/[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 438 by ph10, Sun Sep 6 20:00:47 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When DEBUG is defined, we need the pcre_printint() function, which is also
57    used by pcretest. DEBUG is not defined when building a production library. */
58    
59    #ifdef DEBUG
60    #include "pcre_printint.src"
61    #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
179  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
225    
226  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
227  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
228  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
229    characters are removed, and for [:alpha:] and [:alnum:] the underscore
230    character is removed. The triples in the table consist of the base map offset,
231    second map offset or -1 if no second map, and a non-negative value for map
232    addition or a negative value for map subtraction (if there are two maps). The
233    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
234    remove vertical space characters, 2 => remove underscore. */
235    
236  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
237    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
238    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
239    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
240    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
241    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
242    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
243    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
244    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
245    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
246    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
247    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
248    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
249    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
250    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
251  };  };
252    
253    
254  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
255  are passed to the outside world. */  #define XSTRING(s) STRING(s)
256    
257  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
258    "no error",  are passed to the outside world. Do not ever re-use any error number, because
259    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
260    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
261    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
262    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
263    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    simply count through to the one we want - this isn't a performance issue
265    because these strings are used only when there is a compilation error. */
266    
267    static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
319    "two named groups have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
325  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327      /* 50 */
328      "repeated subpattern is too long\0"    /** DEAD **/
329      "octal value is greater than \\377 (not in UTF-8 mode)\0"
330      "internal error: overran compiling workspace\0"
331      "internal error: previously-checked referenced subpattern not found\0"
332      "DEFINE group contains more than one branch\0"
333      /* 55 */
334      "repeating a DEFINE group is not allowed\0"
335      "inconsistent NEWLINE options\0"
336      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
337      "a numbered reference must not be zero\0"
338      "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 220  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 417  static const unsigned char digitab[] =
417    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
418    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
420    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
421    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
422    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
423    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 451  static const unsigned char ebcdic_charta
451    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
452    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
453    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
454    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
455    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
456    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
457    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 478  static const unsigned char ebcdic_charta
478  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
479    
480  static BOOL  static BOOL
481    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
482      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
483    
484    
485    
486    /*************************************************
487    *            Find an error text                  *
488    *************************************************/
489    
490    /* The error texts are now all in one long string, to save on relocations. As
491    some of the text is of unknown length, we can't use a table of offsets.
492    Instead, just count through the strings. This is not a performance issue
493    because it happens only when there has been a compilation error.
494    
495    Argument:   the error number
496    Returns:    pointer to the error string
497    */
498    
499    static const char *
500    find_error_text(int n)
501    {
502    const char *s = error_texts;
503    for (; n > 0; n--) while (*s++ != 0) {};
504    return s;
505    }
506    
507    
508  /*************************************************  /*************************************************
# Line 342  static BOOL Line 511  static BOOL
511    
512  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
513  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
514  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
515  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
516  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
517    ptr is pointing at the \. On exit, it is on the final character of the escape
518    sequence.
519    
520  Arguments:  Arguments:
521    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 526  Arguments:
526    
527  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
528                   negative => a special escape sequence                   negative => a special escape sequence
529                   on error, errorptr is set                   on error, errorcodeptr is set
530  */  */
531    
532  static int  static int
533  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
534    int options, BOOL isclass)    int options, BOOL isclass)
535  {  {
536  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
537    const uschar *ptr = *ptrptr + 1;
538  int c, i;  int c, i;
539    
540    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
541    ptr--;                            /* Set pointer back to the last byte */
542    
543  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
544    
 c = *(++ptr);  
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
548  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 562  else if ((i = escapes[c - 0x48]) != 0)
562  else  else
563    {    {
564    const uschar *oldptr;    const uschar *oldptr;
565      BOOL braced, negated;
566    
567    switch (c)    switch (c)
568      {      {
569      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580        /* \g must be followed by one of a number of specific things:
581    
582        (1) A number, either plain or braced. If positive, it is an absolute
583        backreference. If negative, it is a relative backreference. This is a Perl
584        5.10 feature.
585    
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596        case CHAR_g:
597        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606          {
607          const uschar *p;
608          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611            {
612            c = -ESC_k;
613            break;
614            }
615          braced = TRUE;
616          ptr++;
617          }
618        else braced = FALSE;
619    
620        if (ptr[1] == CHAR_MINUS)
621          {
622          negated = TRUE;
623          ptr++;
624          }
625        else negated = FALSE;
626    
627        c = 0;
628        while ((digitab[ptr[1]] & ctype_digit) != 0)
629          c = c * 10 + *(++ptr) - CHAR_0;
630    
631        if (c < 0)   /* Integer overflow */
632          {
633          *errorcodeptr = ERR61;
634          break;
635          }
636    
637        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638          {
639          *errorcodeptr = ERR57;
640          break;
641          }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647          }
648    
649        if (negated)
650          {
651          if (c > bracount)
652            {
653            *errorcodeptr = ERR15;
654            break;
655            }
656          c = bracount - (c - 1);
657          }
658    
659        c = -(ESC_REF + c);
660        break;
661    
662      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
663      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
664      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 671  else
671      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683          if (c < 0)    /* Integer overflow */
684            {
685            *errorcodeptr = ERR61;
686            break;
687            }
688        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
689          {          {
690          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 697  else
697      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 442  else Line 705  else
705        }        }
706    
707      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
708      larger first octal digit. */      larger first octal digit. The original code used just to take the least
709        significant 8 bits of octal numbers (I think this is what early Perls used
710      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
711      c -= '0';      than 3 octal digits. */
712      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
713          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
714      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
715        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716            c = c * 8 + *(++ptr) - CHAR_0;
717        if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
720      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
721      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
722        treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        register int count = 0;        int count = 0;
729    
730        c = 0;        c = 0;
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          int cc = *pt++;          register int cc = *pt++;
734            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736  #if !EBCDIC    /* ASCII coding */  
737          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
741          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745        if (*pt == '}')  
746          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
750          break;          break;
751          }          }
752    
753        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
754        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
755        }        }
 #endif  
756    
757      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
758    
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
773    
774      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
775        This coding is ASCII-specific, but then the whole concept of \cx is
776        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
782        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
783        return 0;        break;
784        }        }
785    
786      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
     if (c >= 'a' && c <= 'z') c -= 32;  
788      c ^= 0x40;      c ^= 0x40;
789  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
798      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 828  escape sequence.
828  Argument:  Argument:
829    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
830    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
831      dptr           points to an int that is set to the detailed property value
832    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
833    
834  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
835  */  */
836    
837  static int  static int
838  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
839  {  {
840  int c, i, bot, top;  int c, i, bot, top;
841  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
842  char name[4];  char name[32];
843    
844  c = *(++ptr);  c = *(++ptr);
845  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
846    
847  *negptr = FALSE;  *negptr = FALSE;
848    
849  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
850  preceded by ^ for negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
858      }      }
859    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    if (c !='}')   /* Try to distinguish error cases */    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 619  top = _pcre_utt_size; Line 884  top = _pcre_utt_size;
884    
885  while (bot < top)  while (bot < top)
886    {    {
887    i = (bot + top)/2;    i = (bot + top) >> 1;
888    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
889    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
890        {
891        *dptr = _pcre_utt[i].value;
892        return _pcre_utt[i].type;
893        }
894    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
895    }    }
896    
 UNKNOWN_RETURN:  
897  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
898  *ptrptr = ptr;  *ptrptr = ptr;
899  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 698  read_repeat_counts(const uschar *p, int Line 966  read_repeat_counts(const uschar *p, int
966  int min = 0;  int min = 0;
967  int max = -1;  int max = -1;
968    
969  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
970    an integer overflow. */
971    
972    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973    if (min < 0 || min > 65535)
974      {
975      *errorcodeptr = ERR5;
976      return p;
977      }
978    
979    /* Read the maximum value if there is one, and again do a paranoid on its size.
980    Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988        if (max < 0 || max > 65535)
989          {
990          *errorcodeptr = ERR5;
991          return p;
992          }
993      if (max < min)      if (max < min)
994        {        {
995        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 998  if (*p == '}') max = min; else
998      }      }
999    }    }
1000    
1001  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1002  pointer to the terminating '}'. */  '}'. */
1003    
1004  if (min > 65535 || max > 65535)  *minp = min;
1005    *errorcodeptr = ERR5;  *maxp = max;
1006  else  return p;
1007    }
1008    
1009    
1010    
1011    /*************************************************
1012    *  Subroutine for finding forward reference      *
1013    *************************************************/
1014    
1015    /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018    subpatterns, and counting them. If it finds a named pattern that matches the
1019    name it is given, it returns its number. Alternatively, if the name is NULL, it
1020    returns when it reaches a given numbered subpattern. We know that if (?P< is
1021    encountered, the name will be terminated by '>' because that is checked in the
1022    first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025    Arguments:
1026      ptrptr       address of the current character pointer (updated)
1027      cd           compile background data
1028      name         name to seek, or NULL if seeking a numbered subpattern
1029      lorn         name length, or subpattern number if name is NULL
1030      xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033    Returns:       the number of the named subpattern, or -1 if not found
1034    */
1035    
1036    static int
1037    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1038      BOOL xmode, int *count)
1039    {
1040    uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045    /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          term++;
1104          }
1105        }
1106    }    }
1107  return p;  
1108    /* Past any initial parenthesis handling, scan for parentheses or vertical
1109    bars. */
1110    
1111    for (; *ptr != 0; ptr++)
1112      {
1113      /* Skip over backslashed characters and also entire \Q...\E */
1114    
1115      if (*ptr == CHAR_BACKSLASH)
1116        {
1117        if (*(++ptr) == 0) goto FAIL_EXIT;
1118        if (*ptr == CHAR_Q) for (;;)
1119          {
1120          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1121          if (*ptr == 0) goto FAIL_EXIT;
1122          if (*(++ptr) == CHAR_E) break;
1123          }
1124        continue;
1125        }
1126    
1127      /* Skip over character classes; this logic must be similar to the way they
1128      are handled for real. If the first character is '^', skip it. Also, if the
1129      first few characters (either before or after ^) are \Q\E or \E we skip them
1130      too. This makes for compatibility with Perl. Note the use of STR macros to
1131      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1132    
1133      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134        {
1135        BOOL negate_class = FALSE;
1136        for (;;)
1137          {
1138          if (ptr[1] == CHAR_BACKSLASH)
1139            {
1140            if (ptr[2] == CHAR_E)
1141              ptr+= 2;
1142            else if (strncmp((const char *)ptr+2,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 4;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1149            {
1150            negate_class = TRUE;
1151            ptr++;
1152            }
1153          else break;
1154          }
1155    
1156        /* If the next character is ']', it is a data character that must be
1157        skipped, except in JavaScript compatibility mode. */
1158    
1159        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1160            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1161          ptr++;
1162    
1163        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1164          {
1165          if (*ptr == 0) return -1;
1166          if (*ptr == CHAR_BACKSLASH)
1167            {
1168            if (*(++ptr) == 0) goto FAIL_EXIT;
1169            if (*ptr == CHAR_Q) for (;;)
1170              {
1171              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1172              if (*ptr == 0) goto FAIL_EXIT;
1173              if (*(++ptr) == CHAR_E) break;
1174              }
1175            continue;
1176            }
1177          }
1178        continue;
1179        }
1180    
1181      /* Skip comments in /x mode */
1182    
1183      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1184        {
1185        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1186        if (*ptr == 0) goto FAIL_EXIT;
1187        continue;
1188        }
1189    
1190      /* Check for the special metacharacters */
1191    
1192      if (*ptr == CHAR_LEFT_PARENTHESIS)
1193        {
1194        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1195        if (rc > 0) return rc;
1196        if (*ptr == 0) goto FAIL_EXIT;
1197        }
1198    
1199      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200        {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204        }
1205    
1206      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1207        {
1208        if (*count > hwm_count) hwm_count = *count;
1209        *count = start_count;
1210        }
1211      }
1212    
1213    FAIL_EXIT:
1214    *ptrptr = ptr;
1215    return -1;
1216    }
1217    
1218    
1219    
1220    
1221    /*************************************************
1222    *       Find forward referenced subpattern       *
1223    *************************************************/
1224    
1225    /* This function scans along a pattern's text looking for capturing
1226    subpatterns, and counting them. If it finds a named pattern that matches the
1227    name it is given, it returns its number. Alternatively, if the name is NULL, it
1228    returns when it reaches a given numbered subpattern. This is used for forward
1229    references to subpatterns. We used to be able to start this scan from the
1230    current compiling point, using the current count value from cd->bracount, and
1231    do it all in a single loop, but the addition of the possibility of duplicate
1232    subpattern numbers means that we have to scan from the very start, in order to
1233    take account of such duplicates, and to use a recursive function to keep track
1234    of the different types of group.
1235    
1236    Arguments:
1237      cd           compile background data
1238      name         name to seek, or NULL if seeking a numbered subpattern
1239      lorn         name length, or subpattern number if name is NULL
1240      xmode        TRUE if we are in /x mode
1241    
1242    Returns:       the number of the found subpattern, or -1 if not found
1243    */
1244    
1245    static int
1246    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1247    {
1248    uschar *ptr = (uschar *)cd->start_pattern;
1249    int count = 0;
1250    int rc;
1251    
1252    /* If the pattern does not start with an opening parenthesis, the first call
1253    to find_parens_sub() will scan right to the end (if necessary). However, if it
1254    does start with a parenthesis, find_parens_sub() will return when it hits the
1255    matching closing parens. That is why we have to have a loop. */
1256    
1257    for (;;)
1258      {
1259      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1260      if (rc > 0 || *ptr++ == 0) break;
1261      }
1262    
1263    return rc;
1264  }  }
1265    
1266    
1267    
1268    
1269  /*************************************************  /*************************************************
1270  *      Find first significant op code            *  *      Find first significant op code            *
1271  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1315  for (;;)
1315    
1316      case OP_CALLOUT:      case OP_CALLOUT:
1317      case OP_CREF:      case OP_CREF:
1318      case OP_BRANUMBER:      case OP_RREF:
1319        case OP_DEF:
1320      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1321      break;      break;
1322    
# Line 823  for (;;) Line 1361  for (;;)
1361    {    {
1362    int d;    int d;
1363    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1364    switch (op)    switch (op)
1365      {      {
1366        case OP_CBRA:
1367      case OP_BRA:      case OP_BRA:
1368      case OP_ONCE:      case OP_ONCE:
1369      case OP_COND:      case OP_COND:
1370      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1371      if (d < 0) return d;      if (d < 0) return d;
1372      branchlength += d;      branchlength += d;
1373      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1402  for (;;)
1402      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1403    
1404      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1405      case OP_CREF:      case OP_CREF:
1406        case OP_RREF:
1407        case OP_DEF:
1408      case OP_OPT:      case OP_OPT:
1409      case OP_CALLOUT:      case OP_CALLOUT:
1410      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1422  for (;;)
1422    
1423      case OP_CHAR:      case OP_CHAR:
1424      case OP_CHARNC:      case OP_CHARNC:
1425        case OP_NOT:
1426      branchlength++;      branchlength++;
1427      cc += 2;      cc += 2;
1428  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1429      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1430        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1431  #endif  #endif
1432      break;      break;
1433    
# Line 901  for (;;) Line 1438  for (;;)
1438      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1439      cc += 4;      cc += 4;
1440  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1441      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1442        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1443  #endif  #endif
1444      break;      break;
1445    
1446      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1447      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1448        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1449      cc += 4;      cc += 4;
1450      break;      break;
1451    
# Line 917  for (;;) Line 1453  for (;;)
1453    
1454      case OP_PROP:      case OP_PROP:
1455      case OP_NOTPROP:      case OP_NOTPROP:
1456      cc++;      cc += 2;
1457      /* Fall through */      /* Fall through */
1458    
1459      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1463  for (;;)
1463      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1464      case OP_WORDCHAR:      case OP_WORDCHAR:
1465      case OP_ANY:      case OP_ANY:
1466        case OP_ALLANY:
1467      branchlength++;      branchlength++;
1468      cc++;      cc++;
1469      break;      break;
# Line 998  Returns:      pointer to the opcode for Line 1535  Returns:      pointer to the opcode for
1535  static const uschar *  static const uschar *
1536  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1537  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1538  for (;;)  for (;;)
1539    {    {
1540    register int c = *code;    register int c = *code;
1541    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1542    else if (c > OP_BRA)  
1543      /* XCLASS is used for classes that cannot be represented just by a bit
1544      map. This includes negated single high-valued characters. The length in
1545      the table is zero; the actual length is stored in the compiled code. */
1546    
1547      if (c == OP_XCLASS) code += GET(code, 1);
1548    
1549      /* Handle capturing bracket */
1550    
1551      else if (c == OP_CBRA)
1552      {      {
1553      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1554      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1555      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1556      }      }
1557    
1558      /* Otherwise, we can get the item's length from the table, except that for
1559      repeated character types, we have to test for \p and \P, which have an extra
1560      two bytes of parameters. */
1561    
1562    else    else
1563      {      {
1564      code += _pcre_OP_lengths[c];      switch(c)
1565          {
1566          case OP_TYPESTAR:
1567          case OP_TYPEMINSTAR:
1568          case OP_TYPEPLUS:
1569          case OP_TYPEMINPLUS:
1570          case OP_TYPEQUERY:
1571          case OP_TYPEMINQUERY:
1572          case OP_TYPEPOSSTAR:
1573          case OP_TYPEPOSPLUS:
1574          case OP_TYPEPOSQUERY:
1575          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1576          break;
1577    
1578  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1579          case OP_TYPEMINUPTO:
1580          case OP_TYPEEXACT:
1581          case OP_TYPEPOSUPTO:
1582          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1583          break;
1584          }
1585    
1586      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
     by a multi-byte character. The length in the table is a minimum, so we have  
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
1587    
1588        code += _pcre_OP_lengths[c];
1589    
1590      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1591      a multi-byte character. The length in the table is a minimum, so we have to
1592      arrange to skip the extra bytes. */
1593    
1594    #ifdef SUPPORT_UTF8
1595      if (utf8) switch(c)      if (utf8) switch(c)
1596        {        {
1597        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1599  for (;;)
1599        case OP_EXACT:        case OP_EXACT:
1600        case OP_UPTO:        case OP_UPTO:
1601        case OP_MINUPTO:        case OP_MINUPTO:
1602          case OP_POSUPTO:
1603        case OP_STAR:        case OP_STAR:
1604        case OP_MINSTAR:        case OP_MINSTAR:
1605          case OP_POSSTAR:
1606        case OP_PLUS:        case OP_PLUS:
1607        case OP_MINPLUS:        case OP_MINPLUS:
1608          case OP_POSPLUS:
1609        case OP_QUERY:        case OP_QUERY:
1610        case OP_MINQUERY:        case OP_MINQUERY:
1611        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1612        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;  
1613        break;        break;
1614        }        }
1615    #else
1616        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1617  #endif  #endif
1618      }      }
1619    }    }
# Line 1072  Returns:      pointer to the opcode for Line 1638  Returns:      pointer to the opcode for
1638  static const uschar *  static const uschar *
1639  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1640  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1641  for (;;)  for (;;)
1642    {    {
1643    register int c = *code;    register int c = *code;
1644    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1645    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1646    else if (c > OP_BRA)  
1647      {    /* XCLASS is used for classes that cannot be represented just by a bit
1648      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1649      }    the table is zero; the actual length is stored in the compiled code. */
1650    
1651      if (c == OP_XCLASS) code += GET(code, 1);
1652    
1653      /* Otherwise, we can get the item's length from the table, except that for
1654      repeated character types, we have to test for \p and \P, which have an extra
1655      two bytes of parameters. */
1656    
1657    else    else
1658      {      {
1659      code += _pcre_OP_lengths[c];      switch(c)
1660          {
1661          case OP_TYPESTAR:
1662          case OP_TYPEMINSTAR:
1663          case OP_TYPEPLUS:
1664          case OP_TYPEMINPLUS:
1665          case OP_TYPEQUERY:
1666          case OP_TYPEMINQUERY:
1667          case OP_TYPEPOSSTAR:
1668          case OP_TYPEPOSPLUS:
1669          case OP_TYPEPOSQUERY:
1670          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1671          break;
1672    
1673  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1674          case OP_TYPEUPTO:
1675          case OP_TYPEMINUPTO:
1676          case OP_TYPEEXACT:
1677          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1678          break;
1679          }
1680    
1681        /* Add in the fixed length from the table */
1682    
1683        code += _pcre_OP_lengths[c];
1684    
1685      /* 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
1686      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
1687      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. */  
1688    
1689    #ifdef SUPPORT_UTF8
1690      if (utf8) switch(c)      if (utf8) switch(c)
1691        {        {
1692        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1694  for (;;)
1694        case OP_EXACT:        case OP_EXACT:
1695        case OP_UPTO:        case OP_UPTO:
1696        case OP_MINUPTO:        case OP_MINUPTO:
1697          case OP_POSUPTO:
1698        case OP_STAR:        case OP_STAR:
1699        case OP_MINSTAR:        case OP_MINSTAR:
1700          case OP_POSSTAR:
1701        case OP_PLUS:        case OP_PLUS:
1702        case OP_MINPLUS:        case OP_MINPLUS:
1703          case OP_POSPLUS:
1704        case OP_QUERY:        case OP_QUERY:
1705        case OP_MINQUERY:        case OP_MINQUERY:
1706        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1707        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;  
1708        break;        break;
1709        }        }
1710    #else
1711        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1712  #endif  #endif
1713      }      }
1714    }    }
# Line 1132  for (;;) Line 1721  for (;;)
1721  *************************************************/  *************************************************/
1722    
1723  /* 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
1724  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()
1725  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
1726  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
1727  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1728    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1729    bracket whose current branch will already have been scanned.
1730    
1731  Arguments:  Arguments:
1732    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1740  static BOOL
1740  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1741  {  {
1742  register int c;  register int c;
1743  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);
1744       code < endcode;       code < endcode;
1745       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1746    {    {
# Line 1157  for (code = first_significant_code(code Line 1748  for (code = first_significant_code(code
1748    
1749    c = *code;    c = *code;
1750    
1751    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1752      first_significant_code() with a TRUE final argument. */
1753    
1754      if (c == OP_ASSERT)
1755        {
1756        do code += GET(code, 1); while (*code == OP_ALT);
1757        c = *code;
1758        continue;
1759        }
1760    
1761      /* Groups with zero repeats can of course be empty; skip them. */
1762    
1763      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1764        {
1765        code += _pcre_OP_lengths[c];
1766        do code += GET(code, 1); while (*code == OP_ALT);
1767        c = *code;
1768        continue;
1769        }
1770    
1771      /* For other groups, scan the branches. */
1772    
1773      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1774      {      {
1775      BOOL empty_branch;      BOOL empty_branch;
1776      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1777    
1778      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1779        empty branch, so just skip over the conditional, because it could be empty.
1780        Otherwise, scan the individual branches of the group. */
1781    
1782      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1783        code += GET(code, 1);        code += GET(code, 1);
1784        else
1785          {
1786          empty_branch = FALSE;
1787          do
1788            {
1789            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1790              empty_branch = TRUE;
1791            code += GET(code, 1);
1792            }
1793          while (*code == OP_ALT);
1794          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1795        }        }
1796      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1797      c = *code;      c = *code;
1798        continue;
1799      }      }
1800    
1801    else switch (c)    /* Handle the other opcodes */
1802    
1803      switch (c)
1804      {      {
1805      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1806        cannot be represented just by a bit map. This includes negated single
1807        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1808        actual length is stored in the compiled code, so we must update "code"
1809        here. */
1810    
1811  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1812      case OP_XCLASS:      case OP_XCLASS:
1813      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1814      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1815  #endif  #endif
1816    
# Line 1227  for (code = first_significant_code(code Line 1854  for (code = first_significant_code(code
1854      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1855      case OP_WORDCHAR:      case OP_WORDCHAR:
1856      case OP_ANY:      case OP_ANY:
1857        case OP_ALLANY:
1858      case OP_ANYBYTE:      case OP_ANYBYTE:
1859      case OP_CHAR:      case OP_CHAR:
1860      case OP_CHARNC:      case OP_CHARNC:
1861      case OP_NOT:      case OP_NOT:
1862      case OP_PLUS:      case OP_PLUS:
1863      case OP_MINPLUS:      case OP_MINPLUS:
1864        case OP_POSPLUS:
1865      case OP_EXACT:      case OP_EXACT:
1866      case OP_NOTPLUS:      case OP_NOTPLUS:
1867      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1868        case OP_NOTPOSPLUS:
1869      case OP_NOTEXACT:      case OP_NOTEXACT:
1870      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1871      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1872        case OP_TYPEPOSPLUS:
1873      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1874      return FALSE;      return FALSE;
1875    
1876        /* These are going to continue, as they may be empty, but we have to
1877        fudge the length for the \p and \P cases. */
1878    
1879        case OP_TYPESTAR:
1880        case OP_TYPEMINSTAR:
1881        case OP_TYPEPOSSTAR:
1882        case OP_TYPEQUERY:
1883        case OP_TYPEMINQUERY:
1884        case OP_TYPEPOSQUERY:
1885        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1886        break;
1887    
1888        /* Same for these */
1889    
1890        case OP_TYPEUPTO:
1891        case OP_TYPEMINUPTO:
1892        case OP_TYPEPOSUPTO:
1893        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1894        break;
1895    
1896      /* End of branch */      /* End of branch */
1897    
1898      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1901  for (code = first_significant_code(code
1901      case OP_ALT:      case OP_ALT:
1902      return TRUE;      return TRUE;
1903    
1904      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1905      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1906    
1907  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1908      case OP_STAR:      case OP_STAR:
1909      case OP_MINSTAR:      case OP_MINSTAR:
1910        case OP_POSSTAR:
1911      case OP_QUERY:      case OP_QUERY:
1912      case OP_MINQUERY:      case OP_MINQUERY:
1913        case OP_POSQUERY:
1914        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1915        break;
1916    
1917      case OP_UPTO:      case OP_UPTO:
1918      case OP_MINUPTO:      case OP_MINUPTO:
1919      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
1920        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1921      break;      break;
1922  #endif  #endif
1923      }      }
# Line 1308  return TRUE; Line 1965  return TRUE;
1965  *************************************************/  *************************************************/
1966    
1967  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1968  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
1969  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1970  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1971    
1972    Originally, this function only recognized a sequence of letters between the
1973    terminators, but it seems that Perl recognizes any sequence of characters,
1974    though of course unknown POSIX names are subsequently rejected. Perl gives an
1975    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1976    didn't consider this to be a POSIX class. Likewise for [:1234:].
1977    
1978    The problem in trying to be exactly like Perl is in the handling of escapes. We
1979    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1980    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1981    below handles the special case of \], but does not try to do any other escape
1982    processing. This makes it different from Perl for cases such as [:l\ower:]
1983    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1984    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1985    I think.
1986    
1987  Argument:  Arguments:
1988    ptr      pointer to the initial [    ptr      pointer to the initial [
1989    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1990    
1991  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1992  */  */
1993    
1994  static BOOL  static BOOL
1995  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1996  {  {
1997  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1998  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1999  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2000    {    {
2001    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2002    return TRUE;      {
2003        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2004        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2005          {
2006          *endptr = ptr;
2007          return TRUE;
2008          }
2009        }
2010    }    }
2011  return FALSE;  return FALSE;
2012  }  }
# Line 1355  Returns:     a value representing the na Line 2031  Returns:     a value representing the na
2031  static int  static int
2032  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2033  {  {
2034    const char *pn = posix_names;
2035  register int yield = 0;  register int yield = 0;
2036  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2037    {    {
2038    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2039      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2040      pn += posix_name_lengths[yield] + 1;
2041    yield++;    yield++;
2042    }    }
2043  return -1;  return -1;
# Line 1374  return -1; Line 2052  return -1;
2052  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2053  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2054  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
2055  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
2056  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
2057  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
2058  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
2059  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2060    OP_END.
2061    
2062    This function has been extended with the possibility of forward references for
2063    recursions and subroutine calls. It must also check the list of such references
2064    for the group we are dealing with. If it finds that one of the recursions in
2065    the current group is on this list, it adjusts the offset in the list, not the
2066    value in the reference (which is a group number).
2067    
2068  Arguments:  Arguments:
2069    group      points to the start of the group    group      points to the start of the group
2070    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2071    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2072    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2073      save_hwm   the hwm forward reference pointer at the start of the group
2074    
2075  Returns:     nothing  Returns:     nothing
2076  */  */
2077    
2078  static void  static void
2079  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2080      uschar *save_hwm)
2081  {  {
2082  uschar *ptr = group;  uschar *ptr = group;
2083    
2084  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2085    {    {
2086    int offset = GET(ptr, 1);    int offset;
2087    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2088    
2089      /* See if this recursion is on the forward reference list. If so, adjust the
2090      reference. */
2091    
2092      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2093        {
2094        offset = GET(hc, 0);
2095        if (cd->start_code + offset == ptr + 1)
2096          {
2097          PUT(hc, 0, offset + adjust);
2098          break;
2099          }
2100        }
2101    
2102      /* Otherwise, adjust the recursion offset if it's after the start of this
2103      group. */
2104    
2105      if (hc >= cd->hwm)
2106        {
2107        offset = GET(ptr, 1);
2108        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2109        }
2110    
2111    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2112    }    }
2113  }  }
# Line 1475  Yield:        TRUE when range returned; Line 2186  Yield:        TRUE when range returned;
2186  */  */
2187    
2188  static BOOL  static BOOL
2189  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2190      unsigned int *odptr)
2191  {  {
2192  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2193    
2194  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2195    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2196    
2197  if (c > d) return FALSE;  if (c > d) return FALSE;
2198    
# Line 1492  next = othercase + 1; Line 2201  next = othercase + 1;
2201    
2202  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2203    {    {
2204    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2205    next++;    next++;
2206    }    }
2207    
# Line 1506  return TRUE; Line 2213  return TRUE;
2213  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2214    
2215    
2216    
2217    /*************************************************
2218    *     Check if auto-possessifying is possible    *
2219    *************************************************/
2220    
2221    /* This function is called for unlimited repeats of certain items, to see
2222    whether the next thing could possibly match the repeated item. If not, it makes
2223    sense to automatically possessify the repeated item.
2224    
2225    Arguments:
2226      op_code       the repeated op code
2227      this          data for this item, depends on the opcode
2228      utf8          TRUE in UTF-8 mode
2229      utf8_char     used for utf8 character bytes, NULL if not relevant
2230      ptr           next character in pattern
2231      options       options bits
2232      cd            contains pointers to tables etc.
2233    
2234    Returns:        TRUE if possessifying is wanted
2235    */
2236    
2237    static BOOL
2238    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2239      const uschar *ptr, int options, compile_data *cd)
2240    {
2241    int next;
2242    
2243    /* Skip whitespace and comments in extended mode */
2244    
2245    if ((options & PCRE_EXTENDED) != 0)
2246      {
2247      for (;;)
2248        {
2249        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2250        if (*ptr == CHAR_NUMBER_SIGN)
2251          {
2252          while (*(++ptr) != 0)
2253            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2254          }
2255        else break;
2256        }
2257      }
2258    
2259    /* If the next item is one that we can handle, get its value. A non-negative
2260    value is a character, a negative value is an escape value. */
2261    
2262    if (*ptr == CHAR_BACKSLASH)
2263      {
2264      int temperrorcode = 0;
2265      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2266      if (temperrorcode != 0) return FALSE;
2267      ptr++;    /* Point after the escape sequence */
2268      }
2269    
2270    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2271      {
2272    #ifdef SUPPORT_UTF8
2273      if (utf8) { GETCHARINC(next, ptr); } else
2274    #endif
2275      next = *ptr++;
2276      }
2277    
2278    else return FALSE;
2279    
2280    /* Skip whitespace and comments in extended mode */
2281    
2282    if ((options & PCRE_EXTENDED) != 0)
2283      {
2284      for (;;)
2285        {
2286        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2287        if (*ptr == CHAR_NUMBER_SIGN)
2288          {
2289          while (*(++ptr) != 0)
2290            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2291          }
2292        else break;
2293        }
2294      }
2295    
2296    /* If the next thing is itself optional, we have to give up. */
2297    
2298    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2299      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2300        return FALSE;
2301    
2302    /* Now compare the next item with the previous opcode. If the previous is a
2303    positive single character match, "item" either contains the character or, if
2304    "item" is greater than 127 in utf8 mode, the character's bytes are in
2305    utf8_char. */
2306    
2307    
2308    /* Handle cases when the next item is a character. */
2309    
2310    if (next >= 0) switch(op_code)
2311      {
2312      case OP_CHAR:
2313    #ifdef SUPPORT_UTF8
2314      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2315    #else
2316      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2317    #endif
2318      return item != next;
2319    
2320      /* For CHARNC (caseless character) we must check the other case. If we have
2321      Unicode property support, we can use it to test the other case of
2322      high-valued characters. */
2323    
2324      case OP_CHARNC:
2325    #ifdef SUPPORT_UTF8
2326      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2327    #endif
2328      if (item == next) return FALSE;
2329    #ifdef SUPPORT_UTF8
2330      if (utf8)
2331        {
2332        unsigned int othercase;
2333        if (next < 128) othercase = cd->fcc[next]; else
2334    #ifdef SUPPORT_UCP
2335        othercase = UCD_OTHERCASE((unsigned int)next);
2336    #else
2337        othercase = NOTACHAR;
2338    #endif
2339        return (unsigned int)item != othercase;
2340        }
2341      else
2342    #endif  /* SUPPORT_UTF8 */
2343      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2344    
2345      /* For OP_NOT, "item" must be a single-byte character. */
2346    
2347      case OP_NOT:
2348      if (item == next) return TRUE;
2349      if ((options & PCRE_CASELESS) == 0) return FALSE;
2350    #ifdef SUPPORT_UTF8
2351      if (utf8)
2352        {
2353        unsigned int othercase;
2354        if (next < 128) othercase = cd->fcc[next]; else
2355    #ifdef SUPPORT_UCP
2356        othercase = UCD_OTHERCASE(next);
2357    #else
2358        othercase = NOTACHAR;
2359    #endif
2360        return (unsigned int)item == othercase;
2361        }
2362      else
2363    #endif  /* SUPPORT_UTF8 */
2364      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2365    
2366      case OP_DIGIT:
2367      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2368    
2369      case OP_NOT_DIGIT:
2370      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2371    
2372      case OP_WHITESPACE:
2373      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2374    
2375      case OP_NOT_WHITESPACE:
2376      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2377    
2378      case OP_WORDCHAR:
2379      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2380    
2381      case OP_NOT_WORDCHAR:
2382      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2383    
2384      case OP_HSPACE:
2385      case OP_NOT_HSPACE:
2386      switch(next)
2387        {
2388        case 0x09:
2389        case 0x20:
2390        case 0xa0:
2391        case 0x1680:
2392        case 0x180e:
2393        case 0x2000:
2394        case 0x2001:
2395        case 0x2002:
2396        case 0x2003:
2397        case 0x2004:
2398        case 0x2005:
2399        case 0x2006:
2400        case 0x2007:
2401        case 0x2008:
2402        case 0x2009:
2403        case 0x200A:
2404        case 0x202f:
2405        case 0x205f:
2406        case 0x3000:
2407        return op_code != OP_HSPACE;
2408        default:
2409        return op_code == OP_HSPACE;
2410        }
2411    
2412      case OP_VSPACE:
2413      case OP_NOT_VSPACE:
2414      switch(next)
2415        {
2416        case 0x0a:
2417        case 0x0b:
2418        case 0x0c:
2419        case 0x0d:
2420        case 0x85:
2421        case 0x2028:
2422        case 0x2029:
2423        return op_code != OP_VSPACE;
2424        default:
2425        return op_code == OP_VSPACE;
2426        }
2427    
2428      default:
2429      return FALSE;
2430      }
2431    
2432    
2433    /* Handle the case when the next item is \d, \s, etc. */
2434    
2435    switch(op_code)
2436      {
2437      case OP_CHAR:
2438      case OP_CHARNC:
2439    #ifdef SUPPORT_UTF8
2440      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2441    #endif
2442      switch(-next)
2443        {
2444        case ESC_d:
2445        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2446    
2447        case ESC_D:
2448        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2449    
2450        case ESC_s:
2451        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2452    
2453        case ESC_S:
2454        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2455    
2456        case ESC_w:
2457        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2458    
2459        case ESC_W:
2460        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2461    
2462        case ESC_h:
2463        case ESC_H:
2464        switch(item)
2465          {
2466          case 0x09:
2467          case 0x20:
2468          case 0xa0:
2469          case 0x1680:
2470          case 0x180e:
2471          case 0x2000:
2472          case 0x2001:
2473          case 0x2002:
2474          case 0x2003:
2475          case 0x2004:
2476          case 0x2005:
2477          case 0x2006:
2478          case 0x2007:
2479          case 0x2008:
2480          case 0x2009:
2481          case 0x200A:
2482          case 0x202f:
2483          case 0x205f:
2484          case 0x3000:
2485          return -next != ESC_h;
2486          default:
2487          return -next == ESC_h;
2488          }
2489    
2490        case ESC_v:
2491        case ESC_V:
2492        switch(item)
2493          {
2494          case 0x0a:
2495          case 0x0b:
2496          case 0x0c:
2497          case 0x0d:
2498          case 0x85:
2499          case 0x2028:
2500          case 0x2029:
2501          return -next != ESC_v;
2502          default:
2503          return -next == ESC_v;
2504          }
2505    
2506        default:
2507        return FALSE;
2508        }
2509    
2510      case OP_DIGIT:
2511      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2512             next == -ESC_h || next == -ESC_v;
2513    
2514      case OP_NOT_DIGIT:
2515      return next == -ESC_d;
2516    
2517      case OP_WHITESPACE:
2518      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2519    
2520      case OP_NOT_WHITESPACE:
2521      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2522    
2523      case OP_HSPACE:
2524      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2525    
2526      case OP_NOT_HSPACE:
2527      return next == -ESC_h;
2528    
2529      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2530      case OP_VSPACE:
2531      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2532    
2533      case OP_NOT_VSPACE:
2534      return next == -ESC_v;
2535    
2536      case OP_WORDCHAR:
2537      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2538    
2539      case OP_NOT_WORDCHAR:
2540      return next == -ESC_w || next == -ESC_d;
2541    
2542      default:
2543      return FALSE;
2544      }
2545    
2546    /* Control does not reach here */
2547    }
2548    
2549    
2550    
2551  /*************************************************  /*************************************************
2552  *           Compile one branch                   *  *           Compile one branch                   *
2553  *************************************************/  *************************************************/
2554    
2555  /* 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
2556  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
2557  bits.  bits. This function is used during the pre-compile phase when we are trying
2558    to find out the amount of memory needed, as well as during the real compile
2559    phase. The value of lengthptr distinguishes the two phases.
2560    
2561  Arguments:  Arguments:
2562    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2563    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2564    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2565    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2567  Arguments:
2567    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2568    bcptr          points to current branch chain    bcptr          points to current branch chain
2569    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2570      lengthptr      NULL during the real compile phase
2571                     points to length accumulator during pre-compile phase
2572    
2573  Returns:         TRUE on success  Returns:         TRUE on success
2574                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2575  */  */
2576    
2577  static BOOL  static BOOL
2578  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2579    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2580    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2581  {  {
2582  int repeat_type, op_type;  int repeat_type, op_type;
2583  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 2586  int greedy_default, greedy_non_default;
2586  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2587  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2588  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2589  int options = *optionsptr;  int options = *optionsptr;
2590  int after_manual_callout = 0;  int after_manual_callout = 0;
2591    int length_prevgroup = 0;
2592  register int c;  register int c;
2593  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2594    uschar *last_code = code;
2595    uschar *orig_code = code;
2596  uschar *tempcode;  uschar *tempcode;
2597  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2598  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2600  const uschar *ptr = *ptrptr;
2600  const uschar *tempptr;  const uschar *tempptr;
2601  uschar *previous = NULL;  uschar *previous = NULL;
2602  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2603    uschar *save_hwm = NULL;
2604  uschar classbits[32];  uschar classbits[32];
2605    
2606  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2607  BOOL class_utf8;  BOOL class_utf8;
2608  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2609  uschar *class_utf8data;  uschar *class_utf8data;
2610    uschar *class_utf8data_base;
2611  uschar utf8_char[6];  uschar utf8_char[6];
2612  #else  #else
2613  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2614    uschar *utf8_char = NULL;
2615    #endif
2616    
2617    #ifdef DEBUG
2618    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2619  #endif  #endif
2620    
2621  /* 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 2647  req_caseopt = ((options & PCRE_CASELESS)
2647  for (;; ptr++)  for (;; ptr++)
2648    {    {
2649    BOOL negate_class;    BOOL negate_class;
2650      BOOL should_flip_negation;
2651    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2652    BOOL is_quantifier;    BOOL is_quantifier;
2653      BOOL is_recurse;
2654      BOOL reset_bracount;
2655    int class_charcount;    int class_charcount;
2656    int class_lastchar;    int class_lastchar;
2657    int newoptions;    int newoptions;
2658    int recno;    int recno;
2659      int refsign;
2660    int skipbytes;    int skipbytes;
2661    int subreqbyte;    int subreqbyte;
2662    int subfirstbyte;    int subfirstbyte;
2663      int terminator;
2664    int mclength;    int mclength;
2665    uschar mcbuffer[8];    uschar mcbuffer[8];
2666    
2667    /* Next byte in the pattern */    /* Get next byte in the pattern */
2668    
2669    c = *ptr;    c = *ptr;
2670    
2671      /* If we are in the pre-compile phase, accumulate the length used for the
2672      previous cycle of this loop. */
2673    
2674      if (lengthptr != NULL)
2675        {
2676    #ifdef DEBUG
2677        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2678    #endif
2679        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2680          {
2681          *errorcodeptr = ERR52;
2682          goto FAILED;
2683          }
2684    
2685        /* There is at least one situation where code goes backwards: this is the
2686        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2687        the class is simply eliminated. However, it is created first, so we have to
2688        allow memory for it. Therefore, don't ever reduce the length at this point.
2689        */
2690    
2691        if (code < last_code) code = last_code;
2692    
2693        /* Paranoid check for integer overflow */
2694    
2695        if (OFLOW_MAX - *lengthptr < code - last_code)
2696          {
2697          *errorcodeptr = ERR20;
2698          goto FAILED;
2699          }
2700    
2701        *lengthptr += code - last_code;
2702        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2703    
2704        /* If "previous" is set and it is not at the start of the work space, move
2705        it back to there, in order to avoid filling up the work space. Otherwise,
2706        if "previous" is NULL, reset the current code pointer to the start. */
2707    
2708        if (previous != NULL)
2709          {
2710          if (previous > orig_code)
2711            {
2712            memmove(orig_code, previous, code - previous);
2713            code -= previous - orig_code;
2714            previous = orig_code;
2715            }
2716          }
2717        else code = orig_code;
2718    
2719        /* Remember where this code item starts so we can pick up the length
2720        next time round. */
2721    
2722        last_code = code;
2723        }
2724    
2725      /* In the real compile phase, just check the workspace used by the forward
2726      reference list. */
2727    
2728      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2729        {
2730        *errorcodeptr = ERR52;
2731        goto FAILED;
2732        }
2733    
2734    /* 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 */
2735    
2736    if (inescq && c != 0)    if (inescq && c != 0)
2737      {      {
2738      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2739        {        {
2740        inescq = FALSE;        inescq = FALSE;
2741        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 2745  for (;; ptr++)
2745        {        {
2746        if (previous_callout != NULL)        if (previous_callout != NULL)
2747          {          {
2748          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2749              complete_callout(previous_callout, ptr, cd);
2750          previous_callout = NULL;          previous_callout = NULL;
2751          }          }
2752        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2761  for (;; ptr++)
2761    /* 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
2762    a quantifier. */    a quantifier. */
2763    
2764    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2765      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2766        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2767    
2768    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2769         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2770      {      {
2771      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2772          complete_callout(previous_callout, ptr, cd);
2773      previous_callout = NULL;      previous_callout = NULL;
2774      }      }
2775    
# Line 1653  for (;; ptr++) Line 2778  for (;; ptr++)
2778    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2779      {      {
2780      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2781      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2782        {        {
2783        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2784        on the Macintosh. */          {
2785        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2786        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2787          if (*ptr != 0) continue;
2788    
2789          /* Else fall through to handle end of string */
2790          c = 0;
2791        }        }
2792      }      }
2793    
# Line 1672  for (;; ptr++) Line 2801  for (;; ptr++)
2801    
2802    switch(c)    switch(c)
2803      {      {
2804      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2805        case 0:                        /* The branch terminates at string end */
2806      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2807      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2808      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2809      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2810      *codeptr = code;      *codeptr = code;
2811      *ptrptr = ptr;      *ptrptr = ptr;
2812        if (lengthptr != NULL)
2813          {
2814          if (OFLOW_MAX - *lengthptr < code - last_code)
2815            {
2816            *errorcodeptr = ERR20;
2817            goto FAILED;
2818            }
2819          *lengthptr += code - last_code;   /* To include callout length */
2820          DPRINTF((">> end branch\n"));
2821          }
2822      return TRUE;      return TRUE;
2823    
2824    
2825        /* ===================================================================*/
2826      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2827      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2828    
2829      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2830      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2831        {        {
2832        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2835  for (;; ptr++)
2835      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2836      break;      break;
2837    
2838      case '$':      case CHAR_DOLLAR_SIGN:
2839      previous = NULL;      previous = NULL;
2840      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2841      break;      break;
# Line 1703  for (;; ptr++) Line 2843  for (;; ptr++)
2843      /* 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
2844      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2845    
2846      case '.':      case CHAR_DOT:
2847      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2848      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2849      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2850      previous = code;      previous = code;
2851      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2852      break;      break;
2853    
2854      /* Character classes. If the included characters are all < 255 in value, we  
2855      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2856      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
2857      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
2858      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2859        map as usual, then invert it at the end. However, we use a different opcode
2860        so that data characters > 255 can be handled correctly.
2861    
2862      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2863      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,
2864      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2865      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.
     */  
2866    
2867      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2868        default (Perl) mode, it is treated as a data character. */
2869    
2870        case CHAR_RIGHT_SQUARE_BRACKET:
2871        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2872          {
2873          *errorcodeptr = ERR64;
2874          goto FAILED;
2875          }
2876        goto NORMAL_CHAR;
2877    
2878        case CHAR_LEFT_SQUARE_BRACKET:
2879      previous = code;      previous = code;
2880    
2881      /* 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
2882      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. */
2883    
2884      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2885          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2886            check_posix_syntax(ptr, &tempptr))
2887        {        {
2888        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2889        goto FAILED;        goto FAILED;
2890        }        }
2891    
2892      /* 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,
2893        if the first few characters (either before or after ^) are \Q\E or \E we
2894        skip them too. This makes for compatibility with Perl. */
2895    
2896      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2897        for (;;)
2898        {        {
       negate_class = TRUE;  
2899        c = *(++ptr);        c = *(++ptr);
2900          if (c == CHAR_BACKSLASH)
2901            {
2902            if (ptr[1] == CHAR_E)
2903              ptr++;
2904            else if (strncmp((const char *)ptr+1,
2905                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2906              ptr += 3;
2907            else
2908              break;
2909            }
2910          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2911            negate_class = TRUE;
2912          else break;
2913        }        }
2914      else  
2915        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2916        an initial ']' is taken as a data character -- the code below handles
2917        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2918        [^] must match any character, so generate OP_ALLANY. */
2919    
2920        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2921            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2922        {        {
2923        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2924          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2925          zerofirstbyte = firstbyte;
2926          break;
2927        }        }
2928    
2929        /* If a class contains a negative special such as \S, we need to flip the
2930        negation flag at the end, so that support for characters > 255 works
2931        correctly (they are all included in the class). */
2932    
2933        should_flip_negation = FALSE;
2934    
2935      /* 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
2936      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
2937      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2938    
2939      class_charcount = 0;      class_charcount = 0;
2940      class_lastchar = -1;      class_lastchar = -1;
2941    
2942        /* Initialize the 32-char bit map to all zeros. We build the map in a
2943        temporary bit of memory, in case the class contains only 1 character (less
2944        than 256), because in that case the compiled code doesn't use the bit map.
2945        */
2946    
2947        memset(classbits, 0, 32 * sizeof(uschar));
2948    
2949  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2950      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2951      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2952        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2953  #endif  #endif
2954    
     /* 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));  
   
2955      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2956      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
2957      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. */  
2958    
2959      do      if (c != 0) do
2960        {        {
2961          const uschar *oldptr;
2962    
2963  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2964        if (utf8 && c > 127)        if (utf8 && c > 127)
2965          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2966          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2967          }          }
2968    
2969          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2970          data and reset the pointer. This is so that very large classes that
2971          contain a zillion UTF-8 characters no longer overwrite the work space
2972          (which is on the stack). */
2973    
2974          if (lengthptr != NULL)
2975            {
2976            *lengthptr += class_utf8data - class_utf8data_base;
2977            class_utf8data = class_utf8data_base;
2978            }
2979    
2980  #endif  #endif
2981    
2982        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2983    
2984        if (inescq)        if (inescq)
2985          {          {
2986          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2987            {            {
2988            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2989            ptr++;            ptr++;                            /* Skip the 'E' */
2990            continue;            continue;                         /* Carry on with next */
2991            }            }
2992          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2993          }          }
2994    
2995        /* 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 2998  for (;; ptr++)
2998        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2999        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3000    
3001        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3002            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3003            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3004          {          {
3005          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3006          int posix_class, i;          int posix_class, taboffset, tabopt;
3007          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3008            uschar pbits[32];
3009    
3010          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3011            {            {
3012            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3013            goto FAILED;            goto FAILED;
3014            }            }
3015    
3016          ptr += 2;          ptr += 2;
3017          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3018            {            {
3019            local_negate = TRUE;            local_negate = TRUE;
3020              should_flip_negation = TRUE;  /* Note negative special */
3021            ptr++;            ptr++;
3022            }            }
3023    
# Line 1836  for (;; ptr++) Line 3035  for (;; ptr++)
3035          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3036            posix_class = 0;            posix_class = 0;
3037    
3038          /* 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
3039          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
3040          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
3041          white space chars afterwards. */          result into the bit map that is being built. */
3042    
3043          posix_class *= 3;          posix_class *= 3;
3044          for (i = 0; i < 3; i++)  
3045            /* Copy in the first table (always present) */
3046    
3047            memcpy(pbits, cbits + posix_class_maps[posix_class],
3048              32 * sizeof(uschar));
3049    
3050            /* If there is a second table, add or remove it as required. */
3051    
3052            taboffset = posix_class_maps[posix_class + 1];
3053            tabopt = posix_class_maps[posix_class + 2];
3054    
3055            if (taboffset >= 0)
3056            {            {
3057            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3058            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;  
             }  
3059            else            else
3060              {              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;  
             }  
3061            }            }
3062    
3063            /* Not see if we need to remove any special characters. An option
3064            value of 1 removes vertical space and 2 removes underscore. */
3065    
3066            if (tabopt < 0) tabopt = -tabopt;
3067            if (tabopt == 1) pbits[1] &= ~0x3c;
3068              else if (tabopt == 2) pbits[11] &= 0x7f;
3069    
3070            /* Add the POSIX table or its complement into the main table that is
3071            being built and we are done. */
3072    
3073            if (local_negate)
3074              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3075            else
3076              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3077    
3078          ptr = tempptr + 1;          ptr = tempptr + 1;
3079          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3080          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3081          }          }
3082    
3083        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3084        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
3085        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.
3086        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3087        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  
3088        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3089    
3090        if (c == '\\')        if (c == CHAR_BACKSLASH)
3091          {          {
3092          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3093            if (*errorcodeptr != 0) goto FAILED;
3094    
3095          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 */
3096          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 */
3097            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3098          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3099            {            {
3100            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3101              {              {
3102              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3103              }              }
3104            else inescq = TRUE;            else inescq = TRUE;
3105            continue;            continue;
3106            }            }
3107            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3108    
3109          if (c < 0)          if (c < 0)
3110            {            {
3111            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3112            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3113            switch (-c)  
3114              /* Save time by not doing this in the pre-compile phase. */
3115    
3116              if (lengthptr == NULL) switch (-c)
3117              {              {
3118              case ESC_d:              case ESC_d:
3119              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3120              continue;              continue;
3121    
3122              case ESC_D:              case ESC_D:
3123                should_flip_negation = TRUE;
3124              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3125              continue;              continue;
3126    
# Line 1910  for (;; ptr++) Line 3129  for (;; ptr++)
3129              continue;              continue;
3130    
3131              case ESC_W:              case ESC_W:
3132                should_flip_negation = TRUE;
3133              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3134              continue;              continue;
3135    
# Line 1919  for (;; ptr++) Line 3139  for (;; ptr++)
3139              continue;              continue;
3140    
3141              case ESC_S:              case ESC_S:
3142                should_flip_negation = TRUE;
3143              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3144              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3145              continue;              continue;
3146    
3147  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3148              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3149              case ESC_P:              }
3150    
3151              /* In the pre-compile phase, just do the recognition. */
3152    
3153              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3154                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3155    
3156              /* We need to deal with \H, \h, \V, and \v in both phases because
3157              they use extra memory. */
3158    
3159              if (-c == ESC_h)
3160                {
3161                SETBIT(classbits, 0x09); /* VT */
3162                SETBIT(classbits, 0x20); /* SPACE */
3163                SETBIT(classbits, 0xa0); /* NSBP */
3164    #ifdef SUPPORT_UTF8
3165                if (utf8)
3166                  {
3167                  class_utf8 = TRUE;
3168                  *class_utf8data++ = XCL_SINGLE;
3169                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3170                  *class_utf8data++ = XCL_SINGLE;
3171                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3172                  *class_utf8data++ = XCL_RANGE;
3173                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3174                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3175                  *class_utf8data++ = XCL_SINGLE;
3176                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3177                  *class_utf8data++ = XCL_SINGLE;
3178                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3179                  *class_utf8data++ = XCL_SINGLE;
3180                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3181                  }
3182    #endif
3183                continue;
3184                }
3185    
3186              if (-c == ESC_H)
3187                {
3188                for (c = 0; c < 32; c++)
3189                  {
3190                  int x = 0xff;
3191                  switch (c)
3192                    {
3193                    case 0x09/8: x ^= 1 << (0x09%8); break;
3194                    case 0x20/8: x ^= 1 << (0x20%8); break;
3195                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3196                    default: break;
3197                    }
3198                  classbits[c] |= x;
3199                  }
3200    
3201    #ifdef SUPPORT_UTF8
3202                if (utf8)
3203                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3204                class_utf8 = TRUE;                class_utf8 = TRUE;
3205                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3206                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3207                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3208                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3209                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3210                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3211                  *class_utf8data++ = XCL_RANGE;
3212                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3213                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3214                  *class_utf8data++ = XCL_RANGE;
3215                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3216                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3217                  *class_utf8data++ = XCL_RANGE;
3218                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3219                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3220                  *class_utf8data++ = XCL_RANGE;
3221                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3222                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3223                  *class_utf8data++ = XCL_RANGE;
3224                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3225                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3226                }                }
3227    #endif
3228              continue;              continue;
3229                }
3230    
3231              if (-c == ESC_v)
3232                {
3233                SETBIT(classbits, 0x0a); /* LF */
3234                SETBIT(classbits, 0x0b); /* VT */
3235                SETBIT(classbits, 0x0c); /* FF */
3236                SETBIT(classbits, 0x0d); /* CR */
3237                SETBIT(classbits, 0x85); /* NEL */
3238    #ifdef SUPPORT_UTF8
3239                if (utf8)
3240                  {
3241                  class_utf8 = TRUE;
3242                  *class_utf8data++ = XCL_RANGE;
3243                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3244                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3245                  }
3246  #endif  #endif
3247                continue;
3248                }
3249    
3250              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3251              strict mode. By default, for compatibility with Perl, they are              {
3252              treated as literals. */              for (c = 0; c < 32; c++)
3253                  {
3254                  int x = 0xff;
3255                  switch (c)
3256                    {
3257                    case 0x0a/8: x ^= 1 << (0x0a%8);
3258                                 x ^= 1 << (0x0b%8);
3259                                 x ^= 1 << (0x0c%8);
3260                                 x ^= 1 << (0x0d%8);
3261                                 break;
3262                    case 0x85/8: x ^= 1 << (0x85%8); break;
3263                    default: break;
3264                    }
3265                  classbits[c] |= x;
3266                  }
3267    
3268              default:  #ifdef SUPPORT_UTF8
3269              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3270                {                {
3271                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3272                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3273                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3274                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3275                  *class_utf8data++ = XCL_RANGE;
3276                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3277                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3278                }                }
3279              c = *ptr;              /* The final character */  #endif
3280              class_charcount -= 2;  /* Undo the default count from above */              continue;
3281                }
3282    
3283              /* We need to deal with \P and \p in both phases. */
3284    
3285    #ifdef SUPPORT_UCP
3286              if (-c == ESC_p || -c == ESC_P)
3287                {
3288                BOOL negated;
3289                int pdata;
3290                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3291                if (ptype < 0) goto FAILED;
3292                class_utf8 = TRUE;
3293                *class_utf8data++ = ((-c == ESC_p) != negated)?
3294                  XCL_PROP : XCL_NOTPROP;
3295                *class_utf8data++ = ptype;
3296                *class_utf8data++ = pdata;
3297                class_charcount -= 2;   /* Not a < 256 character */
3298                continue;
3299                }
3300    #endif
3301              /* Unrecognized escapes are faulted if PCRE is running in its
3302              strict mode. By default, for compatibility with Perl, they are
3303              treated as literals. */
3304    
3305              if ((options & PCRE_EXTRA) != 0)
3306                {
3307                *errorcodeptr = ERR7;
3308                goto FAILED;
3309              }              }
3310    
3311              class_charcount -= 2;  /* Undo the default count from above */
3312              c = *ptr;              /* Get the final character and fall through */
3313            }            }
3314    
3315          /* 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
3316          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3317    
3318          }   /* End of backslash handling */          }   /* End of backslash handling */
3319    
3320        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3321        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
3322        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3323          entirely. The code for handling \Q and \E is messy. */
3324    
3325          CHECK_RANGE:
3326          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3327            {
3328            inescq = FALSE;
3329            ptr += 2;
3330            }
3331    
3332          oldptr = ptr;
3333    
3334          /* Remember \r or \n */
3335    
3336          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3337    
3338        if (ptr[1] == '-' && ptr[2] != ']')        /* Check for range */
3339    
3340          if (!inescq && ptr[1] == CHAR_MINUS)
3341          {          {
3342          int d;          int d;
3343          ptr += 2;          ptr += 2;
3344            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3345    
3346            /* If we hit \Q (not followed by \E) at this point, go into escaped
3347            mode. */
3348    
3349            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3350              {
3351              ptr += 2;
3352              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3353                { ptr += 2; continue; }
3354              inescq = TRUE;
3355              break;
3356              }
3357    
3358            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3359              {
3360              ptr = oldptr;
3361              goto LONE_SINGLE_CHARACTER;
3362              }
3363    
3364  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3365          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3374  for (;; ptr++)
3374          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
3375          in such circumstances. */          in such circumstances. */
3376    
3377          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3378            {            {
3379            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3380            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3381    
3382            /* \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
3383            was literal */            special means the '-' was literal */
3384    
3385            if (d < 0)            if (d < 0)
3386              {              {
3387              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3388              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3389                else if (d == -ESC_R) d = CHAR_R; else
3390                {                {
3391                ptr = oldptr - 2;                ptr = oldptr;
3392                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3393                }                }
3394              }              }
3395            }            }
3396    
3397          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3398          the pre-pass. Optimize one-character ranges */          one-character ranges */
3399    
3400            if (d < c)
3401              {
3402              *errorcodeptr = ERR8;
3403              goto FAILED;
3404              }
3405    
3406          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3407    
3408            /* Remember \r or \n */
3409    
3410            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3411    
3412          /* 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
3413          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3414          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 3426  for (;; ptr++)
3426  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3427            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3428              {              {
3429              int occ, ocd;              unsigned int occ, ocd;
3430              int cc = c;              unsigned int cc = c;
3431              int origd = d;              unsigned int origd = d;
3432              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3433                {                {
3434                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3435                      ocd <= (unsigned int)d)
3436                    continue;                          /* Skip embedded ranges */
3437    
3438                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3439                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3440                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3441                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3442                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3443                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3444                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3445                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3446                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3447                  d = ocd;                  d = ocd;
3448                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3490  for (;; ptr++)
3490          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
3491          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3492    
3493          for (; c <= d; c++)          class_charcount += d - c + 1;
3494            class_lastchar = d;
3495    
3496            /* We can save a bit of time by skipping this in the pre-compile. */
3497    
3498            if (lengthptr == NULL) for (; c <= d; c++)
3499            {            {
3500            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3501            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3503  for (;; ptr++)
3503              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3504              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3505              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3506            }            }
3507    
3508          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3526  for (;; ptr++)
3526  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3527          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3528            {            {
3529            int chartype;            unsigned int othercase;
3530            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3531              {              {
3532              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3533              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3552  for (;; ptr++)
3552          }          }
3553        }        }
3554    
3555      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3556      loop. This "while" is the end of the "do" above. */  
3557        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3558    
3559        if (c == 0)                          /* Missing terminating ']' */
3560          {
3561          *errorcodeptr = ERR6;
3562          goto FAILED;
3563          }
3564    
3565    
3566    /* This code has been disabled because it would mean that \s counts as
3567    an explicit \r or \n reference, and that's not really what is wanted. Now
3568    we set the flag only if there is a literal "\r" or "\n" in the class. */
3569    
3570    #if 0
3571        /* Remember whether \r or \n are in this class */
3572    
3573        if (negate_class)
3574          {
3575          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3576          }
3577        else
3578          {
3579          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3580          }
3581    #endif
3582    
     while ((c = *(++ptr)) != ']' || inescq);  
3583    
3584      /* 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
3585      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
3586      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
3587      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3588      single-bytes only. This is an historical hangover. Maybe one day we can  
3589      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3590        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3591        operate on single-bytes only. This is an historical hangover. Maybe one day
3592        we can tidy these opcodes to handle multi-byte characters.
3593    
3594      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
3595      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 3599  for (;; ptr++)
3599      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3600    
3601  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3602      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3603            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3604  #else  #else
3605      if (class_charcount == 1)      if (class_charcount == 1)
3606  #endif  #endif
# Line 2209  for (;; ptr++) Line 3643  for (;; ptr++)
3643      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3644    
3645      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3646      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3647      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3648        the class, so any that were explicitly given as well can be ignored. If
3649        (when there are explicit characters > 255 that must be listed) there are no
3650        characters < 256, we can omit the bitmap in the actual compiled code. */
3651    
3652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3653      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3654        {        {
3655        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3656        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3657        code += LINK_SIZE;        code += LINK_SIZE;
3658        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3659    
3660        /* 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;
3661        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3662    
3663        if (class_charcount > 0)        if (class_charcount > 0)
3664          {          {
3665          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3666            memmove(code + 32, code, class_utf8data - code);
3667          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3668          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;  
3669          }          }
3670          else code = class_utf8data;
3671    
3672        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3673    
# Line 2246  for (;; ptr++) Line 3676  for (;; ptr++)
3676        }        }
3677  #endif  #endif
3678    
3679      /* 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
3680      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
3681      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
3682      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3683    
3684        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3685      if (negate_class)      if (negate_class)
3686        {        {
3687        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3688        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3689        }        }
3690      else      else
3691        {        {
       *code++ = OP_CLASS;  
3692        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3693        }        }
3694      code += 32;      code += 32;
3695      break;      break;
3696    
3697    
3698        /* ===================================================================*/
3699      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3700      has been tested above. */      has been tested above. */
3701    
3702      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3703      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3704      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3705      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3706      goto REPEAT;      goto REPEAT;
3707    
3708      case '*':      case CHAR_ASTERISK:
3709      repeat_min = 0;      repeat_min = 0;
3710      repeat_max = -1;      repeat_max = -1;
3711      goto REPEAT;      goto REPEAT;
3712    
3713      case '+':      case CHAR_PLUS:
3714      repeat_min = 1;      repeat_min = 1;
3715      repeat_max = -1;      repeat_max = -1;
3716      goto REPEAT;      goto REPEAT;
3717    
3718      case '?':      case CHAR_QUESTION_MARK:
3719      repeat_min = 0;      repeat_min = 0;
3720      repeat_max = 1;      repeat_max = 1;
3721    
# Line 2318  for (;; ptr++) Line 3750  for (;; ptr++)
3750      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
3751      repeat type to the non-default. */      repeat type to the non-default. */
3752    
3753      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3754        {        {
3755        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3756        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3757        ptr++;        ptr++;
3758        }        }
3759      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3760        {        {
3761        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3762        ptr++;        ptr++;
3763        }        }
3764      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3765    
     /* 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;  
       }  
   
3766      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3767      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
3768      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 3796  for (;; ptr++)
3796          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3797          }          }
3798    
3799          /* If the repetition is unlimited, it pays to see if the next thing on
3800          the line is something that cannot possibly match this character. If so,
3801          automatically possessifying this item gains some performance in the case
3802          where the match fails. */
3803    
3804          if (!possessive_quantifier &&
3805              repeat_max < 0 &&
3806              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3807                options, cd))
3808            {
3809            repeat_type = 0;    /* Force greedy */
3810            possessive_quantifier = TRUE;
3811            }
3812    
3813        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3814        }        }
3815    
3816      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3817      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-
3818      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3819      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3820        currently used only for single-byte chars. */
3821    
3822      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3823        {        {
3824        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3825        c = previous[1];        c = previous[1];
3826          if (!possessive_quantifier &&
3827              repeat_max < 0 &&
3828              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3829            {
3830            repeat_type = 0;    /* Force greedy */
3831            possessive_quantifier = TRUE;
3832            }
3833        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3834        }        }
3835    
# Line 2403  for (;; ptr++) Line 3843  for (;; ptr++)
3843      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3844        {        {
3845        uschar *oldcode;        uschar *oldcode;
3846        int prop_type;        int prop_type, prop_value;
3847        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3848        c = *previous;        c = *previous;
3849    
3850          if (!possessive_quantifier &&
3851              repeat_max < 0 &&
3852              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3853            {
3854            repeat_type = 0;    /* Force greedy */
3855            possessive_quantifier = TRUE;
3856            }
3857    
3858        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3859        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3860          previous[1] : -1;          {
3861            prop_type = previous[1];
3862            prop_value = previous[2];
3863            }
3864          else prop_type = prop_value = -1;
3865    
3866        oldcode = code;        oldcode = code;
3867        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2419  for (;; ptr++) Line 3871  for (;; ptr++)
3871    
3872        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3873    
3874          /*--------------------------------------------------------------------*/
3875          /* This code is obsolete from release 8.00; the restriction was finally
3876          removed: */
3877    
3878        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3879        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3880    
3881        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3882          /*--------------------------------------------------------------------*/
3883    
3884        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3885    
# Line 2443  for (;; ptr++) Line 3900  for (;; ptr++)
3900          }          }
3901    
3902        /* 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
3903        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3904        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
3905        one less than the maximum. */        one less than the maximum. */
3906    
# Line 2470  for (;; ptr++) Line 3927  for (;; ptr++)
3927    
3928          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3929          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
3930          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3931          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3932          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3933    
# Line 2486  for (;; ptr++) Line 3943  for (;; ptr++)
3943  #endif  #endif
3944              {              {
3945              *code++ = c;              *code++ = c;
3946              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3947                  {
3948                  *code++ = prop_type;
3949                  *code++ = prop_value;
3950                  }
3951              }              }
3952            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3953            }            }
3954    
3955          /* 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
3956          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3957            UPTO is just for 1 instance, we can use QUERY instead. */
3958    
3959          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3960            {            {
# Line 2505  for (;; ptr++) Line 3967  for (;; ptr++)
3967            else            else
3968  #endif  #endif
3969            *code++ = c;            *code++ = c;
3970            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3971                {
3972                *code++ = prop_type;
3973                *code++ = prop_value;
3974                }
3975            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3976            *code++ = OP_UPTO + repeat_type;  
3977            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3978                {
3979                *code++ = OP_QUERY + repeat_type;
3980                }
3981              else
3982                {
3983                *code++ = OP_UPTO + repeat_type;
3984                PUT2INC(code, 0, repeat_max);
3985                }
3986            }            }
3987          }          }
3988    
# Line 2524  for (;; ptr++) Line 3998  for (;; ptr++)
3998  #endif  #endif
3999        *code++ = c;        *code++ = c;
4000    
4001        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
4002        defines the required property. */        define the required property. */
4003    
4004  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4005        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
4006            {
4007            *code++ = prop_type;
4008            *code++ = prop_value;
4009            }
4010  #endif  #endif
4011        }        }
4012    
# Line 2548  for (;; ptr++) Line 4026  for (;; ptr++)
4026          goto END_REPEAT;          goto END_REPEAT;
4027          }          }
4028    
4029          /*--------------------------------------------------------------------*/
4030          /* This code is obsolete from release 8.00; the restriction was finally
4031          removed: */
4032    
4033        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4034        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4035    
4036        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4037          /*--------------------------------------------------------------------*/
4038    
4039        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4040          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 4054  for (;; ptr++)
4054      /* 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
4055      cases. */      cases. */
4056    
4057      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4058               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4059        {        {
4060        register int i;        register int i;
4061        int ketoffset = 0;        int ketoffset = 0;
4062        int len = code - previous;        int len = code - previous;
4063        uschar *bralink = NULL;        uschar *bralink = NULL;
4064    
4065          /* Repeating a DEFINE group is pointless */
4066    
4067          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4068            {
4069            *errorcodeptr = ERR55;
4070            goto FAILED;
4071            }
4072    
4073        /* 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
4074        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
4075        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 4092  for (;; ptr++)
4092    
4093        if (repeat_min == 0)        if (repeat_min == 0)
4094          {          {
4095          /* 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
4096          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4097    
4098          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4099          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4100          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4101          any internal group, because the offset is from the start of the whole          **   goto END_REPEAT;
4102          regex. Temporarily terminate the pattern while doing this. */          **   }
4103    
4104            However, that fails when a group is referenced as a subroutine from
4105            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4106            so that it is skipped on execution. As we don't have a list of which
4107            groups are referenced, we cannot do this selectively.
4108    
4109            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4110            and do no more at this point. However, we do need to adjust any
4111            OP_RECURSE calls inside the group that refer to the group itself or any
4112            internal or forward referenced group, because the offset is from the
4113            start of the whole regex. Temporarily terminate the pattern while doing
4114            this. */
4115    
4116          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4117            {            {
4118            *code = OP_END;            *code = OP_END;
4119            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4120            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4121            code++;            code++;
4122              if (repeat_max == 0)
4123                {
4124                *previous++ = OP_SKIPZERO;
4125                goto END_REPEAT;
4126                }
4127            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4128            }            }
4129    
# Line 2637  for (;; ptr++) Line 4139  for (;; ptr++)
4139            {            {
4140            int offset;            int offset;
4141            *code = OP_END;            *code = OP_END;
4142            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
4143            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
4144            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
4145            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 4159  for (;; ptr++)
4159        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
4160        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
4161        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
4162        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
4163          forward reference subroutine calls in the group, there will be entries on
4164          the workspace list; replicate these with an appropriate increment. */
4165    
4166        else        else
4167          {          {
4168          if (repeat_min > 1)          if (repeat_min > 1)
4169            {            {
4170            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
4171            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
4172              potential integer overflow. */
4173    
4174              if (lengthptr != NULL)
4175                {
4176                int delta = (repeat_min - 1)*length_prevgroup;
4177                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4178                                                                (double)INT_MAX ||
4179                    OFLOW_MAX - *lengthptr < delta)
4180                  {
4181                  *errorcodeptr = ERR20;
4182                  goto FAILED;
4183                  }
4184                *lengthptr += delta;
4185                }
4186    
4187              /* This is compiling for real */
4188    
4189              else
4190              {              {
4191              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4192              code += len;              for (i = 1; i < repeat_min; i++)
4193                  {
4194                  uschar *hc;
4195                  uschar *this_hwm = cd->hwm;
4196                  memcpy(code, previous, len);
4197                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4198                    {
4199                    PUT(cd->hwm, 0, GET(hc, 0) + len);
4200                    cd->hwm += LINK_SIZE;
4201                    }
4202                  save_hwm = this_hwm;
4203                  code += len;
4204                  }
4205              }              }
4206            }            }
4207    
4208          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
4209          }          }
4210    
# Line 2677  for (;; ptr++) Line 4212  for (;; ptr++)
4212        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
4213        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,
4214        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
4215        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
4216          replicate entries on the forward reference list. */
4217    
4218        if (repeat_max >= 0)        if (repeat_max >= 0)
4219          {          {
4220          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
4221            just adjust the length as if we had. For each repetition we must add 1
4222            to the length for BRAZERO and for all but the last repetition we must
4223            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4224            paranoid checks to avoid integer overflow. */
4225    
4226            if (lengthptr != NULL && repeat_max > 0)
4227              {
4228              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4229                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4230              if ((double)repeat_max *
4231                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4232                      > (double)INT_MAX ||
4233                  OFLOW_MAX - *lengthptr < delta)
4234                {
4235                *errorcodeptr = ERR20;
4236                goto FAILED;
4237                }
4238              *lengthptr += delta;
4239              }
4240    
4241            /* This is compiling for real */
4242    
4243            else for (i = repeat_max - 1; i >= 0; i--)
4244            {            {
4245              uschar *hc;
4246              uschar *this_hwm = cd->hwm;
4247    
4248            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
4249    
4250            /* 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 4260  for (;; ptr++)
4260              }              }
4261    
4262            memcpy(code, previous, len);            memcpy(code, previous, len);
4263              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4264                {
4265                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4266                cd->hwm += LINK_SIZE;
4267                }
4268              save_hwm = this_hwm;
4269            code += len;            code += len;
4270            }            }
4271    
# Line 2720  for (;; ptr++) Line 4288  for (;; ptr++)
4288        /* 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
4289        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
4290        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
4291        correct offset was computed above. */        correct offset was computed above.
4292    
4293          Then, when we are doing the actual compile phase, check to see whether
4294          this group is a non-atomic one that could match an empty string. If so,
4295          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4296          that runtime checking can be done. [This check is also applied to
4297          atomic groups at runtime, but in a different way.] */
4298    
4299        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
4300            {
4301            uschar *ketcode = code - ketoffset;
4302            uschar *bracode = ketcode - GET(ketcode, 1);
4303            *ketcode = OP_KETRMAX + repeat_type;
4304            if (lengthptr == NULL && *bracode != OP_ONCE)
4305              {
4306              uschar *scode = bracode;
4307              do
4308                {
4309                if (could_be_empty_branch(scode, ketcode, utf8))
4310                  {
4311                  *bracode += OP_SBRA - OP_BRA;
4312                  break;
4313                  }
4314                scode += GET(scode, 1);
4315                }
4316              while (*scode == OP_ALT);
4317              }
4318            }
4319        }        }
4320    
4321        /* If previous is OP_FAIL, it was generated by an empty class [] in
4322        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4323        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4324        error above. We can just ignore the repeat in JS case. */
4325    
4326        else if (*previous == OP_FAIL) goto END_REPEAT;
4327    
4328      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4329    
4330      else      else
# Line 2733  for (;; ptr++) Line 4333  for (;; ptr++)
4333        goto FAILED;        goto FAILED;
4334        }        }
4335    
4336      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4337      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4338      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4339      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.
4340      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4341        but the special opcodes can optimize it a bit. The repeated item starts at
4342        tempcode, not at previous, which might be the first part of a string whose
4343        (former) last char we repeated.
4344    
4345        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4346        an 'upto' may follow. We skip over an 'exact' item, and then test the
4347        length of what remains before proceeding. */
4348    
4349      if (possessive_quantifier)      if (possessive_quantifier)
4350        {        {
4351        int len = code - tempcode;        int len;
4352        memmove(tempcode + 1+LINK_SIZE, tempcode, len);  
4353        code += 1 + LINK_SIZE;        if (*tempcode == OP_TYPEEXACT)
4354        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4355        tempcode[0] = OP_ONCE;            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4356        *code++ = OP_KET;  
4357        PUTINC(code, 0, len);        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4358        PUT(tempcode, 1, len);          {
4359            tempcode += _pcre_OP_lengths[*tempcode];
4360    #ifdef SUPPORT_UTF8
4361            if (utf8 && tempcode[-1] >= 0xc0)
4362              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4363    #endif
4364            }
4365    
4366          len = code - tempcode;
4367          if (len > 0) switch (*tempcode)
4368            {
4369            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4370            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4371            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4372            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4373    
4374            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4375            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4376            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4377            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4378    
4379            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4380            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4381            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4382            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4383    
4384            default:
4385            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4386            code += 1 + LINK_SIZE;
4387            len += 1 + LINK_SIZE;
4388