/[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 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 411 by ph10, Fri Apr 10 15:40:21 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          }
1104        }
1105    }    }
1106  return p;  
1107    /* Past any initial parenthesis handling, scan for parentheses or vertical
1108    bars. */
1109    
1110    for (; *ptr != 0; ptr++)
1111      {
1112      /* Skip over backslashed characters and also entire \Q...\E */
1113    
1114      if (*ptr == CHAR_BACKSLASH)
1115        {
1116        if (*(++ptr) == 0) goto FAIL_EXIT;
1117        if (*ptr == CHAR_Q) for (;;)
1118          {
1119          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1120          if (*ptr == 0) goto FAIL_EXIT;
1121          if (*(++ptr) == CHAR_E) break;
1122          }
1123        continue;
1124        }
1125    
1126      /* Skip over character classes; this logic must be similar to the way they
1127      are handled for real. If the first character is '^', skip it. Also, if the
1128      first few characters (either before or after ^) are \Q\E or \E we skip them
1129      too. This makes for compatibility with Perl. Note the use of STR macros to
1130      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1131    
1132      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1133        {
1134        BOOL negate_class = FALSE;
1135        for (;;)
1136          {
1137          int c = *(++ptr);
1138          if (c == CHAR_BACKSLASH)
1139            {
1140            if (ptr[1] == CHAR_E)
1141              ptr++;
1142            else if (strncmp((const char *)ptr+1,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 3;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
1149            negate_class = TRUE;
1150          else break;
1151          }
1152    
1153        /* If the next character is ']', it is a data character that must be
1154        skipped, except in JavaScript compatibility mode. */
1155    
1156        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1157            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1158          ptr++;
1159    
1160        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1161          {
1162          if (*ptr == 0) return -1;
1163          if (*ptr == CHAR_BACKSLASH)
1164            {
1165            if (*(++ptr) == 0) goto FAIL_EXIT;
1166            if (*ptr == CHAR_Q) for (;;)
1167              {
1168              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1169              if (*ptr == 0) goto FAIL_EXIT;
1170              if (*(++ptr) == CHAR_E) break;
1171              }
1172            continue;
1173            }
1174          }
1175        continue;
1176        }
1177    
1178      /* Skip comments in /x mode */
1179    
1180      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1181        {
1182        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1183        if (*ptr == 0) goto FAIL_EXIT;
1184        continue;
1185        }
1186    
1187      /* Check for the special metacharacters */
1188    
1189      if (*ptr == CHAR_LEFT_PARENTHESIS)
1190        {
1191        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1192        if (rc > 0) return rc;
1193        if (*ptr == 0) goto FAIL_EXIT;
1194        }
1195    
1196      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1197        {
1198        if (dup_parens && *count < hwm_count) *count = hwm_count;
1199        *ptrptr = ptr;
1200        return -1;
1201        }
1202    
1203      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1204        {
1205        if (*count > hwm_count) hwm_count = *count;
1206        *count = start_count;
1207        }
1208      }
1209    
1210    FAIL_EXIT:
1211    *ptrptr = ptr;
1212    return -1;
1213    }
1214    
1215    
1216    
1217    
1218    /*************************************************
1219    *       Find forward referenced subpattern       *
1220    *************************************************/
1221    
1222    /* This function scans along a pattern's text looking for capturing
1223    subpatterns, and counting them. If it finds a named pattern that matches the
1224    name it is given, it returns its number. Alternatively, if the name is NULL, it
1225    returns when it reaches a given numbered subpattern. This is used for forward
1226    references to subpatterns. We used to be able to start this scan from the
1227    current compiling point, using the current count value from cd->bracount, and
1228    do it all in a single loop, but the addition of the possibility of duplicate
1229    subpattern numbers means that we have to scan from the very start, in order to
1230    take account of such duplicates, and to use a recursive function to keep track
1231    of the different types of group.
1232    
1233    Arguments:
1234      cd           compile background data
1235      name         name to seek, or NULL if seeking a numbered subpattern
1236      lorn         name length, or subpattern number if name is NULL
1237      xmode        TRUE if we are in /x mode
1238    
1239    Returns:       the number of the found subpattern, or -1 if not found
1240    */
1241    
1242    static int
1243    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1244    {
1245    uschar *ptr = (uschar *)cd->start_pattern;
1246    int count = 0;
1247    int rc;
1248    
1249    /* If the pattern does not start with an opening parenthesis, the first call
1250    to find_parens_sub() will scan right to the end (if necessary). However, if it
1251    does start with a parenthesis, find_parens_sub() will return when it hits the
1252    matching closing parens. That is why we have to have a loop. */
1253    
1254    for (;;)
1255      {
1256      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1257      if (rc > 0 || *ptr++ == 0) break;
1258      }
1259    
1260    return rc;
1261  }  }
1262    
1263    
1264    
1265    
1266  /*************************************************  /*************************************************
1267  *      Find first significant op code            *  *      Find first significant op code            *
1268  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1312  for (;;)
1312    
1313      case OP_CALLOUT:      case OP_CALLOUT:
1314      case OP_CREF:      case OP_CREF:
1315      case OP_BRANUMBER:      case OP_RREF:
1316        case OP_DEF:
1317      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1318      break;      break;
1319    
# Line 823  for (;;) Line 1358  for (;;)
1358    {    {
1359    int d;    int d;
1360    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1361    switch (op)    switch (op)
1362      {      {
1363        case OP_CBRA:
1364      case OP_BRA:      case OP_BRA:
1365      case OP_ONCE:      case OP_ONCE:
1366      case OP_COND:      case OP_COND:
1367      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1368      if (d < 0) return d;      if (d < 0) return d;
1369      branchlength += d;      branchlength += d;
1370      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1399  for (;;)
1399      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1400    
1401      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1402      case OP_CREF:      case OP_CREF:
1403        case OP_RREF:
1404        case OP_DEF:
1405      case OP_OPT:      case OP_OPT:
1406      case OP_CALLOUT:      case OP_CALLOUT:
1407      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1419  for (;;)
1419    
1420      case OP_CHAR:      case OP_CHAR:
1421      case OP_CHARNC:      case OP_CHARNC:
1422        case OP_NOT:
1423      branchlength++;      branchlength++;
1424      cc += 2;      cc += 2;
1425  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 910  for (;;) Line 1446  for (;;)
1446    
1447      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1448      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1449        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1450      cc += 4;      cc += 4;
1451      break;      break;
1452    
# Line 917  for (;;) Line 1454  for (;;)
1454    
1455      case OP_PROP:      case OP_PROP:
1456      case OP_NOTPROP:      case OP_NOTPROP:
1457      cc++;      cc += 2;
1458      /* Fall through */      /* Fall through */
1459    
1460      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1464  for (;;)
1464      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1465      case OP_WORDCHAR:      case OP_WORDCHAR:
1466      case OP_ANY:      case OP_ANY:
1467        case OP_ALLANY:
1468      branchlength++;      branchlength++;
1469      cc++;      cc++;
1470      break;      break;
# Line 998  Returns: pointer to the opcode for Line 1536  Returns: pointer to the opcode for
1536  static const uschar *  static const uschar *
1537  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1538  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1539  for (;;)  for (;;)
1540    {    {
1541    register int c = *code;    register int c = *code;
1542    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1543    else if (c > OP_BRA)  
1544      /* XCLASS is used for classes that cannot be represented just by a bit
1545      map. This includes negated single high-valued characters. The length in
1546      the table is zero; the actual length is stored in the compiled code. */
1547    
1548      if (c == OP_XCLASS) code += GET(code, 1);
1549    
1550      /* Handle capturing bracket */
1551    
1552      else if (c == OP_CBRA)
1553      {      {
1554      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1555      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1556      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1557      }      }
1558    
1559      /* Otherwise, we can get the item's length from the table, except that for
1560      repeated character types, we have to test for \p and \P, which have an extra
1561      two bytes of parameters. */
1562    
1563    else    else
1564      {      {
1565      code += _pcre_OP_lengths[c];      switch(c)
1566          {
1567          case OP_TYPESTAR:
1568          case OP_TYPEMINSTAR:
1569          case OP_TYPEPLUS:
1570          case OP_TYPEMINPLUS:
1571          case OP_TYPEQUERY:
1572          case OP_TYPEMINQUERY:
1573          case OP_TYPEPOSSTAR:
1574          case OP_TYPEPOSPLUS:
1575          case OP_TYPEPOSQUERY:
1576          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1577          break;
1578    
1579  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1580          case OP_TYPEMINUPTO:
1581          case OP_TYPEEXACT:
1582          case OP_TYPEPOSUPTO:
1583          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1584          break;
1585          }
1586    
1587      /* 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. */  
1588    
1589        code += _pcre_OP_lengths[c];
1590    
1591      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1592      a multi-byte character. The length in the table is a minimum, so we have to
1593      arrange to skip the extra bytes. */
1594    
1595    #ifdef SUPPORT_UTF8
1596      if (utf8) switch(c)      if (utf8) switch(c)
1597        {        {
1598        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1600  for (;;)
1600        case OP_EXACT:        case OP_EXACT:
1601        case OP_UPTO:        case OP_UPTO:
1602        case OP_MINUPTO:        case OP_MINUPTO:
1603          case OP_POSUPTO:
1604        case OP_STAR:        case OP_STAR:
1605        case OP_MINSTAR:        case OP_MINSTAR:
1606          case OP_POSSTAR:
1607        case OP_PLUS:        case OP_PLUS:
1608        case OP_MINPLUS:        case OP_MINPLUS:
1609          case OP_POSPLUS:
1610        case OP_QUERY:        case OP_QUERY:
1611        case OP_MINQUERY:        case OP_MINQUERY:
1612        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1613        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;  
1614        break;        break;
1615        }        }
1616    #else
1617        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1618  #endif  #endif
1619      }      }
1620    }    }
# Line 1072  Returns: pointer to the opcode for Line 1639  Returns: pointer to the opcode for
1639  static const uschar *  static const uschar *
1640  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1641  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1642  for (;;)  for (;;)
1643    {    {
1644    register int c = *code;    register int c = *code;
1645    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1646    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1647    else if (c > OP_BRA)  
1648      {    /* XCLASS is used for classes that cannot be represented just by a bit
1649      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1650      }    the table is zero; the actual length is stored in the compiled code. */
1651    
1652      if (c == OP_XCLASS) code += GET(code, 1);
1653    
1654      /* Otherwise, we can get the item's length from the table, except that for
1655      repeated character types, we have to test for \p and \P, which have an extra
1656      two bytes of parameters. */
1657    
1658    else    else
1659      {      {
1660      code += _pcre_OP_lengths[c];      switch(c)
1661          {
1662          case OP_TYPESTAR:
1663          case OP_TYPEMINSTAR:
1664          case OP_TYPEPLUS:
1665          case OP_TYPEMINPLUS:
1666          case OP_TYPEQUERY:
1667          case OP_TYPEMINQUERY:
1668          case OP_TYPEPOSSTAR:
1669          case OP_TYPEPOSPLUS:
1670          case OP_TYPEPOSQUERY:
1671          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1672          break;
1673    
1674  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1675          case OP_TYPEUPTO:
1676          case OP_TYPEMINUPTO:
1677          case OP_TYPEEXACT:
1678          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1679          break;
1680          }
1681    
1682        /* Add in the fixed length from the table */
1683    
1684        code += _pcre_OP_lengths[c];
1685    
1686      /* 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
1687      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
1688      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. */  
1689    
1690    #ifdef SUPPORT_UTF8
1691      if (utf8) switch(c)      if (utf8) switch(c)
1692        {        {
1693        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1695  for (;;)
1695        case OP_EXACT:        case OP_EXACT:
1696        case OP_UPTO:        case OP_UPTO:
1697        case OP_MINUPTO:        case OP_MINUPTO:
1698          case OP_POSUPTO:
1699        case OP_STAR:        case OP_STAR:
1700        case OP_MINSTAR:        case OP_MINSTAR:
1701          case OP_POSSTAR:
1702        case OP_PLUS:        case OP_PLUS:
1703        case OP_MINPLUS:        case OP_MINPLUS:
1704          case OP_POSPLUS:
1705        case OP_QUERY:        case OP_QUERY:
1706        case OP_MINQUERY:        case OP_MINQUERY:
1707        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1708        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;  
1709        break;        break;
1710        }        }
1711    #else
1712        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1713  #endif  #endif
1714      }      }
1715    }    }
# Line 1132  for (;;) Line 1722  for (;;)
1722  *************************************************/  *************************************************/
1723    
1724  /* 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
1725  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()
1726  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
1727  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
1728  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1729    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1730    bracket whose current branch will already have been scanned.
1731    
1732  Arguments:  Arguments:
1733    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1741  static BOOL
1741  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1742  {  {
1743  register int c;  register int c;
1744  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);
1745       code < endcode;       code < endcode;
1746       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1747    {    {
# Line 1157  for (code = first_significant_code(code Line 1749  for (code = first_significant_code(code
1749    
1750    c = *code;    c = *code;
1751    
1752    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1753      first_significant_code() with a TRUE final argument. */
1754    
1755      if (c == OP_ASSERT)
1756        {
1757        do code += GET(code, 1); while (*code == OP_ALT);
1758        c = *code;
1759        continue;
1760        }
1761    
1762      /* Groups with zero repeats can of course be empty; skip them. */
1763    
1764      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1765        {
1766        code += _pcre_OP_lengths[c];
1767        do code += GET(code, 1); while (*code == OP_ALT);
1768        c = *code;
1769        continue;
1770        }
1771    
1772      /* For other groups, scan the branches. */
1773    
1774      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1775      {      {
1776      BOOL empty_branch;      BOOL empty_branch;
1777      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1778    
1779      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1780        empty branch, so just skip over the conditional, because it could be empty.
1781        Otherwise, scan the individual branches of the group. */
1782    
1783      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;  
1784        code += GET(code, 1);        code += GET(code, 1);
1785        else
1786          {
1787          empty_branch = FALSE;
1788          do
1789            {
1790            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1791              empty_branch = TRUE;
1792            code += GET(code, 1);
1793            }
1794          while (*code == OP_ALT);
1795          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1796        }        }
1797      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1798      c = *code;      c = *code;
1799        continue;
1800      }      }
1801    
1802    else switch (c)    /* Handle the other opcodes */
1803    
1804      switch (c)
1805      {      {
1806      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1807        cannot be represented just by a bit map. This includes negated single
1808        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1809        actual length is stored in the compiled code, so we must update "code"
1810        here. */
1811    
1812  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1813      case OP_XCLASS:      case OP_XCLASS:
1814      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1815      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1816  #endif  #endif
1817    
# Line 1227  for (code = first_significant_code(code Line 1855  for (code = first_significant_code(code
1855      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1856      case OP_WORDCHAR:      case OP_WORDCHAR:
1857      case OP_ANY:      case OP_ANY:
1858        case OP_ALLANY:
1859      case OP_ANYBYTE:      case OP_ANYBYTE:
1860      case OP_CHAR:      case OP_CHAR:
1861      case OP_CHARNC:      case OP_CHARNC:
1862      case OP_NOT:      case OP_NOT:
1863      case OP_PLUS:      case OP_PLUS:
1864      case OP_MINPLUS:      case OP_MINPLUS:
1865        case OP_POSPLUS:
1866      case OP_EXACT:      case OP_EXACT:
1867      case OP_NOTPLUS:      case OP_NOTPLUS:
1868      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1869        case OP_NOTPOSPLUS:
1870      case OP_NOTEXACT:      case OP_NOTEXACT:
1871      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1872      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1873        case OP_TYPEPOSPLUS:
1874      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1875      return FALSE;      return FALSE;
1876    
1877        /* These are going to continue, as they may be empty, but we have to
1878        fudge the length for the \p and \P cases. */
1879    
1880        case OP_TYPESTAR:
1881        case OP_TYPEMINSTAR:
1882        case OP_TYPEPOSSTAR:
1883        case OP_TYPEQUERY:
1884        case OP_TYPEMINQUERY:
1885        case OP_TYPEPOSQUERY:
1886        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1887        break;
1888    
1889        /* Same for these */
1890    
1891        case OP_TYPEUPTO:
1892        case OP_TYPEMINUPTO:
1893        case OP_TYPEPOSUPTO:
1894        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1895        break;
1896    
1897      /* End of branch */      /* End of branch */
1898    
1899      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1902  for (code = first_significant_code(code
1902      case OP_ALT:      case OP_ALT:
1903      return TRUE;      return TRUE;
1904    
1905      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1906      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1907    
1908  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1909      case OP_STAR:      case OP_STAR:
1910      case OP_MINSTAR:      case OP_MINSTAR:
1911        case OP_POSSTAR:
1912      case OP_QUERY:      case OP_QUERY:
1913      case OP_MINQUERY:      case OP_MINQUERY:
1914        case OP_POSQUERY:
1915      case OP_UPTO:      case OP_UPTO:
1916      case OP_MINUPTO:      case OP_MINUPTO:
1917        case OP_POSUPTO:
1918      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1919      break;      break;
1920  #endif  #endif
# Line 1308  return TRUE; Line 1963  return TRUE;
1963  *************************************************/  *************************************************/
1964    
1965  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1966  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
1967  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1968  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1969    
1970    Originally, this function only recognized a sequence of letters between the
1971    terminators, but it seems that Perl recognizes any sequence of characters,
1972    though of course unknown POSIX names are subsequently rejected. Perl gives an
1973    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1974    didn't consider this to be a POSIX class. Likewise for [:1234:].
1975    
1976    The problem in trying to be exactly like Perl is in the handling of escapes. We
1977    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1978    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1979    below handles the special case of \], but does not try to do any other escape
1980    processing. This makes it different from Perl for cases such as [:l\ower:]
1981    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1982    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1983    I think.
1984    
1985  Argument:  Arguments:
1986    ptr      pointer to the initial [    ptr      pointer to the initial [
1987    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1988    
1989  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1990  */  */
1991    
1992  static BOOL  static BOOL
1993  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1994  {  {
1995  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1996  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1997  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1998    {    {
1999    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2000    return TRUE;      {
2001        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2002        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2003          {
2004          *endptr = ptr;
2005          return TRUE;
2006          }
2007        }
2008    }    }
2009  return FALSE;  return FALSE;
2010  }  }
# Line 1355  Returns: a value representing the na Line 2029  Returns: a value representing the na
2029  static int  static int
2030  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2031  {  {
2032    const char *pn = posix_names;
2033  register int yield = 0;  register int yield = 0;
2034  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2035    {    {
2036    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2037      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2038      pn += posix_name_lengths[yield] + 1;
2039    yield++;    yield++;
2040    }    }
2041  return -1;  return -1;
# Line 1374  return -1; Line 2050  return -1;
2050  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2051  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2052  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
2053  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
2054  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
2055  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
2056  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
2057  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2058    OP_END.
2059    
2060    This function has been extended with the possibility of forward references for
2061    recursions and subroutine calls. It must also check the list of such references
2062    for the group we are dealing with. If it finds that one of the recursions in
2063    the current group is on this list, it adjusts the offset in the list, not the
2064    value in the reference (which is a group number).
2065    
2066  Arguments:  Arguments:
2067    group      points to the start of the group    group      points to the start of the group
2068    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2069    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2070    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2071      save_hwm   the hwm forward reference pointer at the start of the group
2072    
2073  Returns:     nothing  Returns:     nothing
2074  */  */
2075    
2076  static void  static void
2077  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2078      uschar *save_hwm)
2079  {  {
2080  uschar *ptr = group;  uschar *ptr = group;
2081    
2082  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2083    {    {
2084    int offset = GET(ptr, 1);    int offset;
2085    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2086    
2087      /* See if this recursion is on the forward reference list. If so, adjust the
2088      reference. */
2089    
2090      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2091        {
2092        offset = GET(hc, 0);
2093        if (cd->start_code + offset == ptr + 1)
2094          {
2095          PUT(hc, 0, offset + adjust);
2096          break;
2097          }
2098        }
2099    
2100      /* Otherwise, adjust the recursion offset if it's after the start of this
2101      group. */
2102    
2103      if (hc >= cd->hwm)
2104        {
2105        offset = GET(ptr, 1);
2106        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2107        }
2108    
2109    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2110    }    }
2111  }  }
# Line 1475  Yield: TRUE when range returned; Line 2184  Yield: TRUE when range returned;
2184  */  */
2185    
2186  static BOOL  static BOOL
2187  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2188      unsigned int *odptr)
2189  {  {
2190  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2191    
2192  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2193    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2194    
2195  if (c > d) return FALSE;  if (c > d) return FALSE;
2196    
# Line 1492  next = othercase + 1; Line 2199  next = othercase + 1;
2199    
2200  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2201    {    {
2202    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2203    next++;    next++;
2204    }    }
2205    
# Line 1506  return TRUE; Line 2211  return TRUE;
2211  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2212    
2213    
2214    
2215    /*************************************************
2216    *     Check if auto-possessifying is possible    *
2217    *************************************************/
2218    
2219    /* This function is called for unlimited repeats of certain items, to see
2220    whether the next thing could possibly match the repeated item. If not, it makes
2221    sense to automatically possessify the repeated item.
2222    
2223    Arguments:
2224      op_code       the repeated op code
2225      this          data for this item, depends on the opcode
2226      utf8          TRUE in UTF-8 mode
2227      utf8_char     used for utf8 character bytes, NULL if not relevant
2228      ptr           next character in pattern
2229      options       options bits
2230      cd            contains pointers to tables etc.
2231    
2232    Returns:        TRUE if possessifying is wanted
2233    */
2234    
2235    static BOOL
2236    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2237      const uschar *ptr, int options, compile_data *cd)
2238    {
2239    int next;
2240    
2241    /* Skip whitespace and comments in extended mode */
2242    
2243    if ((options & PCRE_EXTENDED) != 0)
2244      {
2245      for (;;)
2246        {
2247        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2248        if (*ptr == CHAR_NUMBER_SIGN)
2249          {
2250          while (*(++ptr) != 0)
2251            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2252          }
2253        else break;
2254        }
2255      }
2256    
2257    /* If the next item is one that we can handle, get its value. A non-negative
2258    value is a character, a negative value is an escape value. */
2259    
2260    if (*ptr == CHAR_BACKSLASH)
2261      {
2262      int temperrorcode = 0;
2263      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2264      if (temperrorcode != 0) return FALSE;
2265      ptr++;    /* Point after the escape sequence */
2266      }
2267    
2268    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2269      {
2270    #ifdef SUPPORT_UTF8
2271      if (utf8) { GETCHARINC(next, ptr); } else
2272    #endif
2273      next = *ptr++;
2274      }
2275    
2276    else return FALSE;
2277    
2278    /* Skip whitespace and comments in extended mode */
2279    
2280    if ((options & PCRE_EXTENDED) != 0)
2281      {
2282      for (;;)
2283        {
2284        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2285        if (*ptr == CHAR_NUMBER_SIGN)
2286          {
2287          while (*(++ptr) != 0)
2288            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2289          }
2290        else break;
2291        }
2292      }
2293    
2294    /* If the next thing is itself optional, we have to give up. */
2295    
2296    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2297      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2298        return FALSE;
2299    
2300    /* Now compare the next item with the previous opcode. If the previous is a
2301    positive single character match, "item" either contains the character or, if
2302    "item" is greater than 127 in utf8 mode, the character's bytes are in
2303    utf8_char. */
2304    
2305    
2306    /* Handle cases when the next item is a character. */
2307    
2308    if (next >= 0) switch(op_code)
2309      {
2310      case OP_CHAR:
2311    #ifdef SUPPORT_UTF8
2312      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2313    #else
2314      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2315    #endif
2316      return item != next;
2317    
2318      /* For CHARNC (caseless character) we must check the other case. If we have
2319      Unicode property support, we can use it to test the other case of
2320      high-valued characters. */
2321    
2322      case OP_CHARNC:
2323    #ifdef SUPPORT_UTF8
2324      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2325    #endif
2326      if (item == next) return FALSE;
2327    #ifdef SUPPORT_UTF8
2328      if (utf8)
2329        {
2330        unsigned int othercase;
2331        if (next < 128) othercase = cd->fcc[next]; else
2332    #ifdef SUPPORT_UCP
2333        othercase = UCD_OTHERCASE((unsigned int)next);
2334    #else
2335        othercase = NOTACHAR;
2336    #endif
2337        return (unsigned int)item != othercase;
2338        }
2339      else
2340    #endif  /* SUPPORT_UTF8 */
2341      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2342    
2343      /* For OP_NOT, "item" must be a single-byte character. */
2344    
2345      case OP_NOT:
2346      if (item == next) return TRUE;
2347      if ((options & PCRE_CASELESS) == 0) return FALSE;
2348    #ifdef SUPPORT_UTF8
2349      if (utf8)
2350        {
2351        unsigned int othercase;
2352        if (next < 128) othercase = cd->fcc[next]; else
2353    #ifdef SUPPORT_UCP
2354        othercase = UCD_OTHERCASE(next);
2355    #else
2356        othercase = NOTACHAR;
2357    #endif
2358        return (unsigned int)item == othercase;
2359        }
2360      else
2361    #endif  /* SUPPORT_UTF8 */
2362      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2363    
2364      case OP_DIGIT:
2365      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2366    
2367      case OP_NOT_DIGIT:
2368      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2369    
2370      case OP_WHITESPACE:
2371      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2372    
2373      case OP_NOT_WHITESPACE:
2374      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2375    
2376      case OP_WORDCHAR:
2377      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2378    
2379      case OP_NOT_WORDCHAR:
2380      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2381    
2382      case OP_HSPACE:
2383      case OP_NOT_HSPACE:
2384      switch(next)
2385        {
2386        case 0x09:
2387        case 0x20:
2388        case 0xa0:
2389        case 0x1680:
2390        case 0x180e:
2391        case 0x2000:
2392        case 0x2001:
2393        case 0x2002:
2394        case 0x2003:
2395        case 0x2004:
2396        case 0x2005:
2397        case 0x2006:
2398        case 0x2007:
2399        case 0x2008:
2400        case 0x2009:
2401        case 0x200A:
2402        case 0x202f:
2403        case 0x205f:
2404        case 0x3000:
2405        return op_code != OP_HSPACE;
2406        default:
2407        return op_code == OP_HSPACE;
2408        }
2409    
2410      case OP_VSPACE:
2411      case OP_NOT_VSPACE:
2412      switch(next)
2413        {
2414        case 0x0a:
2415        case 0x0b:
2416        case 0x0c:
2417        case 0x0d:
2418        case 0x85:
2419        case 0x2028:
2420        case 0x2029:
2421        return op_code != OP_VSPACE;
2422        default:
2423        return op_code == OP_VSPACE;
2424        }
2425    
2426      default:
2427      return FALSE;
2428      }
2429    
2430    
2431    /* Handle the case when the next item is \d, \s, etc. */
2432    
2433    switch(op_code)
2434      {
2435      case OP_CHAR:
2436      case OP_CHARNC:
2437    #ifdef SUPPORT_UTF8
2438      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2439    #endif
2440      switch(-next)
2441        {
2442        case ESC_d:
2443        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2444    
2445        case ESC_D:
2446        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2447    
2448        case ESC_s:
2449        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2450    
2451        case ESC_S:
2452        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2453    
2454        case ESC_w:
2455        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2456    
2457        case ESC_W:
2458        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2459    
2460        case ESC_h:
2461        case ESC_H:
2462        switch(item)
2463          {
2464          case 0x09:
2465          case 0x20:
2466          case 0xa0:
2467          case 0x1680:
2468          case 0x180e:
2469          case 0x2000:
2470          case 0x2001:
2471          case 0x2002:
2472          case 0x2003:
2473          case 0x2004:
2474          case 0x2005:
2475          case 0x2006:
2476          case 0x2007:
2477          case 0x2008:
2478          case 0x2009:
2479          case 0x200A:
2480          case 0x202f:
2481          case 0x205f:
2482          case 0x3000:
2483          return -next != ESC_h;
2484          default:
2485          return -next == ESC_h;
2486          }
2487    
2488        case ESC_v:
2489        case ESC_V:
2490        switch(item)
2491          {
2492          case 0x0a:
2493          case 0x0b:
2494          case 0x0c:
2495          case 0x0d:
2496          case 0x85:
2497          case 0x2028:
2498          case 0x2029:
2499          return -next != ESC_v;
2500          default:
2501          return -next == ESC_v;
2502          }
2503    
2504        default:
2505        return FALSE;
2506        }
2507    
2508      case OP_DIGIT:
2509      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2510             next == -ESC_h || next == -ESC_v;
2511    
2512      case OP_NOT_DIGIT:
2513      return next == -ESC_d;
2514    
2515      case OP_WHITESPACE:
2516      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2517    
2518      case OP_NOT_WHITESPACE:
2519      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2520    
2521      case OP_HSPACE:
2522      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2523    
2524      case OP_NOT_HSPACE:
2525      return next == -ESC_h;
2526    
2527      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2528      case OP_VSPACE:
2529      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2530    
2531      case OP_NOT_VSPACE:
2532      return next == -ESC_v;
2533    
2534      case OP_WORDCHAR:
2535      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2536    
2537      case OP_NOT_WORDCHAR:
2538      return next == -ESC_w || next == -ESC_d;
2539    
2540      default:
2541      return FALSE;
2542      }
2543    
2544    /* Control does not reach here */
2545    }
2546    
2547    
2548    
2549  /*************************************************  /*************************************************
2550  *           Compile one branch                   *  *           Compile one branch                   *
2551  *************************************************/  *************************************************/
2552    
2553  /* 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
2554  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
2555  bits.  bits. This function is used during the pre-compile phase when we are trying
2556    to find out the amount of memory needed, as well as during the real compile
2557    phase. The value of lengthptr distinguishes the two phases.
2558    
2559  Arguments:  Arguments:
2560    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2561    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2562    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2563    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2565  Arguments:
2565    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2566    bcptr          points to current branch chain    bcptr          points to current branch chain
2567    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2568      lengthptr      NULL during the real compile phase
2569                     points to length accumulator during pre-compile phase
2570    
2571  Returns:         TRUE on success  Returns:         TRUE on success
2572                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2573  */  */
2574    
2575  static BOOL  static BOOL
2576  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2577    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2578    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2579  {  {
2580  int repeat_type, op_type;  int repeat_type, op_type;
2581  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 2584  int greedy_default, greedy_non_default;
2584  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2585  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2586  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2587  int options = *optionsptr;  int options = *optionsptr;
2588  int after_manual_callout = 0;  int after_manual_callout = 0;
2589    int length_prevgroup = 0;
2590  register int c;  register int c;
2591  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2592    uschar *last_code = code;
2593    uschar *orig_code = code;
2594  uschar *tempcode;  uschar *tempcode;
2595  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2596  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2598  const uschar *ptr = *ptrptr;
2598  const uschar *tempptr;  const uschar *tempptr;
2599  uschar *previous = NULL;  uschar *previous = NULL;
2600  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2601    uschar *save_hwm = NULL;
2602  uschar classbits[32];  uschar classbits[32];
2603    
2604  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2605  BOOL class_utf8;  BOOL class_utf8;
2606  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2607  uschar *class_utf8data;  uschar *class_utf8data;
2608    uschar *class_utf8data_base;
2609  uschar utf8_char[6];  uschar utf8_char[6];
2610  #else  #else
2611  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2612    uschar *utf8_char = NULL;
2613    #endif
2614    
2615    #ifdef DEBUG
2616    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2617  #endif  #endif
2618    
2619  /* 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 2645  req_caseopt = ((options & PCRE_CASELESS)
2645  for (;; ptr++)  for (;; ptr++)
2646    {    {
2647    BOOL negate_class;    BOOL negate_class;
2648      BOOL should_flip_negation;
2649    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2650    BOOL is_quantifier;    BOOL is_quantifier;
2651      BOOL is_recurse;
2652      BOOL reset_bracount;
2653    int class_charcount;    int class_charcount;
2654    int class_lastchar;    int class_lastchar;
2655    int newoptions;    int newoptions;
2656    int recno;    int recno;
2657      int refsign;
2658    int skipbytes;    int skipbytes;
2659    int subreqbyte;    int subreqbyte;
2660    int subfirstbyte;    int subfirstbyte;
2661      int terminator;
2662    int mclength;    int mclength;
2663    uschar mcbuffer[8];    uschar mcbuffer[8];
2664    
2665    /* Next byte in the pattern */    /* Get next byte in the pattern */
2666    
2667    c = *ptr;    c = *ptr;
2668    
2669      /* If we are in the pre-compile phase, accumulate the length used for the
2670      previous cycle of this loop. */
2671    
2672      if (lengthptr != NULL)
2673        {
2674    #ifdef DEBUG
2675        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2676    #endif
2677        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2678          {
2679          *errorcodeptr = ERR52;
2680          goto FAILED;
2681          }
2682    
2683        /* There is at least one situation where code goes backwards: this is the
2684        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2685        the class is simply eliminated. However, it is created first, so we have to
2686        allow memory for it. Therefore, don't ever reduce the length at this point.
2687        */
2688    
2689        if (code < last_code) code = last_code;
2690    
2691        /* Paranoid check for integer overflow */
2692    
2693        if (OFLOW_MAX - *lengthptr < code - last_code)
2694          {
2695          *errorcodeptr = ERR20;
2696          goto FAILED;
2697          }
2698    
2699        *lengthptr += code - last_code;
2700        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2701    
2702        /* If "previous" is set and it is not at the start of the work space, move
2703        it back to there, in order to avoid filling up the work space. Otherwise,
2704        if "previous" is NULL, reset the current code pointer to the start. */
2705    
2706        if (previous != NULL)
2707          {
2708          if (previous > orig_code)
2709            {
2710            memmove(orig_code, previous, code - previous);
2711            code -= previous - orig_code;
2712            previous = orig_code;
2713            }
2714          }
2715        else code = orig_code;
2716    
2717        /* Remember where this code item starts so we can pick up the length
2718        next time round. */
2719    
2720        last_code = code;
2721        }
2722    
2723      /* In the real compile phase, just check the workspace used by the forward
2724      reference list. */
2725    
2726      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2727        {
2728        *errorcodeptr = ERR52;
2729        goto FAILED;
2730        }
2731    
2732    /* 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 */
2733    
2734    if (inescq && c != 0)    if (inescq && c != 0)
2735      {      {
2736      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2737        {        {
2738        inescq = FALSE;        inescq = FALSE;
2739        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 2743  for (;; ptr++)
2743        {        {
2744        if (previous_callout != NULL)        if (previous_callout != NULL)
2745          {          {
2746          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2747              complete_callout(previous_callout, ptr, cd);
2748          previous_callout = NULL;          previous_callout = NULL;
2749          }          }
2750        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2759  for (;; ptr++)
2759    /* 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
2760    a quantifier. */    a quantifier. */
2761    
2762    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2763      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2764        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2765    
2766    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2767         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2768      {      {
2769      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2770          complete_callout(previous_callout, ptr, cd);
2771      previous_callout = NULL;      previous_callout = NULL;
2772      }      }
2773    
# Line 1653  for (;; ptr++) Line 2776  for (;; ptr++)
2776    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2777      {      {
2778      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2779      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2780        {        {
2781        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2782        on the Macintosh. */          {
2783        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2784        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2785          if (*ptr != 0) continue;
2786    
2787          /* Else fall through to handle end of string */
2788          c = 0;
2789        }        }
2790      }      }
2791    
# Line 1672  for (;; ptr++) Line 2799  for (;; ptr++)
2799    
2800    switch(c)    switch(c)
2801      {      {
2802      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2803        case 0:                        /* The branch terminates at string end */
2804      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2805      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2806      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2807      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2808      *codeptr = code;      *codeptr = code;
2809      *ptrptr = ptr;      *ptrptr = ptr;
2810        if (lengthptr != NULL)
2811          {
2812          if (OFLOW_MAX - *lengthptr < code - last_code)
2813            {
2814            *errorcodeptr = ERR20;
2815            goto FAILED;
2816            }
2817          *lengthptr += code - last_code;   /* To include callout length */
2818          DPRINTF((">> end branch\n"));
2819          }
2820      return TRUE;      return TRUE;
2821    
2822    
2823        /* ===================================================================*/
2824      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2825      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2826    
2827      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2828      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2829        {        {
2830        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2833  for (;; ptr++)
2833      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2834      break;      break;
2835    
2836      case '$':      case CHAR_DOLLAR_SIGN:
2837      previous = NULL;      previous = NULL;
2838      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2839      break;      break;
# Line 1703  for (;; ptr++) Line 2841  for (;; ptr++)
2841      /* 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
2842      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2843    
2844      case '.':      case CHAR_DOT:
2845      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2846      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2847      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2848      previous = code;      previous = code;
2849      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2850      break;      break;
2851    
2852      /* Character classes. If the included characters are all < 255 in value, we  
2853      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2854      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
2855      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
2856      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2857        map as usual, then invert it at the end. However, we use a different opcode
2858        so that data characters > 255 can be handled correctly.
2859    
2860      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2861      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,
2862      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2863      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.
     */  
2864    
2865      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2866        default (Perl) mode, it is treated as a data character. */
2867    
2868        case CHAR_RIGHT_SQUARE_BRACKET:
2869        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2870          {
2871          *errorcodeptr = ERR64;
2872          goto FAILED;
2873          }
2874        goto NORMAL_CHAR;
2875    
2876        case CHAR_LEFT_SQUARE_BRACKET:
2877      previous = code;      previous = code;
2878    
2879      /* 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
2880      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. */
2881    
2882      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2883          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2884            check_posix_syntax(ptr, &tempptr))
2885        {        {
2886        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2887        goto FAILED;        goto FAILED;
2888        }        }
2889    
2890      /* 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,
2891        if the first few characters (either before or after ^) are \Q\E or \E we
2892        skip them too. This makes for compatibility with Perl. */
2893    
2894      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2895        for (;;)
2896        {        {
       negate_class = TRUE;  
2897        c = *(++ptr);        c = *(++ptr);
2898        }        if (c == CHAR_BACKSLASH)
2899      else          {
2900        {          if (ptr[1] == CHAR_E)
2901        negate_class = FALSE;            ptr++;
2902        }          else if (strncmp((const char *)ptr+1,
2903                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2904              ptr += 3;
2905            else
2906              break;
2907            }
2908          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2909            negate_class = TRUE;
2910          else break;
2911          }
2912    
2913        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2914        an initial ']' is taken as a data character -- the code below handles
2915        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2916        [^] must match any character, so generate OP_ALLANY. */
2917    
2918        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2919            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2920          {
2921          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2922          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2923          zerofirstbyte = firstbyte;
2924          break;
2925          }
2926    
2927        /* If a class contains a negative special such as \S, we need to flip the
2928        negation flag at the end, so that support for characters > 255 works
2929        correctly (they are all included in the class). */
2930    
2931        should_flip_negation = FALSE;
2932    
2933      /* 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
2934      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
2935      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2936    
2937      class_charcount = 0;      class_charcount = 0;
2938      class_lastchar = -1;      class_lastchar = -1;
2939    
2940        /* Initialize the 32-char bit map to all zeros. We build the map in a
2941        temporary bit of memory, in case the class contains only 1 character (less
2942        than 256), because in that case the compiled code doesn't use the bit map.
2943        */
2944    
2945        memset(classbits, 0, 32 * sizeof(uschar));
2946    
2947  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2948      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2949      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2950        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2951  #endif  #endif
2952    
     /* 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));  
   
2953      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2954      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
2955      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. */  
2956    
2957      do      if (c != 0) do
2958        {        {
2959          const uschar *oldptr;
2960    
2961  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2962        if (utf8 && c > 127)        if (utf8 && c > 127)
2963          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2964          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2965          }          }
2966    
2967          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2968          data and reset the pointer. This is so that very large classes that
2969          contain a zillion UTF-8 characters no longer overwrite the work space
2970          (which is on the stack). */
2971    
2972          if (lengthptr != NULL)
2973            {
2974            *lengthptr += class_utf8data - class_utf8data_base;
2975            class_utf8data = class_utf8data_base;
2976            }
2977    
2978  #endif  #endif
2979    
2980        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2981    
2982        if (inescq)        if (inescq)
2983          {          {
2984          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2985            {            {
2986            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2987            ptr++;            ptr++;                            /* Skip the 'E' */
2988            continue;            continue;                         /* Carry on with next */
2989            }            }
2990          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2991          }          }
2992    
2993        /* 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 2996  for (;; ptr++)
2996        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2997        5.6 and 5.8 do. */        5.6 and 5.8 do. */
2998    
2999        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3000            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3001            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3002          {          {
3003          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3004          int posix_class, i;          int posix_class, taboffset, tabopt;
3005          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3006            uschar pbits[32];
3007    
3008          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3009            {            {
3010            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3011            goto FAILED;            goto FAILED;
3012            }            }
3013    
3014          ptr += 2;          ptr += 2;
3015          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3016            {            {
3017            local_negate = TRUE;            local_negate = TRUE;
3018              should_flip_negation = TRUE;  /* Note negative special */
3019            ptr++;            ptr++;
3020            }            }
3021    
# Line 1836  for (;; ptr++) Line 3033  for (;; ptr++)
3033          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3034            posix_class = 0;            posix_class = 0;
3035    
3036          /* 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
3037          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
3038          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
3039          white space chars afterwards. */          result into the bit map that is being built. */
3040    
3041          posix_class *= 3;          posix_class *= 3;
3042          for (i = 0; i < 3; i++)  
3043            /* Copy in the first table (always present) */
3044    
3045            memcpy(pbits, cbits + posix_class_maps[posix_class],
3046              32 * sizeof(uschar));
3047    
3048            /* If there is a second table, add or remove it as required. */
3049    
3050            taboffset = posix_class_maps[posix_class + 1];
3051            tabopt = posix_class_maps[posix_class + 2];
3052    
3053            if (taboffset >= 0)
3054            {            {
3055            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3056            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;  
             }  
3057            else            else
3058              {              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;  
             }  
3059            }            }
3060    
3061            /* Not see if we need to remove any special characters. An option
3062            value of 1 removes vertical space and 2 removes underscore. */
3063    
3064            if (tabopt < 0) tabopt = -tabopt;
3065            if (tabopt == 1) pbits[1] &= ~0x3c;
3066              else if (tabopt == 2) pbits[11] &= 0x7f;
3067    
3068            /* Add the POSIX table or its complement into the main table that is
3069            being built and we are done. */
3070    
3071            if (local_negate)
3072              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3073            else
3074              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3075    
3076          ptr = tempptr + 1;          ptr = tempptr + 1;
3077          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3078          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3079          }          }
3080    
3081        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3082        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
3083        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.
3084        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3085        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  
3086        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3087    
3088        if (c == '\\')        if (c == CHAR_BACKSLASH)
3089          {          {
3090          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3091            if (*errorcodeptr != 0) goto FAILED;
3092    
3093          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 */
3094          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 */
3095            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3096          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3097            {            {
3098            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3099              {              {
3100              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3101              }              }
3102            else inescq = TRUE;            else inescq = TRUE;
3103            continue;            continue;
3104            }            }
3105            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3106    
3107          if (c < 0)          if (c < 0)
3108            {            {
3109            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3110            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3111            switch (-c)  
3112              /* Save time by not doing this in the pre-compile phase. */
3113    
3114              if (lengthptr == NULL) switch (-c)
3115              {              {
3116              case ESC_d:              case ESC_d:
3117              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3118              continue;              continue;
3119    
3120              case ESC_D:              case ESC_D:
3121                should_flip_negation = TRUE;
3122              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3123              continue;              continue;
3124    
# Line 1910  for (;; ptr++) Line 3127  for (;; ptr++)
3127              continue;              continue;
3128    
3129              case ESC_W:              case ESC_W:
3130                should_flip_negation = TRUE;
3131              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3132              continue;              continue;
3133    
# Line 1919  for (;; ptr++) Line 3137  for (;; ptr++)
3137              continue;              continue;
3138    
3139              case ESC_S:              case ESC_S:
3140                should_flip_negation = TRUE;
3141              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3142              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3143              continue;              continue;
3144    
3145  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3146              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3147              case ESC_P:              }
3148    
3149              /* In the pre-compile phase, just do the recognition. */
3150    
3151              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3152                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3153    
3154              /* We need to deal with \H, \h, \V, and \v in both phases because
3155              they use extra memory. */
3156    
3157              if (-c == ESC_h)
3158                {
3159                SETBIT(classbits, 0x09); /* VT */
3160                SETBIT(classbits, 0x20); /* SPACE */
3161                SETBIT(classbits, 0xa0); /* NSBP */
3162    #ifdef SUPPORT_UTF8
3163                if (utf8)
3164                  {
3165                  class_utf8 = TRUE;
3166                  *class_utf8data++ = XCL_SINGLE;
3167                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3168                  *class_utf8data++ = XCL_SINGLE;
3169                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3170                  *class_utf8data++ = XCL_RANGE;
3171                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3172                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3173                  *class_utf8data++ = XCL_SINGLE;
3174                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3175                  *class_utf8data++ = XCL_SINGLE;
3176                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3177                  *class_utf8data++ = XCL_SINGLE;
3178                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3179                  }
3180    #endif
3181                continue;
3182                }
3183    
3184              if (-c == ESC_H)
3185                {
3186                for (c = 0; c < 32; c++)
3187                  {
3188                  int x = 0xff;
3189                  switch (c)
3190                    {
3191                    case 0x09/8: x ^= 1 << (0x09%8); break;
3192                    case 0x20/8: x ^= 1 << (0x20%8); break;
3193                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3194                    default: break;
3195                    }
3196                  classbits[c] |= x;
3197                  }
3198    
3199    #ifdef SUPPORT_UTF8
3200                if (utf8)
3201                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3202                class_utf8 = TRUE;                class_utf8 = TRUE;
3203                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3204                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3205                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3206                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3207                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3208                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3209                  *class_utf8data++ = XCL_RANGE;
3210                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3211                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3212                  *class_utf8data++ = XCL_RANGE;
3213                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3214                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3215                  *class_utf8data++ = XCL_RANGE;
3216                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3217                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3218                  *class_utf8data++ = XCL_RANGE;
3219                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3220                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3221                  *class_utf8data++ = XCL_RANGE;
3222                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3223                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3224                }                }
3225    #endif
3226              continue;              continue;
3227                }
3228    
3229              if (-c == ESC_v)
3230                {
3231                SETBIT(classbits, 0x0a); /* LF */
3232                SETBIT(classbits, 0x0b); /* VT */
3233                SETBIT(classbits, 0x0c); /* FF */
3234                SETBIT(classbits, 0x0d); /* CR */
3235                SETBIT(classbits, 0x85); /* NEL */
3236    #ifdef SUPPORT_UTF8
3237                if (utf8)
3238                  {
3239                  class_utf8 = TRUE;
3240                  *class_utf8data++ = XCL_RANGE;
3241                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3242                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3243                  }
3244  #endif  #endif
3245                continue;
3246                }
3247    
3248              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3249              strict mode. By default, for compatibility with Perl, they are              {
3250              treated as literals. */              for (c = 0; c < 32; c++)
3251                  {
3252                  int x = 0xff;
3253                  switch (c)
3254                    {
3255                    case 0x0a/8: x ^= 1 << (0x0a%8);
3256                                 x ^= 1 << (0x0b%8);
3257                                 x ^= 1 << (0x0c%8);
3258                                 x ^= 1 << (0x0d%8);
3259                                 break;
3260                    case 0x85/8: x ^= 1 << (0x85%8); break;
3261                    default: break;
3262                    }
3263                  classbits[c] |= x;
3264                  }
3265    
3266              default:  #ifdef SUPPORT_UTF8
3267              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3268                {                {
3269                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3270                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3271                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3272                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3273                  *class_utf8data++ = XCL_RANGE;
3274                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3275                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3276                }                }
3277              c = *ptr;              /* The final character */  #endif
3278              class_charcount -= 2;  /* Undo the default count from above */              continue;
3279                }
3280    
3281              /* We need to deal with \P and \p in both phases. */
3282    
3283    #ifdef SUPPORT_UCP
3284              if (-c == ESC_p || -c == ESC_P)
3285                {
3286                BOOL negated;
3287                int pdata;
3288                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3289                if (ptype < 0) goto FAILED;
3290                class_utf8 = TRUE;
3291                *class_utf8data++ = ((-c == ESC_p) != negated)?
3292                  XCL_PROP : XCL_NOTPROP;
3293                *class_utf8data++ = ptype;
3294                *class_utf8data++ = pdata;
3295                class_charcount -= 2;   /* Not a < 256 character */
3296                continue;
3297                }
3298    #endif
3299              /* Unrecognized escapes are faulted if PCRE is running in its
3300              strict mode. By default, for compatibility with Perl, they are
3301              treated as literals. */
3302    
3303              if ((options & PCRE_EXTRA) != 0)
3304                {
3305                *errorcodeptr = ERR7;
3306                goto FAILED;
3307              }              }
3308    
3309              class_charcount -= 2;  /* Undo the default count from above */
3310              c = *ptr;              /* Get the final character and fall through */
3311            }            }
3312    
3313          /* 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
3314          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3315    
3316          }   /* End of backslash handling */          }   /* End of backslash handling */
3317    
3318        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3319        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
3320        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3321          entirely. The code for handling \Q and \E is messy. */
3322    
3323          CHECK_RANGE:
3324          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3325            {
3326            inescq = FALSE;
3327            ptr += 2;
3328            }
3329    
3330          oldptr = ptr;
3331    
3332          /* Remember \r or \n */
3333    
3334          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3335    
3336          /* Check for range */
3337    
3338        if (ptr[1] == '-' && ptr[2] != ']')        if (!inescq && ptr[1] == CHAR_MINUS)
3339          {          {
3340          int d;          int d;
3341          ptr += 2;          ptr += 2;
3342            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3343    
3344            /* If we hit \Q (not followed by \E) at this point, go into escaped
3345            mode. */
3346    
3347            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3348              {
3349              ptr += 2;
3350              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3351                { ptr += 2; continue; }
3352              inescq = TRUE;
3353              break;
3354              }
3355    
3356            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3357              {
3358              ptr = oldptr;
3359              goto LONE_SINGLE_CHARACTER;
3360              }
3361    
3362  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3363          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3372  for (;; ptr++)
3372          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
3373          in such circumstances. */          in such circumstances. */
3374    
3375          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3376            {            {
3377            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3378            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3379    
3380            /* \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
3381            was literal */            special means the '-' was literal */
3382    
3383            if (d < 0)            if (d < 0)
3384              {              {
3385              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3386              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3387                else if (d == -ESC_R) d = CHAR_R; else
3388                {                {
3389                ptr = oldptr - 2;                ptr = oldptr;
3390                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3391                }                }
3392              }              }
3393            }            }
3394    
3395          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3396          the pre-pass. Optimize one-character ranges */          one-character ranges */
3397    
3398            if (d < c)
3399              {
3400              *errorcodeptr = ERR8;
3401              goto FAILED;
3402              }
3403    
3404          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3405    
3406            /* Remember \r or \n */
3407    
3408            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3409    
3410          /* 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
3411          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3412          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 3424  for (;; ptr++)
3424  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3425            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3426              {              {
3427              int occ, ocd;              unsigned int occ, ocd;
3428              int cc = c;              unsigned int cc = c;
3429              int origd = d;              unsigned int origd = d;
3430              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3431                {                {
3432                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3433                      ocd <= (unsigned int)d)
3434                    continue;                          /* Skip embedded ranges */
3435    
3436                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3437                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3438                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3439                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3440                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3441                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3442                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3443                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3444                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3445                  d = ocd;                  d = ocd;
3446                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3488  for (;; ptr++)
3488          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
3489          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3490    
3491          for (; c <= d; c++)          class_charcount += d - c + 1;
3492            class_lastchar = d;
3493    
3494            /* We can save a bit of time by skipping this in the pre-compile. */
3495    
3496            if (lengthptr == NULL) for (; c <= d; c++)
3497            {            {
3498            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3499            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3501  for (;; ptr++)
3501              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3502              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3503              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3504            }            }
3505    
3506          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3524  for (;; ptr++)
3524  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3525          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3526            {            {
3527            int chartype;            unsigned int othercase;
3528            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3529              {              {
3530              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3531              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3550  for (;; ptr++)
3550          }          }
3551        }        }
3552    
3553      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3554      loop. This "while" is the end of the "do" above. */  
3555        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3556    
3557        if (c == 0)                          /* Missing terminating ']' */
3558          {
3559          *errorcodeptr = ERR6;
3560          goto FAILED;
3561          }
3562    
3563    
3564    /* This code has been disabled because it would mean that \s counts as
3565    an explicit \r or \n reference, and that's not really what is wanted. Now
3566    we set the flag only if there is a literal "\r" or "\n" in the class. */
3567    
3568    #if 0
3569        /* Remember whether \r or \n are in this class */
3570    
3571        if (negate_class)
3572          {
3573          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3574          }
3575        else
3576          {
3577          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3578          }
3579    #endif
3580    
     while ((c = *(++ptr)) != ']' || inescq);  
3581    
3582      /* 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
3583      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
3584      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
3585      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3586      single-bytes only. This is an historical hangover. Maybe one day we can  
3587      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3588        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3589        operate on single-bytes only. This is an historical hangover. Maybe one day
3590        we can tidy these opcodes to handle multi-byte characters.
3591    
3592      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
3593      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 3597  for (;; ptr++)
3597      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3598    
3599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3600      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3601            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3602  #else  #else
3603      if (class_charcount == 1)      if (class_charcount == 1)
3604  #endif  #endif
# Line 2209  for (;; ptr++) Line 3641  for (;; ptr++)
3641      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3642    
3643      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3644      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3645      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3646        the class, so any that were explicitly given as well can be ignored. If
3647        (when there are explicit characters > 255 that must be listed) there are no
3648        characters < 256, we can omit the bitmap in the actual compiled code. */
3649    
3650  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3651      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3652        {        {
3653        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3654        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3655        code += LINK_SIZE;        code += LINK_SIZE;
3656        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3657    
3658        /* 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;
3659        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3660    
3661        if (class_charcount > 0)        if (class_charcount > 0)
3662          {          {
3663          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3664            memmove(code + 32, code, class_utf8data - code);
3665          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3666          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;  
3667          }          }
3668          else code = class_utf8data;
3669    
3670        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3671    
# Line 2246  for (;; ptr++) Line 3674  for (;; ptr++)
3674        }        }
3675  #endif  #endif
3676    
3677      /* 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
3678      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
3679      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
3680      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3681    
3682        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3683      if (negate_class)      if (negate_class)
3684        {        {
3685        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3686        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3687        }        }
3688      else      else
3689        {        {
       *code++ = OP_CLASS;  
3690        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3691        }        }
3692      code += 32;      code += 32;
3693      break;      break;
3694    
3695    
3696        /* ===================================================================*/
3697      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3698      has been tested above. */      has been tested above. */
3699    
3700      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3701      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3702      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3703      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3704      goto REPEAT;      goto REPEAT;
3705    
3706      case '*':      case CHAR_ASTERISK:
3707      repeat_min = 0;      repeat_min = 0;
3708      repeat_max = -1;      repeat_max = -1;
3709      goto REPEAT;      goto REPEAT;
3710    
3711      case '+':      case CHAR_PLUS:
3712      repeat_min = 1;      repeat_min = 1;
3713      repeat_max = -1;      repeat_max = -1;
3714      goto REPEAT;      goto REPEAT;
3715    
3716      case '?':      case CHAR_QUESTION_MARK:
3717      repeat_min = 0;      repeat_min = 0;
3718      repeat_max = 1;      repeat_max = 1;
3719    
# Line 2318  for (;; ptr++) Line 3748  for (;; ptr++)
3748      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
3749      repeat type to the non-default. */      repeat type to the non-default. */
3750    
3751      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3752        {        {
3753        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3754        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3755        ptr++;        ptr++;
3756        }        }
3757      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3758        {        {
3759        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3760        ptr++;        ptr++;
3761        }        }
3762      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3763    
     /* 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;  
       }  
   
3764      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3765      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
3766      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 3794  for (;; ptr++)
3794          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3795          }          }
3796    
3797          /* If the repetition is unlimited, it pays to see if the next thing on
3798          the line is something that cannot possibly match this character. If so,
3799          automatically possessifying this item gains some performance in the case
3800          where the match fails. */
3801    
3802          if (!possessive_quantifier &&
3803              repeat_max < 0 &&
3804              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3805                options, cd))
3806            {
3807            repeat_type = 0;    /* Force greedy */
3808            possessive_quantifier = TRUE;
3809            }
3810    
3811        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3812        }        }
3813    
3814      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3815      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-
3816      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3817      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3818        currently used only for single-byte chars. */
3819    
3820      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3821        {        {
3822        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3823        c = previous[1];        c = previous[1];
3824          if (!possessive_quantifier &&
3825              repeat_max < 0 &&
3826              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3827            {
3828            repeat_type = 0;    /* Force greedy */
3829            possessive_quantifier = TRUE;
3830            }
3831        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3832        }        }
3833    
# Line 2403  for (;; ptr++) Line 3841  for (;; ptr++)
3841      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3842        {        {
3843        uschar *oldcode;        uschar *oldcode;
3844        int prop_type;        int prop_type, prop_value;
3845        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3846        c = *previous;        c = *previous;
3847    
3848          if (!possessive_quantifier &&
3849              repeat_max < 0 &&
3850              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3851            {
3852            repeat_type = 0;    /* Force greedy */
3853            possessive_quantifier = TRUE;
3854            }
3855    
3856        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3857        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3858          previous[1] : -1;          {
3859            prop_type = previous[1];
3860            prop_value = previous[2];
3861            }
3862          else prop_type = prop_value = -1;
3863    
3864        oldcode = code;        oldcode = code;
3865        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3872  for (;; ptr++)
3872        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3873        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3874    
3875        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3876    
3877        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3878    
# Line 2443  for (;; ptr++) Line 3893  for (;; ptr++)
3893          }          }
3894    
3895        /* 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
3896        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3897        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
3898        one less than the maximum. */        one less than the maximum. */
3899    
# Line 2470  for (;; ptr++) Line 3920  for (;; ptr++)
3920    
3921          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3922          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
3923          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3924          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3925          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3926    
# Line 2486  for (;; ptr++) Line 3936  for (;; ptr++)
3936  #endif  #endif
3937              {              {
3938              *code++ = c;              *code++ = c;
3939              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3940                  {
3941                  *code++ = prop_type;
3942                  *code++ = prop_value;
3943                  }
3944              }              }
3945            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3946            }            }
3947    
3948          /* 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
3949          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3950            UPTO is just for 1 instance, we can use QUERY instead. */
3951    
3952          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3953            {            {
# Line 2505  for (;; ptr++) Line 3960  for (;; ptr++)
3960            else            else
3961  #endif  #endif
3962            *code++ = c;            *code++ = c;
3963            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3964                {
3965                *code++ = prop_type;
3966                *code++ = prop_value;
3967                }
3968            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3969            *code++ = OP_UPTO + repeat_type;  
3970            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3971                {
3972                *code++ = OP_QUERY + repeat_type;
3973                }
3974              else
3975                {
3976                *code++ = OP_UPTO + repeat_type;
3977                PUT2INC(code, 0, repeat_max);
3978                }
3979            }            }
3980          }          }
3981    
# Line 2524  for (;; ptr++) Line 3991  for (;; ptr++)
3991  #endif  #endif
3992        *code++ = c;        *code++ = c;
3993    
3994        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3995        defines the required property. */        define the required property. */
3996    
3997  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3998        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3999            {
4000            *code++ = prop_type;
4001            *code++ = prop_value;
4002            }
4003  #endif  #endif
4004        }        }
4005    
# Line 2551  for (;; ptr++) Line 4022  for (;; ptr++)
4022        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4023        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4024    
4025        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
4026    
4027        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4028          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 4042  for (;; ptr++)
4042      /* 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
4043      cases. */      cases. */
4044    
4045      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4046               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4047        {        {
4048        register int i;        register int i;
4049        int ketoffset = 0;        int ketoffset = 0;
4050        int len = code - previous;        int len = code - previous;
4051        uschar *bralink = NULL;        uschar *bralink = NULL;
4052    
4053          /* Repeating a DEFINE group is pointless */
4054    
4055          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4056            {
4057            *errorcodeptr = ERR55;
4058            goto FAILED;
4059            }
4060    
4061        /* 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
4062        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
4063        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 4080  for (;; ptr++)
4080    
4081        if (repeat_min == 0)        if (repeat_min == 0)
4082          {          {
4083          /* 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
4084          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4085    
4086          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4087          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4088          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4089          any internal group, because the offset is from the start of the whole          **   goto END_REPEAT;
4090          regex. Temporarily terminate the pattern while doing this. */          **   }
4091    
4092            However, that fails when a group is referenced as a subroutine from
4093            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4094            so that it is skipped on execution. As we don't have a list of which
4095            groups are referenced, we cannot do this selectively.
4096    
4097            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4098            and do no more at this point. However, we do need to adjust any
4099            OP_RECURSE calls inside the group that refer to the group itself or any
4100            internal or forward referenced group, because the offset is from the
4101            start of the whole regex. Temporarily terminate the pattern while doing
4102            this. */
4103    
4104          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4105            {            {
4106            *code = OP_END;            *code = OP_END;
4107            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4108            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4109            code++;            code++;
4110              if (repeat_max == 0)
4111                {
4112                *previous++ = OP_SKIPZERO;
4113                goto END_REPEAT;
4114                }
4115            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4116            }            }
4117    
# Line 2637  for (;; ptr++) Line 4127  for (;; ptr++)
4127            {            {
4128            int offset;            int offset;
4129            *code = OP_END;            *code = OP_END;
4130            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
4131            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
4132            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
4133            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 4147  for (;; ptr++)
4147        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
4148        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
4149        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
4150        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
4151          forward reference subroutine calls in the group, there will be entries on
4152          the workspace list; replicate these with an appropriate increment. */
4153    
4154