/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

code/trunk/pcre_compile.c revision 236 by ph10, Tue Sep 11 12:57:06 2007 UTC code/branches/pcre16/pcre_compile.c revision 805 by ph10, Wed Dec 14 16:49:20 2011 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) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 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 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  #include "pcre_printint.c"
63  #endif  #endif
64    
65    
# Line 87  so this number is very generous. Line 89  so this number is very generous.
89  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
90  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
91  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
92  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
93    filled up by repetitions of forward references, for example patterns like
94    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
95    that the workspace is expanded using malloc() in this situation. The value
96    below is therefore a minimum, and we put a maximum on it for safety. The
97    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
98    kicks in at the same number of forward references in all cases. */
99    
100  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
101    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
102    
103    /* The overrun tests check for a slightly smaller size so that they detect the
104    overrun before it actually does run off the end of the data block. */
105    
106    #define WORK_SIZE_SAFETY_MARGIN (100)
107    
108    /* Private flags added to firstchar and reqchar. */
109    
110    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
111    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
112    
113    /* Repeated character flags. */
114    
115    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
116    
117  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
118  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
119  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
120  is invalid. */  is invalid. */
121    
122  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
123    
124    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
125    in UTF-8 mode. */
126    
127  static const short int escapes[] = {  static const short int escapes[] = {
128       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
129       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
130     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
131  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
132  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
133  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
134     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
135  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
136  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
137       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
138         -ESC_D,                  -ESC_E,
139         0,                       -ESC_G,
140         -ESC_H,                  0,
141         0,                       -ESC_K,
142         0,                       0,
143         -ESC_N,                  0,
144         -ESC_P,                  -ESC_Q,
145         -ESC_R,                  -ESC_S,
146         0,                       0,
147         -ESC_V,                  -ESC_W,
148         -ESC_X,                  0,
149         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
150         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
151         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
152         CHAR_GRAVE_ACCENT,       7,
153         -ESC_b,                  0,
154         -ESC_d,                  ESC_e,
155         ESC_f,                   0,
156         -ESC_h,                  0,
157         0,                       -ESC_k,
158         0,                       0,
159         ESC_n,                   0,
160         -ESC_p,                  0,
161         ESC_r,                   -ESC_s,
162         ESC_tee,                 0,
163         -ESC_v,                  -ESC_w,
164         0,                       0,
165         -ESC_z
166  };  };
167    
168  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
169    
170    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
171    
172  static const short int escapes[] = {  static const short int escapes[] = {
173  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
174  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 187  static const short int escapes[] = {
187  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
188  /*  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,
189  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
190  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
191  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
192  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
193  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 197  static const short int escapes[] = {
197  #endif  #endif
198    
199    
200  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
201    searched linearly. Put all the names into a single string, in order to reduce
202    the number of relocations when a shared library is dynamically linked. The
203    string is built from string macros so that it works in UTF-8 mode on EBCDIC
204    platforms. */
205    
206  typedef struct verbitem {  typedef struct verbitem {
207    const char *name;    int   len;                 /* Length of verb name */
208    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
209    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
210  } verbitem;  } verbitem;
211    
212  static verbitem verbs[] = {  static const char verbnames[] =
213    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
214    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
215    { "F",      1, OP_FAIL },    STRING_ACCEPT0
216    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
217    { "PRUNE",  5, OP_PRUNE },    STRING_F0
218    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
219    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
220      STRING_SKIP0
221      STRING_THEN;
222    
223    static const verbitem verbs[] = {
224      { 0, -1,        OP_MARK },
225      { 4, -1,        OP_MARK },
226      { 6, OP_ACCEPT, -1 },
227      { 6, OP_COMMIT, -1 },
228      { 1, OP_FAIL,   -1 },
229      { 4, OP_FAIL,   -1 },
230      { 5, OP_PRUNE,  OP_PRUNE_ARG },
231      { 4, OP_SKIP,   OP_SKIP_ARG  },
232      { 4, OP_THEN,   OP_THEN_ARG  }
233  };  };
234    
235  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
236    
237    
238  /* Tables of names of POSIX character classes and their lengths. The list is  /* Tables of names of POSIX character classes and their lengths. The names are
239  terminated by a zero length entry. The first three must be alpha, lower, upper,  now all in a single string, to reduce the number of relocations when a shared
240  as this is assumed for handling case independence. */  library is dynamically loaded. The list of lengths is terminated by a zero
241    length entry. The first three must be alpha, lower, upper, as this is assumed
242    for handling case independence. */
243    
244    static const char posix_names[] =
245      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
246      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
247      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
248      STRING_word0  STRING_xdigit;
249    
250  static const char *const posix_names[] = {  static const pcre_uint8 posix_name_lengths[] = {
   "alpha", "lower", "upper",  
   "alnum", "ascii", "blank", "cntrl", "digit", "graph",  
   "print", "punct", "space", "word",  "xdigit" };  
   
 static const uschar posix_name_lengths[] = {  
251    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 };
252    
253  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 200  static const int posix_class_maps[] = { Line 277  static const int posix_class_maps[] = {
277    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
278  };  };
279    
280    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
281    substitutes must be in the order of the names, defined above, and there are
282    both positive and negative cases. NULL means no substitute. */
283    
284    #ifdef SUPPORT_UCP
285    static const pcre_uchar string_PNd[]  = {
286      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
287      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
288    static const pcre_uchar string_pNd[]  = {
289      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
290      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    static const pcre_uchar string_PXsp[] = {
292      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
293      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    static const pcre_uchar string_pXsp[] = {
295      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
296      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
297    static const pcre_uchar string_PXwd[] = {
298      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_pXwd[] = {
301      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    
304    static const pcre_uchar *substitutes[] = {
305      string_PNd,           /* \D */
306      string_pNd,           /* \d */
307      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
308      string_pXsp,          /* \s */
309      string_PXwd,          /* \W */
310      string_pXwd           /* \w */
311    };
312    
313    static const pcre_uchar string_pL[] =   {
314      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
315      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
316    static const pcre_uchar string_pLl[] =  {
317      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pLu[] =  {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_pXan[] = {
323      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_h[] =    {
326      CHAR_BACKSLASH, CHAR_h, '\0' };
327    static const pcre_uchar string_pXps[] = {
328      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_PL[] =   {
331      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_PLl[] =  {
334      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_PLu[] =  {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_PXan[] = {
340      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
341      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
342    static const pcre_uchar string_H[] =    {
343      CHAR_BACKSLASH, CHAR_H, '\0' };
344    static const pcre_uchar string_PXps[] = {
345      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
346      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347    
348    static const pcre_uchar *posix_substitutes[] = {
349      string_pL,            /* alpha */
350      string_pLl,           /* lower */
351      string_pLu,           /* upper */
352      string_pXan,          /* alnum */
353      NULL,                 /* ascii */
354      string_h,             /* blank */
355      NULL,                 /* cntrl */
356      string_pNd,           /* digit */
357      NULL,                 /* graph */
358      NULL,                 /* print */
359      NULL,                 /* punct */
360      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
361      string_pXwd,          /* word */
362      NULL,                 /* xdigit */
363      /* Negated cases */
364      string_PL,            /* ^alpha */
365      string_PLl,           /* ^lower */
366      string_PLu,           /* ^upper */
367      string_PXan,          /* ^alnum */
368      NULL,                 /* ^ascii */
369      string_H,             /* ^blank */
370      NULL,                 /* ^cntrl */
371      string_PNd,           /* ^digit */
372      NULL,                 /* ^graph */
373      NULL,                 /* ^print */
374      NULL,                 /* ^punct */
375      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
376      string_PXwd,          /* ^word */
377      NULL                  /* ^xdigit */
378    };
379    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
380    #endif
381    
382  #define STRING(a)  # a  #define STRING(a)  # a
383  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 207  static const int posix_class_maps[] = { Line 385  static const int posix_class_maps[] = {
385  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
386  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
387  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
388  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
389    the number of relocations needed when a shared library is loaded dynamically,
390  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
391    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
392    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
393    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
394    "unrecognized character follows \\",  
395    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
396    substring, so that the whole string ends with \0\0, which can be detected when
397    counting through. */
398    
399    static const char error_texts[] =
400      "no error\0"
401      "\\ at end of pattern\0"
402      "\\c at end of pattern\0"
403      "unrecognized character follows \\\0"
404      "numbers out of order in {} quantifier\0"
405    /* 5 */    /* 5 */
406    "number too big in {} quantifier",    "number too big in {} quantifier\0"
407    "missing terminating ] for character class",    "missing terminating ] for character class\0"
408    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
409    "range out of order in character class",    "range out of order in character class\0"
410    "nothing to repeat",    "nothing to repeat\0"
411    /* 10 */    /* 10 */
412    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
413    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
414    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
415    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
416    "missing )",    "missing )\0"
417    /* 15 */    /* 15 */
418    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
419    "erroffset passed as NULL",    "erroffset passed as NULL\0"
420    "unknown option bit(s) set",    "unknown option bit(s) set\0"
421    "missing ) after comment",    "missing ) after comment\0"
422    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
423    /* 20 */    /* 20 */
424    "regular expression is too large",    "regular expression is too large\0"
425    "failed to get memory",    "failed to get memory\0"
426    "unmatched parentheses",    "unmatched parentheses\0"
427    "internal error: code overflow",    "internal error: code overflow\0"
428    "unrecognized character after (?<",    "unrecognized character after (?<\0"
429    /* 25 */    /* 25 */
430    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
431    "malformed number or name after (?(",    "malformed number or name after (?(\0"
432    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
433    "assertion expected after (?(",    "assertion expected after (?(\0"
434    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
435    /* 30 */    /* 30 */
436    "unknown POSIX class name",    "unknown POSIX class name\0"
437    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
438    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
439    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
440    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
441    /* 35 */    /* 35 */
442    "invalid condition (?(0)",    "invalid condition (?(0)\0"
443    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
444    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
445    "number after (?C is > 255",    "number after (?C is > 255\0"
446    "closing ) for (?C expected",    "closing ) for (?C expected\0"
447    /* 40 */    /* 40 */
448    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
449    "unrecognized character after (?P",    "unrecognized character after (?P\0"
450    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
451    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
452    "invalid UTF-8 string",    "invalid UTF-8 string\0"
453    /* 45 */    /* 45 */
454    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
455    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
456    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
457    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
458    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
459    /* 50 */    /* 50 */
460    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
461    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
462    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
463    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
464    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
465    /* 55 */    /* 55 */
466    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
467    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
468    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
469    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
470    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
471    /* 60 */    /* 60 */
472    "(*VERB) not recognized",    "(*VERB) not recognized\0"
473    "number is too big"    "number is too big\0"
474  };    "subpattern name expected\0"
475      "digit expected after (?+\0"
476      "] is an invalid data character in JavaScript compatibility mode\0"
477      /* 65 */
478      "different names for subpatterns of the same number are not allowed\0"
479      "(*MARK) must have an argument\0"
480      "this version of PCRE is not compiled with PCRE_UCP support\0"
481      "\\c must be followed by an ASCII character\0"
482      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
483      /* 70 */
484      "internal error: unknown opcode in find_fixedlength()\0"
485      "\\N is not supported in a class\0"
486      "too many forward references\0"
487      "disallowed UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
488      ;
489    
490  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
491  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 303  For convenience, we use the same bit def Line 503  For convenience, we use the same bit def
503    
504  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
505    
506  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
507  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
508    into a subtraction and unsigned comparison). */
509    
510    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
511    
512    #ifndef EBCDIC
513    
514    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
515    UTF-8 mode. */
516    
517    static const pcre_uint8 digitab[] =
518    {    {
519    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
520    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 339  static const unsigned char digitab[] = Line 549  static const unsigned char digitab[] =
549    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
550    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
551    
552  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
553  static const unsigned char digitab[] =  
554    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
555    
556    static const pcre_uint8 digitab[] =
557    {    {
558    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
559    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 375  static const unsigned char digitab[] = Line 588  static const unsigned char digitab[] =
588    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
589    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
590    
591  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
592    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
593    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
594    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 414  static const unsigned char ebcdic_charta Line 627  static const unsigned char ebcdic_charta
627  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
628    
629  static BOOL  static BOOL
630    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
631      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
632    
633    
634    
635  /*************************************************  /*************************************************
636    *            Find an error text                  *
637    *************************************************/
638    
639    /* The error texts are now all in one long string, to save on relocations. As
640    some of the text is of unknown length, we can't use a table of offsets.
641    Instead, just count through the strings. This is not a performance issue
642    because it happens only when there has been a compilation error.
643    
644    Argument:   the error number
645    Returns:    pointer to the error string
646    */
647    
648    static const char *
649    find_error_text(int n)
650    {
651    const char *s = error_texts;
652    for (; n > 0; n--)
653      {
654      while (*s++ != 0) {};
655      if (*s == 0) return "Error text not found (please report)";
656      }
657    return s;
658    }
659    
660    
661    /*************************************************
662    *           Expand the workspace                 *
663    *************************************************/
664    
665    /* This function is called during the second compiling phase, if the number of
666    forward references fills the existing workspace, which is originally a block on
667    the stack. A larger block is obtained from malloc() unless the ultimate limit
668    has been reached or the increase will be rather small.
669    
670    Argument: pointer to the compile data block
671    Returns:  0 if all went well, else an error number
672    */
673    
674    static int
675    expand_workspace(compile_data *cd)
676    {
677    pcre_uchar *newspace;
678    int newsize = cd->workspace_size * 2;
679    
680    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
681    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
682        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
683     return ERR72;
684    
685    newspace = (PUBL(malloc))(newsize);
686    if (newspace == NULL) return ERR21;
687    
688    memcpy(newspace, cd->start_workspace, cd->workspace_size);
689    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
690    if (cd->workspace_size > COMPILE_WORK_SIZE)
691      (PUBL(free))((void *)cd->start_workspace);
692    cd->start_workspace = newspace;
693    cd->workspace_size = newsize;
694    return 0;
695    }
696    
697    
698    
699    /*************************************************
700    *            Check for counted repeat            *
701    *************************************************/
702    
703    /* This function is called when a '{' is encountered in a place where it might
704    start a quantifier. It looks ahead to see if it really is a quantifier or not.
705    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
706    where the ddds are digits.
707    
708    Arguments:
709      p         pointer to the first char after '{'
710    
711    Returns:    TRUE or FALSE
712    */
713    
714    static BOOL
715    is_counted_repeat(const pcre_uchar *p)
716    {
717    if (!IS_DIGIT(*p)) return FALSE;
718    p++;
719    while (IS_DIGIT(*p)) p++;
720    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
721    
722    if (*p++ != CHAR_COMMA) return FALSE;
723    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
724    
725    if (!IS_DIGIT(*p)) return FALSE;
726    p++;
727    while (IS_DIGIT(*p)) p++;
728    
729    return (*p == CHAR_RIGHT_CURLY_BRACKET);
730    }
731    
732    
733    
734    /*************************************************
735  *            Handle escapes                      *  *            Handle escapes                      *
736  *************************************************/  *************************************************/
737    
# Line 444  Returns:         zero or positive => a d Line 756  Returns:         zero or positive => a d
756  */  */
757    
758  static int  static int
759  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
760    int options, BOOL isclass)    int options, BOOL isclass)
761  {  {
762  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
763  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
764  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
765    pcre_int32 c;
766    int i;
767    
768  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
769  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 458  ptr--;                            /* Set Line 772  ptr--;                            /* Set
772    
773  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
774    
775  /* 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
776  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.
777  Otherwise further processing may be required. */  Otherwise further processing may be required. */
778    
779  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
780  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
781  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
782    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
783    
784  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
785  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
786    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
787  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
788  #endif  #endif
789    
# Line 475  else if ((i = escapes[c - 0x48]) != 0) Line 791  else if ((i = escapes[c - 0x48]) != 0)
791    
792  else  else
793    {    {
794    const uschar *oldptr;    const pcre_uchar *oldptr;
795    BOOL braced, negated;    BOOL braced, negated;
796    
797    switch (c)    switch (c)
# Line 483  else Line 799  else
799      /* 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
800      error. */      error. */
801    
802      case 'l':      case CHAR_l:
803      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
804      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
805      break;      break;
806    
807      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
808      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
809      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
810      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
811      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
812      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
813            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
814      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
815      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
816        {          {
817        const uschar *p;          c = 0;
818        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
819          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
820        if (*p != 0 && *p != '}')            register int cc = *(++ptr);
821    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
822              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
823              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
824    #else           /* EBCDIC coding */
825              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
826              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
827    #endif
828              }
829            }
830          }
831        else
832          *errorcodeptr = ERR37;
833        break;
834    
835        case CHAR_U:
836        /* In JavaScript, \U is an uppercase U letter. */
837        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
838        break;
839    
840        /* In a character class, \g is just a literal "g". Outside a character
841        class, \g must be followed by one of a number of specific things:
842    
843        (1) A number, either plain or braced. If positive, it is an absolute
844        backreference. If negative, it is a relative backreference. This is a Perl
845        5.10 feature.
846    
847        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
848        is part of Perl's movement towards a unified syntax for back references. As
849        this is synonymous with \k{name}, we fudge it up by pretending it really
850        was \k.
851    
852        (3) For Oniguruma compatibility we also support \g followed by a name or a
853        number either in angle brackets or in single quotes. However, these are
854        (possibly recursive) subroutine calls, _not_ backreferences. Just return
855        the -ESC_g code (cf \k). */
856    
857        case CHAR_g:
858        if (isclass) break;
859        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
860          {
861          c = -ESC_g;
862          break;
863          }
864    
865        /* Handle the Perl-compatible cases */
866    
867        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
868          {
869          const pcre_uchar *p;
870          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
871            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
872          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
873          {          {
874          c = -ESC_k;          c = -ESC_k;
875          break;          break;
# Line 514  else Line 879  else
879        }        }
880      else braced = FALSE;      else braced = FALSE;
881    
882      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
883        {        {
884        negated = TRUE;        negated = TRUE;
885        ptr++;        ptr++;
886        }        }
887      else negated = FALSE;      else negated = FALSE;
888    
889        /* The integer range is limited by the machine's int representation. */
890      c = 0;      c = 0;
891      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
       c = c * 10 + *(++ptr) - '0';  
   
     if (c < 0)  
892        {        {
893          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
894            {
895            c = -1;
896            break;
897            }
898          c = c * 10 + *(++ptr) - CHAR_0;
899          }
900        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
901          {
902          while (IS_DIGIT(ptr[1]))
903            ptr++;
904        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
905        break;        break;
906        }        }
907    
908      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
909        {        {
910        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
911        break;        break;
912        }        }
913    
914        if (c == 0)
915          {
916          *errorcodeptr = ERR58;
917          break;
918          }
919    
920      if (negated)      if (negated)
921        {        {
922        if (c > bracount)        if (c > bracount)
# Line 562  else Line 942  else
942      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
943      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
944    
945      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:
946      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
947    
948      if (!isclass)      if (!isclass)
949        {        {
950        oldptr = ptr;        oldptr = ptr;
951        c -= '0';        /* The integer range is limited by the machine's int representation. */
952        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
953          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
954        if (c < 0)          {
955            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
956              {
957              c = -1;
958              break;
959              }
960            c = c * 10 + *(++ptr) - CHAR_0;
961            }
962          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
963          {          {
964            while (IS_DIGIT(ptr[1]))
965              ptr++;
966          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
967          break;          break;
968          }          }
# Line 588  else Line 978  else
978      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.
979      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
980    
981      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
982        {        {
983        ptr--;        ptr--;
984        c = 0;        c = 0;
# Line 601  else Line 991  else
991      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
992      than 3 octal digits. */      than 3 octal digits. */
993    
994      case '0':      case CHAR_0:
995      c -= '0';      c -= CHAR_0;
996      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
997          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
998      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
999      break;      break;
1000    
1001      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
1002      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1003      treated as a data character. */      If not, { is treated as a data character. */
1004    
1005        case CHAR_x:
1006        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1007          {
1008          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1009          Otherwise it is a lowercase x letter. */
1010          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1011            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1012            {
1013            c = 0;
1014            for (i = 0; i < 2; ++i)
1015              {
1016              register int cc = *(++ptr);
1017    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1018              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1019              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1020    #else           /* EBCDIC coding */
1021              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1022              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1023    #endif
1024              }
1025            }
1026          break;
1027          }
1028    
1029      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1030        {        {
1031        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1032    
1033        c = 0;        c = 0;
1034        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1035          {          {
1036          register int cc = *pt++;          register int cc = *pt++;
1037          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1038    
1039  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1040          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1041          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1042  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1043          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1044          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1045    #endif
1046    
1047    #ifdef COMPILE_PCRE8
1048            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1049    #else
1050    #ifdef COMPILE_PCRE16
1051            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1052    #endif
1053  #endif  #endif
1054          }          }
1055    
1056        if (*pt == '}')        if (c < 0)
1057          {          {
1058          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1059            *errorcodeptr = ERR34;
1060            }
1061    
1062          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1063            {
1064            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1065          ptr = pt;          ptr = pt;
1066          break;          break;
1067          }          }
# Line 648  else Line 1073  else
1073      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1074    
1075      c = 0;      c = 0;
1076      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1077        {        {
1078        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1079        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1080  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1081        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1082        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1083  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1084        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1085        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1086  #endif  #endif
1087        }        }
1088      break;      break;
1089    
1090      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1091      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1092        coding is ASCII-specific, but then the whole concept of \cx is
1093      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1094    
1095      case 'c':      case CHAR_c:
1096      c = *(++ptr);      c = *(++ptr);
1097      if (c == 0)      if (c == 0)
1098        {        {
1099        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1100        break;        break;
1101        }        }
1102    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1103  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1104      if (c >= 'a' && c <= 'z') c -= 32;        {
1105          *errorcodeptr = ERR68;
1106          break;
1107          }
1108        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1109      c ^= 0x40;      c ^= 0x40;
1110  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1111      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1112      c ^= 0xC0;      c ^= 0xC0;
1113  #endif  #endif
1114      break;      break;
1115    
1116      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1117      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1118      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
1119      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
1120      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1121    
1122      default:      default:
1123      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 1130  else
1130      }      }
1131    }    }
1132    
1133    /* Perl supports \N{name} for character names, as well as plain \N for "not
1134    newline". PCRE does not support \N{name}. However, it does support
1135    quantification such as \N{2,3}. */
1136    
1137    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1138         !is_counted_repeat(ptr+2))
1139      *errorcodeptr = ERR37;
1140    
1141    /* If PCRE_UCP is set, we change the values for \d etc. */
1142    
1143    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1144      c -= (ESC_DU - ESC_D);
1145    
1146    /* Set the pointer to the final character before returning. */
1147    
1148  *ptrptr = ptr;  *ptrptr = ptr;
1149  return c;  return c;
1150  }  }
# Line 726  Returns:         type value from ucp_typ Line 1171  Returns:         type value from ucp_typ
1171  */  */
1172    
1173  static int  static int
1174  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1175  {  {
1176  int c, i, bot, top;  int c, i, bot, top;
1177  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1178  char name[32];  pcre_uchar name[32];
1179    
1180  c = *(++ptr);  c = *(++ptr);
1181  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 740  if (c == 0) goto ERROR_RETURN; Line 1185  if (c == 0) goto ERROR_RETURN;
1185  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1186  negation. */  negation. */
1187    
1188  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1189    {    {
1190    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1191      {      {
1192      *negptr = TRUE;      *negptr = TRUE;
1193      ptr++;      ptr++;
# Line 751  if (c == '{') Line 1196  if (c == '{')
1196      {      {
1197      c = *(++ptr);      c = *(++ptr);
1198      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1199      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1200      name[i] = c;      name[i] = c;
1201      }      }
1202    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1203    name[i] = 0;    name[i] = 0;
1204    }    }
1205    
# Line 771  else Line 1216  else
1216  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1217    
1218  bot = 0;  bot = 0;
1219  top = _pcre_utt_size;  top = PRIV(utt_size);
1220    
1221  while (bot < top)  while (bot < top)
1222    {    {
1223    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1224    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1225    if (c == 0)    if (c == 0)
1226      {      {
1227      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1228      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1229      }      }
1230    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1231    }    }
# Line 800  return -1; Line 1245  return -1;
1245    
1246    
1247  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1248  *         Read repeat counts                     *  *         Read repeat counts                     *
1249  *************************************************/  *************************************************/
1250    
# Line 851  Returns:         pointer to '}' on succe Line 1263  Returns:         pointer to '}' on succe
1263                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1264  */  */
1265    
1266  static const uschar *  static const pcre_uchar *
1267  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1268  {  {
1269  int min = 0;  int min = 0;
1270  int max = -1;  int max = -1;
# Line 860  int max = -1; Line 1272  int max = -1;
1272  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1273  an integer overflow. */  an integer overflow. */
1274    
1275  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1276  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1277    {    {
1278    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1282  if (min < 0 || min > 65535)
1282  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1283  Also, max must not be less than min. */  Also, max must not be less than min. */
1284    
1285  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1286    {    {
1287    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1288      {      {
1289      max = 0;      max = 0;
1290      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1291      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1292        {        {
1293        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1312  return p;
1312    
1313    
1314  /*************************************************  /*************************************************
1315  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1316  *************************************************/  *************************************************/
1317    
1318  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1319    top-level call starts at the beginning of the pattern. All other calls must
1320    start at a parenthesis. It scans along a pattern's text looking for capturing
1321  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1322  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1323  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1324  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1325  be terminated by '>' because that is checked in the first pass.  
1326    This function was originally called only from the second pass, in which we know
1327    that if (?< or (?' or (?P< is encountered, the name will be correctly
1328    terminated because that is checked in the first pass. There is now one call to
1329    this function in the first pass, to check for a recursive back reference by
1330    name (so that we can make the whole group atomic). In this case, we need check
1331    only up to the current position in the pattern, and that is still OK because
1332    and previous occurrences will have been checked. To make this work, the test
1333    for "end of pattern" is a check against cd->end_pattern in the main loop,
1334    instead of looking for a binary zero. This means that the special first-pass
1335    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1336    processing items within the loop are OK, because afterwards the main loop will
1337    terminate.)
1338    
1339  Arguments:  Arguments:
1340    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1341    count        current count of capturing parens so far encountered    cd           compile background data
1342    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1343    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1344    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1345      utf          TRUE if we are in UTF-8 / UTF-16 mode
1346      count        pointer to the current capturing subpattern number (updated)
1347    
1348  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1349  */  */
1350    
1351  static int  static int
1352  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1353    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1354  {  {
1355  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1356    int start_count = *count;
1357    int hwm_count = start_count;
1358    BOOL dup_parens = FALSE;
1359    
1360    /* If the first character is a parenthesis, check on the type of group we are
1361    dealing with. The very first call may not start with a parenthesis. */
1362    
1363  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1364    {    {
1365    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1366    
1367      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1368    
1369      /* Handle a normal, unnamed capturing parenthesis. */
1370    
1371      else if (ptr[1] != CHAR_QUESTION_MARK)
1372        {
1373        *count += 1;
1374        if (name == NULL && *count == lorn) return *count;
1375        ptr++;
1376        }
1377    
1378      /* All cases now have (? at the start. Remember when we are in a group
1379      where the parenthesis numbers are duplicated. */
1380    
1381      else if (ptr[2] == CHAR_VERTICAL_LINE)
1382        {
1383        ptr += 3;
1384        dup_parens = TRUE;
1385        }
1386    
1387      /* Handle comments; all characters are allowed until a ket is reached. */
1388    
1389      else if (ptr[2] == CHAR_NUMBER_SIGN)
1390        {
1391        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1392        goto FAIL_EXIT;
1393        }
1394    
1395      /* Handle a condition. If it is an assertion, just carry on so that it
1396      is processed as normal. If not, skip to the closing parenthesis of the
1397      condition (there can't be any nested parens). */
1398    
1399      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1400        {
1401        ptr += 2;
1402        if (ptr[1] != CHAR_QUESTION_MARK)
1403          {
1404          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1405          if (*ptr != 0) ptr++;
1406          }
1407        }
1408    
1409      /* Start with (? but not a condition. */
1410    
1411      else
1412        {
1413        ptr += 2;
1414        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1415    
1416        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1417    
1418        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1419            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1420          {
1421          int term;
1422          const pcre_uchar *thisname;
1423          *count += 1;
1424          if (name == NULL && *count == lorn) return *count;
1425          term = *ptr++;
1426          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1427          thisname = ptr;
1428          while (*ptr != term) ptr++;
1429          if (name != NULL && lorn == ptr - thisname &&
1430              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1431            return *count;
1432          term++;
1433          }
1434        }
1435      }
1436    
1437    /* Past any initial parenthesis handling, scan for parentheses or vertical
1438    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1439    first-pass call when this value is temporarily adjusted to stop at the current
1440    position. So DO NOT change this to a test for binary zero. */
1441    
1442    for (; ptr < cd->end_pattern; ptr++)
1443      {
1444    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1445    
1446    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1447      {      {
1448      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1449      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1450        {        {
1451        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1452        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1453        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1454        }        }
1455      continue;      continue;
1456      }      }
1457    
1458    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1459      are handled for real. If the first character is '^', skip it. Also, if the
1460      first few characters (either before or after ^) are \Q\E or \E we skip them
1461      too. This makes for compatibility with Perl. Note the use of STR macros to
1462      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1463    
1464    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1465      {      {
1466      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1467        for (;;)
1468          {
1469          if (ptr[1] == CHAR_BACKSLASH)
1470            {
1471            if (ptr[2] == CHAR_E)
1472              ptr+= 2;
1473            else if (STRNCMP_UC_C8(ptr + 2,
1474                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1475              ptr += 4;
1476            else
1477              break;
1478            }
1479          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1480            {
1481            negate_class = TRUE;
1482            ptr++;
1483            }
1484          else break;
1485          }
1486    
1487        /* If the next character is ']', it is a data character that must be
1488        skipped, except in JavaScript compatibility mode. */
1489    
1490        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1491            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1492          ptr++;
1493    
1494        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1495        {        {
1496        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1497        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1498          {          {
1499          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1500          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1501            {            {
1502            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1503            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1504            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1505            }            }
1506          continue;          continue;
1507          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1511  for (; *ptr != 0; ptr++)
1511    
1512    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1513    
1514    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1515      {      {
1516      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1517      if (*ptr == 0) return -1;      while (*ptr != 0)
1518          {
1519          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1520          ptr++;
1521    #ifdef SUPPORT_UTF
1522          if (utf) FORWARDCHAR(ptr);
1523    #endif
1524          }
1525        if (*ptr == 0) goto FAIL_EXIT;
1526      continue;      continue;
1527      }      }
1528    
1529    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1530    
1531    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1532      {      {
1533      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1534      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1535      continue;      if (*ptr == 0) goto FAIL_EXIT;
1536      }      }
1537    
1538    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1539    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1540        if (dup_parens && *count < hwm_count) *count = hwm_count;
1541        goto FAIL_EXIT;
1542        }
1543    
1544      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1545        {
1546        if (*count > hwm_count) hwm_count = *count;
1547        *count = start_count;
1548        }
1549      }
1550    
1551    FAIL_EXIT:
1552    *ptrptr = ptr;
1553    return -1;
1554    }
1555    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1556    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1557    
   count++;  
1558    
1559    if (name == NULL && count == lorn) return count;  /*************************************************
1560    term = *ptr++;  *       Find forward referenced subpattern       *
1561    if (term == '<') term = '>';  *************************************************/
1562    thisname = ptr;  
1563    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1564    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1565        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1566      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1567    references to subpatterns. We used to be able to start this scan from the
1568    current compiling point, using the current count value from cd->bracount, and
1569    do it all in a single loop, but the addition of the possibility of duplicate
1570    subpattern numbers means that we have to scan from the very start, in order to
1571    take account of such duplicates, and to use a recursive function to keep track
1572    of the different types of group.
1573    
1574    Arguments:
1575      cd           compile background data
1576      name         name to seek, or NULL if seeking a numbered subpattern
1577      lorn         name length, or subpattern number if name is NULL
1578      xmode        TRUE if we are in /x mode
1579      utf          TRUE if we are in UTF-8 / UTF-16 mode
1580    
1581    Returns:       the number of the found subpattern, or -1 if not found
1582    */
1583    
1584    static int
1585    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1586      BOOL utf)
1587    {
1588    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1589    int count = 0;
1590    int rc;
1591    
1592    /* If the pattern does not start with an opening parenthesis, the first call
1593    to find_parens_sub() will scan right to the end (if necessary). However, if it
1594    does start with a parenthesis, find_parens_sub() will return when it hits the
1595    matching closing parens. That is why we have to have a loop. */
1596    
1597    for (;;)
1598      {
1599      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1600      if (rc > 0 || *ptr++ == 0) break;
1601    }    }
1602    
1603  return -1;  return rc;
1604  }  }
1605    
1606    
1607    
1608    
1609  /*************************************************  /*************************************************
1610  *      Find first significant op code            *  *      Find first significant op code            *
1611  *************************************************/  *************************************************/
1612    
1613  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1614  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1615  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1616  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1617  assertions, and also the \b assertion; for others it does not.  does not.
1618    
1619  Arguments:  Arguments:
1620    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1621    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1622    
1623  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1624  */  */
1625    
1626  static const uschar*  static const pcre_uchar*
1627  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1628  {  {
1629  for (;;)  for (;;)
1630    {    {
1631    switch ((int)*code)    switch ((int)*code)
1632      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1633      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1634      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1635      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1636      if (!skipassert) return code;      if (!skipassert) return code;
1637      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1638      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1639      break;      break;
1640    
1641      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1060  for (;;) Line 1645  for (;;)
1645    
1646      case OP_CALLOUT:      case OP_CALLOUT:
1647      case OP_CREF:      case OP_CREF:
1648        case OP_NCREF:
1649      case OP_RREF:      case OP_RREF:
1650        case OP_NRREF:
1651      case OP_DEF:      case OP_DEF:
1652      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1653      break;      break;
1654    
1655      default:      default:
# Line 1076  for (;;) Line 1663  for (;;)
1663    
1664    
1665  /*************************************************  /*************************************************
1666  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1667  *************************************************/  *************************************************/
1668    
1669  /* 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,
1670  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.
1671  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
1672    temporarily terminated with OP_END when this function is called.
1673    
1674    This function is called when a backward assertion is encountered, so that if it
1675    fails, the error message can point to the correct place in the pattern.
1676    However, we cannot do this when the assertion contains subroutine calls,
1677    because they can be forward references. We solve this by remembering this case
1678    and doing the check at the end; a flag specifies which mode we are running in.
1679    
1680  Arguments:  Arguments:
1681    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1682    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1683      atend    TRUE if called when the pattern is complete
1684  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1685               or -2 if \C was encountered  
1686    Returns:   the fixed length,
1687                 or -1 if there is no fixed length,
1688                 or -2 if \C was encountered (in UTF-8 mode only)
1689                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1690                 or -4 if an unknown opcode was encountered (internal error)
1691  */  */
1692    
1693  static int  static int
1694  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1695  {  {
1696  int length = -1;  int length = -1;
1697    
1698  register int branchlength = 0;  register int branchlength = 0;
1699  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1700    
1701  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1702  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1105  branch, check the length against that of Line 1704  branch, check the length against that of
1704  for (;;)  for (;;)
1705    {    {
1706    int d;    int d;
1707      pcre_uchar *ce, *cs;
1708    register int op = *cc;    register int op = *cc;
1709    switch (op)    switch (op)
1710      {      {
1711        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1712        OP_BRA (normal non-capturing bracket) because the other variants of these
1713        opcodes are all concerned with unlimited repeated groups, which of course
1714        are not of fixed length. */
1715    
1716      case OP_CBRA:      case OP_CBRA:
1717      case OP_BRA:      case OP_BRA:
1718      case OP_ONCE:      case OP_ONCE:
1719        case OP_ONCE_NC:
1720      case OP_COND:      case OP_COND:
1721      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1722      if (d < 0) return d;      if (d < 0) return d;
1723      branchlength += d;      branchlength += d;
1724      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1725      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1726      break;      break;
1727    
1728      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1729      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1730      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1731        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1732        because they all imply an unlimited repeat. */
1733    
1734      case OP_ALT:      case OP_ALT:
1735      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1736      case OP_END:      case OP_END:
1737        case OP_ACCEPT:
1738        case OP_ASSERT_ACCEPT:
1739      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1740        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1741      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1135  for (;;) Line 1743  for (;;)
1743      branchlength = 0;      branchlength = 0;
1744      break;      break;
1745    
1746        /* A true recursion implies not fixed length, but a subroutine call may
1747        be OK. If the subroutine is a forward reference, we can't deal with
1748        it until the end of the pattern, so return -3. */
1749    
1750        case OP_RECURSE:
1751        if (!atend) return -3;
1752        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1753        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1754        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1755        d = find_fixedlength(cs + 2, utf, atend, cd);
1756        if (d < 0) return d;
1757        branchlength += d;
1758        cc += 1 + LINK_SIZE;
1759        break;
1760    
1761      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1762    
1763      case OP_ASSERT:      case OP_ASSERT:
# Line 1146  for (;;) Line 1769  for (;;)
1769    
1770      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1771    
1772      case OP_REVERSE:      case OP_MARK:
1773        case OP_PRUNE_ARG:
1774        case OP_SKIP_ARG:
1775        case OP_THEN_ARG:
1776        cc += cc[1] + PRIV(OP_lengths)[*cc];
1777        break;
1778    
1779        case OP_CALLOUT:
1780        case OP_CIRC:
1781        case OP_CIRCM:
1782        case OP_CLOSE:
1783        case OP_COMMIT:
1784      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1785      case OP_DEF:      case OP_DEF:
1786      case OP_OPT:      case OP_DOLL:
1787      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1788      case OP_EOD:      case OP_EOD:
1789      case OP_EODN:      case OP_EODN:
1790      case OP_CIRC:      case OP_FAIL:
1791      case OP_DOLL:      case OP_NCREF:
1792        case OP_NRREF:
1793      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1794        case OP_PRUNE:
1795        case OP_REVERSE:
1796        case OP_RREF:
1797        case OP_SET_SOM:
1798        case OP_SKIP:
1799        case OP_SOD:
1800        case OP_SOM:
1801        case OP_THEN:
1802      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1803      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1804      break;      break;
1805    
1806      /* Handle literal characters */      /* Handle literal characters */
1807    
1808      case OP_CHAR:      case OP_CHAR:
1809      case OP_CHARNC:      case OP_CHARI:
1810      case OP_NOT:      case OP_NOT:
1811        case OP_NOTI:
1812      branchlength++;      branchlength++;
1813      cc += 2;      cc += 2;
1814  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1815      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1816  #endif  #endif
1817      break;      break;
1818    
# Line 1182  for (;;) Line 1820  for (;;)
1820      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1821    
1822      case OP_EXACT:      case OP_EXACT:
1823        case OP_EXACTI:
1824        case OP_NOTEXACT:
1825        case OP_NOTEXACTI:
1826      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1827      cc += 4;      cc += 2 + IMM2_SIZE;
1828  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1829      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1830  #endif  #endif
1831      break;      break;
1832    
1833      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1834      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1835      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1836      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1837      break;      break;
1838    
1839      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1205  for (;;) Line 1843  for (;;)
1843      cc += 2;      cc += 2;
1844      /* Fall through */      /* Fall through */
1845    
1846        case OP_HSPACE:
1847        case OP_VSPACE:
1848        case OP_NOT_HSPACE:
1849        case OP_NOT_VSPACE:
1850      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1851      case OP_DIGIT:      case OP_DIGIT:
1852      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1212  for (;;) Line 1854  for (;;)
1854      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1855      case OP_WORDCHAR:      case OP_WORDCHAR:
1856      case OP_ANY:      case OP_ANY:
1857        case OP_ALLANY:
1858      branchlength++;      branchlength++;
1859      cc++;      cc++;
1860      break;      break;
1861    
1862      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1863        otherwise \C is coded as OP_ALLANY. */
1864    
1865      case OP_ANYBYTE:      case OP_ANYBYTE:
1866      return -2;      return -2;
1867    
1868      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1869    
1870  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1871      case OP_XCLASS:      case OP_XCLASS:
1872      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1873      /* Fall through */      /* Fall through */
1874  #endif  #endif
1875    
1876      case OP_CLASS:      case OP_CLASS:
1877      case OP_NCLASS:      case OP_NCLASS:
1878      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1879    
1880      switch (*cc)      switch (*cc)
1881        {        {
1882          case OP_CRPLUS:
1883          case OP_CRMINPLUS:
1884        case OP_CRSTAR:        case OP_CRSTAR:
1885        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1886        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1243  for (;;) Line 1889  for (;;)
1889    
1890        case OP_CRRANGE:        case OP_CRRANGE:
1891        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1892        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1893        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1894        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1895        break;        break;
1896    
1897        default:        default:
# Line 1255  for (;;) Line 1901  for (;;)
1901    
1902      /* Anything else is variable length */      /* Anything else is variable length */
1903    
1904      default:      case OP_ANYNL:
1905        case OP_BRAMINZERO:
1906        case OP_BRAPOS:
1907        case OP_BRAPOSZERO:
1908        case OP_BRAZERO:
1909        case OP_CBRAPOS:
1910        case OP_EXTUNI:
1911        case OP_KETRMAX:
1912        case OP_KETRMIN:
1913        case OP_KETRPOS:
1914        case OP_MINPLUS:
1915        case OP_MINPLUSI:
1916        case OP_MINQUERY:
1917        case OP_MINQUERYI:
1918        case OP_MINSTAR:
1919        case OP_MINSTARI:
1920        case OP_MINUPTO:
1921        case OP_MINUPTOI:
1922        case OP_NOTMINPLUS:
1923        case OP_NOTMINPLUSI:
1924        case OP_NOTMINQUERY:
1925        case OP_NOTMINQUERYI:
1926        case OP_NOTMINSTAR:
1927        case OP_NOTMINSTARI:
1928        case OP_NOTMINUPTO:
1929        case OP_NOTMINUPTOI:
1930        case OP_NOTPLUS:
1931        case OP_NOTPLUSI:
1932        case OP_NOTPOSPLUS:
1933        case OP_NOTPOSPLUSI:
1934        case OP_NOTPOSQUERY:
1935        case OP_NOTPOSQUERYI:
1936        case OP_NOTPOSSTAR:
1937        case OP_NOTPOSSTARI:
1938        case OP_NOTPOSUPTO:
1939        case OP_NOTPOSUPTOI:
1940        case OP_NOTQUERY:
1941        case OP_NOTQUERYI:
1942        case OP_NOTSTAR:
1943        case OP_NOTSTARI:
1944        case OP_NOTUPTO:
1945        case OP_NOTUPTOI:
1946        case OP_PLUS:
1947        case OP_PLUSI:
1948        case OP_POSPLUS:
1949        case OP_POSPLUSI:
1950        case OP_POSQUERY:
1951        case OP_POSQUERYI:
1952        case OP_POSSTAR:
1953        case OP_POSSTARI:
1954        case OP_POSUPTO:
1955        case OP_POSUPTOI:
1956        case OP_QUERY:
1957        case OP_QUERYI:
1958        case OP_REF:
1959        case OP_REFI:
1960        case OP_SBRA:
1961        case OP_SBRAPOS:
1962        case OP_SCBRA:
1963        case OP_SCBRAPOS:
1964        case OP_SCOND:
1965        case OP_SKIPZERO:
1966        case OP_STAR:
1967        case OP_STARI:
1968        case OP_TYPEMINPLUS:
1969        case OP_TYPEMINQUERY:
1970        case OP_TYPEMINSTAR:
1971        case OP_TYPEMINUPTO:
1972        case OP_TYPEPLUS:
1973        case OP_TYPEPOSPLUS:
1974        case OP_TYPEPOSQUERY:
1975        case OP_TYPEPOSSTAR:
1976        case OP_TYPEPOSUPTO:
1977        case OP_TYPEQUERY:
1978        case OP_TYPESTAR:
1979        case OP_TYPEUPTO:
1980        case OP_UPTO:
1981        case OP_UPTOI:
1982      return -1;      return -1;
1983    
1984        /* Catch unrecognized opcodes so that when new ones are added they
1985        are not forgotten, as has happened in the past. */
1986    
1987        default:
1988        return -4;
1989      }      }
1990    }    }
1991  /* Control never gets here */  /* Control never gets here */
# Line 1266  for (;;) Line 1995  for (;;)
1995    
1996    
1997  /*************************************************  /*************************************************
1998  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1999  *************************************************/  *************************************************/
2000    
2001  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2002  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2003    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2004    so that it can be called from pcre_study() when finding the minimum matching
2005    length.
2006    
2007  Arguments:  Arguments:
2008    code        points to start of expression    code        points to start of expression
2009    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2010    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2011    
2012  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
2013  */  */
2014    
2015  static const uschar *  const pcre_uchar *
2016  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2017  {  {
2018  for (;;)  for (;;)
2019    {    {
2020    register int c = *code;    register int c = *code;
2021    
2022    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2023    
2024    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1294  for (;;) Line 2027  for (;;)
2027    
2028    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2029    
2030      /* Handle recursion */
2031    
2032      else if (c == OP_REVERSE)
2033        {
2034        if (number < 0) return (pcre_uchar *)code;
2035        code += PRIV(OP_lengths)[c];
2036        }
2037    
2038    /* Handle capturing bracket */    /* Handle capturing bracket */
2039    
2040    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2041               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2042      {      {
2043      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2044      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2045      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2046      }      }
2047    
2048    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2049    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2050    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2051      must add in its length. */
2052    
2053    else    else
2054      {      {
# Line 1327  for (;;) Line 2070  for (;;)
2070        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2071        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2072        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2073        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2074            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2075          break;
2076    
2077          case OP_MARK:
2078          case OP_PRUNE_ARG:
2079          case OP_SKIP_ARG:
2080          code += code[1];
2081          break;
2082    
2083          case OP_THEN_ARG:
2084          code += code[1];
2085        break;        break;
2086        }        }
2087    
2088      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2089    
2090      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2091    
2092    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2093    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2094    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2095    
2096  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2097      if (utf8) switch(c)      if (utf) switch(c)
2098        {        {
2099        case OP_CHAR:        case OP_CHAR:
2100        case OP_CHARNC:        case OP_CHARI:
2101        case OP_EXACT:        case OP_EXACT:
2102          case OP_EXACTI:
2103        case OP_UPTO:        case OP_UPTO:
2104          case OP_UPTOI:
2105        case OP_MINUPTO:        case OP_MINUPTO:
2106          case OP_MINUPTOI:
2107        case OP_POSUPTO:        case OP_POSUPTO:
2108          case OP_POSUPTOI:
2109        case OP_STAR:        case OP_STAR:
2110          case OP_STARI:
2111        case OP_MINSTAR:        case OP_MINSTAR:
2112          case OP_MINSTARI:
2113        case OP_POSSTAR:        case OP_POSSTAR:
2114          case OP_POSSTARI:
2115        case OP_PLUS:        case OP_PLUS:
2116          case OP_PLUSI:
2117        case OP_MINPLUS:        case OP_MINPLUS:
2118          case OP_MINPLUSI:
2119        case OP_POSPLUS:        case OP_POSPLUS:
2120          case OP_POSPLUSI:
2121        case OP_QUERY:        case OP_QUERY:
2122          case OP_QUERYI:
2123        case OP_MINQUERY:        case OP_MINQUERY:
2124          case OP_MINQUERYI:
2125        case OP_POSQUERY:        case OP_POSQUERY:
2126        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2127          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2128        break;        break;
2129        }        }
2130    #else
2131        (void)(utf);  /* Keep compiler happy by referencing function argument */
2132  #endif  #endif
2133      }      }
2134    }    }
# Line 1376  instance of OP_RECURSE. Line 2145  instance of OP_RECURSE.
2145    
2146  Arguments:  Arguments:
2147    code        points to start of expression    code        points to start of expression
2148    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2149    
2150  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2151  */  */
2152    
2153  static const uschar *  static const pcre_uchar *
2154  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2155  {  {
2156  for (;;)  for (;;)
2157    {    {
# Line 1398  for (;;) Line 2167  for (;;)
2167    
2168    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2169    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2170    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2171      must add in its length. */
2172    
2173    else    else
2174      {      {
# Line 1420  for (;;) Line 2190  for (;;)
2190        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2191        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2192        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2193        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2194            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2195          break;
2196    
2197          case OP_MARK:
2198          case OP_PRUNE_ARG:
2199          case OP_SKIP_ARG:
2200          code += code[1];
2201          break;
2202    
2203          case OP_THEN_ARG:
2204          code += code[1];
2205        break;        break;
2206        }        }
2207    
2208      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2209    
2210      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2211    
2212      /* 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
2213      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
2214      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2215    
2216  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2217      if (utf8) switch(c)      if (utf) switch(c)
2218        {        {
2219        case OP_CHAR:        case OP_CHAR:
2220        case OP_CHARNC:        case OP_CHARI:
2221        case OP_EXACT:        case OP_EXACT:
2222          case OP_EXACTI:
2223        case OP_UPTO:        case OP_UPTO:
2224          case OP_UPTOI:
2225        case OP_MINUPTO:        case OP_MINUPTO:
2226          case OP_MINUPTOI:
2227        case OP_POSUPTO:        case OP_POSUPTO:
2228          case OP_POSUPTOI:
2229        case OP_STAR:        case OP_STAR:
2230          case OP_STARI:
2231        case OP_MINSTAR:        case OP_MINSTAR:
2232          case OP_MINSTARI:
2233        case OP_POSSTAR:        case OP_POSSTAR:
2234          case OP_POSSTARI:
2235        case OP_PLUS:        case OP_PLUS:
2236          case OP_PLUSI:
2237        case OP_MINPLUS:        case OP_MINPLUS:
2238          case OP_MINPLUSI:
2239        case OP_POSPLUS:        case OP_POSPLUS:
2240          case OP_POSPLUSI:
2241        case OP_QUERY:        case OP_QUERY:
2242          case OP_QUERYI:
2243        case OP_MINQUERY:        case OP_MINQUERY:
2244          case OP_MINQUERYI:
2245        case OP_POSQUERY:        case OP_POSQUERY:
2246        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2247          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2248        break;        break;
2249        }        }
2250    #else
2251        (void)(utf);  /* Keep compiler happy by referencing function argument */
2252  #endif  #endif
2253      }      }
2254    }    }
# Line 1468  for (;;) Line 2264  for (;;)
2264  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2265  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
2266  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2267  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2268  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2269    bracket whose current branch will already have been scanned.
2270    
2271  Arguments:  Arguments:
2272    code        points to start of search    code        points to start of search
2273    endcode     points to where to stop    endcode     points to where to stop
2274    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2275      cd          contains pointers to tables etc.
2276    
2277  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2278  */  */
2279    
2280  static BOOL  static BOOL
2281  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2282      BOOL utf, compile_data *cd)
2283  {  {
2284  register int c;  register int c;
2285  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2286       code < endcode;       code < endcode;
2287       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2288    {    {
2289    const uschar *ccode;    const pcre_uchar *ccode;
2290    
2291    c = *code;    c = *code;
2292    
2293    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2294      first_significant_code() with a TRUE final argument. */
2295    
2296    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2297      {      {
     code += _pcre_OP_lengths[c];  
2298      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2299      c = *code;      c = *code;
2300      continue;      continue;
2301      }      }
2302    
2303    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2304      implies a backward reference subroutine call, we can scan it. If it's a
2305      forward reference subroutine call, we can't. To detect forward reference
2306      we have to scan up the list that is kept in the workspace. This function is
2307      called only when doing the real compile, not during the pre-compile that
2308      measures the size of the compiled pattern. */
2309    
2310    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2311      {      {
2312        const pcre_uchar *scode;
2313      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2314    
2315      /* Scan a closed bracket */      /* Test for forward reference */
2316    
2317        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2318          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2319    
2320        /* Not a forward reference, test for completed backward reference */
2321    
2322      empty_branch = FALSE;      empty_branch = FALSE;
2323        scode = cd->start_code + GET(code, 1);
2324        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2325    
2326        /* Completed backwards reference */
2327    
2328      do      do
2329        {        {
2330        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2331            {
2332          empty_branch = TRUE;          empty_branch = TRUE;
2333            break;
2334            }
2335          scode += GET(scode, 1);
2336          }
2337        while (*scode == OP_ALT);
2338    
2339        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2340        continue;
2341        }
2342    
2343      /* Groups with zero repeats can of course be empty; skip them. */
2344    
2345      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2346          c == OP_BRAPOSZERO)
2347        {
2348        code += PRIV(OP_lengths)[c];
2349        do code += GET(code, 1); while (*code == OP_ALT);
2350        c = *code;
2351        continue;
2352        }
2353    
2354      /* A nested group that is already marked as "could be empty" can just be
2355      skipped. */
2356    
2357      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2358          c == OP_SCBRA || c == OP_SCBRAPOS)
2359        {
2360        do code += GET(code, 1); while (*code == OP_ALT);
2361        c = *code;
2362        continue;
2363        }
2364    
2365      /* For other groups, scan the branches. */
2366    
2367      if (c == OP_BRA  || c == OP_BRAPOS ||
2368          c == OP_CBRA || c == OP_CBRAPOS ||
2369          c == OP_ONCE || c == OP_ONCE_NC ||
2370          c == OP_COND)
2371        {
2372        BOOL empty_branch;
2373        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2374    
2375        /* If a conditional group has only one branch, there is a second, implied,
2376        empty branch, so just skip over the conditional, because it could be empty.
2377        Otherwise, scan the individual branches of the group. */
2378    
2379        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2380        code += GET(code, 1);        code += GET(code, 1);
2381        else
2382          {
2383          empty_branch = FALSE;
2384          do
2385            {
2386            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2387              empty_branch = TRUE;
2388            code += GET(code, 1);
2389            }
2390          while (*code == OP_ALT);
2391          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2392        }        }
2393      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2394      c = *code;      c = *code;
2395      continue;      continue;
2396      }      }
# Line 1529  for (code = first_significant_code(code Line 2401  for (code = first_significant_code(code
2401      {      {
2402      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2403      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2404      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2405      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2406      here. */      here. */
2407    
2408  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2409      case OP_XCLASS:      case OP_XCLASS:
2410      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2411      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1541  for (code = first_significant_code(code Line 2413  for (code = first_significant_code(code
2413    
2414      case OP_CLASS:      case OP_CLASS:
2415      case OP_NCLASS:      case OP_NCLASS:
2416      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2417    
2418  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2419      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2420  #endif  #endif
2421    
# Line 1579  for (code = first_significant_code(code Line 2451  for (code = first_significant_code(code
2451      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2452      case OP_WORDCHAR:      case OP_WORDCHAR:
2453      case OP_ANY:      case OP_ANY:
2454        case OP_ALLANY:
2455      case OP_ANYBYTE:      case OP_ANYBYTE:
2456      case OP_CHAR:      case OP_CHAR:
2457      case OP_CHARNC:      case OP_CHARI:
2458      case OP_NOT:      case OP_NOT:
2459        case OP_NOTI:
2460      case OP_PLUS:      case OP_PLUS:
2461      case OP_MINPLUS:      case OP_MINPLUS:
2462      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1614  for (code = first_significant_code(code Line 2488  for (code = first_significant_code(code
2488      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2489      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2490      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2491      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2492          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2493      break;      break;
2494    
2495      /* End of branch */      /* End of branch */
# Line 1622  for (code = first_significant_code(code Line 2497  for (code = first_significant_code(code
2497      case OP_KET:      case OP_KET:
2498      case OP_KETRMAX:      case OP_KETRMAX:
2499      case OP_KETRMIN:      case OP_KETRMIN:
2500        case OP_KETRPOS:
2501      case OP_ALT:      case OP_ALT:
2502      return TRUE;      return TRUE;
2503    
2504      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2505      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2506    
2507  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2508      case OP_STAR:      case OP_STAR:
2509        case OP_STARI:
2510      case OP_MINSTAR:      case OP_MINSTAR:
2511        case OP_MINSTARI:
2512      case OP_POSSTAR:      case OP_POSSTAR:
2513        case OP_POSSTARI:
2514      case OP_QUERY:      case OP_QUERY:
2515        case OP_QUERYI:
2516      case OP_MINQUERY:      case OP_MINQUERY:
2517        case OP_MINQUERYI:
2518      case OP_POSQUERY:      case OP_POSQUERY:
2519        case OP_POSQUERYI:
2520        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2521        break;
2522    
2523      case OP_UPTO:      case OP_UPTO:
2524        case OP_UPTOI:
2525      case OP_MINUPTO:      case OP_MINUPTO:
2526        case OP_MINUPTOI:
2527      case OP_POSUPTO:      case OP_POSUPTO:
2528      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2529        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2530      break;      break;
2531  #endif  #endif
2532    
2533        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2534        string. */
2535    
2536        case OP_MARK:
2537        case OP_PRUNE_ARG:
2538        case OP_SKIP_ARG:
2539        code += code[1];
2540        break;
2541    
2542        case OP_THEN_ARG:
2543        code += code[1];
2544        break;
2545    
2546        /* None of the remaining opcodes are required to match a character. */
2547    
2548        default:
2549        break;
2550      }      }
2551    }    }
2552    
# Line 1657  return TRUE; Line 2563  return TRUE;
2563  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2564  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2565  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2566    This function is called only during the real compile, not during the
2567    pre-compile.
2568    
2569  Arguments:  Arguments:
2570    code        points to start of the recursion    code        points to start of the recursion
2571    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2572    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2573    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2574      cd          pointers to tables etc
2575    
2576  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2577  */  */
2578    
2579  static BOOL  static BOOL
2580  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2581    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2582  {  {
2583  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2584    {    {
2585    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2586        return FALSE;
2587    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2588    }    }
2589  return TRUE;  return TRUE;
# Line 1686  return TRUE; Line 2596  return TRUE;
2596  *************************************************/  *************************************************/
2597    
2598  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2599  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
2600  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2601  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2602    
2603    Originally, this function only recognized a sequence of letters between the
2604    terminators, but it seems that Perl recognizes any sequence of characters,
2605    though of course unknown POSIX names are subsequently rejected. Perl gives an
2606    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2607    didn't consider this to be a POSIX class. Likewise for [:1234:].
2608    
2609    The problem in trying to be exactly like Perl is in the handling of escapes. We
2610    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2611    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2612    below handles the special case of \], but does not try to do any other escape
2613    processing. This makes it different from Perl for cases such as [:l\ower:]
2614    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2615    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2616    I think.
2617    
2618    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2619    It seems that the appearance of a nested POSIX class supersedes an apparent
2620    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2621    a digit.
2622    
2623    In Perl, unescaped square brackets may also appear as part of class names. For
2624    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2625    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2626    seem right at all. PCRE does not allow closing square brackets in POSIX class
2627    names.
2628    
2629  Argument:  Arguments:
2630    ptr      pointer to the initial [    ptr      pointer to the initial [
2631    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2632    
2633  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2634  */  */
2635    
2636  static BOOL  static BOOL
2637  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2638  {  {
2639  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2640  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2641  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2642    {    {
2643    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2644    return TRUE;      ptr++;
2645      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2646      else
2647        {
2648        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2649          {
2650          *endptr = ptr;
2651          return TRUE;
2652          }
2653        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2654             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2655              ptr[1] == CHAR_EQUALS_SIGN) &&
2656            check_posix_syntax(ptr, endptr))
2657          return FALSE;
2658        }
2659    }    }
2660  return FALSE;  return FALSE;
2661  }  }
# Line 1731  Returns:     a value representing the na Line 2678  Returns:     a value representing the na
2678  */  */
2679    
2680  static int  static int
2681  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2682  {  {
2683    const char *pn = posix_names;
2684  register int yield = 0;  register int yield = 0;
2685  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2686    {    {
2687    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2688      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2689      pn += posix_name_lengths[yield] + 1;
2690    yield++;    yield++;
2691    }    }
2692  return -1;  return -1;
# Line 1752  return -1; Line 2701  return -1;
2701  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2702  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2703  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
2704  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
2705  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
2706  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
2707  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2708  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2709    OP_END.
2710    
2711  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2712  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1767  value in the reference (which is a group Line 2717  value in the reference (which is a group
2717  Arguments:  Arguments:
2718    group      points to the start of the group    group      points to the start of the group
2719    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2720    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2721    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2722    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm   the hwm forward reference pointer at the start of the group
2723    
# Line 1775  Returns:     nothing Line 2725  Returns:     nothing
2725  */  */
2726    
2727  static void  static void
2728  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2729    uschar *save_hwm)    pcre_uchar *save_hwm)
2730  {  {
2731  uschar *ptr = group;  pcre_uchar *ptr = group;
2732    
2733  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2734    {    {
2735    int offset;    int offset;
2736    uschar *hc;    pcre_uchar *hc;
2737    
2738    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2739    reference. */    reference. */
# Line 1828  Arguments: Line 2778  Arguments:
2778  Returns:         new code pointer  Returns:         new code pointer
2779  */  */
2780    
2781  static uschar *  static pcre_uchar *
2782  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2783  {  {
2784  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2785  *code++ = 255;  *code++ = 255;
2786  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2787  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2788  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2789  }  }
2790    
2791    
# Line 1857  Returns:             nothing Line 2807  Returns:             nothing
2807  */  */
2808    
2809  static void  static void
2810  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2811  {  {
2812  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2813  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2814  }  }
2815    
# Line 1891  get_othercase_range(unsigned int *cptr, Line 2841  get_othercase_range(unsigned int *cptr,
2841  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2842    
2843  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2844    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2845    
2846  if (c > d) return FALSE;  if (c > d) return FALSE;
2847    
# Line 1900  next = othercase + 1; Line 2850  next = othercase + 1;
2850    
2851  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2852    {    {
2853    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2854    next++;    next++;
2855    }    }
2856    
# Line 1909  for (++c; c <= d; c++) Line 2859  for (++c; c <= d; c++)
2859    
2860  return TRUE;  return TRUE;
2861  }  }
2862    
2863    
2864    
2865    /*************************************************
2866    *        Check a character and a property        *
2867    *************************************************/
2868    
2869    /* This function is called by check_auto_possessive() when a property item
2870    is adjacent to a fixed character.
2871    
2872    Arguments:
2873      c            the character
2874      ptype        the property type
2875      pdata        the data for the type
2876      negated      TRUE if it's a negated property (\P or \p{^)
2877    
2878    Returns:       TRUE if auto-possessifying is OK
2879    */
2880    
2881    static BOOL
2882    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2883    {
2884    const ucd_record *prop = GET_UCD(c);
2885    switch(ptype)
2886      {
2887      case PT_LAMP:
2888      return (prop->chartype == ucp_Lu ||
2889              prop->chartype == ucp_Ll ||
2890              prop->chartype == ucp_Lt) == negated;
2891    
2892      case PT_GC:
2893      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2894    
2895      case PT_PC:
2896      return (pdata == prop->chartype) == negated;
2897    
2898      case PT_SC:
2899      return (pdata == prop->script) == negated;
2900    
2901      /* These are specials */
2902    
2903      case PT_ALNUM:
2904      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2905              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2906    
2907      case PT_SPACE:    /* Perl space */
2908      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2909              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2910              == negated;
2911    
2912      case PT_PXSPACE:  /* POSIX space */
2913      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2914              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2915              c == CHAR_FF || c == CHAR_CR)
2916              == negated;
2917    
2918      case PT_WORD:
2919      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2920              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2921              c == CHAR_UNDERSCORE) == negated;
2922      }
2923    return FALSE;
2924    }
2925  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2926    
2927    
# Line 1922  whether the next thing could possibly ma Line 2935  whether the next thing could possibly ma
2935  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2936    
2937  Arguments:  Arguments:
2938    op_code       the repeated op code    previous      pointer to the repeated opcode
2939    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2940    ptr           next character in pattern    ptr           next character in pattern
2941    options       options bits    options       options bits
2942    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1934  Returns:        TRUE if possessifying is Line 2945  Returns:        TRUE if possessifying is
2945  */  */
2946    
2947  static BOOL  static BOOL
2948  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2949    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2950  {  {
2951  int next;  pcre_int32 c, next;
2952    int op_code = *previous++;
2953    
2954  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2955    
# Line 1945  if ((options & PCRE_EXTENDED) != 0) Line 2957  if ((options & PCRE_EXTENDED) != 0)
2957    {    {
2958    for (;;)    for (;;)
2959      {      {
2960      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2961      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2962        {        {
2963        while (*(++ptr) != 0)        ptr++;
2964          while (*ptr != 0)
2965            {
2966          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2967            ptr++;
2968    #ifdef SUPPORT_UTF
2969            if (utf) FORWARDCHAR(ptr);
2970    #endif
2971            }
2972        }        }
2973      else break;      else break;
2974      }      }
# Line 1958  if ((options & PCRE_EXTENDED) != 0) Line 2977  if ((options & PCRE_EXTENDED) != 0)
2977  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2978  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2979    
2980  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2981    {    {
2982    int temperrorcode = 0;    int temperrorcode = 0;
2983    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2984    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2985    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2986    }    }
2987    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2988    {    {
2989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2990    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2991  #endif  #endif
2992    next = *ptr++;    next = *ptr++;
2993    }    }
   
2994  else return FALSE;  else return FALSE;
2995    
2996  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1982  if ((options & PCRE_EXTENDED) != 0) Line 2999  if ((options & PCRE_EXTENDED) != 0)
2999    {    {
3000    for (;;)    for (;;)
3001      {      {
3002      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3003      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3004        {        {
3005        while (*(++ptr) != 0)        ptr++;
3006          while (*ptr != 0)
3007            {
3008          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3009            ptr++;
3010    #ifdef SUPPORT_UTF
3011            if (utf) FORWARDCHAR(ptr);
3012    #endif
3013            }
3014        }        }
3015      else break;      else break;
3016      }      }
# Line 1994  if ((options & PCRE_EXTENDED) != 0) Line 3018  if ((options & PCRE_EXTENDED) != 0)
3018    
3019  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3020    
3021  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3022    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3023        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
3024    
3025  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3026    the next item is a character. */
3027    
3028  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3029    {    {
3030    case OP_CHAR:    case OP_CHAR:
3031  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3032    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3033    #else
3034      c = *previous;
3035  #endif  #endif
3036    return item != next;    return c != next;
3037    
3038    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
3039    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
3040    high-valued characters. */    high-valued characters. */
3041    
3042    case OP_CHARNC:    case OP_CHARI:
3043  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3044    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3045  #endif  #else
3046    if (item == next) return FALSE;    c = *previous;
3047  #ifdef SUPPORT_UTF8  #endif
3048    if (utf8)    if (c == next) return FALSE;
3049    #ifdef SUPPORT_UTF
3050      if (utf)
3051      {      {
3052      unsigned int othercase;      unsigned int othercase;
3053      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3054  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3055      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3056  #else  #else
3057      othercase = NOTACHAR;      othercase = NOTACHAR;
3058  #endif  #endif
3059      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3060      }      }
3061    else    else
3062  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3063    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
3064    
3065    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3066      opcodes are not used for multi-byte characters, because they are coded using
3067      an XCLASS instead. */
3068    
3069    case OP_NOT:    case OP_NOT:
3070    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
3071    if (item == next) return TRUE;  
3072    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
3073  #ifdef SUPPORT_UTF8    if ((c = *previous) == next) return TRUE;
3074    if (utf8)  #ifdef SUPPORT_UTF
3075      if (utf)
3076      {      {
3077      unsigned int othercase;      unsigned int othercase;
3078      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3079  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3080      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3081  #else  #else
3082      othercase = NOTACHAR;      othercase = NOTACHAR;
3083  #endif  #endif
3084      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3085      }      }
3086    else    else
3087  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3088    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
3089    
3090      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3091      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3092    
3093    case OP_DIGIT:    case OP_DIGIT:
3094    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2101  if (next >= 0) switch(op_code) Line 3131  if (next >= 0) switch(op_code)
3131      case 0x202f:      case 0x202f:
3132      case 0x205f:      case 0x205f:
3133      case 0x3000:      case 0x3000:
3134      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3135      default:      default:
3136      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3137      }      }
3138    
3139      case OP_ANYNL:
3140    case OP_VSPACE:    case OP_VSPACE:
3141    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3142    switch(next)    switch(next)
# Line 2117  if (next >= 0) switch(op_code) Line 3148  if (next >= 0) switch(op_code)
3148      case 0x85:      case 0x85:
3149      case 0x2028:      case 0x2028:
3150      case 0x2029:      case 0x2029:
3151      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3152      default:      default:
3153      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3154      }      }
3155    
3156    #ifdef SUPPORT_UCP
3157      case OP_PROP:
3158      return check_char_prop(next, previous[0], previous[1], FALSE);
3159    
3160      case OP_NOTPROP:
3161      return check_char_prop(next, previous[0], previous[1], TRUE);
3162    #endif
3163    
3164    default:    default:
3165    return FALSE;    return FALSE;
3166    }    }
3167    
3168    
3169  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3170    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3171    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3172    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3173    replaced by OP_PROP codes when PCRE_UCP is set. */
3174    
3175  switch(op_code)  switch(op_code)
3176    {    {
3177    case OP_CHAR:    case OP_CHAR:
3178    case OP_CHARNC:    case OP_CHARI:
3179  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3180    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3181    #else
3182      c = *previous;
3183  #endif  #endif
3184    switch(-next)    switch(-next)
3185      {      {
3186      case ESC_d:      case ESC_d:
3187      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3188    
3189      case ESC_D:      case ESC_D:
3190      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3191    
3192      case ESC_s:      case ESC_s:
3193      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3194    
3195      case ESC_S:      case ESC_S:
3196      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3197    
3198      case ESC_w:      case ESC_w:
3199      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3200    
3201      case ESC_W:      case ESC_W:
3202      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3203    
3204      case ESC_h:      case ESC_h:
3205      case ESC_H:      case ESC_H:
3206      switch(item)      switch(c)
3207        {        {
3208        case 0x09:        case 0x09:
3209        case 0x20:        case 0x20:
# Line 2186  switch(op_code) Line 3231  switch(op_code)
3231    
3232      case ESC_v:      case ESC_v:
3233      case ESC_V:      case ESC_V:
3234      switch(item)      switch(c)
3235        {        {
3236        case 0x0a:        case 0x0a:
3237        case 0x0b:        case 0x0b:
# Line 2200  switch(op_code) Line 3245  switch(op_code)
3245        return -next == ESC_v;        return -next == ESC_v;
3246        }        }
3247    
3248        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3249        their substitutions and process them. The result will always be either
3250        -ESC_p or -ESC_P. Then fall through to process those values. */
3251    
3252    #ifdef SUPPORT_UCP
3253        case ESC_du:
3254        case ESC_DU:
3255        case ESC_wu:
3256        case ESC_WU:
3257        case ESC_su:
3258        case ESC_SU:
3259          {
3260          int temperrorcode = 0;
3261          ptr = substitutes[-next - ESC_DU];
3262          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3263          if (temperrorcode != 0) return FALSE;
3264          ptr++;    /* For compatibility */
3265          }
3266        /* Fall through */
3267    
3268        case ESC_p:
3269        case ESC_P:
3270          {
3271          int ptype, pdata, errorcodeptr;
3272          BOOL negated;
3273    
3274          ptr--;      /* Make ptr point at the p or P */
3275          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3276          if (ptype < 0) return FALSE;
3277          ptr++;      /* Point past the final curly ket */
3278    
3279          /* If the property item is optional, we have to give up. (When generated
3280          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3281          to the original \d etc. At this point, ptr will point to a zero byte. */
3282    
3283          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3284            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3285              return FALSE;
3286    
3287          /* Do the property check. */
3288    
3289          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3290          }
3291    #endif
3292    
3293      default:      default:
3294      return FALSE;      return FALSE;
3295      }      }
3296    
3297      /* In principle, support for Unicode properties should be integrated here as
3298      well. It means re-organizing the above code so as to get hold of the property
3299      values before switching on the op-code. However, I wonder how many patterns
3300      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3301      these op-codes are never generated.) */
3302    
3303    case OP_DIGIT:    case OP_DIGIT:
3304    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3305           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3306    
3307    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3308    return next == -ESC_d;    return next == -ESC_d;
3309    
3310    case OP_WHITESPACE:    case OP_WHITESPACE:
3311    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3312    
3313    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3314    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3315    
3316    case OP_HSPACE:    case OP_HSPACE:
3317    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3318             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3319    
3320    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3321    return next == -ESC_h;    return next == -ESC_h;
3322    
3323    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3324      case OP_ANYNL:
3325    case OP_VSPACE:    case OP_VSPACE:
3326    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3327    
3328    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3329    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3330    
3331    case OP_WORDCHAR:    case OP_WORDCHAR:
3332    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3333             next == -ESC_v || next == -ESC_R;
3334    
3335    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3336    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2260  Arguments: Line 3359  Arguments:
3359    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3360    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3361    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3362    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3363    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3364    bcptr          points to current branch chain    bcptr          points to current branch chain
3365      cond_depth     conditional nesting depth
3366    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3367    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3368                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2272  Returns:         TRUE on success Line 3372  Returns:         TRUE on success
3372  */  */
3373    
3374  static BOOL  static BOOL
3375  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3376    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3377      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3378    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3379  {  {
3380  int repeat_type, op_type;  int repeat_type, op_type;
3381  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3382  int bravalue = 0;  int bravalue = 0;
3383  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3384  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3385  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3386  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3387  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3388  int after_manual_callout = 0;  int after_manual_callout = 0;
3389  int length_prevgroup = 0;  int length_prevgroup = 0;
3390  register int c;  register int c;
3391  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3392  uschar *last_code = code;  pcre_uchar *last_code = code;
3393  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3394  uschar *tempcode;  pcre_uchar *tempcode;
3395  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3396  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3397  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3398  const uschar *tempptr;  const pcre_uchar *tempptr;
3399  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3400  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3401  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3402  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3403    pcre_uint8 classbits[32];
3404  #ifdef SUPPORT_UTF8  
3405  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3406  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3407  uschar *class_utf8data;  dynamically as we process the pattern. */
3408  uschar utf8_char[6];  
3409    #ifdef SUPPORT_UTF
3410    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3411    BOOL utf = (options & PCRE_UTF8) != 0;
3412    pcre_uchar utf_chars[6];
3413  #else  #else
3414  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3415  uschar *utf8_char = NULL;  #endif
3416    
3417    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3418    
3419    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3420    BOOL xclass;
3421    pcre_uchar *class_uchardata;
3422    pcre_uchar *class_uchardata_base;
3423  #endif  #endif
3424    
3425  #ifdef DEBUG  #ifdef PCRE_DEBUG
3426  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3427  #endif  #endif
3428    
# Line 2321  greedy_non_default = greedy_default ^ 1; Line 3433  greedy_non_default = greedy_default ^ 1;
3433    
3434  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3435  matching encountered yet". It gets changed to REQ_NONE if we hit something that  matching encountered yet". It gets changed to REQ_NONE if we hit something that
3436  matches a non-fixed char first char; reqbyte just remains unset if we never  matches a non-fixed char first char; reqchar just remains unset if we never
3437  find one.  find one.
3438    
3439  When we hit a repeat whose minimum is zero, we may have to adjust these values  When we hit a repeat whose minimum is zero, we may have to adjust these values
3440  to take the zero repeat into account. This is implemented by setting them to  to take the zero repeat into account. This is implemented by setting them to
3441  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3442  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3443    
3444  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3445    
3446  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3447  according to the current setting of the caseless flag. REQ_CASELESS is a bit  or zero, according to the current setting of the caseless flag. The
3448  value > 255. It is added into the firstbyte or reqbyte variables to record the  REQ_CASELESS leaves the lower 28 bit empty. It is added into the
3449  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3450    value. This is used only for ASCII characters. */
3451    
3452  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3453    
3454  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3455    
3456  for (;; ptr++)  for (;; ptr++)
3457    {    {
3458    BOOL negate_class;    BOOL negate_class;
3459      BOOL should_flip_negation;
3460    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3461    BOOL is_quantifier;    BOOL is_quantifier;
3462    BOOL is_recurse;    BOOL is_recurse;
3463    BOOL reset_bracount;    BOOL reset_bracount;
3464    int class_charcount;    int class_has_8bitchar;
3465    int class_lastchar;    int class_single_char;
3466    int newoptions;    int newoptions;
3467    int recno;    int recno;
3468    int refsign;    int refsign;
3469    int skipbytes;    int skipbytes;
3470    int subreqbyte;    int subreqchar;
3471    int subfirstbyte;    int subfirstchar;
3472    int terminator;    int terminator;
3473    int mclength;    int mclength;
3474    uschar mcbuffer[8];    int tempbracount;
3475      pcre_uchar mcbuffer[8];
3476    
3477    /* Get next byte in the pattern */    /* Get next character in the pattern */
3478    
3479    c = *ptr;    c = *ptr;
3480    
3481      /* If we are at the end of a nested substitution, revert to the outer level
3482      string. Nesting only happens one level deep. */
3483    
3484      if (c == 0 && nestptr != NULL)
3485        {
3486        ptr = nestptr;
3487        nestptr = NULL;
3488        c = *ptr;
3489        }
3490    
3491    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3492    previous cycle of this loop. */    previous cycle of this loop. */
3493    
3494    if (lengthptr != NULL)    if (lengthptr != NULL)
3495      {      {
3496  #ifdef DEBUG  #ifdef PCRE_DEBUG
3497      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3498  #endif  #endif
3499      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3500            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3501        {        {
3502        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3503        goto FAILED;        goto FAILED;
# Line 2393  for (;; ptr++) Line 3519  for (;; ptr++)
3519        goto FAILED;        goto FAILED;
3520        }        }
3521    
3522      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3523      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3524          (int)(code - last_code), c, c));
3525    
3526      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3527      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
3528      if "previous" is NULL, reset the current code pointer to the start. */      if "previous" is NULL, reset the current code pointer to the start. */
# Line 2404  for (;; ptr++) Line 3531  for (;; ptr++)
3531        {        {
3532        if (previous > orig_code)        if (previous > orig_code)
3533          {          {
3534          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3535          code -= previous - orig_code;          code -= previous - orig_code;
3536          previous = orig_code;          previous = orig_code;
3537          }          }
# Line 2420  for (;; ptr++) Line 3547  for (;; ptr++)
3547    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3548    reference list. */    reference list. */
3549    
3550    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3551               WORK_SIZE_SAFETY_MARGIN)
3552      {      {
3553      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3554      goto FAILED;      goto FAILED;
# Line 2430  for (;; ptr++) Line 3558  for (;; ptr++)
3558    
3559    if (inescq && c != 0)    if (inescq && c != 0)
3560      {      {
3561      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3562        {        {
3563        inescq = FALSE;        inescq = FALSE;
3564        ptr++;        ptr++;
# Line 2456  for (;; ptr++) Line 3584  for (;; ptr++)
3584    /* 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
3585    a quantifier. */    a quantifier. */
3586    
3587    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3588      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3589        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3590    
3591    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3592         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2467  for (;; ptr++) Line 3596  for (;; ptr++)
3596      previous_callout = NULL;      previous_callout = NULL;
3597      }      }
3598    
3599    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3600    
3601    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3602      {      {
3603      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3604      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3605        {        {
3606        while (*(++ptr) != 0)        ptr++;
3607          while (*ptr != 0)
3608          {          {
3609          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3610            ptr++;
3611    #ifdef SUPPORT_UTF
3612            if (utf) FORWARDCHAR(ptr);
3613    #endif
3614          }          }
3615        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3616    
# Line 2497  for (;; ptr++) Line 3631  for (;; ptr++)
3631      {      {
3632      /* ===================================================================*/      /* ===================================================================*/
3633      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3634      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3635      case ')':      case CHAR_RIGHT_PARENTHESIS:
3636      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3637      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3638      *codeptr = code;      *codeptr = code;
3639      *ptrptr = ptr;      *ptrptr = ptr;
3640      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2510  for (;; ptr++) Line 3644  for (;; ptr++)
3644          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3645          goto FAILED;          goto FAILED;
3646          }          }
3647        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3648        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3649        }        }
3650      return TRUE;      return TRUE;
# Line 2520  for (;; ptr++) Line 3654  for (;; ptr++)
3654      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3655      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3656    
3657      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3658        previous = NULL;
3659      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3660        {        {
3661        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3662          *code++ = OP_CIRCM;
3663        }        }
3664      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3665      break;      break;
3666    
3667      case '$':      case CHAR_DOLLAR_SIGN:
3668      previous = NULL;      previous = NULL;
3669      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3670      break;      break;
3671    
3672      /* 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
3673      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3674    
3675      case '.':      case CHAR_DOT:
3676      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3677      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3678      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3679      previous = code;      previous = code;
3680      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3681      break;      break;
3682    
3683    
# Line 2557  for (;; ptr++) Line 3692  for (;; ptr++)
3692      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,
3693      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3694      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.
     */  
3695    
3696      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3697        default (Perl) mode, it is treated as a data character. */
3698    
3699        case CHAR_RIGHT_SQUARE_BRACKET:
3700        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3701          {
3702          *errorcodeptr = ERR64;
3703          goto FAILED;
3704          }
3705        goto NORMAL_CHAR;
3706    
3707        case CHAR_LEFT_SQUARE_BRACKET:
3708      previous = code;      previous = code;
3709    
3710      /* 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
3711      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. */
3712    
3713      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3714          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3715            check_posix_syntax(ptr, &tempptr))
3716        {        {
3717        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3718        goto FAILED;        goto FAILED;
3719        }        }
3720    
# Line 2580  for (;; ptr++) Line 3726  for (;; ptr++)
3726      for (;;)      for (;;)
3727        {        {
3728        c = *(++ptr);        c = *(++ptr);
3729        if (c == '\\')        if (c == CHAR_BACKSLASH)
3730          {          {
3731          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3732            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3733              else break;          else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
3734              ptr += 3;
3735            else
3736              break;
3737          }          }
3738        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3739          negate_class = TRUE;          negate_class = TRUE;
3740        else break;        else break;
3741        }        }
3742    
3743      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3744      of just a single character (as long as it's < 256). However, For higher      an initial ']' is taken as a data character -- the code below handles
3745      valued UTF-8 characters, we don't yet do any optimization. */      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3746        [^] must match any character, so generate OP_ALLANY. */
3747    
3748        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3749            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3750          {
3751          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3752          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3753          zerofirstchar = firstchar;
3754          break;
3755          }
3756    
3757        /* If a class contains a negative special such as \S, we need to flip the
3758        negation flag at the end, so that support for characters > 255 works
3759        correctly (they are all included in the class). */
3760    
3761        should_flip_negation = FALSE;
3762    
3763        /* For optimization purposes, we track some properties of the class.
3764        class_has_8bitchar will be non-zero, if the class contains at least one
3765        < 256 character. class_single_char will be 1 if the class contains only
3766        a single character. */
3767    
3768      class_charcount = 0;      class_has_8bitchar = 0;
3769      class_lastchar = -1;      class_single_char = 0;
3770    
3771      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
3772      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains only 1 character (less
3773      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3774      */      */
3775    
3776      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3777    
3778  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3779      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3780      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3781        class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3782  #endif  #endif
3783    
3784      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2616  for (;; ptr++) Line 3787  for (;; ptr++)
3787    
3788      if (c != 0) do      if (c != 0) do
3789        {        {
3790        const uschar *oldptr;        const pcre_uchar *oldptr;
3791    
3792  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3793        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3794          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3795          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3796          }          }
3797  #endif  #endif
3798    
3799    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3800          /* In the pre-compile phase, accumulate the length of any extra
3801          data and reset the pointer. This is so that very large classes that
3802          contain a zillion > 255 characters no longer overwrite the work space
3803          (which is on the stack). */
3804    
3805          if (lengthptr != NULL)
3806            {
3807            *lengthptr += class_uchardata - class_uchardata_base;
3808            class_uchardata = class_uchardata_base;
3809            }
3810    #endif
3811    
3812        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3813    
3814        if (inescq)        if (inescq)
3815          {          {
3816          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3817            {            {
3818            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3819            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2644  for (;; ptr++) Line 3828  for (;; ptr++)
3828        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3829        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3830    
3831        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3832            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3833            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3834          {          {
3835          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3836          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3837          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3838          uschar pbits[32];          pcre_uint8 pbits[32];
3839    
3840          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3841            {            {
3842            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3843            goto FAILED;            goto FAILED;
3844            }            }
3845    
3846          ptr += 2;          ptr += 2;
3847          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3848            {            {
3849            local_negate = TRUE;            local_negate = TRUE;
3850              should_flip_negation = TRUE;  /* Note negative special */
3851            ptr++;            ptr++;
3852            }            }
3853    
3854          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3855          if (posix_class < 0)          if (posix_class < 0)
3856            {            {
3857            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2680  for (;; ptr++) Line 3865  for (;; ptr++)
3865          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3866            posix_class = 0;            posix_class = 0;
3867    
3868          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3869          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3870          subtract bits that may be in the main map already. At the end we or the  
3871          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3872            if ((options & PCRE_UCP) != 0)
3873              {
3874              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3875              if (posix_substitutes[pc] != NULL)
3876                {
3877                nestptr = tempptr + 1;
3878                ptr = posix_substitutes[pc] - 1;
3879                continue;
3880                }
3881              }
3882    #endif
3883            /* In the non-UCP case, we build the bit map for the POSIX class in a
3884            chunk of local store because we may be adding and subtracting from it,
3885            and we don't want to subtract bits that may be in the main map already.
3886            At the end we or the result into the bit map that is being built. */
3887    
3888          posix_class *= 3;          posix_class *= 3;
3889    
3890          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3891    
3892          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3893            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3894    
3895          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3896    
# Line 2721  for (;; ptr++) Line 3921  for (;; ptr++)
3921            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3922    
3923          ptr = tempptr + 1;          ptr = tempptr + 1;
3924          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3925            class_has_8bitchar = 1;
3926            /* Every class contains at least two characters. */
3927            class_single_char = 2;
3928          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3929          }          }
3930    
3931        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3932        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3933        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3934        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so
3935        to 'or' into the one we are building. We assume they have more than one        speculatively set both class_has_8bitchar and class_single_char bigger
3936        character in them, so set class_charcount bigger than one. */        than one. Unrecognized escapes fall through and are either treated
3937          as literal characters (by default), or are faulted if
3938          PCRE_EXTRA is set. */
3939    
3940        if (c == '\\')        if (c == CHAR_BACKSLASH)
3941          {          {
3942          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3943 &