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

Diff of /code/trunk/pcre_compile.c

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

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 507 by ph10, Wed Mar 10 16:08:01 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         0,                       0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;
192      int   op;
193    } verbitem;
194    
195    static const char verbnames[] =
196      STRING_ACCEPT0
197      STRING_COMMIT0
198      STRING_F0
199      STRING_FAIL0
200      STRING_PRUNE0
201      STRING_SKIP0
202      STRING_THEN;
203    
204    static const verbitem verbs[] = {
205      { 6, OP_ACCEPT },
206      { 6, OP_COMMIT },
207      { 1, OP_FAIL },
208      { 4, OP_FAIL },
209      { 5, OP_PRUNE },
210      { 4, OP_SKIP  },
211      { 4, OP_THEN  }
212    };
213    
214    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
215    
216    
217    /* Tables of names of POSIX character classes and their lengths. The names are
218    now all in a single string, to reduce the number of relocations when a shared
219    library is dynamically loaded. The list of lengths is terminated by a zero
220    length entry. The first three must be alpha, lower, upper, as this is assumed
221    for handling case independence. */
222    
223    static const char posix_names[] =
224      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
225      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
226      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
227      STRING_word0  STRING_xdigit;
228    
229  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
230    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 };
231    
232  /* 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
233  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
234  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
235    characters are removed, and for [:alpha:] and [:alnum:] the underscore
236    character is removed. The triples in the table consist of the base map offset,
237    second map offset or -1 if no second map, and a non-negative value for map
238    addition or a negative value for map subtraction (if there are two maps). The
239    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
240    remove vertical space characters, 2 => remove underscore. */
241    
242  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
243    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
244    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
245    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
246    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
247    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
248    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
249    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
250    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
251    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
252    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
253    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
254    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
255    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
256    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
257  };  };
258    
259    
260  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
261  are passed to the outside world. */  #define XSTRING(s) STRING(s)
262    
263  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
264    "no error",  are passed to the outside world. Do not ever re-use any error number, because
265    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
266    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
267    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
268    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
269    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
270    simply count through to the one we want - this isn't a performance issue
271    because these strings are used only when there is a compilation error.
272    
273    Each substring ends with \0 to insert a null character. This includes the final
274    substring, so that the whole string ends with \0\0, which can be detected when
275    counting through. */
276    
277    static const char error_texts[] =
278      "no error\0"
279      "\\ at end of pattern\0"
280      "\\c at end of pattern\0"
281      "unrecognized character follows \\\0"
282      "numbers out of order in {} quantifier\0"
283    /* 5 */    /* 5 */
284    "number too big in {} quantifier",    "number too big in {} quantifier\0"
285    "missing terminating ] for character class",    "missing terminating ] for character class\0"
286    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
287    "range out of order in character class",    "range out of order in character class\0"
288    "nothing to repeat",    "nothing to repeat\0"
289    /* 10 */    /* 10 */
290    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
291    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
292    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
293    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
294    "missing )",    "missing )\0"
295    /* 15 */    /* 15 */
296    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
297    "erroffset passed as NULL",    "erroffset passed as NULL\0"
298    "unknown option bit(s) set",    "unknown option bit(s) set\0"
299    "missing ) after comment",    "missing ) after comment\0"
300    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
301    /* 20 */    /* 20 */
302    "regular expression too large",    "regular expression is too large\0"
303    "failed to get memory",    "failed to get memory\0"
304    "unmatched parentheses",    "unmatched parentheses\0"
305    "internal error: code overflow",    "internal error: code overflow\0"
306    "unrecognized character after (?<",    "unrecognized character after (?<\0"
307    /* 25 */    /* 25 */
308    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
309    "malformed number after (?(",    "malformed number or name after (?(\0"
310    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
311    "assertion expected after (?(",    "assertion expected after (?(\0"
312    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
313    /* 30 */    /* 30 */
314    "unknown POSIX class name",    "unknown POSIX class name\0"
315    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
316    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
317    "spare error",    "spare error\0"  /** DEAD **/
318    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
319    /* 35 */    /* 35 */
320    "invalid condition (?(0)",    "invalid condition (?(0)\0"
321    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
322    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
323    "number after (?C is > 255",    "number after (?C is > 255\0"
324    "closing ) for (?C expected",    "closing ) for (?C expected\0"
325    /* 40 */    /* 40 */
326    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
327    "unrecognized character after (?P",    "unrecognized character after (?P\0"
328    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
329    "two named groups have the same name",    "two named subpatterns have the same name\0"
330    "invalid UTF-8 string",    "invalid UTF-8 string\0"
331    /* 45 */    /* 45 */
332    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
333    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
334    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
335  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
336      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
337      /* 50 */
338      "repeated subpattern is too long\0"    /** DEAD **/
339      "octal value is greater than \\377 (not in UTF-8 mode)\0"
340      "internal error: overran compiling workspace\0"
341      "internal error: previously-checked referenced subpattern not found\0"
342      "DEFINE group contains more than one branch\0"
343      /* 55 */
344      "repeating a DEFINE group is not allowed\0"
345      "inconsistent NEWLINE options\0"
346      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
347      "a numbered reference must not be zero\0"
348      "(*VERB) with an argument is not supported\0"
349      /* 60 */
350      "(*VERB) not recognized\0"
351      "number is too big\0"
352      "subpattern name expected\0"
353      "digit expected after (?+\0"
354      "] is an invalid data character in JavaScript compatibility mode\0"
355      /* 65 */
356      "different names for subpatterns of the same number are not allowed\0";
357    
358  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
359  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 371  For convenience, we use the same bit def
371    
372  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
373    
374  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
375    
376    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
377    UTF-8 mode. */
378    
379  static const unsigned char digitab[] =  static const unsigned char digitab[] =
380    {    {
381    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 411  static const unsigned char digitab[] =
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
412    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
413    
414  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
415    
416    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
417    
418  static const unsigned char digitab[] =  static const unsigned char digitab[] =
419    {    {
420    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 428  static const unsigned char digitab[] =
428    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
429    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
430    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
431    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
432    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
433    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
434    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 462  static const unsigned char ebcdic_charta
462    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
463    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
464    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
465    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
467    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
468    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 489  static const unsigned char ebcdic_charta
489  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
490    
491  static BOOL  static BOOL
492    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
493      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
494    
495    
496    
497  /*************************************************  /*************************************************
498    *            Find an error text                  *
499    *************************************************/
500    
501    /* The error texts are now all in one long string, to save on relocations. As
502    some of the text is of unknown length, we can't use a table of offsets.
503    Instead, just count through the strings. This is not a performance issue
504    because it happens only when there has been a compilation error.
505    
506    Argument:   the error number
507    Returns:    pointer to the error string
508    */
509    
510    static const char *
511    find_error_text(int n)
512    {
513    const char *s = error_texts;
514    for (; n > 0; n--)
515      {
516      while (*s++ != 0) {};
517      if (*s == 0) return "Error text not found (please report)";
518      }
519    return s;
520    }
521    
522    
523    /*************************************************
524  *            Handle escapes                      *  *            Handle escapes                      *
525  *************************************************/  *************************************************/
526    
527  /* 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
528  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
529  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
530  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
531  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,
532    ptr is pointing at the \. On exit, it is on the final character of the escape
533    sequence.
534    
535  Arguments:  Arguments:
536    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 541  Arguments:
541    
542  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
543                   negative => a special escape sequence                   negative => a special escape sequence
544                   on error, errorptr is set                   on error, errorcodeptr is set
545  */  */
546    
547  static int  static int
548  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
549    int options, BOOL isclass)    int options, BOOL isclass)
550  {  {
551  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
552    const uschar *ptr = *ptrptr + 1;
553  int c, i;  int c, i;
554    
555    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
556    ptr--;                            /* Set pointer back to the last byte */
557    
558  /* 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. */
559    
 c = *(++ptr);  
560  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
561    
562  /* 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
563  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.
564  Otherwise further processing may be required. */  Otherwise further processing may be required. */
565    
566  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
567  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
568  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
569    
570  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
571  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
572  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
573  #endif  #endif
574    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 577  else if ((i = escapes[c - 0x48]) != 0)
577  else  else
578    {    {
579    const uschar *oldptr;    const uschar *oldptr;
580      BOOL braced, negated;
581    
582    switch (c)    switch (c)
583      {      {
584      /* 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
585      error. */      error. */
586    
587      case 'l':      case CHAR_l:
588      case 'L':      case CHAR_L:
589      case 'N':      case CHAR_N:
590      case 'u':      case CHAR_u:
591      case 'U':      case CHAR_U:
592      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
593      break;      break;
594    
595        /* \g must be followed by one of a number of specific things:
596    
597        (1) A number, either plain or braced. If positive, it is an absolute
598        backreference. If negative, it is a relative backreference. This is a Perl
599        5.10 feature.
600    
601        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
602        is part of Perl's movement towards a unified syntax for back references. As
603        this is synonymous with \k{name}, we fudge it up by pretending it really
604        was \k.
605    
606        (3) For Oniguruma compatibility we also support \g followed by a name or a
607        number either in angle brackets or in single quotes. However, these are
608        (possibly recursive) subroutine calls, _not_ backreferences. Just return
609        the -ESC_g code (cf \k). */
610    
611        case CHAR_g:
612        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
613          {
614          c = -ESC_g;
615          break;
616          }
617    
618        /* Handle the Perl-compatible cases */
619    
620        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
621          {
622          const uschar *p;
623          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
624            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
625          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
626            {
627            c = -ESC_k;
628            break;
629            }
630          braced = TRUE;
631          ptr++;
632          }
633        else braced = FALSE;
634    
635        if (ptr[1] == CHAR_MINUS)
636          {
637          negated = TRUE;
638          ptr++;
639          }
640        else negated = FALSE;
641    
642        c = 0;
643        while ((digitab[ptr[1]] & ctype_digit) != 0)
644          c = c * 10 + *(++ptr) - CHAR_0;
645    
646        if (c < 0)   /* Integer overflow */
647          {
648          *errorcodeptr = ERR61;
649          break;
650          }
651    
652        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
653          {
654          *errorcodeptr = ERR57;
655          break;
656          }
657    
658        if (c == 0)
659          {
660          *errorcodeptr = ERR58;
661          break;
662          }
663    
664        if (negated)
665          {
666          if (c > bracount)
667            {
668            *errorcodeptr = ERR15;
669            break;
670            }
671          c = bracount - (c - 1);
672          }
673    
674        c = -(ESC_REF + c);
675        break;
676    
677      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
678      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
679      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 686  else
686      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
687      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
688    
689      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:
690      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
691    
692      if (!isclass)      if (!isclass)
693        {        {
694        oldptr = ptr;        oldptr = ptr;
695        c -= '0';        c -= CHAR_0;
696        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
697          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
698          if (c < 0)    /* Integer overflow */
699            {
700            *errorcodeptr = ERR61;
701            break;
702            }
703        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
704          {          {
705          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 712  else
712      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.
713      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
714    
715      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
716        {        {
717        ptr--;        ptr--;
718        c = 0;        c = 0;
# Line 442  else Line 720  else
720        }        }
721    
722      /* \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
723      larger first octal digit. */      larger first octal digit. The original code used just to take the least
724        significant 8 bits of octal numbers (I think this is what early Perls used
725      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
726      c -= '0';      than 3 octal digits. */
727      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
728          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
729      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
730        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
731            c = c * 8 + *(++ptr) - CHAR_0;
732        if (!utf8 && c > 255) *errorcodeptr = ERR51;
733      break;      break;
734    
735      /* \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
736      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
737        treated as a data character. */
738    
739      case 'x':      case CHAR_x:
740  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
741        {        {
742        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
743        register int count = 0;        int count = 0;
744    
745        c = 0;        c = 0;
746        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
747          {          {
748          int cc = *pt++;          register int cc = *pt++;
749            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
750          count++;          count++;
751  #if !EBCDIC    /* ASCII coding */  
752          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
753          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
754  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
755          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
756          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
757            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
758  #endif  #endif
759          }          }
760        if (*pt == '}')  
761          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
762          {          {
763          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
764          ptr = pt;          ptr = pt;
765          break;          break;
766          }          }
767    
768        /* 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
769        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
770        }        }
 #endif  
771    
772      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
773    
774      c = 0;      c = 0;
775      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
776        {        {
777        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
778        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
779  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
780        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
781        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
782  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
783        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
784        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
785  #endif  #endif
786        }        }
787      break;      break;
788    
789      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
790        This coding is ASCII-specific, but then the whole concept of \cx is
791        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
792    
793      case 'c':      case CHAR_c:
794      c = *(++ptr);      c = *(++ptr);
795      if (c == 0)      if (c == 0)
796        {        {
797        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
798        return 0;        break;
799        }        }
800    
801      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
802      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;  
803      c ^= 0x40;      c ^= 0x40;
804  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
805      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
806      c ^= 0xC0;      c ^= 0xC0;
807  #endif  #endif
808      break;      break;
809    
810      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
811      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
812      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
813      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
814      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
815    
816      default:      default:
817      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 843  escape sequence.
843  Argument:  Argument:
844    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
845    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
846      dptr           points to an int that is set to the detailed property value
847    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
848    
849  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
850  */  */
851    
852  static int  static int
853  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
854  {  {
855  int c, i, bot, top;  int c, i, bot, top;
856  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
857  char name[4];  char name[32];
858    
859  c = *(++ptr);  c = *(++ptr);
860  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
861    
862  *negptr = FALSE;  *negptr = FALSE;
863    
864  /* \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
865  preceded by ^ for negation. */  negation. */
866    
867  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
868    {    {
869    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
870      {      {
871      *negptr = TRUE;      *negptr = TRUE;
872      ptr++;      ptr++;
873      }      }
874    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
875      {      {
876      c = *(++ptr);      c = *(++ptr);
877      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
878      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
879      name[i] = c;      name[i] = c;
880      }      }
881    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;  
     }  
882    name[i] = 0;    name[i] = 0;
883    }    }
884    
# Line 619  top = _pcre_utt_size; Line 899  top = _pcre_utt_size;
899    
900  while (bot < top)  while (bot < top)
901    {    {
902    i = (bot + top)/2;    i = (bot + top) >> 1;
903    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
904    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
905        {
906        *dptr = _pcre_utt[i].value;
907        return _pcre_utt[i].type;
908        }
909    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
910    }    }
911    
 UNKNOWN_RETURN:  
912  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
913  *ptrptr = ptr;  *ptrptr = ptr;
914  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 943  is_counted_repeat(const uschar *p)
943  {  {
944  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
945  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
946  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
947    
948  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
949  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
950    
951  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
952  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
953    
954  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
955  }  }
956    
957    
# Line 698  read_repeat_counts(const uschar *p, int Line 981  read_repeat_counts(const uschar *p, int
981  int min = 0;  int min = 0;
982  int max = -1;  int max = -1;
983    
984  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
985    an integer overflow. */
986    
987    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
988    if (min < 0 || min > 65535)
989      {
990      *errorcodeptr = ERR5;
991      return p;
992      }
993    
994    /* Read the maximum value if there is one, and again do a paranoid on its size.
995    Also, max must not be less than min. */
996    
997  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
998    {    {
999    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1000      {      {
1001      max = 0;      max = 0;
1002      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1003        if (max < 0 || max > 65535)
1004          {
1005          *errorcodeptr = ERR5;
1006          return p;
1007          }
1008      if (max < min)      if (max < min)
1009        {        {
1010        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1013  if (*p == '}') max = min; else
1013      }      }
1014    }    }
1015    
1016  /* 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
1017  pointer to the terminating '}'. */  '}'. */
1018    
1019  if (min > 65535 || max > 65535)  *minp = min;
1020    *errorcodeptr = ERR5;  *maxp = max;
1021  else  return p;
1022    }
1023    
1024    
1025    
1026    /*************************************************
1027    *  Subroutine for finding forward reference      *
1028    *************************************************/
1029    
1030    /* This recursive function is called only from find_parens() below. The
1031    top-level call starts at the beginning of the pattern. All other calls must
1032    start at a parenthesis. It scans along a pattern's text looking for capturing
1033    subpatterns, and counting them. If it finds a named pattern that matches the
1034    name it is given, it returns its number. Alternatively, if the name is NULL, it
1035    returns when it reaches a given numbered subpattern. We know that if (?P< is
1036    encountered, the name will be terminated by '>' because that is checked in the
1037    first pass. Recursion is used to keep track of subpatterns that reset the
1038    capturing group numbers - the (?| feature.
1039    
1040    Arguments:
1041      ptrptr       address of the current character pointer (updated)
1042      cd           compile background data
1043      name         name to seek, or NULL if seeking a numbered subpattern
1044      lorn         name length, or subpattern number if name is NULL
1045      xmode        TRUE if we are in /x mode
1046      count        pointer to the current capturing subpattern number (updated)
1047    
1048    Returns:       the number of the named subpattern, or -1 if not found
1049    */
1050    
1051    static int
1052    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1053      BOOL xmode, int *count)
1054    {
1055    uschar *ptr = *ptrptr;
1056    int start_count = *count;
1057    int hwm_count = start_count;
1058    BOOL dup_parens = FALSE;
1059    
1060    /* If the first character is a parenthesis, check on the type of group we are
1061    dealing with. The very first call may not start with a parenthesis. */
1062    
1063    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1064    {    {
1065    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1066    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1067        {
1068        ptr += 3;
1069        dup_parens = TRUE;
1070        }
1071    
1072      /* Handle a normal, unnamed capturing parenthesis */
1073    
1074      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1075        {
1076        *count += 1;
1077        if (name == NULL && *count == lorn) return *count;
1078        ptr++;
1079        }
1080    
1081      /* Handle a condition. If it is an assertion, just carry on so that it
1082      is processed as normal. If not, skip to the closing parenthesis of the
1083      condition (there can't be any nested parens. */
1084    
1085      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1086        {
1087        ptr += 2;
1088        if (ptr[1] != CHAR_QUESTION_MARK)
1089          {
1090          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1091          if (*ptr != 0) ptr++;
1092          }
1093        }
1094    
1095      /* We have either (? or (* and not a condition */
1096    
1097      else
1098        {
1099        ptr += 2;
1100        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1101    
1102        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1103    
1104        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1105            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1106          {
1107          int term;
1108          const uschar *thisname;
1109          *count += 1;
1110          if (name == NULL && *count == lorn) return *count;
1111          term = *ptr++;
1112          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1113          thisname = ptr;
1114          while (*ptr != term) ptr++;
1115          if (name != NULL && lorn == ptr - thisname &&
1116              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1117            return *count;
1118          term++;
1119          }
1120        }
1121    }    }
1122  return p;  
1123    /* Past any initial parenthesis handling, scan for parentheses or vertical
1124    bars. */
1125    
1126    for (; *ptr != 0; ptr++)
1127      {
1128      /* Skip over backslashed characters and also entire \Q...\E */
1129    
1130      if (*ptr == CHAR_BACKSLASH)
1131        {
1132        if (*(++ptr) == 0) goto FAIL_EXIT;
1133        if (*ptr == CHAR_Q) for (;;)
1134          {
1135          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1136          if (*ptr == 0) goto FAIL_EXIT;
1137          if (*(++ptr) == CHAR_E) break;
1138          }
1139        continue;
1140        }
1141    
1142      /* Skip over character classes; this logic must be similar to the way they
1143      are handled for real. If the first character is '^', skip it. Also, if the
1144      first few characters (either before or after ^) are \Q\E or \E we skip them
1145      too. This makes for compatibility with Perl. Note the use of STR macros to
1146      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1147    
1148      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1149        {
1150        BOOL negate_class = FALSE;
1151        for (;;)
1152          {
1153          if (ptr[1] == CHAR_BACKSLASH)
1154            {
1155            if (ptr[2] == CHAR_E)
1156              ptr+= 2;
1157            else if (strncmp((const char *)ptr+2,
1158                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1159              ptr += 4;
1160            else
1161              break;
1162            }
1163          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1164            {
1165            negate_class = TRUE;
1166            ptr++;
1167            }
1168          else break;
1169          }
1170    
1171        /* If the next character is ']', it is a data character that must be
1172        skipped, except in JavaScript compatibility mode. */
1173    
1174        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1175            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1176          ptr++;
1177    
1178        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1179          {
1180          if (*ptr == 0) return -1;
1181          if (*ptr == CHAR_BACKSLASH)
1182            {
1183            if (*(++ptr) == 0) goto FAIL_EXIT;
1184            if (*ptr == CHAR_Q) for (;;)
1185              {
1186              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1187              if (*ptr == 0) goto FAIL_EXIT;
1188              if (*(++ptr) == CHAR_E) break;
1189              }
1190            continue;
1191            }
1192          }
1193        continue;
1194        }
1195    
1196      /* Skip comments in /x mode */
1197    
1198      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1199        {
1200        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1201        if (*ptr == 0) goto FAIL_EXIT;
1202        continue;
1203        }
1204    
1205      /* Check for the special metacharacters */
1206    
1207      if (*ptr == CHAR_LEFT_PARENTHESIS)
1208        {
1209        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1210        if (rc > 0) return rc;
1211        if (*ptr == 0) goto FAIL_EXIT;
1212        }
1213    
1214      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1215        {
1216        if (dup_parens && *count < hwm_count) *count = hwm_count;
1217        *ptrptr = ptr;
1218        return -1;
1219        }
1220    
1221      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1222        {
1223        if (*count > hwm_count) hwm_count = *count;
1224        *count = start_count;
1225        }
1226      }
1227    
1228    FAIL_EXIT:
1229    *ptrptr = ptr;
1230    return -1;
1231    }
1232    
1233    
1234    
1235    
1236    /*************************************************
1237    *       Find forward referenced subpattern       *
1238    *************************************************/
1239    
1240    /* This function scans along a pattern's text looking for capturing
1241    subpatterns, and counting them. If it finds a named pattern that matches the
1242    name it is given, it returns its number. Alternatively, if the name is NULL, it
1243    returns when it reaches a given numbered subpattern. This is used for forward
1244    references to subpatterns. We used to be able to start this scan from the
1245    current compiling point, using the current count value from cd->bracount, and
1246    do it all in a single loop, but the addition of the possibility of duplicate
1247    subpattern numbers means that we have to scan from the very start, in order to
1248    take account of such duplicates, and to use a recursive function to keep track
1249    of the different types of group.
1250    
1251    Arguments:
1252      cd           compile background data
1253      name         name to seek, or NULL if seeking a numbered subpattern
1254      lorn         name length, or subpattern number if name is NULL
1255      xmode        TRUE if we are in /x mode
1256    
1257    Returns:       the number of the found subpattern, or -1 if not found
1258    */
1259    
1260    static int
1261    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1262    {
1263    uschar *ptr = (uschar *)cd->start_pattern;
1264    int count = 0;
1265    int rc;
1266    
1267    /* If the pattern does not start with an opening parenthesis, the first call
1268    to find_parens_sub() will scan right to the end (if necessary). However, if it
1269    does start with a parenthesis, find_parens_sub() will return when it hits the
1270    matching closing parens. That is why we have to have a loop. */
1271    
1272    for (;;)
1273      {
1274      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1275      if (rc > 0 || *ptr++ == 0) break;
1276      }
1277    
1278    return rc;
1279  }  }
1280    
1281    
1282    
1283    
1284  /*************************************************  /*************************************************
1285  *      Find first significant op code            *  *      Find first significant op code            *
1286  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1330  for (;;)
1330    
1331      case OP_CALLOUT:      case OP_CALLOUT:
1332      case OP_CREF:      case OP_CREF:
1333      case OP_BRANUMBER:      case OP_NCREF:
1334        case OP_RREF:
1335        case OP_NRREF:
1336        case OP_DEF:
1337      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1338      break;      break;
1339    
# Line 793  for (;;) Line 1348  for (;;)
1348    
1349    
1350  /*************************************************  /*************************************************
1351  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1352  *************************************************/  *************************************************/
1353    
1354  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1355  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1356  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1357    temporarily terminated with OP_END when this function is called.
1358    
1359    This function is called when a backward assertion is encountered, so that if it
1360    fails, the error message can point to the correct place in the pattern.
1361    However, we cannot do this when the assertion contains subroutine calls,
1362    because they can be forward references. We solve this by remembering this case
1363    and doing the check at the end; a flag specifies which mode we are running in.
1364    
1365  Arguments:  Arguments:
1366    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1367    options  the compiling options    options  the compiling options
1368      atend    TRUE if called when the pattern is complete
1369      cd       the "compile data" structure
1370    
1371  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1372                 or -1 if there is no fixed length,
1373               or -2 if \C was encountered               or -2 if \C was encountered
1374                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1375  */  */
1376    
1377  static int  static int
1378  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1379  {  {
1380  int length = -1;  int length = -1;
1381    
# Line 822  branch, check the length against that of Line 1388  branch, check the length against that of
1388  for (;;)  for (;;)
1389    {    {
1390    int d;    int d;
1391      uschar *ce, *cs;
1392    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1393    switch (op)    switch (op)
1394      {      {
1395        case OP_CBRA:
1396      case OP_BRA:      case OP_BRA:
1397      case OP_ONCE:      case OP_ONCE:
1398      case OP_COND:      case OP_COND:
1399      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1400      if (d < 0) return d;      if (d < 0) return d;
1401      branchlength += d;      branchlength += d;
1402      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 853  for (;;) Line 1419  for (;;)
1419      branchlength = 0;      branchlength = 0;
1420      break;      break;
1421    
1422        /* A true recursion implies not fixed length, but a subroutine call may
1423        be OK. If the subroutine is a forward reference, we can't deal with
1424        it until the end of the pattern, so return -3. */
1425    
1426        case OP_RECURSE:
1427        if (!atend) return -3;
1428        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1429        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1430        if (cc > cs && cc < ce) return -1;                /* Recursion */
1431        d = find_fixedlength(cs + 2, options, atend, cd);
1432        if (d < 0) return d;
1433        branchlength += d;
1434        cc += 1 + LINK_SIZE;
1435        break;
1436    
1437      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1438    
1439      case OP_ASSERT:      case OP_ASSERT:
# Line 865  for (;;) Line 1446  for (;;)
1446      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1447    
1448      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1449      case OP_CREF:      case OP_CREF:
1450        case OP_NCREF:
1451        case OP_RREF:
1452        case OP_NRREF:
1453        case OP_DEF:
1454      case OP_OPT:      case OP_OPT:
1455      case OP_CALLOUT:      case OP_CALLOUT:
1456      case OP_SOD:      case OP_SOD:
1457      case OP_SOM:      case OP_SOM:
1458        case OP_SET_SOM:
1459      case OP_EOD:      case OP_EOD:
1460      case OP_EODN:      case OP_EODN:
1461      case OP_CIRC:      case OP_CIRC:
# Line 884  for (;;) Line 1469  for (;;)
1469    
1470      case OP_CHAR:      case OP_CHAR:
1471      case OP_CHARNC:      case OP_CHARNC:
1472        case OP_NOT:
1473      branchlength++;      branchlength++;
1474      cc += 2;      cc += 2;
1475  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1476      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1477        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1478  #endif  #endif
1479      break;      break;
1480    
# Line 901  for (;;) Line 1485  for (;;)
1485      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1486      cc += 4;      cc += 4;
1487  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1488      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1489        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1490  #endif  #endif
1491      break;      break;
1492    
1493      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1494      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1495        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1496      cc += 4;      cc += 4;
1497      break;      break;
1498    
# Line 917  for (;;) Line 1500  for (;;)
1500    
1501      case OP_PROP:      case OP_PROP:
1502      case OP_NOTPROP:      case OP_NOTPROP:
1503      cc++;      cc += 2;
1504      /* Fall through */      /* Fall through */
1505    
1506      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1510  for (;;)
1510      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1511      case OP_WORDCHAR:      case OP_WORDCHAR:
1512      case OP_ANY:      case OP_ANY:
1513        case OP_ALLANY:
1514      branchlength++;      branchlength++;
1515      cc++;      cc++;
1516      break;      break;
# Line 981  for (;;) Line 1565  for (;;)
1565    
1566    
1567  /*************************************************  /*************************************************
1568  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1569  *************************************************/  *************************************************/
1570    
1571  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1572  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1573    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1574    so that it can be called from pcre_study() when finding the minimum matching
1575    length.
1576    
1577  Arguments:  Arguments:
1578    code        points to start of expression    code        points to start of expression
1579    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1580    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1581    
1582  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1583  */  */
1584    
1585  static const uschar *  const uschar *
1586  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1587  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1588  for (;;)  for (;;)
1589    {    {
1590    register int c = *code;    register int c = *code;
1591    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1592    else if (c > OP_BRA)  
1593      /* XCLASS is used for classes that cannot be represented just by a bit
1594      map. This includes negated single high-valued characters. The length in
1595      the table is zero; the actual length is stored in the compiled code. */
1596    
1597      if (c == OP_XCLASS) code += GET(code, 1);
1598    
1599      /* Handle recursion */
1600    
1601      else if (c == OP_REVERSE)
1602      {      {
1603      int n = c - OP_BRA;      if (number < 0) return (uschar *)code;
1604      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      code += _pcre_OP_lengths[c];
     if (n == number) return (uschar *)code;  
     code += _pcre_OP_lengths[OP_BRA];  
1605      }      }
1606    else  
1607      /* Handle capturing bracket */
1608    
1609      else if (c == OP_CBRA)
1610      {      {
1611        int n = GET2(code, 1+LINK_SIZE);
1612        if (n == number) return (uschar *)code;
1613      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1614        }
1615    
1616  #ifdef SUPPORT_UTF8    /* Otherwise, we can get the item's length from the table, except that for
1617      repeated character types, we have to test for \p and \P, which have an extra
1618      two bytes of parameters. */
1619    
1620      /* In UTF-8 mode, opcodes that are followed by a character may be followed    else
1621      by a multi-byte character. The length in the table is a minimum, so we have      {
1622      to scan along to skip the extra bytes. All opcodes are less than 128, so we      switch(c)
1623      can use relatively efficient code. */        {
1624          case OP_TYPESTAR:
1625          case OP_TYPEMINSTAR:
1626          case OP_TYPEPLUS:
1627          case OP_TYPEMINPLUS:
1628          case OP_TYPEQUERY:
1629          case OP_TYPEMINQUERY:
1630          case OP_TYPEPOSSTAR:
1631          case OP_TYPEPOSPLUS:
1632          case OP_TYPEPOSQUERY:
1633          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1634          break;
1635    
1636          case OP_TYPEUPTO:
1637          case OP_TYPEMINUPTO:
1638          case OP_TYPEEXACT:
1639          case OP_TYPEPOSUPTO:
1640          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1641          break;
1642          }
1643    
1644        /* Add in the fixed length from the table */
1645    
1646        code += _pcre_OP_lengths[c];
1647    
1648      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1649      a multi-byte character. The length in the table is a minimum, so we have to
1650      arrange to skip the extra bytes. */
1651    
1652    #ifdef SUPPORT_UTF8
1653      if (utf8) switch(c)      if (utf8) switch(c)
1654        {        {
1655        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1657  for (;;)
1657        case OP_EXACT:        case OP_EXACT:
1658        case OP_UPTO:        case OP_UPTO:
1659        case OP_MINUPTO:        case OP_MINUPTO:
1660          case OP_POSUPTO:
1661        case OP_STAR:        case OP_STAR:
1662        case OP_MINSTAR:        case OP_MINSTAR:
1663          case OP_POSSTAR:
1664        case OP_PLUS:        case OP_PLUS:
1665        case OP_MINPLUS:        case OP_MINPLUS:
1666          case OP_POSPLUS:
1667        case OP_QUERY:        case OP_QUERY:
1668        case OP_MINQUERY:        case OP_MINQUERY:
1669        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1670        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;  
1671        break;        break;
1672        }        }
1673    #else
1674        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1675  #endif  #endif
1676      }      }
1677    }    }
# Line 1072  Returns: pointer to the opcode for Line 1696  Returns: pointer to the opcode for
1696  static const uschar *  static const uschar *
1697  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1698  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1703    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1704    else if (c > OP_BRA)  
1705      {    /* XCLASS is used for classes that cannot be represented just by a bit
1706      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1707      }    the table is zero; the actual length is stored in the compiled code. */
1708    
1709      if (c == OP_XCLASS) code += GET(code, 1);
1710    
1711      /* Otherwise, we can get the item's length from the table, except that for
1712      repeated character types, we have to test for \p and \P, which have an extra
1713      two bytes of parameters. */
1714    
1715    else    else
1716      {      {
1717      code += _pcre_OP_lengths[c];      switch(c)
1718          {
1719          case OP_TYPESTAR:
1720          case OP_TYPEMINSTAR:
1721          case OP_TYPEPLUS:
1722          case OP_TYPEMINPLUS:
1723          case OP_TYPEQUERY:
1724          case OP_TYPEMINQUERY:
1725          case OP_TYPEPOSSTAR:
1726          case OP_TYPEPOSPLUS:
1727          case OP_TYPEPOSQUERY:
1728          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1729          break;
1730    
1731  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1732          case OP_TYPEUPTO:
1733          case OP_TYPEMINUPTO:
1734          case OP_TYPEEXACT:
1735          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1736          break;
1737          }
1738    
1739        /* Add in the fixed length from the table */
1740    
1741        code += _pcre_OP_lengths[c];
1742    
1743      /* 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
1744      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
1745      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. */  
1746    
1747    #ifdef SUPPORT_UTF8
1748      if (utf8) switch(c)      if (utf8) switch(c)
1749        {        {
1750        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1752  for (;;)
1752        case OP_EXACT:        case OP_EXACT:
1753        case OP_UPTO:        case OP_UPTO:
1754        case OP_MINUPTO:        case OP_MINUPTO:
1755          case OP_POSUPTO:
1756        case OP_STAR:        case OP_STAR:
1757        case OP_MINSTAR:        case OP_MINSTAR:
1758          case OP_POSSTAR:
1759        case OP_PLUS:        case OP_PLUS:
1760        case OP_MINPLUS:        case OP_MINPLUS:
1761          case OP_POSPLUS:
1762        case OP_QUERY:        case OP_QUERY:
1763        case OP_MINQUERY:        case OP_MINQUERY:
1764        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1765        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;  
1766        break;        break;
1767        }        }
1768    #else
1769        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1770  #endif  #endif
1771      }      }
1772    }    }
# Line 1132  for (;;) Line 1779  for (;;)
1779  *************************************************/  *************************************************/
1780    
1781  /* 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
1782  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()
1783  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
1784  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
1785  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1786    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1787    bracket whose current branch will already have been scanned.
1788    
1789  Arguments:  Arguments:
1790    code        points to start of search    code        points to start of search
1791    endcode     points to where to stop    endcode     points to where to stop
1792    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1793      cd          contains pointers to tables etc.
1794    
1795  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1796  */  */
1797    
1798  static BOOL  static BOOL
1799  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1800      compile_data *cd)
1801  {  {
1802  register int c;  register int c;
1803  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);
1804       code < endcode;       code < endcode;
1805       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1806    {    {
# Line 1157  for (code = first_significant_code(code Line 1808  for (code = first_significant_code(code
1808    
1809    c = *code;    c = *code;
1810    
1811    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1812      first_significant_code() with a TRUE final argument. */
1813    
1814      if (c == OP_ASSERT)
1815      {      {
1816      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1817      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1818        continue;
1819        }
1820    
1821      /* Groups with zero repeats can of course be empty; skip them. */
1822    
1823      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1824        {
1825        code += _pcre_OP_lengths[c];
1826        do code += GET(code, 1); while (*code == OP_ALT);
1827        c = *code;
1828        continue;
1829        }
1830    
1831      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1832      implies a subroutine call, we can scan it. */
1833    
1834      empty_branch = FALSE;    if (c == OP_RECURSE)
1835        {
1836        BOOL empty_branch = FALSE;
1837        const uschar *scode = cd->start_code + GET(code, 1);
1838        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1839      do      do
1840        {        {
1841        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1842            {
1843          empty_branch = TRUE;          empty_branch = TRUE;
1844            break;
1845            }
1846          scode += GET(scode, 1);
1847          }
1848        while (*scode == OP_ALT);
1849        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1850        continue;
1851        }
1852    
1853      /* For other groups, scan the branches. */
1854    
1855      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1856        {
1857        BOOL empty_branch;
1858        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1859    
1860        /* If a conditional group has only one branch, there is a second, implied,
1861        empty branch, so just skip over the conditional, because it could be empty.
1862        Otherwise, scan the individual branches of the group. */
1863    
1864        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1865        code += GET(code, 1);        code += GET(code, 1);
1866        else
1867          {
1868          empty_branch = FALSE;
1869          do
1870            {
1871            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1872              empty_branch = TRUE;
1873            code += GET(code, 1);
1874            }
1875          while (*code == OP_ALT);
1876          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1877        }        }
1878      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1879      c = *code;      c = *code;
1880        continue;
1881      }      }
1882    
1883    else switch (c)    /* Handle the other opcodes */
1884    
1885      switch (c)
1886      {      {
1887      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1888        cannot be represented just by a bit map. This includes negated single
1889        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1890        actual length is stored in the compiled code, so we must update "code"
1891        here. */
1892    
1893  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1894      case OP_XCLASS:      case OP_XCLASS:
1895      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1896      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1897  #endif  #endif
1898    
# Line 1227  for (code = first_significant_code(code Line 1936  for (code = first_significant_code(code
1936      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1937      case OP_WORDCHAR:      case OP_WORDCHAR:
1938      case OP_ANY:      case OP_ANY:
1939        case OP_ALLANY:
1940      case OP_ANYBYTE:      case OP_ANYBYTE:
1941      case OP_CHAR:      case OP_CHAR:
1942      case OP_CHARNC:      case OP_CHARNC:
1943      case OP_NOT:      case OP_NOT:
1944      case OP_PLUS:      case OP_PLUS:
1945      case OP_MINPLUS:      case OP_MINPLUS:
1946        case OP_POSPLUS:
1947      case OP_EXACT:      case OP_EXACT:
1948      case OP_NOTPLUS:      case OP_NOTPLUS:
1949      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1950        case OP_NOTPOSPLUS:
1951      case OP_NOTEXACT:      case OP_NOTEXACT:
1952      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1953      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1954        case OP_TYPEPOSPLUS:
1955      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1956      return FALSE;      return FALSE;
1957    
1958        /* These are going to continue, as they may be empty, but we have to
1959        fudge the length for the \p and \P cases. */
1960    
1961        case OP_TYPESTAR:
1962        case OP_TYPEMINSTAR:
1963        case OP_TYPEPOSSTAR:
1964        case OP_TYPEQUERY:
1965        case OP_TYPEMINQUERY:
1966        case OP_TYPEPOSQUERY:
1967        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1968        break;
1969    
1970        /* Same for these */
1971    
1972        case OP_TYPEUPTO:
1973        case OP_TYPEMINUPTO:
1974        case OP_TYPEPOSUPTO:
1975        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1976        break;
1977    
1978      /* End of branch */      /* End of branch */
1979    
1980      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1983  for (code = first_significant_code(code
1983      case OP_ALT:      case OP_ALT:
1984      return TRUE;      return TRUE;
1985    
1986      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1987      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1988    
1989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1990      case OP_STAR:      case OP_STAR:
1991      case OP_MINSTAR:      case OP_MINSTAR:
1992        case OP_POSSTAR:
1993      case OP_QUERY:      case OP_QUERY:
1994      case OP_MINQUERY:      case OP_MINQUERY:
1995        case OP_POSQUERY:
1996        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1997        break;
1998    
1999      case OP_UPTO:      case OP_UPTO:
2000      case OP_MINUPTO:      case OP_MINUPTO:
2001      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2002        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2003      break;      break;
2004  #endif  #endif
2005    
2006        /* None of the remaining opcodes are required to match a character. */
2007    
2008        default:
2009        break;
2010      }      }
2011    }    }
2012    
# Line 1285  Arguments: Line 2029  Arguments:
2029    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2030    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2031    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2032      cd          pointers to tables etc
2033    
2034  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2035  */  */
2036    
2037  static BOOL  static BOOL
2038  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2039    BOOL utf8)    BOOL utf8, compile_data *cd)
2040  {  {
2041  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2042    {    {
2043    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2044        return FALSE;
2045    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2046    }    }
2047  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2054  return TRUE;
2054  *************************************************/  *************************************************/
2055    
2056  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2057  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
2058  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2059  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2060    
2061    Originally, this function only recognized a sequence of letters between the
2062    terminators, but it seems that Perl recognizes any sequence of characters,
2063    though of course unknown POSIX names are subsequently rejected. Perl gives an
2064    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2065    didn't consider this to be a POSIX class. Likewise for [:1234:].
2066    
2067    The problem in trying to be exactly like Perl is in the handling of escapes. We
2068    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2069    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2070    below handles the special case of \], but does not try to do any other escape
2071    processing. This makes it different from Perl for cases such as [:l\ower:]
2072    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2073    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2074    I think.
2075    
2076  Argument:  Arguments:
2077    ptr      pointer to the initial [    ptr      pointer to the initial [
2078    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2079    
2080  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2081  */  */
2082    
2083  static BOOL  static BOOL
2084  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2085  {  {
2086  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2087  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2088  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2089    {    {
2090    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2091    return TRUE;      {
2092        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2093        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2094          {
2095          *endptr = ptr;
2096          return TRUE;
2097          }
2098        }
2099    }    }
2100  return FALSE;  return FALSE;
2101  }  }
# Line 1355  Returns: a value representing the na Line 2120  Returns: a value representing the na
2120  static int  static int
2121  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2122  {  {
2123    const char *pn = posix_names;
2124  register int yield = 0;  register int yield = 0;
2125  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2126    {    {
2127    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2128      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2129      pn += posix_name_lengths[yield] + 1;
2130    yield++;    yield++;
2131    }    }
2132  return -1;  return -1;
# Line 1374  return -1; Line 2141  return -1;
2141  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2142  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2143  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
2144  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
2145  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
2146  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
2147  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
2148  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2149    OP_END.
2150    
2151    This function has been extended with the possibility of forward references for
2152    recursions and subroutine calls. It must also check the list of such references
2153    for the group we are dealing with. If it finds that one of the recursions in
2154    the current group is on this list, it adjusts the offset in the list, not the
2155    value in the reference (which is a group number).
2156    
2157  Arguments:  Arguments:
2158    group      points to the start of the group    group      points to the start of the group
2159    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2160    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2161    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2162      save_hwm   the hwm forward reference pointer at the start of the group
2163    
2164  Returns:     nothing  Returns:     nothing
2165  */  */
2166    
2167  static void  static void
2168  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2169      uschar *save_hwm)
2170  {  {
2171  uschar *ptr = group;  uschar *ptr = group;
2172    
2173  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2174    {    {
2175    int offset = GET(ptr, 1);    int offset;
2176    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2177    
2178      /* See if this recursion is on the forward reference list. If so, adjust the
2179      reference. */
2180    
2181      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2182        {
2183        offset = GET(hc, 0);
2184        if (cd->start_code + offset == ptr + 1)
2185          {
2186          PUT(hc, 0, offset + adjust);
2187          break;
2188          }
2189        }
2190    
2191      /* Otherwise, adjust the recursion offset if it's after the start of this
2192      group. */
2193    
2194      if (hc >= cd->hwm)
2195        {
2196        offset = GET(ptr, 1);
2197        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2198        }
2199    
2200    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2201    }    }
2202  }  }
# Line 1475  Yield: TRUE when range returned; Line 2275  Yield: TRUE when range returned;
2275  */  */
2276    
2277  static BOOL  static BOOL
2278  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2279      unsigned int *odptr)
2280  {  {
2281  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2282    
2283  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2284    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2285    
2286  if (c > d) return FALSE;  if (c > d) return FALSE;
2287    
# Line 1492  next = othercase + 1; Line 2290  next = othercase + 1;
2290    
2291  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2292    {    {
2293    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2294    next++;    next++;
2295    }    }
2296    
# Line 1506  return TRUE; Line 2302  return TRUE;
2302  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2303    
2304    
2305    
2306    /*************************************************
2307    *     Check if auto-possessifying is possible    *
2308    *************************************************/
2309    
2310    /* This function is called for unlimited repeats of certain items, to see
2311    whether the next thing could possibly match the repeated item. If not, it makes
2312    sense to automatically possessify the repeated item.
2313    
2314    Arguments:
2315      op_code       the repeated op code
2316      this          data for this item, depends on the opcode
2317      utf8          TRUE in UTF-8 mode
2318      utf8_char     used for utf8 character bytes, NULL if not relevant
2319      ptr           next character in pattern
2320      options       options bits
2321      cd            contains pointers to tables etc.
2322    
2323    Returns:        TRUE if possessifying is wanted
2324    */
2325    
2326    static BOOL
2327    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2328      const uschar *ptr, int options, compile_data *cd)
2329    {
2330    int next;
2331    
2332    /* Skip whitespace and comments in extended mode */
2333    
2334    if ((options & PCRE_EXTENDED) != 0)
2335      {
2336      for (;;)
2337        {
2338        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2339        if (*ptr == CHAR_NUMBER_SIGN)
2340          {
2341          while (*(++ptr) != 0)
2342            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2343          }
2344        else break;
2345        }
2346      }
2347    
2348    /* If the next item is one that we can handle, get its value. A non-negative
2349    value is a character, a negative value is an escape value. */
2350    
2351    if (*ptr == CHAR_BACKSLASH)
2352      {
2353      int temperrorcode = 0;
2354      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2355      if (temperrorcode != 0) return FALSE;
2356      ptr++;    /* Point after the escape sequence */
2357      }
2358    
2359    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2360      {
2361    #ifdef SUPPORT_UTF8
2362      if (utf8) { GETCHARINC(next, ptr); } else
2363    #endif
2364      next = *ptr++;
2365      }
2366    
2367    else return FALSE;
2368    
2369    /* Skip whitespace and comments in extended mode */
2370    
2371    if ((options & PCRE_EXTENDED) != 0)
2372      {
2373      for (;;)
2374        {
2375        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2376        if (*ptr == CHAR_NUMBER_SIGN)
2377          {
2378          while (*(++ptr) != 0)
2379            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2380          }
2381        else break;
2382        }
2383      }
2384    
2385    /* If the next thing is itself optional, we have to give up. */
2386    
2387    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2388      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2389        return FALSE;
2390    
2391    /* Now compare the next item with the previous opcode. If the previous is a
2392    positive single character match, "item" either contains the character or, if
2393    "item" is greater than 127 in utf8 mode, the character's bytes are in
2394    utf8_char. */
2395    
2396    
2397    /* Handle cases when the next item is a character. */
2398    
2399    if (next >= 0) switch(op_code)
2400      {
2401      case OP_CHAR:
2402    #ifdef SUPPORT_UTF8
2403      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2404    #else
2405      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2406    #endif
2407      return item != next;
2408    
2409      /* For CHARNC (caseless character) we must check the other case. If we have
2410      Unicode property support, we can use it to test the other case of
2411      high-valued characters. */
2412    
2413      case OP_CHARNC:
2414    #ifdef SUPPORT_UTF8
2415      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2416    #endif
2417      if (item == next) return FALSE;
2418    #ifdef SUPPORT_UTF8
2419      if (utf8)
2420        {
2421        unsigned int othercase;
2422        if (next < 128) othercase = cd->fcc[next]; else
2423    #ifdef SUPPORT_UCP
2424        othercase = UCD_OTHERCASE((unsigned int)next);
2425    #else
2426        othercase = NOTACHAR;
2427    #endif
2428        return (unsigned int)item != othercase;
2429        }
2430      else
2431    #endif  /* SUPPORT_UTF8 */
2432      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2433    
2434      /* For OP_NOT, "item" must be a single-byte character. */
2435    
2436      case OP_NOT:
2437      if (item == next) return TRUE;
2438      if ((options & PCRE_CASELESS) == 0) return FALSE;
2439    #ifdef SUPPORT_UTF8
2440      if (utf8)
2441        {
2442        unsigned int othercase;
2443        if (next < 128) othercase = cd->fcc[next]; else
2444    #ifdef SUPPORT_UCP
2445        othercase = UCD_OTHERCASE(next);
2446    #else
2447        othercase = NOTACHAR;
2448    #endif
2449        return (unsigned int)item == othercase;
2450        }
2451      else
2452    #endif  /* SUPPORT_UTF8 */
2453      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2454    
2455      case OP_DIGIT:
2456      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2457    
2458      case OP_NOT_DIGIT:
2459      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2460    
2461      case OP_WHITESPACE:
2462      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2463    
2464      case OP_NOT_WHITESPACE:
2465      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2466    
2467      case OP_WORDCHAR:
2468      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2469    
2470      case OP_NOT_WORDCHAR:
2471      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2472    
2473      case OP_HSPACE:
2474      case OP_NOT_HSPACE:
2475      switch(next)
2476        {
2477        case 0x09:
2478        case 0x20:
2479        case 0xa0:
2480        case 0x1680:
2481        case 0x180e:
2482        case 0x2000:
2483        case 0x2001:
2484        case 0x2002:
2485        case 0x2003:
2486        case 0x2004:
2487        case 0x2005:
2488        case 0x2006:
2489        case 0x2007:
2490        case 0x2008:
2491        case 0x2009:
2492        case 0x200A:
2493        case 0x202f:
2494        case 0x205f:
2495        case 0x3000:
2496        return op_code != OP_HSPACE;
2497        default:
2498        return op_code == OP_HSPACE;
2499        }
2500    
2501      case OP_VSPACE:
2502      case OP_NOT_VSPACE:
2503      switch(next)
2504        {
2505        case 0x0a:
2506        case 0x0b:
2507        case 0x0c:
2508        case 0x0d:
2509        case 0x85:
2510        case 0x2028:
2511        case 0x2029:
2512        return op_code != OP_VSPACE;
2513        default:
2514        return op_code == OP_VSPACE;
2515        }
2516    
2517      default:
2518      return FALSE;
2519      }
2520    
2521    
2522    /* Handle the case when the next item is \d, \s, etc. */
2523    
2524    switch(op_code)
2525      {
2526      case OP_CHAR:
2527      case OP_CHARNC:
2528    #ifdef SUPPORT_UTF8
2529      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2530    #endif
2531      switch(-next)
2532        {
2533        case ESC_d:
2534        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2535    
2536        case ESC_D:
2537        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2538    
2539        case ESC_s:
2540        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2541    
2542        case ESC_S:
2543        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2544    
2545        case ESC_w:
2546        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2547    
2548        case ESC_W:
2549        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2550    
2551        case ESC_h:
2552        case ESC_H:
2553        switch(item)
2554          {
2555          case 0x09:
2556          case 0x20:
2557          case 0xa0:
2558          case 0x1680:
2559          case 0x180e:
2560          case 0x2000:
2561          case 0x2001:
2562          case 0x2002:
2563          case 0x2003:
2564          case 0x2004:
2565          case 0x2005:
2566          case 0x2006:
2567          case 0x2007:
2568          case 0x2008:
2569          case 0x2009:
2570          case 0x200A:
2571          case 0x202f:
2572          case 0x205f:
2573          case 0x3000:
2574          return -next != ESC_h;
2575          default:
2576          return -next == ESC_h;
2577          }
2578    
2579        case ESC_v:
2580        case ESC_V:
2581        switch(item)
2582          {
2583          case 0x0a:
2584          case 0x0b:
2585          case 0x0c:
2586          case 0x0d:
2587          case 0x85:
2588          case 0x2028:
2589          case 0x2029:
2590          return -next != ESC_v;
2591          default:
2592          return -next == ESC_v;
2593          }
2594    
2595        default:
2596        return FALSE;
2597        }
2598    
2599      case OP_DIGIT:
2600      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2601             next == -ESC_h || next == -ESC_v;
2602    
2603      case OP_NOT_DIGIT:
2604      return next == -ESC_d;
2605    
2606      case OP_WHITESPACE:
2607      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2608    
2609      case OP_NOT_WHITESPACE:
2610      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2611    
2612      case OP_HSPACE:
2613      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2614    
2615      case OP_NOT_HSPACE:
2616      return next == -ESC_h;
2617    
2618      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2619      case OP_VSPACE:
2620      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2621    
2622      case OP_NOT_VSPACE:
2623      return next == -ESC_v;
2624    
2625      case OP_WORDCHAR:
2626      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2627    
2628      case OP_NOT_WORDCHAR:
2629      return next == -ESC_w || next == -ESC_d;
2630    
2631      default:
2632      return FALSE;
2633      }
2634    
2635    /* Control does not reach here */
2636    }
2637    
2638    
2639    
2640  /*************************************************  /*************************************************
2641  *           Compile one branch                   *  *           Compile one branch                   *
2642  *************************************************/  *************************************************/
2643    
2644  /* 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
2645  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
2646  bits.  bits. This function is used during the pre-compile phase when we are trying
2647    to find out the amount of memory needed, as well as during the real compile
2648    phase. The value of lengthptr distinguishes the two phases.
2649    
2650  Arguments:  Arguments:
2651    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2652    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2653    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2654    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2656  Arguments:
2656    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2657    bcptr          points to current branch chain    bcptr          points to current branch chain
2658    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2659      lengthptr      NULL during the real compile phase
2660                     points to length accumulator during pre-compile phase
2661    
2662  Returns:         TRUE on success  Returns:         TRUE on success
2663                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2664  */  */
2665    
2666  static BOOL  static BOOL
2667  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2668    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2669    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2670  {  {
2671  int repeat_type, op_type;  int repeat_type, op_type;
2672  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 2675  int greedy_default, greedy_non_default;
2675  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2676  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2677  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2678  int options = *optionsptr;  int options = *optionsptr;
2679  int after_manual_callout = 0;  int after_manual_callout = 0;
2680    int length_prevgroup = 0;
2681  register int c;  register int c;
2682  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2683    uschar *last_code = code;
2684    uschar *orig_code = code;
2685  uschar *tempcode;  uschar *tempcode;
2686  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2687  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2689  const uschar *ptr = *ptrptr;
2689  const uschar *tempptr;  const uschar *tempptr;
2690  uschar *previous = NULL;  uschar *previous = NULL;
2691  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2692    uschar *save_hwm = NULL;
2693  uschar classbits[32];  uschar classbits[32];
2694    
2695  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2696  BOOL class_utf8;  BOOL class_utf8;
2697  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2698  uschar *class_utf8data;  uschar *class_utf8data;
2699    uschar *class_utf8data_base;
2700  uschar utf8_char[6];  uschar utf8_char[6];
2701  #else  #else
2702  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2703    uschar *utf8_char = NULL;
2704    #endif
2705    
2706    #ifdef PCRE_DEBUG
2707    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2708  #endif  #endif
2709    
2710  /* 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 2736  req_caseopt = ((options & PCRE_CASELESS)
2736  for (;; ptr++)  for (;; ptr++)
2737    {    {
2738    BOOL negate_class;    BOOL negate_class;
2739      BOOL should_flip_negation;
2740    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2741    BOOL is_quantifier;    BOOL is_quantifier;
2742      BOOL is_recurse;
2743      BOOL reset_bracount;
2744    int class_charcount;    int class_charcount;
2745    int class_lastchar;    int class_lastchar;
2746    int newoptions;    int newoptions;
2747    int recno;    int recno;
2748      int refsign;
2749    int skipbytes;    int skipbytes;
2750    int subreqbyte;    int subreqbyte;
2751    int subfirstbyte;    int subfirstbyte;
2752      int terminator;
2753    int mclength;    int mclength;
2754    uschar mcbuffer[8];    uschar mcbuffer[8];
2755    
2756    /* Next byte in the pattern */    /* Get next byte in the pattern */
2757    
2758    c = *ptr;    c = *ptr;
2759    
2760    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If we are in the pre-compile phase, accumulate the length used for the
2761      previous cycle of this loop. */
2762    
2763    if (inescq && c != 0)    if (lengthptr != NULL)
2764      {      {
2765      if (c == '\\' && ptr[1] == 'E')  #ifdef PCRE_DEBUG
2766        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2767    #endif
2768        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2769        {        {
2770        inescq = FALSE;        *errorcodeptr = ERR52;
2771        ptr++;        goto FAILED;
       continue;  
2772        }        }
2773      else  
2774        {      /* There is at least one situation where code goes backwards: this is the
2775        if (previous_callout != NULL)      case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2776          {      the class is simply eliminated. However, it is created first, so we have to
2777          complete_callout(previous_callout, ptr, cd);      allow memory for it. Therefore, don't ever reduce the length at this point.
2778          previous_callout = NULL;      */
2779    
2780        if (code < last_code) code = last_code;
2781    
2782        /* Paranoid check for integer overflow */
2783    
2784        if (OFLOW_MAX - *lengthptr < code - last_code)
2785          {
2786          *errorcodeptr = ERR20;
2787          goto FAILED;
2788          }
2789    
2790        *lengthptr += code - last_code;
2791        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2792    
2793        /* If "previous" is set and it is not at the start of the work space, move
2794        it back to there, in order to avoid filling up the work space. Otherwise,
2795        if "previous" is NULL, reset the current code pointer to the start. */
2796    
2797        if (previous != NULL)
2798          {
2799          if (previous > orig_code)
2800            {
2801            memmove(orig_code, previous, code - previous);
2802            code -= previous - orig_code;
2803            previous = orig_code;
2804            }
2805          }
2806        else code = orig_code;
2807    
2808        /* Remember where this code item starts so we can pick up the length
2809        next time round. */
2810    
2811        last_code = code;
2812        }
2813    
2814      /* In the real compile phase, just check the workspace used by the forward
2815      reference list. */
2816    
2817      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2818        {
2819        *errorcodeptr = ERR52;
2820        goto FAILED;
2821        }
2822    
2823      /* If in \Q...\E, check for the end; if not, we have a literal */
2824    
2825      if (inescq && c != 0)
2826        {
2827        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2828          {
2829          inescq = FALSE;
2830          ptr++;
2831          continue;
2832          }
2833        else
2834          {
2835          if (previous_callout != NULL)
2836            {
2837            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2838              complete_callout(previous_callout, ptr, cd);
2839            previous_callout = NULL;
2840          }          }
2841        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
2842          {          {
# Line 1638  for (;; ptr++) Line 2850  for (;; ptr++)
2850    /* 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
2851    a quantifier. */    a quantifier. */
2852    
2853    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2854      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2855        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2856    
2857    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2858         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2859      {      {
2860      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2861          complete_callout(previous_callout, ptr, cd);
2862      previous_callout = NULL;      previous_callout = NULL;
2863      }      }
2864    
# Line 1653  for (;; ptr++) Line 2867  for (;; ptr++)
2867    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2868      {      {
2869      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2870      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2871        {        {
2872        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2873        on the Macintosh. */          {
2874        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2875        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2876          if (*ptr != 0) continue;
2877    
2878          /* Else fall through to handle end of string */
2879          c = 0;
2880        }        }
2881      }      }
2882    
# Line 1672  for (;; ptr++) Line 2890  for (;; ptr++)
2890    
2891    switch(c)    switch(c)
2892      {      {
2893      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2894        case 0:                        /* The branch terminates at string end */
2895      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
2896      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
2897      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2898      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2899      *codeptr = code;      *codeptr = code;
2900      *ptrptr = ptr;      *ptrptr = ptr;
2901        if (lengthptr != NULL)
2902          {
2903          if (OFLOW_MAX - *lengthptr < code - last_code)
2904            {
2905            *errorcodeptr = ERR20;
2906            goto FAILED;
2907            }
2908          *lengthptr += code - last_code;   /* To include callout length */
2909          DPRINTF((">> end branch\n"));
2910          }
2911      return TRUE;      return TRUE;
2912    
2913    
2914        /* ===================================================================*/
2915      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2916      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2917    
2918      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2919      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2920        {        {
2921        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 2924  for (;; ptr++)
2924      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2925      break;      break;
2926    
2927      case '$':      case CHAR_DOLLAR_SIGN:
2928      previous = NULL;      previous = NULL;
2929      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2930      break;      break;
# Line 1703  for (;; ptr++) Line 2932  for (;; ptr++)
2932      /* 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
2933      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2934    
2935      case '.':      case CHAR_DOT:
2936      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2937      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2938      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2939      previous = code;      previous = code;
2940      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2941      break;      break;
2942    
2943      /* Character classes. If the included characters are all < 255 in value, we  
2944      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2945      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
2946      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
2947      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2948        map as usual, then invert it at the end. However, we use a different opcode
2949        so that data characters > 255 can be handled correctly.
2950    
2951      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2952      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,
2953      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2954      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.
     */  
2955    
2956      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2957        default (Perl) mode, it is treated as a data character. */
2958    
2959        case CHAR_RIGHT_SQUARE_BRACKET:
2960        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2961          {
2962          *errorcodeptr = ERR64;
2963          goto FAILED;
2964          }
2965        goto NORMAL_CHAR;
2966    
2967        case CHAR_LEFT_SQUARE_BRACKET:
2968      previous = code;      previous = code;
2969    
2970      /* 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
2971      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. */
2972    
2973      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2974          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2975            check_posix_syntax(ptr, &tempptr))
2976        {        {
2977        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2978        goto FAILED;        goto FAILED;
2979        }        }
2980    
2981      /* 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,
2982        if the first few characters (either before or after ^) are \Q\E or \E we
2983        skip them too. This makes for compatibility with Perl. */
2984    
2985      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2986        for (;;)
2987        {        {
       negate_class = TRUE;  
2988        c = *(++ptr);        c = *(++ptr);
2989          if (c == CHAR_BACKSLASH)
2990            {
2991            if (ptr[1] == CHAR_E)
2992              ptr++;
2993            else if (strncmp((const char *)ptr+1,
2994                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2995              ptr += 3;
2996            else
2997              break;
2998            }
2999          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3000            negate_class = TRUE;
3001          else break;
3002        }        }
3003      else  
3004        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3005        an initial ']' is taken as a data character -- the code below handles
3006        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3007        [^] must match any character, so generate OP_ALLANY. */
3008    
3009        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3010            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3011        {        {
3012        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3013          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3014          zerofirstbyte = firstbyte;
3015          break;
3016        }        }
3017    
3018        /* If a class contains a negative special such as \S, we need to flip the
3019        negation flag at the end, so that support for characters > 255 works
3020        correctly (they are all included in the class). */
3021    
3022        should_flip_negation = FALSE;
3023    
3024      /* 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
3025      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
3026      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3027    
3028      class_charcount = 0;      class_charcount = 0;
3029      class_lastchar = -1;      class_lastchar = -1;
3030    
3031        /* Initialize the 32-char bit map to all zeros. We build the map in a
3032        temporary bit of memory, in case the class contains only 1 character (less
3033        than 256), because in that case the compiled code doesn't use the bit map.
3034        */
3035    
3036        memset(classbits, 0, 32 * sizeof(uschar));
3037    
3038  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3039      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3040      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3041        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3042  #endif  #endif
3043    
     /* 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));  
   
3044      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3045      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
3046      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. */  
3047    
3048      do      if (c != 0) do
3049        {        {
3050          const uschar *oldptr;
3051    
3052  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3053        if (utf8 && c > 127)        if (utf8 && c > 127)
3054          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3055          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3056          }          }
3057    
3058          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3059          data and reset the pointer. This is so that very large classes that
3060          contain a zillion UTF-8 characters no longer overwrite the work space
3061          (which is on the stack). */
3062    
3063          if (lengthptr != NULL)
3064            {
3065            *lengthptr += class_utf8data - class_utf8data_base;
3066            class_utf8data = class_utf8data_base;
3067            }
3068    
3069  #endif  #endif
3070    
3071        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3072    
3073        if (inescq)        if (inescq)
3074          {          {
3075          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3076            {            {
3077            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3078            ptr++;            ptr++;                            /* Skip the 'E' */
3079            continue;            continue;                         /* Carry on with next */
3080            }            }
3081          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3082          }          }
3083    
3084        /* 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 3087  for (;; ptr++)
3087        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3088        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3089    
3090        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3091            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3092            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3093          {          {
3094          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3095          int posix_class, i;          int posix_class, taboffset, tabopt;
3096          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3097            uschar pbits[32];
3098    
3099          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3100            {            {
3101            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3102            goto FAILED;            goto FAILED;
3103            }            }
3104    
3105          ptr += 2;          ptr += 2;
3106          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3107            {            {
3108            local_negate = TRUE;            local_negate = TRUE;
3109              should_flip_negation = TRUE;  /* Note negative special */
3110            ptr++;            ptr++;
3111            }            }
3112    
# Line 1836  for (;; ptr++) Line 3124  for (;; ptr++)
3124          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3125            posix_class = 0;            posix_class = 0;
3126    
3127          /* 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
3128          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
3129          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
3130          white space chars afterwards. */          result into the bit map that is being built. */
3131    
3132          posix_class *= 3;          posix_class *= 3;
3133          for (i = 0; i < 3; i++)  
3134            /* Copy in the first table (always present) */
3135    
3136            memcpy(pbits, cbits + posix_class_maps[posix_class],
3137              32 * sizeof(uschar));
3138    
3139            /* If there is a second table, add or remove it as required. */
3140    
3141            taboffset = posix_class_maps[posix_class + 1];
3142            tabopt = posix_class_maps[posix_class + 2];
3143    
3144            if (taboffset >= 0)
3145            {            {
3146            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3147            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;  
             }  
3148            else            else
3149              {              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;  
             }  
3150            }            }
3151    
3152            /* Not see if we need to remove any special characters. An option
3153            value of 1 removes vertical space and 2 removes underscore. */
3154    
3155            if (tabopt < 0) tabopt = -tabopt;
3156            if (tabopt == 1) pbits[1] &= ~0x3c;
3157              else if (tabopt == 2) pbits[11] &= 0x7f;
3158    
3159            /* Add the POSIX table or its complement into the main table that is
3160            being built and we are done. */
3161    
3162            if (local_negate)
3163              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3164            else
3165              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3166    
3167          ptr = tempptr + 1;          ptr = tempptr + 1;
3168          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3169          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3170          }          }
3171    
3172        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3173        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
3174        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.
3175        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3176        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  
3177        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3178    
3179        if (c == '\\')        if (c == CHAR_BACKSLASH)
3180          {          {
3181          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3182            if (*errorcodeptr != 0) goto FAILED;
3183    
3184          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 */
3185          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 */
3186            else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3187          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3188            {            {
3189            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3190              {              {
3191              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3192              }              }
3193            else inescq = TRUE;            else inescq = TRUE;
3194            continue;            continue;
3195            }            }
3196            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3197    
3198          if (c < 0)          if (c < 0)
3199            {            {
3200            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3201            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3202            switch (-c)  
3203              /* Save time by not doing this in the pre-compile phase. */
3204    
3205              if (lengthptr == NULL) switch (-c)
3206              {              {
3207              case ESC_d:              case ESC_d:
3208              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3209              continue;              continue;
3210    
3211              case ESC_D:              case ESC_D:
3212                should_flip_negation = TRUE;
3213              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3214              continue;              continue;
3215    
# Line 1910  for (;; ptr++) Line 3218  for (;; ptr++)
3218              continue;              continue;
3219    
3220              case ESC_W:              case ESC_W:
3221                should_flip_negation = TRUE;
3222              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3223              continue;              continue;
3224    
# Line 1919  for (;; ptr++) Line 3228  for (;; ptr++)
3228              continue;              continue;
3229    
3230              case ESC_S:              case ESC_S:
3231                should_flip_negation = TRUE;
3232              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3233              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3234              continue;              continue;
3235    
3236  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
3237              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
3238              case ESC_P:              }
3239    
3240              /* In the pre-compile phase, just do the recognition. */
3241    
3242              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3243                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3244    
3245              /* We need to deal with \H, \h, \V, and \v in both phases because
3246              they use extra memory. */
3247    
3248              if (-c == ESC_h)
3249                {
3250                SETBIT(classbits, 0x09); /* VT */
3251                SETBIT(classbits, 0x20); /* SPACE */
3252                SETBIT(classbits, 0xa0); /* NSBP */
3253    #ifdef SUPPORT_UTF8
3254                if (utf8)
3255                  {
3256                  class_utf8 = TRUE;
3257                  *class_utf8data++ = XCL_SINGLE;
3258                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3259                  *class_utf8data++ = XCL_SINGLE;
3260                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3261                  *class_utf8data++ = XCL_RANGE;
3262                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3263                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3264                  *class_utf8data++ = XCL_SINGLE;
3265                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3266                  *class_utf8data++ = XCL_SINGLE;
3267                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3268                  *class_utf8data++ = XCL_SINGLE;
3269                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3270                  }
3271    #endif
3272                continue;
3273                }
3274    
3275              if (-c == ESC_H)
3276                {
3277                for (c = 0; c < 32; c++)
3278                  {
3279                  int x = 0xff;
3280                  switch (c)
3281                    {
3282                    case 0x09/8: x ^= 1 << (0x09%8); break;
3283                    case 0x20/8: x ^= 1 << (0x20%8); break;
3284                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3285                    default: break;
3286                    }
3287                  classbits[c] |= x;
3288                  }
3289    
3290    #ifdef SUPPORT_UTF8
3291                if (utf8)
3292                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
3293                class_utf8 = TRUE;                class_utf8 = TRUE;
3294                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_RANGE;
3295                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3296                *class_utf8data++ = property;                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3297                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
3298                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3299                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3300                  *class_utf8data++ = XCL_RANGE;
3301                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3302                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3303                  *class_utf8data++ = XCL_RANGE;
3304                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3305                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3306                  *class_utf8data++ = XCL_RANGE;
3307                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3308                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3309                  *class_utf8data++ = XCL_RANGE;
3310                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3311                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3312                  *class_utf8data++ = XCL_RANGE;
3313                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3314                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3315                }                }
3316    #endif
3317              continue;              continue;
3318                }
3319    
3320              if (-c == ESC_v)
3321                {
3322                SETBIT(classbits, 0x0a); /* LF */
3323                SETBIT(classbits, 0x0b); /* VT */
3324                SETBIT(classbits, 0x0c); /* FF */
3325                SETBIT(classbits, 0x0d); /* CR */
3326                SETBIT(classbits, 0x85); /* NEL */
3327    #ifdef SUPPORT_UTF8
3328                if (utf8)
3329                  {
3330                  class_utf8 = TRUE;
3331                  *class_utf8data++ = XCL_RANGE;
3332                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3333                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3334                  }
3335  #endif  #endif
3336                continue;
3337                }
3338    
3339              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_V)
3340              strict mode. By default, for compatibility with Perl, they are              {
3341              treated as literals. */              for (c = 0; c < 32; c++)
3342                  {
3343                  int x = 0xff;
3344                  switch (c)
3345                    {
3346                    case 0x0a/8: x ^= 1 << (0x0a%8);
3347                                 x ^= 1 << (0x0b%8);
3348                                 x ^= 1 << (0x0c%8);
3349                                 x ^= 1 << (0x0d%8);
3350                                 break;
3351                    case 0x85/8: x ^= 1 << (0x85%8); break;
3352                    default: break;
3353                    }
3354                  classbits[c] |= x;
3355                  }
3356    
3357              default:  #ifdef SUPPORT_UTF8
3358              if ((options & PCRE_EXTRA) != 0)              if (utf8)
3359                {                {
3360                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3361                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3362                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3363                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3364                  *class_utf8data++ = XCL_RANGE;
3365                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3366                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3367                }                }
3368              c = *ptr;              /* The final character */  #endif
3369              class_charcount -= 2;  /* Undo the default count from above */              continue;
3370                }
3371    
3372              /* We need to deal with \P and \p in both phases. */
3373    
3374    #ifdef SUPPORT_UCP
3375              if (-c == ESC_p || -c == ESC_P)
3376                {
3377                BOOL negated;
3378                int pdata;
3379                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3380                if (ptype < 0) goto FAILED;
3381                class_utf8 = TRUE;
3382                *class_utf8data++ = ((-c == ESC_p) != negated)?
3383                  XCL_PROP : XCL_NOTPROP;
3384                *class_utf8data++ = ptype;
3385                *class_utf8data++ = pdata;
3386                class_charcount -= 2;   /* Not a < 256 character */
3387                continue;
3388                }
3389    #endif
3390              /* Unrecognized escapes are faulted if PCRE is running in its
3391              strict mode. By default, for compatibility with Perl, they are
3392              treated as literals. */
3393    
3394              if ((options & PCRE_EXTRA) != 0)
3395                {
3396                *errorcodeptr = ERR7;
3397                goto FAILED;
3398              }              }
3399    
3400              class_charcount -= 2;  /* Undo the default count from above */
3401              c = *ptr;              /* Get the final character and fall through */
3402            }            }
3403    
3404          /* 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
3405          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3406    
3407          }   /* End of backslash handling */          }   /* End of backslash handling */
3408    
3409        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3410        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
3411        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3412          entirely. The code for handling \Q and \E is messy. */
3413    
3414        if (ptr[1] == '-' && ptr[2] != ']')        CHECK_RANGE:
3415          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3416            {
3417            inescq = FALSE;
3418            ptr += 2;
3419            }
3420    
3421          oldptr = ptr;
3422    
3423          /* Remember \r or \n */
3424    
3425          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3426    
3427          /* Check for range */
3428    
3429          if (!inescq && ptr[1] == CHAR_MINUS)
3430          {          {
3431          int d;          int d;
3432          ptr += 2;          ptr += 2;
3433            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3434    
3435            /* If we hit \Q (not followed by \E) at this point, go into escaped
3436            mode. */
3437    
3438            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3439              {
3440              ptr += 2;
3441              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3442                { ptr += 2; continue; }
3443              inescq = TRUE;
3444              break;
3445              }
3446    
3447            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3448              {
3449              ptr = oldptr;
3450              goto LONE_SINGLE_CHARACTER;
3451              }
3452    
3453  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3454          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3463  for (;; ptr++)
3463          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
3464          in such circumstances. */          in such circumstances. */
3465    
3466          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3467            {            {
3468            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3469            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3470    
3471            /* \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
3472            was literal */            special means the '-' was literal */
3473    
3474            if (d < 0)            if (d < 0)
3475              {              {
3476              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3477              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = CHAR_X;
3478                else if (d == -ESC_R) d = CHAR_R; else
3479                {                {
3480                ptr = oldptr - 2;                ptr = oldptr;
3481                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3482                }                }
3483              }              }
3484            }            }
3485    
3486          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3487          the pre-pass. Optimize one-character ranges */          one-character ranges */
3488    
3489            if (d < c)
3490              {
3491              *errorcodeptr = ERR8;
3492              goto FAILED;
3493              }
3494    
3495          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3496    
3497            /* Remember \r or \n */
3498    
3499            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3500    
3501          /* 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
3502          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3503          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 3515  for (;; ptr++)
3515  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3516            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3517              {              {
3518              int occ, ocd;              unsigned int occ, ocd;
3519              int cc = c;              unsigned int cc = c;
3520              int origd = d;              unsigned int origd = d;
3521              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3522                {                {
3523                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3524                      ocd <= (unsigned int)d)
3525                    continue;                          /* Skip embedded ranges */
3526    
3527                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3528                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3529                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3530                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3531                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3532                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3533                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3534                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3535                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3536                  d = ocd;                  d = ocd;
3537                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3579  for (;; ptr++)
3579          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
3580          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3581    
3582          for (; c <= d; c++)          class_charcount += d - c + 1;
3583            class_lastchar = d;
3584    
3585            /* We can save a bit of time by skipping this in the pre-compile. */
3586    
3587            if (lengthptr == NULL) for (; c <= d; c++)
3588            {            {
3589            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3590            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3592  for (;; ptr++)
3592              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3593              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3594              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3595            }            }
3596    
3597          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3615  for (;; ptr++)
3615  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3616          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3617            {            {
3618            int chartype;            unsigned int othercase;
3619            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3620              {              {
3621              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3622              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3641  for (;; ptr++)
3641          }          }
3642        }        }
3643    
3644      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3645      loop. This "while" is the end of the "do" above. */  
3646        while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3647    
3648        if (c == 0)                          /* Missing terminating ']' */
3649          {
3650          *errorcodeptr = ERR6;
3651          goto FAILED;
3652          }
3653    
3654    
3655    /* This code has been disabled because it would mean that \s counts as
3656    an explicit \r or \n reference, and that's not really what is wanted. Now
3657    we set the flag only if there is a literal "\r" or "\n" in the class. */
3658    
3659    #if 0
3660        /* Remember whether \r or \n are in this class */
3661    
3662        if (negate_class)
3663          {
3664          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3665          }
3666        else
3667          {
3668          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3669          }
3670    #endif
3671    
     while ((c = *(++ptr)) != ']' || inescq);  
3672    
3673      /* 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
3674      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
3675      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
3676      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3677      single-bytes only. This is an historical hangover. Maybe one day we can  
3678      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3679        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3680        operate on single-bytes only. This is an historical hangover. Maybe one day
3681        we can tidy these opcodes to handle multi-byte characters.
3682    
3683      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
3684      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 3688  for (;; ptr++)
3688      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3689    
3690  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3691      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3692            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3693  #else  #else
3694      if (class_charcount == 1)      if (class_charcount == 1)
3695  #endif  #endif
# Line 2209  for (;; ptr++) Line 3732  for (;; ptr++)
3732      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3733    
3734      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3735      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3736      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3737        the class, so any that were explicitly given as well can be ignored. If
3738        (when there are explicit characters > 255 that must be listed) there are no
3739        characters < 256, we can omit the bitmap in the actual compiled code. */
3740    
3741  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3742      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3743        {        {
3744        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3745        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3746        code += LINK_SIZE;        code += LINK_SIZE;
3747        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3748    
3749        /* 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;
3750        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3751    
3752        if (class_charcount > 0)        if (class_charcount > 0)
3753          {          {
3754          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3755            memmove(code + 32, code, class_utf8data - code);
3756          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3757          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;  
3758          }          }
3759          else code = class_utf8data;
3760    
3761        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3762    
# Line 2246  for (;; ptr++) Line 3765  for (;; ptr++)
3765        }        }
3766  #endif  #endif
3767    
3768      /* 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
3769      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
3770      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
3771      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3772    
3773        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3774      if (negate_class)      if (negate_class)
3775        {        {
3776        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3777        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3778        }        }
3779      else      else
3780        {        {
       *code++ = OP_CLASS;  
3781        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3782        }        }
3783      code += 32;      code += 32;
3784      break;      break;
3785    
3786    
3787        /* ===================================================================*/
3788      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3789      has been tested above. */      has been tested above. */
3790    
3791      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3792      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3793      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3794      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3795      goto REPEAT;      goto REPEAT;
3796    
3797      case '*':      case CHAR_ASTERISK:
3798      repeat_min = 0;      repeat_min = 0;
3799      repeat_max = -1;      repeat_max = -1;
3800      goto REPEAT;      goto REPEAT;
3801    
3802      case '+':      case CHAR_PLUS:
3803      repeat_min = 1;      repeat_min = 1;
3804      repeat_max = -1;      repeat_max = -1;
3805      goto REPEAT;      goto REPEAT;
3806    
3807      case '?':      case CHAR_QUESTION_MARK:
3808      repeat_min = 0;      repeat_min = 0;
3809      repeat_max = 1;      repeat_max = 1;
3810    
# Line 2318  for (;; ptr++) Line 3839  for (;; ptr++)
3839      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
3840      repeat type to the non-default. */      repeat type to the non-default. */
3841    
3842      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3843        {        {
3844        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3845        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3846        ptr++;        ptr++;
3847        }        }
3848      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3849        {        {
3850        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3851        ptr++;        ptr++;
3852        }        }
3853      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3854    
     /* 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;  
       }  
   
3855      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3856      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
3857      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 3885  for (;; ptr++)
3885          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3886          }          }
3887    
3888          /* If the repetition is unlimited, it pays to see if the next thing on
3889          the line is something that cannot possibly match this character. If so,
3890          automatically possessifying this item gains some performance in the case
3891          where the match fails. */
3892    
3893          if (!possessive_quantifier &&
3894              repeat_max < 0 &&
3895              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3896                options, cd))
3897            {
3898            repeat_type = 0;    /* Force greedy */
3899            possessive_quantifier = TRUE;
3900            }
3901    
3902        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3903        }        }
3904    
3905      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3906      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-
3907      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3908      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3909        currently used only for single-byte chars. */
3910    
3911      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3912        {        {
3913        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3914        c = previous[1];        c = previous[1];
3915          if (!possessive_quantifier &&
3916              repeat_max < 0 &&
3917              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3918            {
3919            repeat_type = 0;    /* Force greedy */
3920            possessive_quantifier = TRUE;
3921            }
3922        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3923        }        }
3924    
# Line 2403  for (;; ptr++) Line 3932  for (;; ptr++)
3932      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3933        {        {
3934        uschar *oldcode;        uschar *oldcode;
3935        int prop_type;        int prop_type, prop_value;
3936        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3937        c = *previous;        c = *previous;
3938    
3939          if (!possessive_quantifier &&
3940              repeat_max < 0 &&
3941              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3942            {
3943            repeat_type = 0;    /* Force greedy */
3944            possessive_quantifier = TRUE;
3945            }
3946    
3947        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3948        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3949          previous[1] : -1;          {
3950            prop_type = previous[1];
3951            prop_value = previous[2];
3952            }
3953          else prop_type = prop_value = -1;
3954    
3955        oldcode = code;        oldcode = code;
3956        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2419  for (;; ptr++) Line 3960  for (;; ptr++)
3960    
3961        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3962    
3963          /*--------------------------------------------------------------------*/
3964          /* This code is obsolete from release 8.00; the restriction was finally
3965          removed: */
3966    
3967        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe