prop_value;
3823                  }
3824              }              }
3825            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3826            }            }
3827    
3828          /* 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
3829          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3830            UPTO is just for 1 instance, we can use QUERY instead. */
3831    
3832          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3833            {            {
# Line 2505  for (;; ptr++) Line 3840  for (;; ptr++)
3840            else            else
3841  #endif  #endif
3842            *code++ = c;            *code++ = c;
3843            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3844                {
3845                *code++ = prop_type;
3846                *code++ = prop_value;
3847                }
3848            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3849            *code++ = OP_UPTO + repeat_type;  
3850            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3851                {
3852                *code++ = OP_QUERY + repeat_type;
3853                }
3854              else
3855                {
3856                *code++ = OP_UPTO + repeat_type;
3857                PUT2INC(code, 0, repeat_max);
3858                }
3859            }            }
3860          }          }
3861    
# Line 2524  for (;; ptr++) Line 3871  for (;; ptr++)
3871  #endif  #endif
3872        *code++ = c;        *code++ = c;
3873    
3874        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3875        defines the required property. */        define the required property. */
3876    
3877  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3878        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3879            {
3880            *code++ = prop_type;
3881            *code++ = prop_value;
3882            }
3883  #endif  #endif
3884        }        }
3885    
# Line 2551  for (;; ptr++) Line 3902  for (;; ptr++)
3902        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3903        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3904    
3905        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3906    
3907        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3908          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 3922  for (;; ptr++)
3922      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
3923      cases. */      cases. */
3924    
3925      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3926               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3927        {        {
3928        register int i;        register int i;
3929        int ketoffset = 0;        int ketoffset = 0;
3930        int len = code - previous;        int len = code - previous;
3931        uschar *bralink = NULL;        uschar *bralink = NULL;
3932    
3933          /* Repeating a DEFINE group is pointless */
3934    
3935          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3936            {
3937            *errorcodeptr = ERR55;
3938            goto FAILED;
3939            }
3940    
3941        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
3942        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
3943        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2601  for (;; ptr++) Line 3960  for (;; ptr++)
3960    
3961        if (repeat_min == 0)        if (repeat_min == 0)
3962          {          {
3963          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
3964          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3965    
3966          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3967          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3968          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3969          any internal group, because the offset is from the start of the whole          **   goto END_REPEAT;
3970          regex. Temporarily terminate the pattern while doing this. */          **   }
3971    
3972            However, that fails when a group is referenced as a subroutine from
3973            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3974            so that it is skipped on execution. As we don't have a list of which
3975            groups are referenced, we cannot do this selectively.
3976    
3977            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3978            and do no more at this point. However, we do need to adjust any
3979            OP_RECURSE calls inside the group that refer to the group itself or any
3980            internal or forward referenced group, because the offset is from the
3981            start of the whole regex. Temporarily terminate the pattern while doing
3982            this. */
3983    
3984          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3985            {            {
3986            *code = OP_END;            *code = OP_END;
3987            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3988            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3989            code++;            code++;
3990              if (repeat_max == 0)
3991                {
3992                *previous++ = OP_SKIPZERO;
3993                goto END_REPEAT;
3994                }
3995            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3996            }            }
3997    
# Line 2637  for (;; ptr++) Line 4007  for (;; ptr++)
4007            {            {
4008            int offset;            int offset;
4009            *code = OP_END;            *code = OP_END;
4010            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
4011            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
4012            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
4013            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 4027  for (;; ptr++)
4027        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
4028        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
4029        copies that we need. If we set a first char from the group, and didn't        copies that we need. If we set a first char from the group, and didn't
4030        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
4031          forward reference subroutine calls in the group, there will be entries on
4032          the workspace list; replicate these with an appropriate increment. */
4033    
4034        else        else
4035          {          {
4036          if (repeat_min > 1)          if (repeat_min > 1)
4037            {            {
4038            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
4039            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
4040              potential integer overflow. */
4041    
4042              if (lengthptr != NULL)
4043                {
4044                int delta = (repeat_min - 1)*length_prevgroup;
4045                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4046                                                                (double)INT_MAX ||
4047                    OFLOW_MAX - *lengthptr < delta)
4048                  {
4049                  *errorcodeptr = ERR20;
4050                  goto FAILED;
4051                  }
4052                *lengthptr += delta;
4053                }
4054    
4055              /* This is compiling for real */
4056    
4057              else
4058              {              {
4059              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4060              code += len;              for (i = 1; i < repeat_min; i++)
4061                  {
4062                  uschar *hc;
4063                  uschar *this_hwm = cd->hwm;
4064                  memcpy(code, previous, len);
4065                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4066                    {
4067                    PUT(cd->hwm, 0, GET(hc, 0) + len);
4068                    cd->hwm += LINK_SIZE;
4069                    }
4070                  save_hwm = this_hwm;
4071                  code += len;
4072                  }
4073              }              }
4074            }            }
4075    
4076          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
4077          }          }
4078    
# Line 2677  for (;; ptr++) Line 4080  for (;; ptr++)
4080        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
4081        remembering the bracket starts on a stack. In the case of a zero minimum,        remembering the bracket starts on a stack. In the case of a zero minimum,
4082        the first one was set up above. In all cases the repeat_max now specifies        the first one was set up above. In all cases the repeat_max now specifies
4083        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
4084          replicate entries on the forward reference list. */
4085    
4086        if (repeat_max >= 0)        if (repeat_max >= 0)
4087          {          {
4088          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
4089            just adjust the length as if we had. For each repetition we must add 1
4090            to the length for BRAZERO and for all but the last repetition we must
4091            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4092            paranoid checks to avoid integer overflow. */
4093    
4094            if (lengthptr != NULL && repeat_max > 0)
4095              {
4096              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4097                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4098              if ((double)repeat_max *
4099                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4100                      > (double)INT_MAX ||
4101                  OFLOW_MAX - *lengthptr < delta)
4102                {
4103                *errorcodeptr = ERR20;
4104                goto FAILED;
4105                }
4106              *lengthptr += delta;
4107              }
4108    
4109            /* This is compiling for real */
4110    
4111            else for (i = repeat_max - 1; i >= 0; i--)
4112            {            {
4113              uschar *hc;
4114              uschar *this_hwm = cd->hwm;
4115    
4116            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
4117    
4118            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2698  for (;; ptr++) Line 4128  for (;; ptr++)
4128              }              }
4129    
4130            memcpy(code, previous, len);            memcpy(code, previous, len);
4131              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4132                {
4133                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4134                cd->hwm += LINK_SIZE;
4135                }
4136              save_hwm = this_hwm;
4137            code += len;            code += len;
4138            }            }
4139    
# Line 2720  for (;; ptr++) Line 4156  for (;; ptr++)
4156        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. We
4157        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
4158        don't know if there's been an options resetting after the ket. The        don't know if there's been an options resetting after the ket. The
4159        correct offset was computed above. */        correct offset was computed above.
4160    
4161          Then, when we are doing the actual compile phase, check to see whether
4162          this group is a non-atomic one that could match an empty string. If so,
4163          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4164          that runtime checking can be done. [This check is also applied to
4165          atomic groups at runtime, but in a different way.] */
4166    
4167        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
4168            {
4169            uschar *ketcode = code - ketoffset;
4170            uschar *bracode = ketcode - GET(ketcode, 1);
4171            *ketcode = OP_KETRMAX + repeat_type;
4172            if (lengthptr == NULL && *bracode != OP_ONCE)
4173              {
4174              uschar *scode = bracode;
4175              do
4176                {
4177                if (could_be_empty_branch(scode, ketcode, utf8))
4178                  {
4179                  *bracode += OP_SBRA - OP_BRA;
4180                  break;
4181                  }
4182                scode += GET(scode, 1);
4183                }
4184              while (*scode == OP_ALT);
4185              }
4186            }
4187        }        }
4188    
4189        /* If previous is OP_FAIL, it was generated by an empty class [] in
4190        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4191        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4192        error above. We can just ignore the repeat in JS case. */
4193    
4194        else if (*previous == OP_FAIL) goto END_REPEAT;
4195    
4196      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4197    
4198      else      else
# Line 2733  for (;; ptr++) Line 4201  for (;; ptr++)
4201        goto FAILED;        goto FAILED;
4202        }        }
4203    
4204      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4205      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4206      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4207      which might be the first part of a string whose (former) last char we      anything else, we wrap the entire repeated item inside OP_ONCE brackets.
4208      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4209        but the special opcodes can optimize it a bit. The repeated item starts at
4210        tempcode, not at previous, which might be the first part of a string whose
4211        (former) last char we repeated.
4212    
4213        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4214        an 'upto' may follow. We skip over an 'exact' item, and then test the
4215        length of what remains before proceeding. */
4216    
4217      if (possessive_quantifier)      if (possessive_quantifier)
4218        {        {
4219        int len = code - tempcode;        int len;
4220        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4221        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4222        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4223        tempcode[0] = OP_ONCE;            ((*tempcode == OP_TYPEEXACT &&
4224        *code++ = OP_KET;               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4225        PUTINC(code, 0, len);        len = code - tempcode;
4226        PUT(tempcode, 1, len);        if (len > 0) switch (*tempcode)
4227            {
4228            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4229            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4230            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4231            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4232    
4233            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4234            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4235            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4236            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4237    
4238            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4239            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4240            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4241            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4242    
4243            default:
4244            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4245            code += 1 + LINK_SIZE;
4246            len += 1 + LINK_SIZE;
4247            tempcode[0] = OP_ONCE;
4248            *code++ = OP_KET;
4249            PUTINC(code, 0, len);
4250            PUT(tempcode, 1, len);
4251            break;
4252            }
4253        }        }
4254    
4255      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 2761  for (;; ptr++) Line 4262  for (;; ptr++)
4262      break;