/[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 802 by ph10, Tue Dec 13 09:52:20 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_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 87  so this number is very generous. Line 88  so this number is very generous.
88  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
89  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
90  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
92    filled up by repetitions of forward references, for example patterns like
93    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
94    that the workspace is expanded using malloc() in this situation. The value
95    below is therefore a minimum, and we put a maximum on it for safety. The
96    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
97    kicks in at the same number of forward references in all cases. */
98    
99  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
100    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
101    
102    /* The overrun tests check for a slightly smaller size so that they detect the
103    overrun before it actually does run off the end of the data block. */
104    
105    #define WORK_SIZE_SAFETY_MARGIN (100)
106    
107    /* Private flags added to firstchar and reqchar. */
108    
109    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
110    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
111    
112    /* Repeated character flags. */
113    
114    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
115    
116  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
117  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
118  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
119  is invalid. */  is invalid. */
120    
121  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
122    
123    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
124    in UTF-8 mode. */
125    
126  static const short int escapes[] = {  static const short int escapes[] = {
127       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
128       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
129     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
130  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
131  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
132  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
133     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
134  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
135  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
136       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
137         -ESC_D,                  -ESC_E,
138         0,                       -ESC_G,
139         -ESC_H,                  0,
140         0,                       -ESC_K,
141         0,                       0,
142         -ESC_N,                  0,
143         -ESC_P,                  -ESC_Q,
144         -ESC_R,                  -ESC_S,
145         0,                       0,
146         -ESC_V,                  -ESC_W,
147         -ESC_X,                  0,
148         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
149         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
150         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
151         CHAR_GRAVE_ACCENT,       7,
152         -ESC_b,                  0,
153         -ESC_d,                  ESC_e,
154         ESC_f,                   0,
155         -ESC_h,                  0,
156         0,                       -ESC_k,
157         0,                       0,
158         ESC_n,                   0,
159         -ESC_p,                  0,
160         ESC_r,                   -ESC_s,
161         ESC_tee,                 0,
162         -ESC_v,                  -ESC_w,
163         0,                       0,
164         -ESC_z
165  };  };
166    
167  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
168    
169    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
170    
171  static const short int escapes[] = {  static const short int escapes[] = {
172  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
173  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 186  static const short int escapes[] = {
186  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
187  /*  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,
188  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
189  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
190  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
191  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
192  /*  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 196  static const short int escapes[] = {
196  #endif  #endif
197    
198    
199  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
200    searched linearly. Put all the names into a single string, in order to reduce
201    the number of relocations when a shared library is dynamically linked. The
202    string is built from string macros so that it works in UTF-8 mode on EBCDIC
203    platforms. */
204    
205  typedef struct verbitem {  typedef struct verbitem {
206    const char *name;    int   len;                 /* Length of verb name */
207    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
208    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
209  } verbitem;  } verbitem;
210    
211  static verbitem verbs[] = {  static const char verbnames[] =
212    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
213    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
214    { "F",      1, OP_FAIL },    STRING_ACCEPT0
215    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
216    { "PRUNE",  5, OP_PRUNE },    STRING_F0
217    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
218    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
219      STRING_SKIP0
220      STRING_THEN;
221    
222    static const verbitem verbs[] = {
223      { 0, -1,        OP_MARK },
224      { 4, -1,        OP_MARK },
225      { 6, OP_ACCEPT, -1 },
226      { 6, OP_COMMIT, -1 },
227      { 1, OP_FAIL,   -1 },
228      { 4, OP_FAIL,   -1 },
229      { 5, OP_PRUNE,  OP_PRUNE_ARG },
230      { 4, OP_SKIP,   OP_SKIP_ARG  },
231      { 4, OP_THEN,   OP_THEN_ARG  }
232  };  };
233    
234  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
235    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
236    
237  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
238    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
239    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
240    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
241    for handling case independence. */
242    
243    static const char posix_names[] =
244      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
245      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
246      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
247      STRING_word0  STRING_xdigit;
248    
249  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
250    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 };
251    
252  /* 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 276  static const int posix_class_maps[] = {
276    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
277  };  };
278    
279    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
280    substitutes must be in the order of the names, defined above, and there are
281    both positive and negative cases. NULL means no substitute. */
282    
283    #ifdef SUPPORT_UCP
284    static const pcre_uchar string_PNd[]  = {
285      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
286      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
287    static const pcre_uchar string_pNd[]  = {
288      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
289      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
290    static const pcre_uchar string_PXsp[] = {
291      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
292      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
293    static const pcre_uchar string_pXsp[] = {
294      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
295      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
296    static const pcre_uchar string_PXwd[] = {
297      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
298      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
299    static const pcre_uchar string_pXwd[] = {
300      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
301      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
302    
303    static const pcre_uchar *substitutes[] = {
304      string_PNd,           /* \D */
305      string_pNd,           /* \d */
306      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
307      string_pXsp,          /* \s */
308      string_PXwd,          /* \W */
309      string_pXwd           /* \w */
310    };
311    
312    static const pcre_uchar string_pL[] =   {
313      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_pLl[] =  {
316      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pLu[] =  {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    static const pcre_uchar string_pXan[] = {
322      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
323      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324    static const pcre_uchar string_h[] =    {
325      CHAR_BACKSLASH, CHAR_h, '\0' };
326    static const pcre_uchar string_pXps[] = {
327      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
328      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
329    static const pcre_uchar string_PL[] =   {
330      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
331      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
332    static const pcre_uchar string_PLl[] =  {
333      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
334      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
335    static const pcre_uchar string_PLu[] =  {
336      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
337      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
338    static const pcre_uchar string_PXan[] = {
339      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
340      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
341    static const pcre_uchar string_H[] =    {
342      CHAR_BACKSLASH, CHAR_H, '\0' };
343    static const pcre_uchar string_PXps[] = {
344      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *posix_substitutes[] = {
348      string_pL,            /* alpha */
349      string_pLl,           /* lower */
350      string_pLu,           /* upper */
351      string_pXan,          /* alnum */
352      NULL,                 /* ascii */
353      string_h,             /* blank */
354      NULL,                 /* cntrl */
355      string_pNd,           /* digit */
356      NULL,                 /* graph */
357      NULL,                 /* print */
358      NULL,                 /* punct */
359      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
360      string_pXwd,          /* word */
361      NULL,                 /* xdigit */
362      /* Negated cases */
363      string_PL,            /* ^alpha */
364      string_PLl,           /* ^lower */
365      string_PLu,           /* ^upper */
366      string_PXan,          /* ^alnum */
367      NULL,                 /* ^ascii */
368      string_H,             /* ^blank */
369      NULL,                 /* ^cntrl */
370      string_PNd,           /* ^digit */
371      NULL,                 /* ^graph */
372      NULL,                 /* ^print */
373      NULL,                 /* ^punct */
374      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
375      string_PXwd,          /* ^word */
376      NULL                  /* ^xdigit */
377    };
378    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
379    #endif
380    
381  #define STRING(a)  # a  #define STRING(a)  # a
382  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 207  static const int posix_class_maps[] = { Line 384  static const int posix_class_maps[] = {
384  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
385  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
386  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
387  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
388    the number of relocations needed when a shared library is loaded dynamically,
389  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
390    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
391    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
392    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
393    "unrecognized character follows \\",  
394    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
395    substring, so that the whole string ends with \0\0, which can be detected when
396    counting through. */
397    
398    static const char error_texts[] =
399      "no error\0"
400      "\\ at end of pattern\0"
401      "\\c at end of pattern\0"
402      "unrecognized character follows \\\0"
403      "numbers out of order in {} quantifier\0"
404    /* 5 */    /* 5 */
405    "number too big in {} quantifier",    "number too big in {} quantifier\0"
406    "missing terminating ] for character class",    "missing terminating ] for character class\0"
407    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
408    "range out of order in character class",    "range out of order in character class\0"
409    "nothing to repeat",    "nothing to repeat\0"
410    /* 10 */    /* 10 */
411    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
412    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
413    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
414    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
415    "missing )",    "missing )\0"
416    /* 15 */    /* 15 */
417    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
418    "erroffset passed as NULL",    "erroffset passed as NULL\0"
419    "unknown option bit(s) set",    "unknown option bit(s) set\0"
420    "missing ) after comment",    "missing ) after comment\0"
421    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
422    /* 20 */    /* 20 */
423    "regular expression is too large",    "regular expression is too large\0"
424    "failed to get memory",    "failed to get memory\0"
425    "unmatched parentheses",    "unmatched parentheses\0"
426    "internal error: code overflow",    "internal error: code overflow\0"
427    "unrecognized character after (?<",    "unrecognized character after (?<\0"
428    /* 25 */    /* 25 */
429    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
430    "malformed number or name after (?(",    "malformed number or name after (?(\0"
431    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
432    "assertion expected after (?(",    "assertion expected after (?(\0"
433    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
434    /* 30 */    /* 30 */
435    "unknown POSIX class name",    "unknown POSIX class name\0"
436    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
437    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
438    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
439    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
440    /* 35 */    /* 35 */
441    "invalid condition (?(0)",    "invalid condition (?(0)\0"
442    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
443    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
444    "number after (?C is > 255",    "number after (?C is > 255\0"
445    "closing ) for (?C expected",    "closing ) for (?C expected\0"
446    /* 40 */    /* 40 */
447    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
448    "unrecognized character after (?P",    "unrecognized character after (?P\0"
449    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
450    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
451    "invalid UTF-8 string",    "invalid UTF-8 string\0"
452    /* 45 */    /* 45 */
453    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
454    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
455    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
456    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
457    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
458    /* 50 */    /* 50 */
459    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
460    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
461    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
462    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
463    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
464    /* 55 */    /* 55 */
465    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
466    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
467    "\\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"
468    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
469    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
470    /* 60 */    /* 60 */
471    "(*VERB) not recognized",    "(*VERB) not recognized\0"
472    "number is too big"    "number is too big\0"
473  };    "subpattern name expected\0"
474      "digit expected after (?+\0"
475      "] is an invalid data character in JavaScript compatibility mode\0"
476      /* 65 */
477      "different names for subpatterns of the same number are not allowed\0"
478      "(*MARK) must have an argument\0"
479      "this version of PCRE is not compiled with PCRE_UCP support\0"
480      "\\c must be followed by an ASCII character\0"
481      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
482      /* 70 */
483      "internal error: unknown opcode in find_fixedlength()\0"
484      "\\N is not supported in a class\0"
485      "too many forward references\0"
486      "disallowed UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
487      ;
488    
489  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
490  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 502  For convenience, we use the same bit def
502    
503  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
504    
505  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
506  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
507    into a subtraction and unsigned comparison). */
508    
509    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
510    
511    #ifndef EBCDIC
512    
513    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
514    UTF-8 mode. */
515    
516    static const pcre_uint8 digitab[] =
517    {    {
518    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
519    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 548  static const unsigned char digitab[] =
548    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
549    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
550    
551  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
552  static const unsigned char digitab[] =  
553    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
554    
555    static const pcre_uint8 digitab[] =
556    {    {
557    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
558    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 587  static const unsigned char digitab[] =
587    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
588    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
589    
590  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
591    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
592    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
593    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 626  static const unsigned char ebcdic_charta
626  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
627    
628  static BOOL  static BOOL
629    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
630      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
631    
632    
633    
634  /*************************************************  /*************************************************
635    *            Find an error text                  *
636    *************************************************/
637    
638    /* The error texts are now all in one long string, to save on relocations. As
639    some of the text is of unknown length, we can't use a table of offsets.
640    Instead, just count through the strings. This is not a performance issue
641    because it happens only when there has been a compilation error.
642    
643    Argument:   the error number
644    Returns:    pointer to the error string
645    */
646    
647    static const char *
648    find_error_text(int n)
649    {
650    const char *s = error_texts;
651    for (; n > 0; n--)
652      {
653      while (*s++ != 0) {};
654      if (*s == 0) return "Error text not found (please report)";
655      }
656    return s;
657    }
658    
659    
660    /*************************************************
661    *           Expand the workspace                 *
662    *************************************************/
663    
664    /* This function is called during the second compiling phase, if the number of
665    forward references fills the existing workspace, which is originally a block on
666    the stack. A larger block is obtained from malloc() unless the ultimate limit
667    has been reached or the increase will be rather small.
668    
669    Argument: pointer to the compile data block
670    Returns:  0 if all went well, else an error number
671    */
672    
673    static int
674    expand_workspace(compile_data *cd)
675    {
676    pcre_uchar *newspace;
677    int newsize = cd->workspace_size * 2;
678    
679    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
680    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
681        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
682     return ERR72;
683    
684    newspace = (pcre_malloc)(newsize);
685    if (newspace == NULL) return ERR21;
686    
687    memcpy(newspace, cd->start_workspace, cd->workspace_size);
688    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
689    if (cd->workspace_size > COMPILE_WORK_SIZE)
690      (pcre_free)((void *)cd->start_workspace);
691    cd->start_workspace = newspace;
692    cd->workspace_size = newsize;
693    return 0;
694    }
695    
696    
697    
698    /*************************************************
699    *            Check for counted repeat            *
700    *************************************************/
701    
702    /* This function is called when a '{' is encountered in a place where it might
703    start a quantifier. It looks ahead to see if it really is a quantifier or not.
704    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
705    where the ddds are digits.
706    
707    Arguments:
708      p         pointer to the first char after '{'
709    
710    Returns:    TRUE or FALSE
711    */
712    
713    static BOOL
714    is_counted_repeat(const pcre_uchar *p)
715    {
716    if (!IS_DIGIT(*p)) return FALSE;
717    p++;
718    while (IS_DIGIT(*p)) p++;
719    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
720    
721    if (*p++ != CHAR_COMMA) return FALSE;
722    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
723    
724    if (!IS_DIGIT(*p)) return FALSE;
725    p++;
726    while (IS_DIGIT(*p)) p++;
727    
728    return (*p == CHAR_RIGHT_CURLY_BRACKET);
729    }
730    
731    
732    
733    /*************************************************
734  *            Handle escapes                      *  *            Handle escapes                      *
735  *************************************************/  *************************************************/
736    
# Line 444  Returns:         zero or positive => a d Line 755  Returns:         zero or positive => a d
755  */  */
756    
757  static int  static int
758  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
759    int options, BOOL isclass)    int options, BOOL isclass)
760  {  {
761  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
762  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
763  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
764    pcre_int32 c;
765    int i;
766    
767  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
768  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 458  ptr--;                            /* Set Line 771  ptr--;                            /* Set
771    
772  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
773    
774  /* 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
775  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.
776  Otherwise further processing may be required. */  Otherwise further processing may be required. */
777    
778  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
779  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
780  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
781    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
782    
783  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
784  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
785    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
786  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
787  #endif  #endif
788    
# Line 475  else if ((i = escapes[c - 0x48]) != 0) Line 790  else if ((i = escapes[c - 0x48]) != 0)
790    
791  else  else
792    {    {
793    const uschar *oldptr;    const pcre_uchar *oldptr;
794    BOOL braced, negated;    BOOL braced, negated;
795    
796    switch (c)    switch (c)
# Line 483  else Line 798  else
798      /* 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
799      error. */      error. */
800    
801      case 'l':      case CHAR_l:
802      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
803      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
804      break;      break;
805    
806      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
807      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
808      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
809      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
810      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
811      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
812            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
813      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
814      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
815        {          {
816        const uschar *p;          c = 0;
817        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
818          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
819        if (*p != 0 && *p != '}')            register int cc = *(++ptr);
820    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
821              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
822              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
823    #else           /* EBCDIC coding */
824              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
825              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
826    #endif
827              }
828            }
829          }
830        else
831          *errorcodeptr = ERR37;
832        break;
833    
834        case CHAR_U:
835        /* In JavaScript, \U is an uppercase U letter. */
836        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
837        break;
838    
839        /* In a character class, \g is just a literal "g". Outside a character
840        class, \g must be followed by one of a number of specific things:
841    
842        (1) A number, either plain or braced. If positive, it is an absolute
843        backreference. If negative, it is a relative backreference. This is a Perl
844        5.10 feature.
845    
846        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
847        is part of Perl's movement towards a unified syntax for back references. As
848        this is synonymous with \k{name}, we fudge it up by pretending it really
849        was \k.
850    
851        (3) For Oniguruma compatibility we also support \g followed by a name or a
852        number either in angle brackets or in single quotes. However, these are
853        (possibly recursive) subroutine calls, _not_ backreferences. Just return
854        the -ESC_g code (cf \k). */
855    
856        case CHAR_g:
857        if (isclass) break;
858        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
859          {
860          c = -ESC_g;
861          break;
862          }
863    
864        /* Handle the Perl-compatible cases */
865    
866        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
867          {
868          const pcre_uchar *p;
869          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
870            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
871          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
872          {          {
873          c = -ESC_k;          c = -ESC_k;
874          break;          break;
# Line 514  else Line 878  else
878        }        }
879      else braced = FALSE;      else braced = FALSE;
880    
881      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
882        {        {
883        negated = TRUE;        negated = TRUE;
884        ptr++;        ptr++;
885        }        }
886      else negated = FALSE;      else negated = FALSE;
887    
888        /* The integer range is limited by the machine's int representation. */
889      c = 0;      c = 0;
890      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
       c = c * 10 + *(++ptr) - '0';  
   
     if (c < 0)  
891        {        {
892          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
893            {
894            c = -1;
895            break;
896            }
897          c = c * 10 + *(++ptr) - CHAR_0;
898          }
899        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
900          {
901          while (IS_DIGIT(ptr[1]))
902            ptr++;
903        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
904        break;        break;
905        }        }
906    
907      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
908        {        {
909        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
910        break;        break;
911        }        }
912    
913        if (c == 0)
914          {
915          *errorcodeptr = ERR58;
916          break;
917          }
918    
919      if (negated)      if (negated)
920        {        {
921        if (c > bracount)        if (c > bracount)
# Line 562  else Line 941  else
941      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
942      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
943    
944      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:
945      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
946    
947      if (!isclass)      if (!isclass)
948        {        {
949        oldptr = ptr;        oldptr = ptr;
950        c -= '0';        /* The integer range is limited by the machine's int representation. */
951        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
952          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
       if (c < 0)  
953          {          {
954            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
955              {
956              c = -1;
957              break;
958              }
959            c = c * 10 + *(++ptr) - CHAR_0;
960            }
961          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
962            {
963            while (IS_DIGIT(ptr[1]))
964              ptr++;
965          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
966          break;          break;
967          }          }
# Line 588  else Line 977  else
977      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.
978      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
979    
980      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
981        {        {
982        ptr--;        ptr--;
983        c = 0;        c = 0;
# Line 601  else Line 990  else
990      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
991      than 3 octal digits. */      than 3 octal digits. */
992    
993      case '0':      case CHAR_0:
994      c -= '0';      c -= CHAR_0;
995      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
996          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
997      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
998      break;      break;
999    
1000      /* \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
1001      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.
1002      treated as a data character. */      If not, { is treated as a data character. */
1003    
1004        case CHAR_x:
1005        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1006          {
1007          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1008          Otherwise it is a lowercase x letter. */
1009          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1010            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1011            {
1012            c = 0;
1013            for (i = 0; i < 2; ++i)
1014              {
1015              register int cc = *(++ptr);
1016    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1018              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1019    #else           /* EBCDIC coding */
1020              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1021              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1022    #endif
1023              }
1024            }
1025          break;
1026          }
1027    
1028      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1029        {        {
1030        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1031    
1032        c = 0;        c = 0;
1033        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1034          {          {
1035          register int cc = *pt++;          register int cc = *pt++;
1036          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1037    
1038  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1039          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1040          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1041  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1042          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1043          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1044    #endif
1045    
1046    #ifdef COMPILE_PCRE8
1047            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1048    #else
1049    #ifdef COMPILE_PCRE16
1050            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1051    #endif
1052  #endif  #endif
1053          }          }
1054    
1055        if (*pt == '}')        if (c < 0)
1056            {
1057            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1058            *errorcodeptr = ERR34;
1059            }
1060    
1061          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1062          {          {
1063          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1064          ptr = pt;          ptr = pt;
1065          break;          break;
1066          }          }
# Line 648  else Line 1072  else
1072      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1073    
1074      c = 0;      c = 0;
1075      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1076        {        {
1077        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1078        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1079  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1080        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1081        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1082  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1083        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1084        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1085  #endif  #endif
1086        }        }
1087      break;      break;
1088    
1089      /* 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.
1090      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
1091        coding is ASCII-specific, but then the whole concept of \cx is
1092      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1093    
1094      case 'c':      case CHAR_c:
1095      c = *(++ptr);      c = *(++ptr);
1096      if (c == 0)      if (c == 0)
1097        {        {
1098        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1099        break;        break;
1100        }        }
1101    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1102  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1103      if (c >= 'a' && c <= 'z') c -= 32;        {
1104          *errorcodeptr = ERR68;
1105          break;
1106          }
1107        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1108      c ^= 0x40;      c ^= 0x40;
1109  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1110      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1111      c ^= 0xC0;      c ^= 0xC0;
1112  #endif  #endif
1113      break;      break;
1114    
1115      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1116      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1117      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
1118      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
1119      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1120    
1121      default:      default:
1122      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 1129  else
1129      }      }
1130    }    }
1131    
1132    /* Perl supports \N{name} for character names, as well as plain \N for "not
1133    newline". PCRE does not support \N{name}. However, it does support
1134    quantification such as \N{2,3}. */
1135    
1136    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1137         !is_counted_repeat(ptr+2))
1138      *errorcodeptr = ERR37;
1139    
1140    /* If PCRE_UCP is set, we change the values for \d etc. */
1141    
1142    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1143      c -= (ESC_DU - ESC_D);
1144    
1145    /* Set the pointer to the final character before returning. */
1146    
1147  *ptrptr = ptr;  *ptrptr = ptr;
1148  return c;  return c;
1149  }  }
# Line 726  Returns:         type value from ucp_typ Line 1170  Returns:         type value from ucp_typ
1170  */  */
1171    
1172  static int  static int
1173  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1174  {  {
1175  int c, i, bot, top;  int c, i, bot, top;
1176  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1177  char name[32];  pcre_uchar name[32];
1178    
1179  c = *(++ptr);  c = *(++ptr);
1180  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 740  if (c == 0) goto ERROR_RETURN; Line 1184  if (c == 0) goto ERROR_RETURN;
1184  /* \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
1185  negation. */  negation. */
1186    
1187  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1188    {    {
1189    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1190      {      {
1191      *negptr = TRUE;      *negptr = TRUE;
1192      ptr++;      ptr++;
# Line 751  if (c == '{') Line 1195  if (c == '{')
1195      {      {
1196      c = *(++ptr);      c = *(++ptr);
1197      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1198      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1199      name[i] = c;      name[i] = c;
1200      }      }
1201    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1202    name[i] = 0;    name[i] = 0;
1203    }    }
1204    
# Line 771  else Line 1215  else
1215  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1216    
1217  bot = 0;  bot = 0;
1218  top = _pcre_utt_size;  top = PRIV(utt_size);
1219    
1220  while (bot < top)  while (bot < top)
1221    {    {
1222    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1223    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1224    if (c == 0)    if (c == 0)
1225      {      {
1226      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1227      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1228      }      }
1229    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1230    }    }
# Line 800  return -1; Line 1244  return -1;
1244    
1245    
1246  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1247  *         Read repeat counts                     *  *         Read repeat counts                     *
1248  *************************************************/  *************************************************/
1249    
# Line 851  Returns:         pointer to '}' on succe Line 1262  Returns:         pointer to '}' on succe
1262                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1263  */  */
1264    
1265  static const uschar *  static const pcre_uchar *
1266  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)
1267  {  {
1268  int min = 0;  int min = 0;
1269  int max = -1;  int max = -1;
# Line 860  int max = -1; Line 1271  int max = -1;
1271  /* 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
1272  an integer overflow. */  an integer overflow. */
1273    
1274  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1275  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1276    {    {
1277    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1281  if (min < 0 || min > 65535)
1281  /* 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.
1282  Also, max must not be less than min. */  Also, max must not be less than min. */
1283    
1284  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1285    {    {
1286    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1287      {      {
1288      max = 0;      max = 0;
1289      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1290      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1291        {        {
1292        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1311  return p;
1311    
1312    
1313  /*************************************************  /*************************************************
1314  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1315  *************************************************/  *************************************************/
1316    
1317  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1318    top-level call starts at the beginning of the pattern. All other calls must
1319    start at a parenthesis. It scans along a pattern's text looking for capturing
1320  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
1321  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
1322  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
1323  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1324  be terminated by '>' because that is checked in the first pass.  
1325    This function was originally called only from the second pass, in which we know
1326    that if (?< or (?' or (?P< is encountered, the name will be correctly
1327    terminated because that is checked in the first pass. There is now one call to
1328    this function in the first pass, to check for a recursive back reference by
1329    name (so that we can make the whole group atomic). In this case, we need check
1330    only up to the current position in the pattern, and that is still OK because
1331    and previous occurrences will have been checked. To make this work, the test
1332    for "end of pattern" is a check against cd->end_pattern in the main loop,
1333    instead of looking for a binary zero. This means that the special first-pass
1334    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1335    processing items within the loop are OK, because afterwards the main loop will
1336    terminate.)
1337    
1338  Arguments:  Arguments:
1339    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1340    count        current count of capturing parens so far encountered    cd           compile background data
1341    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1342    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1343    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1344      utf          TRUE if we are in UTF-8 / UTF-16 mode
1345      count        pointer to the current capturing subpattern number (updated)
1346    
1347  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1348  */  */
1349    
1350  static int  static int
1351  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,
1352    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1353  {  {
1354  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1355    int start_count = *count;
1356    int hwm_count = start_count;
1357    BOOL dup_parens = FALSE;
1358    
1359    /* If the first character is a parenthesis, check on the type of group we are
1360    dealing with. The very first call may not start with a parenthesis. */
1361    
1362  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1363    {    {
1364    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1365    
1366      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1367    
1368      /* Handle a normal, unnamed capturing parenthesis. */
1369    
1370      else if (ptr[1] != CHAR_QUESTION_MARK)
1371        {
1372        *count += 1;
1373        if (name == NULL && *count == lorn) return *count;
1374        ptr++;
1375        }
1376    
1377      /* All cases now have (? at the start. Remember when we are in a group
1378      where the parenthesis numbers are duplicated. */
1379    
1380      else if (ptr[2] == CHAR_VERTICAL_LINE)
1381        {
1382        ptr += 3;
1383        dup_parens = TRUE;
1384        }
1385    
1386      /* Handle comments; all characters are allowed until a ket is reached. */
1387    
1388      else if (ptr[2] == CHAR_NUMBER_SIGN)
1389        {
1390        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1391        goto FAIL_EXIT;
1392        }
1393    
1394      /* Handle a condition. If it is an assertion, just carry on so that it
1395      is processed as normal. If not, skip to the closing parenthesis of the
1396      condition (there can't be any nested parens). */
1397    
1398      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1399        {
1400        ptr += 2;
1401        if (ptr[1] != CHAR_QUESTION_MARK)
1402          {
1403          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1404          if (*ptr != 0) ptr++;
1405          }
1406        }
1407    
1408      /* Start with (? but not a condition. */
1409    
1410      else
1411        {
1412        ptr += 2;
1413        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1414    
1415        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1416    
1417        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1418            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1419          {
1420          int term;
1421          const pcre_uchar *thisname;
1422          *count += 1;
1423          if (name == NULL && *count == lorn) return *count;
1424          term = *ptr++;
1425          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1426          thisname = ptr;
1427          while (*ptr != term) ptr++;
1428          if (name != NULL && lorn == ptr - thisname &&
1429              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1430            return *count;
1431          term++;
1432          }
1433        }
1434      }
1435    
1436    /* Past any initial parenthesis handling, scan for parentheses or vertical
1437    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1438    first-pass call when this value is temporarily adjusted to stop at the current
1439    position. So DO NOT change this to a test for binary zero. */
1440    
1441    for (; ptr < cd->end_pattern; ptr++)
1442      {
1443    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1444    
1445    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1446      {      {
1447      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1448      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1449        {        {
1450        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1451        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1452        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1453        }        }
1454      continue;      continue;
1455      }      }
1456    
1457    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1458      are handled for real. If the first character is '^', skip it. Also, if the
1459      first few characters (either before or after ^) are \Q\E or \E we skip them
1460      too. This makes for compatibility with Perl. Note the use of STR macros to
1461      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1462    
1463    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1464      {      {
1465      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1466        for (;;)
1467          {
1468          if (ptr[1] == CHAR_BACKSLASH)
1469            {
1470            if (ptr[2] == CHAR_E)
1471              ptr+= 2;
1472            else if (STRNCMP_UC_C8(ptr + 2,
1473                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1474              ptr += 4;
1475            else
1476              break;
1477            }
1478          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1479            {
1480            negate_class = TRUE;
1481            ptr++;
1482            }
1483          else break;
1484          }
1485    
1486        /* If the next character is ']', it is a data character that must be
1487        skipped, except in JavaScript compatibility mode. */
1488    
1489        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1490            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1491          ptr++;
1492    
1493        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1494        {        {
1495        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1496        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1497          {          {
1498          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1499          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1500            {            {
1501            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1502            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1503            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1504            }            }
1505          continue;          continue;
1506          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1510  for (; *ptr != 0; ptr++)
1510    
1511    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1512    
1513    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1514      {      {
1515      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1516      if (*ptr == 0) return -1;      while (*ptr != 0)
1517          {
1518          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1519          ptr++;
1520    #ifdef SUPPORT_UTF
1521          if (utf) FORWARDCHAR(ptr);
1522    #endif
1523          }
1524        if (*ptr == 0) goto FAIL_EXIT;
1525      continue;      continue;
1526      }      }
1527    
1528    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1529    
1530    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1531      {      {
1532      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1533      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1534      continue;      if (*ptr == 0) goto FAIL_EXIT;
1535        }
1536    
1537      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1538        {
1539        if (dup_parens && *count < hwm_count) *count = hwm_count;
1540        goto FAIL_EXIT;
1541        }
1542    
1543      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1544        {
1545        if (*count > hwm_count) hwm_count = *count;
1546        *count = start_count;
1547      }      }
1548      }
1549    
1550    ptr += 2;  FAIL_EXIT:
1551    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *ptrptr = ptr;
1552    return -1;
1553    }
1554    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1555    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1556    
   count++;  
1557    
1558    if (name == NULL && count == lorn) return count;  /*************************************************
1559    term = *ptr++;  *       Find forward referenced subpattern       *
1560    if (term == '<') term = '>';  *************************************************/
1561    thisname = ptr;  
1562    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1563    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1564        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1565      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1566    references to subpatterns. We used to be able to start this scan from the
1567    current compiling point, using the current count value from cd->bracount, and
1568    do it all in a single loop, but the addition of the possibility of duplicate
1569    subpattern numbers means that we have to scan from the very start, in order to
1570    take account of such duplicates, and to use a recursive function to keep track
1571    of the different types of group.
1572    
1573    Arguments:
1574      cd           compile background data
1575      name         name to seek, or NULL if seeking a numbered subpattern
1576      lorn         name length, or subpattern number if name is NULL
1577      xmode        TRUE if we are in /x mode
1578      utf          TRUE if we are in UTF-8 / UTF-16 mode
1579    
1580    Returns:       the number of the found subpattern, or -1 if not found
1581    */
1582    
1583    static int
1584    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1585      BOOL utf)
1586    {
1587    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1588    int count = 0;
1589    int rc;
1590    
1591    /* If the pattern does not start with an opening parenthesis, the first call
1592    to find_parens_sub() will scan right to the end (if necessary). However, if it
1593    does start with a parenthesis, find_parens_sub() will return when it hits the
1594    matching closing parens. That is why we have to have a loop. */
1595    
1596    for (;;)
1597      {
1598      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1599      if (rc > 0 || *ptr++ == 0) break;
1600    }    }
1601    
1602  return -1;  return rc;
1603  }  }
1604    
1605    
1606    
1607    
1608  /*************************************************  /*************************************************
1609  *      Find first significant op code            *  *      Find first significant op code            *
1610  *************************************************/  *************************************************/
1611    
1612  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1613  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
1614  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
1615  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
1616  assertions, and also the \b assertion; for others it does not.  does not.
1617    
1618  Arguments:  Arguments:
1619    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  
1620    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1621    
1622  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1623  */  */
1624    
1625  static const uschar*  static const pcre_uchar*
1626  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1627  {  {
1628  for (;;)  for (;;)
1629    {    {
1630    switch ((int)*code)    switch ((int)*code)
1631      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1632      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1633      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1634      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1635      if (!skipassert) return code;      if (!skipassert) return code;
1636      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1637      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1638      break;      break;
1639    
1640      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1060  for (;;) Line 1644  for (;;)
1644    
1645      case OP_CALLOUT:      case OP_CALLOUT:
1646      case OP_CREF:      case OP_CREF:
1647        case OP_NCREF:
1648      case OP_RREF:      case OP_RREF:
1649        case OP_NRREF:
1650      case OP_DEF:      case OP_DEF:
1651      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1652      break;      break;
1653    
1654      default:      default:
# Line 1076  for (;;) Line 1662  for (;;)
1662    
1663    
1664  /*************************************************  /*************************************************
1665  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1666  *************************************************/  *************************************************/
1667    
1668  /* 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,
1669  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.
1670  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
1671    temporarily terminated with OP_END when this function is called.
1672    
1673    This function is called when a backward assertion is encountered, so that if it
1674    fails, the error message can point to the correct place in the pattern.
1675    However, we cannot do this when the assertion contains subroutine calls,
1676    because they can be forward references. We solve this by remembering this case
1677    and doing the check at the end; a flag specifies which mode we are running in.
1678    
1679  Arguments:  Arguments:
1680    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1681    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1682      atend    TRUE if called when the pattern is complete
1683  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1684               or -2 if \C was encountered  
1685    Returns:   the fixed length,
1686                 or -1 if there is no fixed length,
1687                 or -2 if \C was encountered (in UTF-8 mode only)
1688                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1689                 or -4 if an unknown opcode was encountered (internal error)
1690  */  */
1691    
1692  static int  static int
1693  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1694  {  {
1695  int length = -1;  int length = -1;
1696    
1697  register int branchlength = 0;  register int branchlength = 0;
1698  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1699    
1700  /* 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
1701  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 1703  branch, check the length against that of
1703  for (;;)  for (;;)
1704    {    {
1705    int d;    int d;
1706      pcre_uchar *ce, *cs;
1707    register int op = *cc;    register int op = *cc;
1708    switch (op)    switch (op)
1709      {      {
1710        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1711        OP_BRA (normal non-capturing bracket) because the other variants of these
1712        opcodes are all concerned with unlimited repeated groups, which of course
1713        are not of fixed length. */
1714    
1715      case OP_CBRA:      case OP_CBRA:
1716      case OP_BRA:      case OP_BRA:
1717      case OP_ONCE:      case OP_ONCE:
1718        case OP_ONCE_NC:
1719      case OP_COND:      case OP_COND:
1720      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1721      if (d < 0) return d;      if (d < 0) return d;
1722      branchlength += d;      branchlength += d;
1723      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1724      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1725      break;      break;
1726    
1727      /* 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.
1728      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
1729      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
1730        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1731        because they all imply an unlimited repeat. */
1732    
1733      case OP_ALT:      case OP_ALT:
1734      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1735      case OP_END:      case OP_END:
1736        case OP_ACCEPT:
1737        case OP_ASSERT_ACCEPT:
1738      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1739        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1740      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1135  for (;;) Line 1742  for (;;)
1742      branchlength = 0;      branchlength = 0;
1743      break;      break;
1744    
1745        /* A true recursion implies not fixed length, but a subroutine call may
1746        be OK. If the subroutine is a forward reference, we can't deal with
1747        it until the end of the pattern, so return -3. */
1748    
1749        case OP_RECURSE:
1750        if (!atend) return -3;
1751        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1752        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1753        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1754        d = find_fixedlength(cs + 2, utf, atend, cd);
1755        if (d < 0) return d;
1756        branchlength += d;
1757        cc += 1 + LINK_SIZE;
1758        break;
1759    
1760      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1761    
1762      case OP_ASSERT:      case OP_ASSERT:
# Line 1146  for (;;) Line 1768  for (;;)
1768    
1769      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1770    
1771      case OP_REVERSE:      case OP_MARK:
1772        case OP_PRUNE_ARG:
1773        case OP_SKIP_ARG:
1774        case OP_THEN_ARG:
1775        cc += cc[1] + PRIV(OP_lengths)[*cc];
1776        break;
1777    
1778        case OP_CALLOUT:
1779        case OP_CIRC:
1780        case OP_CIRCM:
1781        case OP_CLOSE:
1782        case OP_COMMIT:
1783      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1784      case OP_DEF:      case OP_DEF:
1785      case OP_OPT:      case OP_DOLL:
1786      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1787      case OP_EOD:      case OP_EOD:
1788      case OP_EODN:      case OP_EODN:
1789      case OP_CIRC:      case OP_FAIL:
1790      case OP_DOLL:      case OP_NCREF:
1791        case OP_NRREF:
1792      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1793        case OP_PRUNE:
1794        case OP_REVERSE:
1795        case OP_RREF:
1796        case OP_SET_SOM:
1797        case OP_SKIP:
1798        case OP_SOD:
1799        case OP_SOM:
1800        case OP_THEN:
1801      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1802      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1803      break;      break;
1804    
1805      /* Handle literal characters */      /* Handle literal characters */
1806    
1807      case OP_CHAR:      case OP_CHAR:
1808      case OP_CHARNC:      case OP_CHARI:
1809      case OP_NOT:      case OP_NOT:
1810        case OP_NOTI:
1811      branchlength++;      branchlength++;
1812      cc += 2;      cc += 2;
1813  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1814      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1815  #endif  #endif
1816      break;      break;
1817    
# Line 1182  for (;;) Line 1819  for (;;)
1819      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1820    
1821      case OP_EXACT:      case OP_EXACT:
1822        case OP_EXACTI:
1823        case OP_NOTEXACT:
1824        case OP_NOTEXACTI:
1825      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1826      cc += 4;      cc += 2 + IMM2_SIZE;
1827  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1828      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1829  #endif  #endif
1830      break;      break;
1831    
1832      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1833      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1834      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;
1835      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1836      break;      break;
1837    
1838      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1205  for (;;) Line 1842  for (;;)
1842      cc += 2;      cc += 2;
1843      /* Fall through */      /* Fall through */
1844    
1845        case OP_HSPACE:
1846        case OP_VSPACE:
1847        case OP_NOT_HSPACE:
1848        case OP_NOT_VSPACE:
1849      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1850      case OP_DIGIT:      case OP_DIGIT:
1851      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1212  for (;;) Line 1853  for (;;)
1853      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1854      case OP_WORDCHAR:      case OP_WORDCHAR:
1855      case OP_ANY:      case OP_ANY:
1856        case OP_ALLANY:
1857      branchlength++;      branchlength++;
1858      cc++;      cc++;
1859      break;      break;
1860    
1861      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1862        otherwise \C is coded as OP_ALLANY. */
1863    
1864      case OP_ANYBYTE:      case OP_ANYBYTE:
1865      return -2;      return -2;
1866    
1867      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1868    
1869  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1870      case OP_XCLASS:      case OP_XCLASS:
1871      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1872      /* Fall through */      /* Fall through */
1873  #endif  #endif
1874    
1875      case OP_CLASS:      case OP_CLASS:
1876      case OP_NCLASS:      case OP_NCLASS:
1877      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1878    
1879      switch (*cc)      switch (*cc)
1880        {        {
1881          case OP_CRPLUS:
1882          case OP_CRMINPLUS:
1883        case OP_CRSTAR:        case OP_CRSTAR:
1884        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1885        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1243  for (;;) Line 1888  for (;;)
1888    
1889        case OP_CRRANGE:        case OP_CRRANGE:
1890        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1891        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1892        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1893        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1894        break;        break;
1895    
1896        default:        default:
# Line 1255  for (;;) Line 1900  for (;;)
1900    
1901      /* Anything else is variable length */      /* Anything else is variable length */
1902    
1903      default:      case OP_ANYNL:
1904        case OP_BRAMINZERO:
1905        case OP_BRAPOS:
1906        case OP_BRAPOSZERO:
1907        case OP_BRAZERO:
1908        case OP_CBRAPOS:
1909        case OP_EXTUNI:
1910        case OP_KETRMAX:
1911        case OP_KETRMIN:
1912        case OP_KETRPOS:
1913        case OP_MINPLUS:
1914        case OP_MINPLUSI:
1915        case OP_MINQUERY:
1916        case OP_MINQUERYI:
1917        case OP_MINSTAR:
1918        case OP_MINSTARI:
1919        case OP_MINUPTO:
1920        case OP_MINUPTOI:
1921        case OP_NOTMINPLUS:
1922        case OP_NOTMINPLUSI:
1923        case OP_NOTMINQUERY:
1924        case OP_NOTMINQUERYI:
1925        case OP_NOTMINSTAR:
1926        case OP_NOTMINSTARI:
1927        case OP_NOTMINUPTO:
1928        case OP_NOTMINUPTOI:
1929        case OP_NOTPLUS:
1930        case OP_NOTPLUSI:
1931        case OP_NOTPOSPLUS:
1932        case OP_NOTPOSPLUSI:
1933        case OP_NOTPOSQUERY:
1934        case OP_NOTPOSQUERYI:
1935        case OP_NOTPOSSTAR:
1936        case OP_NOTPOSSTARI:
1937        case OP_NOTPOSUPTO:
1938        case OP_NOTPOSUPTOI:
1939        case OP_NOTQUERY:
1940        case OP_NOTQUERYI:
1941        case OP_NOTSTAR:
1942        case OP_NOTSTARI:
1943        case OP_NOTUPTO:
1944        case OP_NOTUPTOI:
1945        case OP_PLUS:
1946        case OP_PLUSI:
1947        case OP_POSPLUS:
1948        case OP_POSPLUSI:
1949        case OP_POSQUERY:
1950        case OP_POSQUERYI:
1951        case OP_POSSTAR:
1952        case OP_POSSTARI:
1953        case OP_POSUPTO:
1954        case OP_POSUPTOI:
1955        case OP_QUERY:
1956        case OP_QUERYI:
1957        case OP_REF:
1958        case OP_REFI:
1959        case OP_SBRA:
1960        case OP_SBRAPOS:
1961        case OP_SCBRA:
1962        case OP_SCBRAPOS:
1963        case OP_SCOND:
1964        case OP_SKIPZERO:
1965        case OP_STAR:
1966        case OP_STARI:
1967        case OP_TYPEMINPLUS:
1968        case OP_TYPEMINQUERY:
1969        case OP_TYPEMINSTAR:
1970        case OP_TYPEMINUPTO:
1971        case OP_TYPEPLUS:
1972        case OP_TYPEPOSPLUS:
1973        case OP_TYPEPOSQUERY:
1974        case OP_TYPEPOSSTAR:
1975        case OP_TYPEPOSUPTO:
1976        case OP_TYPEQUERY:
1977        case OP_TYPESTAR:
1978        case OP_TYPEUPTO:
1979        case OP_UPTO:
1980        case OP_UPTOI:
1981      return -1;      return -1;
1982    
1983        /* Catch unrecognized opcodes so that when new ones are added they
1984        are not forgotten, as has happened in the past. */
1985    
1986        default:
1987        return -4;
1988      }      }
1989    }    }
1990  /* Control never gets here */  /* Control never gets here */
# Line 1266  for (;;) Line 1994  for (;;)
1994    
1995    
1996  /*************************************************  /*************************************************
1997  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1998  *************************************************/  *************************************************/
1999    
2000  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2001  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2002    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2003    so that it can be called from pcre_study() when finding the minimum matching
2004    length.
2005    
2006  Arguments:  Arguments:
2007    code        points to start of expression    code        points to start of expression
2008    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2009    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2010    
2011  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
2012  */  */
2013    
2014  static const uschar *  const pcre_uchar *
2015  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2016  {  {
2017  for (;;)  for (;;)
2018    {    {
2019    register int c = *code;    register int c = *code;
2020    
2021    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2022    
2023    /* 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 2026  for (;;)
2026    
2027    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2028    
2029      /* Handle recursion */
2030    
2031      else if (c == OP_REVERSE)
2032        {
2033        if (number < 0) return (pcre_uchar *)code;
2034        code += PRIV(OP_lengths)[c];
2035        }
2036    
2037    /* Handle capturing bracket */    /* Handle capturing bracket */
2038    
2039    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2040               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2041      {      {
2042      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2043      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2044      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2045      }      }
2046    
2047    /* 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
2048    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
2049    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2050      must add in its length. */
2051    
2052    else    else
2053      {      {
# Line 1327  for (;;) Line 2069  for (;;)
2069        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2070        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2071        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2072        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2073            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2074          break;
2075    
2076          case OP_MARK:
2077          case OP_PRUNE_ARG:
2078          case OP_SKIP_ARG:
2079          code += code[1];
2080          break;
2081    
2082          case OP_THEN_ARG:
2083          code += code[1];
2084        break;        break;
2085        }        }
2086    
2087      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2088    
2089      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2090    
2091    /* 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
2092    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
2093    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2094    
2095  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2096      if (utf8) switch(c)      if (utf) switch(c)
2097        {        {
2098        case OP_CHAR:        case OP_CHAR:
2099        case OP_CHARNC:        case OP_CHARI:
2100        case OP_EXACT:        case OP_EXACT:
2101          case OP_EXACTI:
2102        case OP_UPTO:        case OP_UPTO:
2103          case OP_UPTOI:
2104        case OP_MINUPTO:        case OP_MINUPTO:
2105          case OP_MINUPTOI:
2106        case OP_POSUPTO:        case OP_POSUPTO:
2107          case OP_POSUPTOI:
2108        case OP_STAR:        case OP_STAR:
2109          case OP_STARI:
2110        case OP_MINSTAR:        case OP_MINSTAR:
2111          case OP_MINSTARI:
2112        case OP_POSSTAR:        case OP_POSSTAR:
2113          case OP_POSSTARI:
2114        case OP_PLUS:        case OP_PLUS:
2115          case OP_PLUSI:
2116        case OP_MINPLUS:        case OP_MINPLUS:
2117          case OP_MINPLUSI:
2118        case OP_POSPLUS:        case OP_POSPLUS:
2119          case OP_POSPLUSI:
2120        case OP_QUERY:        case OP_QUERY:
2121          case OP_QUERYI:
2122        case OP_MINQUERY:        case OP_MINQUERY:
2123          case OP_MINQUERYI:
2124        case OP_POSQUERY:        case OP_POSQUERY:
2125        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2126          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2127        break;        break;
2128        }        }
2129    #else
2130        (void)(utf);  /* Keep compiler happy by referencing function argument */
2131  #endif  #endif
2132      }      }
2133    }    }
# Line 1376  instance of OP_RECURSE. Line 2144  instance of OP_RECURSE.
2144    
2145  Arguments:  Arguments:
2146    code        points to start of expression    code        points to start of expression
2147    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2148    
2149  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
2150  */  */
2151    
2152  static const uschar *  static const pcre_uchar *
2153  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2154  {  {
2155  for (;;)  for (;;)
2156    {    {
# Line 1398  for (;;) Line 2166  for (;;)
2166    
2167    /* 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
2168    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
2169    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2170      must add in its length. */
2171    
2172    else    else
2173      {      {
# Line 1420  for (;;) Line 2189  for (;;)
2189        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2190        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2191        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2192        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2193            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2194          break;
2195    
2196          case OP_MARK:
2197          case OP_PRUNE_ARG:
2198          case OP_SKIP_ARG:
2199          code += code[1];
2200          break;
2201    
2202          case OP_THEN_ARG:
2203          code += code[1];
2204        break;        break;
2205        }        }
2206    
2207      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2208    
2209      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2210    
2211      /* 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
2212      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
2213      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2214    
2215  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2216      if (utf8) switch(c)      if (utf) switch(c)
2217        {        {
2218        case OP_CHAR:        case OP_CHAR:
2219        case OP_CHARNC:        case OP_CHARI:
2220        case OP_EXACT:        case OP_EXACT:
2221          case OP_EXACTI:
2222        case OP_UPTO:        case OP_UPTO:
2223          case OP_UPTOI:
2224        case OP_MINUPTO:        case OP_MINUPTO:
2225          case OP_MINUPTOI:
2226        case OP_POSUPTO:        case OP_POSUPTO:
2227          case OP_POSUPTOI:
2228        case OP_STAR:        case OP_STAR:
2229          case OP_STARI:
2230        case OP_MINSTAR:        case OP_MINSTAR:
2231          case OP_MINSTARI:
2232        case OP_POSSTAR:        case OP_POSSTAR:
2233          case OP_POSSTARI:
2234        case OP_PLUS:        case OP_PLUS:
2235          case OP_PLUSI:
2236        case OP_MINPLUS:        case OP_MINPLUS:
2237          case OP_MINPLUSI:
2238        case OP_POSPLUS:        case OP_POSPLUS:
2239          case OP_POSPLUSI:
2240        case OP_QUERY:        case OP_QUERY:
2241          case OP_QUERYI:
2242        case OP_MINQUERY:        case OP_MINQUERY:
2243          case OP_MINQUERYI:
2244        case OP_POSQUERY:        case OP_POSQUERY:
2245        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2246          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2247        break;        break;
2248        }        }
2249    #else
2250        (void)(utf);  /* Keep compiler happy by referencing function argument */
2251  #endif  #endif
2252      }      }
2253    }    }
# Line 1468  for (;;) Line 2263  for (;;)
2263  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()
2264  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
2265  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2266  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
2267  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
2268    bracket whose current branch will already have been scanned.
2269    
2270  Arguments:  Arguments:
2271    code        points to start of search    code        points to start of search
2272    endcode     points to where to stop    endcode     points to where to stop
2273    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2274      cd          contains pointers to tables etc.
2275    
2276  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2277  */  */
2278    
2279  static BOOL  static BOOL
2280  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2281      BOOL utf, compile_data *cd)
2282  {  {
2283  register int c;  register int c;
2284  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2285       code < endcode;       code < endcode;
2286       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2287    {    {
2288    const uschar *ccode;    const pcre_uchar *ccode;
2289    
2290    c = *code;    c = *code;
2291    
2292    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2293      first_significant_code() with a TRUE final argument. */
2294    
2295    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2296      {      {
     code += _pcre_OP_lengths[c];  
2297      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2298      c = *code;      c = *code;
2299      continue;      continue;
2300      }      }
2301    
2302    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2303      implies a backward reference subroutine call, we can scan it. If it's a
2304      forward reference subroutine call, we can't. To detect forward reference
2305      we have to scan up the list that is kept in the workspace. This function is
2306      called only when doing the real compile, not during the pre-compile that
2307      measures the size of the compiled pattern. */
2308    
2309    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2310      {      {
2311        const pcre_uchar *scode;
2312      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2313    
2314      /* Scan a closed bracket */      /* Test for forward reference */
2315    
2316        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2317          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2318    
2319        /* Not a forward reference, test for completed backward reference */
2320    
2321      empty_branch = FALSE;      empty_branch = FALSE;
2322        scode = cd->start_code + GET(code, 1);
2323        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2324    
2325        /* Completed backwards reference */
2326    
2327      do      do
2328        {        {
2329        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2330            {
2331          empty_branch = TRUE;          empty_branch = TRUE;
2332            break;
2333            }
2334          scode += GET(scode, 1);
2335          }
2336        while (*scode == OP_ALT);
2337    
2338        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2339        continue;
2340        }
2341    
2342      /* Groups with zero repeats can of course be empty; skip them. */
2343    
2344      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2345          c == OP_BRAPOSZERO)
2346        {
2347        code += PRIV(OP_lengths)[c];
2348        do code += GET(code, 1); while (*code == OP_ALT);
2349        c = *code;
2350        continue;
2351        }
2352    
2353      /* A nested group that is already marked as "could be empty" can just be
2354      skipped. */
2355    
2356      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2357          c == OP_SCBRA || c == OP_SCBRAPOS)
2358        {
2359        do code += GET(code, 1); while (*code == OP_ALT);
2360        c = *code;
2361        continue;
2362        }
2363    
2364      /* For other groups, scan the branches. */
2365    
2366      if (c == OP_BRA  || c == OP_BRAPOS ||
2367          c == OP_CBRA || c == OP_CBRAPOS ||
2368          c == OP_ONCE || c == OP_ONCE_NC ||
2369          c == OP_COND)
2370        {
2371        BOOL empty_branch;
2372        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2373    
2374        /* If a conditional group has only one branch, there is a second, implied,
2375        empty branch, so just skip over the conditional, because it could be empty.
2376        Otherwise, scan the individual branches of the group. */
2377    
2378        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2379        code += GET(code, 1);        code += GET(code, 1);
2380        else
2381          {
2382          empty_branch = FALSE;
2383          do
2384            {
2385            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2386              empty_branch = TRUE;
2387            code += GET(code, 1);
2388            }
2389          while (*code == OP_ALT);
2390          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2391        }        }
2392      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2393      c = *code;      c = *code;
2394      continue;      continue;
2395      }      }
# Line 1529  for (code = first_significant_code(code Line 2400  for (code = first_significant_code(code
2400      {      {
2401      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2402      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2403      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2404      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"
2405      here. */      here. */
2406    
2407  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2408      case OP_XCLASS:      case OP_XCLASS:
2409      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2410      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1541  for (code = first_significant_code(code Line 2412  for (code = first_significant_code(code
2412    
2413      case OP_CLASS:      case OP_CLASS:
2414      case OP_NCLASS:      case OP_NCLASS:
2415      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2416    
2417  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2418      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2419  #endif  #endif
2420    
# Line 1579  for (code = first_significant_code(code Line 2450  for (code = first_significant_code(code
2450      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2451      case OP_WORDCHAR:      case OP_WORDCHAR:
2452      case OP_ANY:      case OP_ANY:
2453        case OP_ALLANY:
2454      case OP_ANYBYTE:      case OP_ANYBYTE:
2455      case OP_CHAR:      case OP_CHAR:
2456      case OP_CHARNC:      case OP_CHARI:
2457      case OP_NOT:      case OP_NOT:
2458        case OP_NOTI:
2459      case OP_PLUS:      case OP_PLUS:
2460      case OP_MINPLUS:      case OP_MINPLUS:
2461      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1614  for (code = first_significant_code(code Line 2487  for (code = first_significant_code(code
2487      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2488      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2489      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2490      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2491          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2492      break;      break;
2493    
2494      /* End of branch */      /* End of branch */
# Line 1622  for (code = first_significant_code(code Line 2496  for (code = first_significant_code(code
2496      case OP_KET:      case OP_KET:
2497      case OP_KETRMAX:      case OP_KETRMAX:
2498      case OP_KETRMIN:      case OP_KETRMIN:
2499        case OP_KETRPOS:
2500      case OP_ALT:      case OP_ALT:
2501      return TRUE;      return TRUE;
2502    
2503      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2504      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2505    
2506  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2507      case OP_STAR:      case OP_STAR:
2508        case OP_STARI:
2509      case OP_MINSTAR:      case OP_MINSTAR:
2510        case OP_MINSTARI:
2511      case OP_POSSTAR:      case OP_POSSTAR:
2512        case OP_POSSTARI:
2513      case OP_QUERY:      case OP_QUERY:
2514        case OP_QUERYI:
2515      case OP_MINQUERY:      case OP_MINQUERY:
2516        case OP_MINQUERYI:
2517      case OP_POSQUERY:      case OP_POSQUERY:
2518        case OP_POSQUERYI:
2519        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2520        break;
2521    
2522      case OP_UPTO:      case OP_UPTO:
2523        case OP_UPTOI:
2524      case OP_MINUPTO:      case OP_MINUPTO:
2525        case OP_MINUPTOI:
2526      case OP_POSUPTO:      case OP_POSUPTO:
2527      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2528        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2529      break;      break;
2530  #endif  #endif
2531    
2532        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2533        string. */
2534    
2535        case OP_MARK:
2536        case OP_PRUNE_ARG:
2537        case OP_SKIP_ARG:
2538        code += code[1];
2539        break;
2540    
2541        case OP_THEN_ARG:
2542        code += code[1];
2543        break;
2544    
2545        /* None of the remaining opcodes are required to match a character. */
2546    
2547        default:
2548        break;
2549      }      }
2550    }    }
2551    
# Line 1657  return TRUE; Line 2562  return TRUE;
2562  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
2563  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,
2564  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.
2565    This function is called only during the real compile, not during the
2566    pre-compile.
2567    
2568  Arguments:  Arguments:
2569    code        points to start of the recursion    code        points to start of the recursion
2570    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2571    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2572    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2573      cd          pointers to tables etc
2574    
2575  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2576  */  */
2577    
2578  static BOOL  static BOOL
2579  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2580    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2581  {  {
2582  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2583    {    {
2584    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2585        return FALSE;
2586    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2587    }    }
2588  return TRUE;  return TRUE;
# Line 1686  return TRUE; Line 2595  return TRUE;
2595  *************************************************/  *************************************************/
2596    
2597  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2598  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
2599  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2600  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2601    
2602    Originally, this function only recognized a sequence of letters between the
2603    terminators, but it seems that Perl recognizes any sequence of characters,
2604    though of course unknown POSIX names are subsequently rejected. Perl gives an
2605    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2606    didn't consider this to be a POSIX class. Likewise for [:1234:].
2607    
2608    The problem in trying to be exactly like Perl is in the handling of escapes. We
2609    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2610    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2611    below handles the special case of \], but does not try to do any other escape
2612    processing. This makes it different from Perl for cases such as [:l\ower:]
2613    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2614    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2615    I think.
2616    
2617    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2618    It seems that the appearance of a nested POSIX class supersedes an apparent
2619    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2620    a digit.
2621    
2622    In Perl, unescaped square brackets may also appear as part of class names. For
2623    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2624    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2625    seem right at all. PCRE does not allow closing square brackets in POSIX class
2626    names.
2627    
2628  Argument:  Arguments:
2629    ptr      pointer to the initial [    ptr      pointer to the initial [
2630    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2631    
2632  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2633  */  */
2634    
2635  static BOOL  static BOOL
2636  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2637  {  {
2638  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2639  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2640  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2641    {    {
2642    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2643    return TRUE;      ptr++;
2644      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2645      else
2646        {
2647        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2648          {
2649          *endptr = ptr;
2650          return TRUE;
2651          }
2652        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2653             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2654              ptr[1] == CHAR_EQUALS_SIGN) &&
2655            check_posix_syntax(ptr, endptr))
2656          return FALSE;
2657        }
2658    }    }
2659  return FALSE;  return FALSE;
2660  }  }
# Line 1731  Returns:     a value representing the na Line 2677  Returns:     a value representing the na
2677  */  */
2678    
2679  static int  static int
2680  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2681  {  {
2682    const char *pn = posix_names;
2683  register int yield = 0;  register int yield = 0;
2684  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2685    {    {
2686    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2687      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2688      pn += posix_name_lengths[yield] + 1;
2689    yield++;    yield++;
2690    }    }
2691  return -1;  return -1;
# Line 1752  return -1; Line 2700  return -1;
2700  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2701  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2702  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
2703  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
2704  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
2705  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
2706  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
2707  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2708    OP_END.
2709    
2710  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2711  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 2716  value in the reference (which is a group
2716  Arguments:  Arguments:
2717    group      points to the start of the group    group      points to the start of the group
2718    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2719    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2720    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2721    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
2722    
# Line 1775  Returns:     nothing Line 2724  Returns:     nothing
2724  */  */
2725    
2726  static void  static void
2727  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2728    uschar *save_hwm)    pcre_uchar *save_hwm)
2729  {  {
2730  uschar *ptr = group;  pcre_uchar *ptr = group;
2731    
2732  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2733    {    {
2734    int offset;    int offset;
2735    uschar *hc;    pcre_uchar *hc;
2736    
2737    /* 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
2738    reference. */    reference. */
# Line 1828  Arguments: Line 2777  Arguments:
2777  Returns:         new code pointer  Returns:         new code pointer
2778  */  */
2779    
2780  static uschar *  static pcre_uchar *
2781  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2782  {  {
2783  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2784  *code++ = 255;  *code++ = 255;
2785  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2786  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2787  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2788  }  }
2789    
2790    
# Line 1857  Returns:             nothing Line 2806  Returns:             nothing
2806  */  */
2807    
2808  static void  static void
2809  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2810  {  {
2811  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2812  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2813  }  }
2814    
# Line 1891  get_othercase_range(unsigned int *cptr, Line 2840  get_othercase_range(unsigned int *cptr,
2840  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2841    
2842  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2843    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2844    
2845  if (c > d) return FALSE;  if (c > d) return FALSE;
2846    
# Line 1900  next = othercase + 1; Line 2849  next = othercase + 1;
2849    
2850  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2851    {    {
2852    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2853    next++;    next++;
2854    }    }
2855    
# Line 1909  for (++c; c <= d; c++) Line 2858  for (++c; c <= d; c++)
2858    
2859  return TRUE;  return TRUE;
2860  }  }
2861    
2862    
2863    
2864    /*************************************************
2865    *        Check a character and a property        *
2866    *************************************************/
2867    
2868    /* This function is called by check_auto_possessive() when a property item
2869    is adjacent to a fixed character.
2870    
2871    Arguments:
2872      c            the character
2873      ptype        the property type
2874      pdata        the data for the type
2875      negated      TRUE if it's a negated property (\P or \p{^)
2876    
2877    Returns:       TRUE if auto-possessifying is OK
2878    */
2879    
2880    static BOOL
2881    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2882    {
2883    const ucd_record *prop = GET_UCD(c);
2884    switch(ptype)
2885      {
2886      case PT_LAMP:
2887      return (prop->chartype == ucp_Lu ||
2888              prop->chartype == ucp_Ll ||
2889              prop->chartype == ucp_Lt) == negated;
2890    
2891      case PT_GC:
2892      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2893    
2894      case PT_PC:
2895      return (pdata == prop->chartype) == negated;
2896    
2897      case PT_SC:
2898      return (pdata == prop->script) == negated;
2899    
2900      /* These are specials */
2901    
2902      case PT_ALNUM:
2903      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2904              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2905    
2906      case PT_SPACE:    /* Perl space */
2907      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2908              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2909              == negated;
2910    
2911      case PT_PXSPACE:  /* POSIX space */
2912      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2913              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2914              c == CHAR_FF || c == CHAR_CR)
2915              == negated;
2916    
2917      case PT_WORD:
2918      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2919              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2920              c == CHAR_UNDERSCORE) == negated;
2921      }
2922    return FALSE;
2923    }
2924  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2925    
2926    
# Line 1922  whether the next thing could possibly ma Line 2934  whether the next thing could possibly ma
2934  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2935    
2936  Arguments:  Arguments:
2937    op_code       the repeated op code    previous      pointer to the repeated opcode
2938    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  
2939    ptr           next character in pattern    ptr           next character in pattern
2940    options       options bits    options       options bits
2941    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1934  Returns:        TRUE if possessifying is Line 2944  Returns:        TRUE if possessifying is
2944  */  */
2945    
2946  static BOOL  static BOOL
2947  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2948    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2949  {  {
2950  int next;  pcre_int32 c, next;
2951    int op_code = *previous++;
2952    
2953  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2954    
# Line 1946  if ((options & PCRE_EXTENDED) != 0) Line 2957  if ((options & PCRE_EXTENDED) != 0)
2957    for (;;)    for (;;)
2958      {      {
2959      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2960      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2961        {        {
2962        while (*(++ptr) != 0)        ptr++;
2963          while (*ptr != 0)
2964            {
2965          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2966            ptr++;
2967    #ifdef SUPPORT_UTF
2968            if (utf) FORWARDCHAR(ptr);
2969    #endif
2970            }
2971        }        }
2972      else break;      else break;
2973      }      }
# Line 1958  if ((options & PCRE_EXTENDED) != 0) Line 2976  if ((options & PCRE_EXTENDED) != 0)
2976  /* 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
2977  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2978    
2979  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2980    {    {
2981    int temperrorcode = 0;    int temperrorcode = 0;
2982    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2983    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2984    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2985    }    }
2986    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2987    {    {
2988  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2989    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2990  #endif  #endif
2991    next = *ptr++;    next = *ptr++;
2992    }    }
   
2993  else return FALSE;  else return FALSE;
2994    
2995  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1983  if ((options & PCRE_EXTENDED) != 0) Line 2999  if ((options & PCRE_EXTENDED) != 0)
2999    for (;;)    for (;;)
3000      {      {
3001      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3002      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3003        {        {
3004        while (*(++ptr) != 0)        ptr++;
3005          while (*ptr != 0)
3006            {
3007          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3008            ptr++;
3009    #ifdef SUPPORT_UTF
3010            if (utf) FORWARDCHAR(ptr);
3011    #endif
3012            }
3013        }        }
3014      else break;      else break;
3015      }      }
# Line 1994  if ((options & PCRE_EXTENDED) != 0) Line 3017  if ((options & PCRE_EXTENDED) != 0)
3017    
3018  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3019    
3020  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3021    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3022        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. */  
   
3023    
3024  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3025    the next item is a character. */
3026    
3027  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3028    {    {
3029    case OP_CHAR:    case OP_CHAR:
3030  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3031    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3032    #else
3033      c = *previous;
3034  #endif  #endif
3035    return item != next;    return c != next;
3036    
3037    /* 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
3038    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
3039    high-valued characters. */    high-valued characters. */
3040    
3041    case OP_CHARNC:    case OP_CHARI:
3042  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3043    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3044  #endif  #else
3045    if (item == next) return FALSE;    c = *previous;
3046  #ifdef SUPPORT_UTF8  #endif
3047    if (utf8)    if (c == next) return FALSE;
3048    #ifdef SUPPORT_UTF
3049      if (utf)
3050      {      {
3051      unsigned int othercase;      unsigned int othercase;
3052      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3053  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3054      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3055  #else  #else
3056      othercase = NOTACHAR;      othercase = NOTACHAR;
3057  #endif  #endif
3058      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3059      }      }
3060    else    else
3061  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3062    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
3063    
3064    /* 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
3065      opcodes are not used for multi-byte characters, because they are coded using
3066      an XCLASS instead. */
3067    
3068    case OP_NOT:    case OP_NOT:
3069    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
3070    if (item == next) return TRUE;  
3071    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
3072  #ifdef SUPPORT_UTF8    if ((c = *previous) == next) return TRUE;
3073    if (utf8)  #ifdef SUPPORT_UTF
3074      if (utf)
3075      {      {
3076      unsigned int othercase;      unsigned int othercase;
3077      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3078  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3079      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3080  #else  #else
3081      othercase = NOTACHAR;      othercase = NOTACHAR;
3082  #endif  #endif
3083      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3084      }      }
3085    else    else
3086  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3087    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
3088    
3089      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3090      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3091    
3092    case OP_DIGIT:    case OP_DIGIT:
3093    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 3130  if (next >= 0) switch(op_code)
3130      case 0x202f:      case 0x202f:
3131      case 0x205f:      case 0x205f:
3132      case 0x3000:      case 0x3000:
3133      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3134      default:      default:
3135      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3136      }      }
3137    
3138      case OP_ANYNL:
3139    case OP_VSPACE:    case OP_VSPACE:
3140    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3141    switch(next)    switch(next)
# Line 2117  if (next >= 0) switch(op_code) Line 3147  if (next >= 0) switch(op_code)
3147      case 0x85:      case 0x85:
3148      case 0x2028:      case 0x2028:
3149      case 0x2029:      case 0x2029:
3150      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3151      default:      default:
3152      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3153      }      }
3154    
3155    #ifdef SUPPORT_UCP
3156      case OP_PROP:
3157      return check_char_prop(next, previous[0], previous[1], FALSE);
3158    
3159      case OP_NOTPROP:
3160      return check_char_prop(next, previous[0], previous[1], TRUE);
3161    #endif
3162    
3163    default:    default:
3164    return FALSE;    return FALSE;
3165    }    }
3166    
3167    
3168  /* 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
3169    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3170    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3171    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3172    replaced by OP_PROP codes when PCRE_UCP is set. */
3173    
3174  switch(op_code)  switch(op_code)
3175    {    {
3176    case OP_CHAR:    case OP_CHAR:
3177    case OP_CHARNC:    case OP_CHARI:
3178  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3179    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3180    #else
3181      c = *previous;
3182  #endif  #endif
3183    switch(-next)    switch(-next)
3184      {      {
3185      case ESC_d:      case ESC_d:
3186      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3187    
3188      case ESC_D:      case ESC_D:
3189      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3190    
3191      case ESC_s:      case ESC_s:
3192      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3193    
3194      case ESC_S:      case ESC_S:
3195      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3196    
3197      case ESC_w:      case ESC_w:
3198      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3199    
3200      case ESC_W:      case ESC_W:
3201      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3202    
3203      case ESC_h:      case ESC_h:
3204      case ESC_H:      case ESC_H:
3205      switch(item)      switch(c)
3206        {        {
3207        case 0x09:        case 0x09:
3208        case 0x20:        case 0x20:
# Line 2186  switch(op_code) Line 3230  switch(op_code)
3230    
3231      case ESC_v:      case ESC_v:
3232      case ESC_V:      case ESC_V:
3233      switch(item)      switch(c)
3234        {        {
3235        case 0x0a:        case 0x0a:
3236        case 0x0b:        case 0x0b:
# Line 2200  switch(op_code) Line 3244  switch(op_code)
3244        return -next == ESC_v;        return -next == ESC_v;
3245        }        }
3246    
3247        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3248        their substitutions and process them. The result will always be either
3249        -ESC_p or -ESC_P. Then fall through to process those values. */
3250    
3251    #ifdef SUPPORT_UCP
3252        case ESC_du:
3253        case ESC_DU:
3254        case ESC_wu:
3255        case ESC_WU:
3256        case ESC_su:
3257        case ESC_SU:
3258          {
3259          int temperrorcode = 0;
3260          ptr = substitutes[-next - ESC_DU];
3261          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3262          if (temperrorcode != 0) return FALSE;
3263          ptr++;    /* For compatibility */
3264          }
3265        /* Fall through */
3266    
3267        case ESC_p:
3268        case ESC_P:
3269          {
3270          int ptype, pdata, errorcodeptr;
3271          BOOL negated;
3272    
3273          ptr--;      /* Make ptr point at the p or P */
3274          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3275          if (ptype < 0) return FALSE;
3276          ptr++;      /* Point past the final curly ket */
3277    
3278          /* If the property item is optional, we have to give up. (When generated
3279          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3280          to the original \d etc. At this point, ptr will point to a zero byte. */
3281    
3282          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3283            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3284              return FALSE;
3285    
3286          /* Do the property check. */
3287    
3288          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3289          }
3290    #endif
3291    
3292      default:      default:
3293      return FALSE;      return FALSE;
3294      }      }
3295    
3296      /* In principle, support for Unicode properties should be integrated here as
3297      well. It means re-organizing the above code so as to get hold of the property
3298      values before switching on the op-code. However, I wonder how many patterns
3299      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3300      these op-codes are never generated.) */
3301    
3302    case OP_DIGIT:    case OP_DIGIT:
3303    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3304           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3305    
3306    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3307    return next == -ESC_d;    return next == -ESC_d;
3308    
3309    case OP_WHITESPACE:    case OP_WHITESPACE:
3310    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3311    
3312    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3313    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3314    
3315    case OP_HSPACE:    case OP_HSPACE:
3316    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3317             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3318    
3319    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3320    return next == -ESC_h;    return next == -ESC_h;
3321    
3322    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3323      case OP_ANYNL:
3324    case OP_VSPACE:    case OP_VSPACE:
3325    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3326    
3327    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3328    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3329    
3330    case OP_WORDCHAR:    case OP_WORDCHAR:
3331    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3332             next == -ESC_v || next == -ESC_R;
3333    
3334    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3335    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2260  Arguments: Line 3358  Arguments:
3358    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3359    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3360    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3361    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3362    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3363    bcptr          points to current branch chain    bcptr          points to current branch chain
3364      cond_depth     conditional nesting depth
3365    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3366    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3367                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2272  Returns:         TRUE on success Line 3371  Returns:         TRUE on success
3371  */  */
3372    
3373  static BOOL  static BOOL
3374  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3375    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3376      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3377    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3378  {  {
3379  int repeat_type, op_type;  int repeat_type, op_type;
3380  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3381  int bravalue = 0;  int bravalue = 0;
3382  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3383  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3384  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3385  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3386  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3387  int after_manual_callout = 0;  int after_manual_callout = 0;
3388  int length_prevgroup = 0;  int length_prevgroup = 0;
3389  register int c;  register int c;
3390  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3391  uschar *last_code = code;  pcre_uchar *last_code = code;
3392  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3393  uschar *tempcode;  pcre_uchar *tempcode;
3394  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3395  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3396  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3397  const uschar *tempptr;  const pcre_uchar *tempptr;
3398  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3399  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3400  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3401  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3402    pcre_uint8 classbits[32];
3403  #ifdef SUPPORT_UTF8  
3404  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3405  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3406  uschar *class_utf8data;  dynamically as we process the pattern. */
3407  uschar utf8_char[6];  
3408    #ifdef SUPPORT_UTF
3409    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3410    BOOL utf = (options & PCRE_UTF8) != 0;
3411    pcre_uchar utf_chars[6];
3412  #else  #else
3413  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3414  uschar *utf8_char = NULL;  #endif
3415    
3416    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3417    
3418    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3419    BOOL xclass;
3420    pcre_uchar *class_uchardata;
3421    pcre_uchar *class_uchardata_base;
3422  #endif  #endif
3423    
3424  #ifdef DEBUG  #ifdef PCRE_DEBUG
3425  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3426  #endif  #endif
3427    
# Line 2321  greedy_non_default = greedy_default ^ 1; Line 3432  greedy_non_default = greedy_default ^ 1;
3432    
3433  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3434  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
3435  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
3436  find one.  find one.
3437    
3438  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
3439  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
3440  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3441  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3442    
3443  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3444    
3445  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3446  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
3447  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
3448  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3449    value. This is used only for ASCII characters. */
3450    
3451  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3452    
3453  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3454    
3455  for (;; ptr++)  for (;; ptr++)
3456    {    {
3457    BOOL negate_class;    BOOL negate_class;
3458      BOOL should_flip_negation;
3459    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3460    BOOL is_quantifier;    BOOL is_quantifier;
3461    BOOL is_recurse;    BOOL is_recurse;
3462    BOOL reset_bracount;    BOOL reset_bracount;
3463    int class_charcount;    int class_has_8bitchar;
3464      int class_single_char;
3465    int class_lastchar;    int class_lastchar;
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 ((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      class_charcount = 0;      /* 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_has_8bitchar = 0;
3769        class_single_char = 0;
3770      class_lastchar = -1;      class_lastchar = -1;
3771    
3772      /* 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
# Line 2603  for (;; ptr++) Line 3774  for (;; ptr++)
3774      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.
3775      */      */
3776    
3777      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3778    
3779  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3780      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3781      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3782        class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3783  #endif  #endif
3784    
3785      /* 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 3788  for (;; ptr++)
3788    
3789      if (c != 0) do      if (c != 0) do
3790        {        {
3791        const uschar *oldptr;        const pcre_uchar *oldptr;
3792    
3793  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3794        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3795          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3796          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3797          }          }
3798  #endif  #endif
3799    
3800    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3801          /* In the pre-compile phase, accumulate the length of any extra
3802          data and reset the pointer. This is so that very large classes that
3803          contain a zillion > 255 characters no longer overwrite the work space
3804          (which is on the stack). */
3805    
3806          if (lengthptr != NULL)
3807            {
3808            *lengthptr += class_uchardata - class_uchardata_base;
3809            class_uchardata = class_uchardata_base;
3810            }
3811    #endif
3812    
3813        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3814    
3815        if (inescq)        if (inescq)
3816          {          {
3817          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3818            {            {
3819            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3820            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2644  for (;; ptr++) Line 3829  for (;; ptr++)
3829        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3830        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3831    
3832        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3833            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3834            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3835          {          {
3836          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3837          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3838          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3839          uschar pbits[32];          pcre_uint8 pbits[32];
3840    
3841          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3842            {            {
3843            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3844            goto FAILED;            goto FAILED;
3845            }            }
3846    
3847          ptr += 2;          ptr += 2;
3848          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3849            {            {
3850            local_negate = TRUE;            local_negate = TRUE;
3851              should_flip_negation = TRUE;  /* Note negative special */
3852            ptr++;            ptr++;
3853            }            }
3854    
3855          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3856          if (posix_class < 0)          if (posix_class < 0)
3857            {            {
3858            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2680  for (;; ptr++) Line 3866  for (;; ptr++)
3866          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3867            posix_class = 0;            posix_class = 0;
3868    
3869          /* 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
3870          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3871          subtract bits that may be in the main map already. At the end we or the  
3872          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3873            if ((options & PCRE_UCP) != 0)
3874              {
3875              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3876              if (posix_substitutes[pc] != NULL)
3877                {
3878                nestptr = tempptr + 1;
3879                ptr = posix_substitutes[pc] - 1;
3880                continue;
3881                }
3882              }
3883    #endif
3884            /* In the non-UCP case, we build the bit map for the POSIX class in a
3885            chunk of local store because we may be adding and subtracting from it,
3886            and we don't want to subtract bits that may be in the main map already.
3887            At the end we or the result into the bit map that is being built. */
3888    
3889          posix_class *= 3;          posix_class *= 3;
3890    
3891          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3892    
3893          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3894            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3895    
3896          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3897    
# Line 2721  for (;; ptr++) Line 3922  for (;; ptr++)
3922            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3923    
3924          ptr = tempptr + 1;          ptr = tempptr + 1;
3925          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3926            class_has_8bitchar = 1;
3927            /* Every class contains at least two characters. */
3928            class_single_char = 2;
3929          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3930          }          }
3931    
3932        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3933        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
3934        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
3935        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so
3936        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
3937        character in them, so set class_charcount bigger than one. */        than one. Unrecognized escapes fall through and are either treated
3938          as literal characters (by default), or are faulted if
3939          PCRE_EXTRA is set. */
3940    
3941        if (c == '\\')        if (c == CHAR_BACKSLASH)
3942          {          {
3943          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3944          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3945