/[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 218 by ph10, Thu Aug 16 10:13:23 2007 UTC code/branches/pcre16/pcre_compile.c revision 812 by zherczeg, Mon Dec 19 21:23:42 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 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
# Line 87  so this number is very generous. Line 92  so this number is very generous.
92  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
93  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
94  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
95  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
96    filled up by repetitions of forward references, for example patterns like
97    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
98    that the workspace is expanded using malloc() in this situation. The value
99    below is therefore a minimum, and we put a maximum on it for safety. The
100    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
101    kicks in at the same number of forward references in all cases. */
102    
103    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
104    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
105    
106    /* The overrun tests check for a slightly smaller size so that they detect the
107    overrun before it actually does run off the end of the data block. */
108    
109    #define WORK_SIZE_SAFETY_MARGIN (100)
110    
111  #define COMPILE_WORK_SIZE (4096)  /* Private flags added to firstchar and reqchar. */
112    
113    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
114    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
115    
116    /* Repeated character flags. */
117    
118    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
119    
120  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
121  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
122  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
123  is invalid. */  is invalid. */
124    
125  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
126    
127    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
128    in UTF-8 mode. */
129    
130  static const short int escapes[] = {  static const short int escapes[] = {
131       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
132       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
133     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
134  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
135  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
136  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
137     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
138  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
139  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
140       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
141         -ESC_D,                  -ESC_E,
142         0,                       -ESC_G,
143         -ESC_H,                  0,
144         0,                       -ESC_K,
145         0,                       0,
146         -ESC_N,                  0,
147         -ESC_P,                  -ESC_Q,
148         -ESC_R,                  -ESC_S,
149         0,                       0,
150         -ESC_V,                  -ESC_W,
151         -ESC_X,                  0,
152         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
153         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
154         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
155         CHAR_GRAVE_ACCENT,       7,
156         -ESC_b,                  0,
157         -ESC_d,                  ESC_e,
158         ESC_f,                   0,
159         -ESC_h,                  0,
160         0,                       -ESC_k,
161         0,                       0,
162         ESC_n,                   0,
163         -ESC_p,                  0,
164         ESC_r,                   -ESC_s,
165         ESC_tee,                 0,
166         -ESC_v,                  -ESC_w,
167         0,                       0,
168         -ESC_z
169  };  };
170    
171  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
172    
173    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
174    
175  static const short int escapes[] = {  static const short int escapes[] = {
176  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
177  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 190  static const short int escapes[] = {
190  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
191  /*  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,
192  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
193  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
194  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
195  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
196  /*  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 200  static const short int escapes[] = {
200  #endif  #endif
201    
202    
203  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
204    searched linearly. Put all the names into a single string, in order to reduce
205    the number of relocations when a shared library is dynamically linked. The
206    string is built from string macros so that it works in UTF-8 mode on EBCDIC
207    platforms. */
208    
209  typedef struct verbitem {  typedef struct verbitem {
210    const char *name;    int   len;                 /* Length of verb name */
211    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
212    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
213  } verbitem;  } verbitem;
214    
215  static verbitem verbs[] = {  static const char verbnames[] =
216    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
217    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
218    { "F",      1, OP_FAIL },    STRING_ACCEPT0
219    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
220    { "PRUNE",  5, OP_PRUNE },    STRING_F0
221    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
222    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
223      STRING_SKIP0
224      STRING_THEN;
225    
226    static const verbitem verbs[] = {
227      { 0, -1,        OP_MARK },
228      { 4, -1,        OP_MARK },
229      { 6, OP_ACCEPT, -1 },
230      { 6, OP_COMMIT, -1 },
231      { 1, OP_FAIL,   -1 },
232      { 4, OP_FAIL,   -1 },
233      { 5, OP_PRUNE,  OP_PRUNE_ARG },
234      { 4, OP_SKIP,   OP_SKIP_ARG  },
235      { 4, OP_THEN,   OP_THEN_ARG  }
236  };  };
237    
238  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
239    
 /* 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. */  
240    
241  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
242    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
243    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
244    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
245    for handling case independence. */
246    
247    static const char posix_names[] =
248      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
249      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
250      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
251      STRING_word0  STRING_xdigit;
252    
253  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
254    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 };
255    
256  /* 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 280  static const int posix_class_maps[] = {
280    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
281  };  };
282    
283    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
284    substitutes must be in the order of the names, defined above, and there are
285    both positive and negative cases. NULL means no substitute. */
286    
287    #ifdef SUPPORT_UCP
288    static const pcre_uchar string_PNd[]  = {
289      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
290      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    static const pcre_uchar string_pNd[]  = {
292      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
293      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    static const pcre_uchar string_PXsp[] = {
295      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
296      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
297    static const pcre_uchar string_pXsp[] = {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_PXwd[] = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pXwd[] = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    
307    static const pcre_uchar *substitutes[] = {
308      string_PNd,           /* \D */
309      string_pNd,           /* \d */
310      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
311      string_pXsp,          /* \s */
312      string_PXwd,          /* \W */
313      string_pXwd           /* \w */
314    };
315    
316    static const pcre_uchar string_pL[] =   {
317      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pLl[] =  {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_pLu[] =  {
323      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_pXan[] = {
326      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    static const pcre_uchar string_h[] =    {
329      CHAR_BACKSLASH, CHAR_h, '\0' };
330    static const pcre_uchar string_pXps[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_PL[] =   {
334      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_PLl[] =  {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_PLu[] =  {
340      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
341      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
342    static const pcre_uchar string_PXan[] = {
343      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_H[] =    {
346      CHAR_BACKSLASH, CHAR_H, '\0' };
347    static const pcre_uchar string_PXps[] = {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    
351    static const pcre_uchar *posix_substitutes[] = {
352      string_pL,            /* alpha */
353      string_pLl,           /* lower */
354      string_pLu,           /* upper */
355      string_pXan,          /* alnum */
356      NULL,                 /* ascii */
357      string_h,             /* blank */
358      NULL,                 /* cntrl */
359      string_pNd,           /* digit */
360      NULL,                 /* graph */
361      NULL,                 /* print */
362      NULL,                 /* punct */
363      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
364      string_pXwd,          /* word */
365      NULL,                 /* xdigit */
366      /* Negated cases */
367      string_PL,            /* ^alpha */
368      string_PLl,           /* ^lower */
369      string_PLu,           /* ^upper */
370      string_PXan,          /* ^alnum */
371      NULL,                 /* ^ascii */
372      string_H,             /* ^blank */
373      NULL,                 /* ^cntrl */
374      string_PNd,           /* ^digit */
375      NULL,                 /* ^graph */
376      NULL,                 /* ^print */
377      NULL,                 /* ^punct */
378      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
379      string_PXwd,          /* ^word */
380      NULL                  /* ^xdigit */
381    };
382    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
383    #endif
384    
385  #define STRING(a)  # a  #define STRING(a)  # a
386  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 207  static const int posix_class_maps[] = { Line 388  static const int posix_class_maps[] = {
388  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
389  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
390  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
391  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
392    the number of relocations needed when a shared library is loaded dynamically,
393  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
394    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
395    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
396    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
397    "unrecognized character follows \\",  
398    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
399    substring, so that the whole string ends with \0\0, which can be detected when
400    counting through. */
401    
402    static const char error_texts[] =
403      "no error\0"
404      "\\ at end of pattern\0"
405      "\\c at end of pattern\0"
406      "unrecognized character follows \\\0"
407      "numbers out of order in {} quantifier\0"
408    /* 5 */    /* 5 */
409    "number too big in {} quantifier",    "number too big in {} quantifier\0"
410    "missing terminating ] for character class",    "missing terminating ] for character class\0"
411    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
412    "range out of order in character class",    "range out of order in character class\0"
413    "nothing to repeat",    "nothing to repeat\0"
414    /* 10 */    /* 10 */
415    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
416    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
417    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
418    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
419    "missing )",    "missing )\0"
420    /* 15 */    /* 15 */
421    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
422    "erroffset passed as NULL",    "erroffset passed as NULL\0"
423    "unknown option bit(s) set",    "unknown option bit(s) set\0"
424    "missing ) after comment",    "missing ) after comment\0"
425    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
426    /* 20 */    /* 20 */
427    "regular expression is too large",    "regular expression is too large\0"
428    "failed to get memory",    "failed to get memory\0"
429    "unmatched parentheses",    "unmatched parentheses\0"
430    "internal error: code overflow",    "internal error: code overflow\0"
431    "unrecognized character after (?<",    "unrecognized character after (?<\0"
432    /* 25 */    /* 25 */
433    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
434    "malformed number or name after (?(",    "malformed number or name after (?(\0"
435    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
436    "assertion expected after (?(",    "assertion expected after (?(\0"
437    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
438    /* 30 */    /* 30 */
439    "unknown POSIX class name",    "unknown POSIX class name\0"
440    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
441    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
442    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
443    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
444    /* 35 */    /* 35 */
445    "invalid condition (?(0)",    "invalid condition (?(0)\0"
446    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
447    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
448    "number after (?C is > 255",    "number after (?C is > 255\0"
449    "closing ) for (?C expected",    "closing ) for (?C expected\0"
450    /* 40 */    /* 40 */
451    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
452    "unrecognized character after (?P",    "unrecognized character after (?P\0"
453    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
454    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
455    "invalid UTF-8 string",    "invalid UTF-8 string\0"
456    /* 45 */    /* 45 */
457    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
458    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
459    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
460    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
461    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
462    /* 50 */    /* 50 */
463    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
464    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
465    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
466    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
467    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
468    /* 55 */    /* 55 */
469    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
470    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
471    "\\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"
472    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
473    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
474    /* 60 */    /* 60 */
475    "(*VERB) not recognized",    "(*VERB) not recognized\0"
476    "number is too big"    "number is too big\0"
477  };    "subpattern name expected\0"
478      "digit expected after (?+\0"
479      "] is an invalid data character in JavaScript compatibility mode\0"
480      /* 65 */
481      "different names for subpatterns of the same number are not allowed\0"
482      "(*MARK) must have an argument\0"
483      "this version of PCRE is not compiled with PCRE_UCP support\0"
484      "\\c must be followed by an ASCII character\0"
485      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
486      /* 70 */
487      "internal error: unknown opcode in find_fixedlength()\0"
488      "\\N is not supported in a class\0"
489      "too many forward references\0"
490      "disallowed UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
491      ;
492    
493  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
494  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 506  For convenience, we use the same bit def
506    
507  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
508    
509  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
510  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
511    into a subtraction and unsigned comparison). */
512    
513    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
514    
515    #ifndef EBCDIC
516    
517    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
518    UTF-8 mode. */
519    
520    static const pcre_uint8 digitab[] =
521    {    {
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
523    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 552  static const unsigned char digitab[] =
552    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
553    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
554    
555  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
556  static const unsigned char digitab[] =  
557    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
558    
559    static const pcre_uint8 digitab[] =
560    {    {
561    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
562    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 591  static const unsigned char digitab[] =
591    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
592    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
593    
594  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
595    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
596    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
597    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 630  static const unsigned char ebcdic_charta
630  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
631    
632  static BOOL  static BOOL
633    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
634      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
635    
636    
637    
638  /*************************************************  /*************************************************
639    *            Find an error text                  *
640    *************************************************/
641    
642    /* The error texts are now all in one long string, to save on relocations. As
643    some of the text is of unknown length, we can't use a table of offsets.
644    Instead, just count through the strings. This is not a performance issue
645    because it happens only when there has been a compilation error.
646    
647    Argument:   the error number
648    Returns:    pointer to the error string
649    */
650    
651    static const char *
652    find_error_text(int n)
653    {
654    const char *s = error_texts;
655    for (; n > 0; n--)
656      {
657      while (*s++ != 0) {};
658      if (*s == 0) return "Error text not found (please report)";
659      }
660    return s;
661    }
662    
663    
664    /*************************************************
665    *           Expand the workspace                 *
666    *************************************************/
667    
668    /* This function is called during the second compiling phase, if the number of
669    forward references fills the existing workspace, which is originally a block on
670    the stack. A larger block is obtained from malloc() unless the ultimate limit
671    has been reached or the increase will be rather small.
672    
673    Argument: pointer to the compile data block
674    Returns:  0 if all went well, else an error number
675    */
676    
677    static int
678    expand_workspace(compile_data *cd)
679    {
680    pcre_uchar *newspace;
681    int newsize = cd->workspace_size * 2;
682    
683    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
684    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
685        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
686     return ERR72;
687    
688    newspace = (PUBL(malloc))(newsize);
689    if (newspace == NULL) return ERR21;
690    
691    memcpy(newspace, cd->start_workspace, cd->workspace_size);
692    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
693    if (cd->workspace_size > COMPILE_WORK_SIZE)
694      (PUBL(free))((void *)cd->start_workspace);
695    cd->start_workspace = newspace;
696    cd->workspace_size = newsize;
697    return 0;
698    }
699    
700    
701    
702    /*************************************************
703    *            Check for counted repeat            *
704    *************************************************/
705    
706    /* This function is called when a '{' is encountered in a place where it might
707    start a quantifier. It looks ahead to see if it really is a quantifier or not.
708    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
709    where the ddds are digits.
710    
711    Arguments:
712      p         pointer to the first char after '{'
713    
714    Returns:    TRUE or FALSE
715    */
716    
717    static BOOL
718    is_counted_repeat(const pcre_uchar *p)
719    {
720    if (!IS_DIGIT(*p)) return FALSE;
721    p++;
722    while (IS_DIGIT(*p)) p++;
723    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
724    
725    if (*p++ != CHAR_COMMA) return FALSE;
726    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
727    
728    if (!IS_DIGIT(*p)) return FALSE;
729    p++;
730    while (IS_DIGIT(*p)) p++;
731    
732    return (*p == CHAR_RIGHT_CURLY_BRACKET);
733    }
734    
735    
736    
737    /*************************************************
738  *            Handle escapes                      *  *            Handle escapes                      *
739  *************************************************/  *************************************************/
740    
# Line 444  Returns:         zero or positive => a d Line 759  Returns:         zero or positive => a d
759  */  */
760    
761  static int  static int
762  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
763    int options, BOOL isclass)    int options, BOOL isclass)
764  {  {
765  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
766  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
767  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
768    pcre_int32 c;
769    int i;
770    
771  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
772  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 458  ptr--;                            /* Set Line 775  ptr--;                            /* Set
775    
776  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
777    
778  /* 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
779  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.
780  Otherwise further processing may be required. */  Otherwise further processing may be required. */
781    
782  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
783  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
784  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
785    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
786    
787  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
788  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
789    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
790  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
791  #endif  #endif
792    
# Line 475  else if ((i = escapes[c - 0x48]) != 0) Line 794  else if ((i = escapes[c - 0x48]) != 0)
794    
795  else  else
796    {    {
797    const uschar *oldptr;    const pcre_uchar *oldptr;
798    BOOL braced, negated;    BOOL braced, negated;
799    
800    switch (c)    switch (c)
# Line 483  else Line 802  else
802      /* 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
803      error. */      error. */
804    
805      case 'l':      case CHAR_l:
806      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
807      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
808      break;      break;
809    
810      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
811      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
812      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
813      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
814      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
815      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
816            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
817      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
818      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
819        {          {
820        const uschar *p;          c = 0;
821        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
822          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
823        if (*p != 0 && *p != '}')            register int cc = *(++ptr);
824    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
825              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
826              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
827    #else           /* EBCDIC coding */
828              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
829              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
830    #endif
831              }
832            }
833          }
834        else
835          *errorcodeptr = ERR37;
836        break;
837    
838        case CHAR_U:
839        /* In JavaScript, \U is an uppercase U letter. */
840        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
841        break;
842    
843        /* In a character class, \g is just a literal "g". Outside a character
844        class, \g must be followed by one of a number of specific things:
845    
846        (1) A number, either plain or braced. If positive, it is an absolute
847        backreference. If negative, it is a relative backreference. This is a Perl
848        5.10 feature.
849    
850        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
851        is part of Perl's movement towards a unified syntax for back references. As
852        this is synonymous with \k{name}, we fudge it up by pretending it really
853        was \k.
854    
855        (3) For Oniguruma compatibility we also support \g followed by a name or a
856        number either in angle brackets or in single quotes. However, these are
857        (possibly recursive) subroutine calls, _not_ backreferences. Just return
858        the -ESC_g code (cf \k). */
859    
860        case CHAR_g:
861        if (isclass) break;
862        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
863          {
864          c = -ESC_g;
865          break;
866          }
867    
868        /* Handle the Perl-compatible cases */
869    
870        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
871          {
872          const pcre_uchar *p;
873          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
874            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
875          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
876          {          {
877          c = -ESC_k;          c = -ESC_k;
878          break;          break;
# Line 514  else Line 882  else
882        }        }
883      else braced = FALSE;      else braced = FALSE;
884    
885      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
886        {        {
887        negated = TRUE;        negated = TRUE;
888        ptr++;        ptr++;
889        }        }
890      else negated = FALSE;      else negated = FALSE;
891    
892        /* The integer range is limited by the machine's int representation. */
893      c = 0;      c = 0;
894      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
895        c = c * 10 + *(++ptr) - '0';        {
896          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
897      if (c < 0)          {
898            c = -1;
899            break;
900            }
901          c = c * 10 + *(++ptr) - CHAR_0;
902          }
903        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
904        {        {
905          while (IS_DIGIT(ptr[1]))
906            ptr++;
907        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
908        break;        break;
909        }        }
910    
911      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
912        {        {
913        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
914        break;        break;
915        }        }
916    
917        if (c == 0)
918          {
919          *errorcodeptr = ERR58;
920          break;
921          }
922    
923      if (negated)      if (negated)
924        {        {
925        if (c > bracount)        if (c > bracount)
# Line 562  else Line 945  else
945      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
946      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
947    
948      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:
949      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
950    
951      if (!isclass)      if (!isclass)
952        {        {
953        oldptr = ptr;        oldptr = ptr;
954        c -= '0';        /* The integer range is limited by the machine's int representation. */
955        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
956          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
957        if (c < 0)          {
958            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
959              {
960              c = -1;
961              break;
962              }
963            c = c * 10 + *(++ptr) - CHAR_0;
964            }
965          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
966          {          {
967            while (IS_DIGIT(ptr[1]))
968              ptr++;
969          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
970          break;          break;
971          }          }
972        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
973          {          {
974          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 588  else Line 981  else
981      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.
982      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
983    
984      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
985        {        {
986        ptr--;        ptr--;
987        c = 0;        c = 0;
# Line 601  else Line 994  else
994      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
995      than 3 octal digits. */      than 3 octal digits. */
996    
997      case '0':      case CHAR_0:
998      c -= '0';      c -= CHAR_0;
999      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1000          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1001      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
1002      break;      break;
1003    
1004      /* \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
1005      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.
1006      treated as a data character. */      If not, { is treated as a data character. */
1007    
1008        case CHAR_x:
1009        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1010          {
1011          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1012          Otherwise it is a lowercase x letter. */
1013          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1014            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1015            {
1016            c = 0;
1017            for (i = 0; i < 2; ++i)
1018              {
1019              register int cc = *(++ptr);
1020    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1021              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1022              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1023    #else           /* EBCDIC coding */
1024              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1025              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1026    #endif
1027              }
1028            }
1029          break;
1030          }
1031    
1032      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1033        {        {
1034        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1035    
1036        c = 0;        c = 0;
1037        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1038          {          {
1039          register int cc = *pt++;          register int cc = *pt++;
1040          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1041    
1042  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1044          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1045  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1046          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1047          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048    #endif
1049    
1050    #ifdef COMPILE_PCRE8
1051            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1052    #else
1053    #ifdef COMPILE_PCRE16
1054            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1055    #endif
1056  #endif  #endif
1057          }          }
1058    
1059        if (*pt == '}')        if (c < 0)
1060            {
1061            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1062            *errorcodeptr = ERR34;
1063            }
1064    
1065          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1066          {          {
1067          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1068          ptr = pt;          ptr = pt;
1069          break;          break;
1070          }          }
# Line 648  else Line 1076  else
1076      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1077    
1078      c = 0;      c = 0;
1079      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1080        {        {
1081        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1082        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1083  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1084        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1085        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1086  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1087        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1088        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1089  #endif  #endif
1090        }        }
1091      break;      break;
1092    
1093      /* 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.
1094      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
1095        coding is ASCII-specific, but then the whole concept of \cx is
1096      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1097    
1098      case 'c':      case CHAR_c:
1099      c = *(++ptr);      c = *(++ptr);
1100      if (c == 0)      if (c == 0)
1101        {        {
1102        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1103        break;        break;
1104        }        }
1105    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1106  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1107      if (c >= 'a' && c <= 'z') c -= 32;        {
1108          *errorcodeptr = ERR68;
1109          break;
1110          }
1111        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1112      c ^= 0x40;      c ^= 0x40;
1113  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1114      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1115      c ^= 0xC0;      c ^= 0xC0;
1116  #endif  #endif
1117      break;      break;
1118    
1119      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1120      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1121      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
1122      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
1123      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1124    
1125      default:      default:
1126      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 1133  else
1133      }      }
1134    }    }
1135    
1136    /* Perl supports \N{name} for character names, as well as plain \N for "not
1137    newline". PCRE does not support \N{name}. However, it does support
1138    quantification such as \N{2,3}. */
1139    
1140    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1141         !is_counted_repeat(ptr+2))
1142      *errorcodeptr = ERR37;
1143    
1144    /* If PCRE_UCP is set, we change the values for \d etc. */
1145    
1146    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1147      c -= (ESC_DU - ESC_D);
1148    
1149    /* Set the pointer to the final character before returning. */
1150    
1151  *ptrptr = ptr;  *ptrptr = ptr;
1152  return c;  return c;
1153  }  }
# Line 726  Returns:         type value from ucp_typ Line 1174  Returns:         type value from ucp_typ
1174  */  */
1175    
1176  static int  static int
1177  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1178  {  {
1179  int c, i, bot, top;  int c, i, bot, top;
1180  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1181  char name[32];  pcre_uchar name[32];
1182    
1183  c = *(++ptr);  c = *(++ptr);
1184  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 740  if (c == 0) goto ERROR_RETURN; Line 1188  if (c == 0) goto ERROR_RETURN;
1188  /* \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
1189  negation. */  negation. */
1190    
1191  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1192    {    {
1193    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1194      {      {
1195      *negptr = TRUE;      *negptr = TRUE;
1196      ptr++;      ptr++;
1197      }      }
1198    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1199      {      {
1200      c = *(++ptr);      c = *(++ptr);
1201      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1202      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1203      name[i] = c;      name[i] = c;
1204      }      }
1205    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1206    name[i] = 0;    name[i] = 0;
1207    }    }
1208    
# Line 771  else Line 1219  else
1219  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1220    
1221  bot = 0;  bot = 0;
1222  top = _pcre_utt_size;  top = PRIV(utt_size);
1223    
1224  while (bot < top)  while (bot < top)
1225    {    {
1226    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1227    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1228    if (c == 0)    if (c == 0)
1229      {      {
1230      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1231      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1232      }      }
1233    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1234    }    }
# Line 800  return -1; Line 1248  return -1;
1248    
1249    
1250  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1251  *         Read repeat counts                     *  *         Read repeat counts                     *
1252  *************************************************/  *************************************************/
1253    
# Line 851  Returns:         pointer to '}' on succe Line 1266  Returns:         pointer to '}' on succe
1266                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1267  */  */
1268    
1269  static const uschar *  static const pcre_uchar *
1270  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)
1271  {  {
1272  int min = 0;  int min = 0;
1273  int max = -1;  int max = -1;
# Line 860  int max = -1; Line 1275  int max = -1;
1275  /* 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
1276  an integer overflow. */  an integer overflow. */
1277    
1278  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1279  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1280    {    {
1281    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1285  if (min < 0 || min > 65535)
1285  /* 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.
1286  Also, max must not be less than min. */  Also, max must not be less than min. */
1287    
1288  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1289    {    {
1290    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1291      {      {
1292      max = 0;      max = 0;
1293      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1294      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1295        {        {
1296        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1315  return p;
1315    
1316    
1317  /*************************************************  /*************************************************
1318  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1319  *************************************************/  *************************************************/
1320    
1321  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1322    top-level call starts at the beginning of the pattern. All other calls must
1323    start at a parenthesis. It scans along a pattern's text looking for capturing
1324  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
1325  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
1326  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
1327  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1328  be terminated by '>' because that is checked in the first pass.  
1329    This function was originally called only from the second pass, in which we know
1330    that if (?< or (?' or (?P< is encountered, the name will be correctly
1331    terminated because that is checked in the first pass. There is now one call to
1332    this function in the first pass, to check for a recursive back reference by
1333    name (so that we can make the whole group atomic). In this case, we need check
1334    only up to the current position in the pattern, and that is still OK because
1335    and previous occurrences will have been checked. To make this work, the test
1336    for "end of pattern" is a check against cd->end_pattern in the main loop,
1337    instead of looking for a binary zero. This means that the special first-pass
1338    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1339    processing items within the loop are OK, because afterwards the main loop will
1340    terminate.)
1341    
1342  Arguments:  Arguments:
1343    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1344    count        current count of capturing parens so far encountered    cd           compile background data
1345    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1346    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1347    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1348      utf          TRUE if we are in UTF-8 / UTF-16 mode
1349      count        pointer to the current capturing subpattern number (updated)
1350    
1351  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1352  */  */
1353    
1354  static int  static int
1355  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,
1356    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1357  {  {
1358  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1359    int start_count = *count;
1360    int hwm_count = start_count;
1361    BOOL dup_parens = FALSE;
1362    
1363  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1364    dealing with. The very first call may not start with a parenthesis. */
1365    
1366    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1367    {    {
1368    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1369    
1370      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1371    
1372      /* Handle a normal, unnamed capturing parenthesis. */
1373    
1374      else if (ptr[1] != CHAR_QUESTION_MARK)
1375        {
1376        *count += 1;
1377        if (name == NULL && *count == lorn) return *count;
1378        ptr++;
1379        }
1380    
1381      /* All cases now have (? at the start. Remember when we are in a group
1382      where the parenthesis numbers are duplicated. */
1383    
1384      else if (ptr[2] == CHAR_VERTICAL_LINE)
1385        {
1386        ptr += 3;
1387        dup_parens = TRUE;
1388        }
1389    
1390      /* Handle comments; all characters are allowed until a ket is reached. */
1391    
1392      else if (ptr[2] == CHAR_NUMBER_SIGN)
1393        {
1394        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1395        goto FAIL_EXIT;
1396        }
1397    
1398      /* Handle a condition. If it is an assertion, just carry on so that it
1399      is processed as normal. If not, skip to the closing parenthesis of the
1400      condition (there can't be any nested parens). */
1401    
1402      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1403        {
1404        ptr += 2;
1405        if (ptr[1] != CHAR_QUESTION_MARK)
1406          {
1407          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1408          if (*ptr != 0) ptr++;
1409          }
1410        }
1411    
1412      /* Start with (? but not a condition. */
1413    
1414      else
1415        {
1416        ptr += 2;
1417        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1418    
1419        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1420    
1421        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1422            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1423          {
1424          int term;
1425          const pcre_uchar *thisname;
1426          *count += 1;
1427          if (name == NULL && *count == lorn) return *count;
1428          term = *ptr++;
1429          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1430          thisname = ptr;
1431          while (*ptr != term) ptr++;
1432          if (name != NULL && lorn == ptr - thisname &&
1433              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1434            return *count;
1435          term++;
1436          }
1437        }
1438      }
1439    
1440    /* Past any initial parenthesis handling, scan for parentheses or vertical
1441    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1442    first-pass call when this value is temporarily adjusted to stop at the current
1443    position. So DO NOT change this to a test for binary zero. */
1444    
1445    for (; ptr < cd->end_pattern; ptr++)
1446      {
1447    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1448    
1449    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1450      {      {
1451      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1452      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1453        {        {
1454        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1455        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1456        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1457        }        }
1458      continue;      continue;
1459      }      }
1460    
1461    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1462      are handled for real. If the first character is '^', skip it. Also, if the
1463      first few characters (either before or after ^) are \Q\E or \E we skip them
1464      too. This makes for compatibility with Perl. Note the use of STR macros to
1465      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1466    
1467    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1468      {      {
1469      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1470        for (;;)
1471        {        {
1472        if (*ptr == 0) return -1;        if (ptr[1] == CHAR_BACKSLASH)
       if (*ptr == '\\')  
1473          {          {
1474          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1475          if (*ptr == 'Q') for (;;)            ptr+= 2;
1476            else if (STRNCMP_UC_C8(ptr + 2,
1477                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1478              ptr += 4;
1479            else
1480              break;
1481            }
1482          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1483            {
1484            negate_class = TRUE;
1485            ptr++;
1486            }
1487          else break;
1488          }
1489    
1490        /* If the next character is ']', it is a data character that must be
1491        skipped, except in JavaScript compatibility mode. */
1492    
1493        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1494            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1495          ptr++;
1496    
1497        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1498          {
1499          if (*ptr == 0) return -1;
1500          if (*ptr == CHAR_BACKSLASH)
1501            {
1502            if (*(++ptr) == 0) goto FAIL_EXIT;
1503            if (*ptr == CHAR_Q) for (;;)
1504            {            {
1505            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1506            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1507            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1508            }            }
1509          continue;          continue;
1510          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1514  for (; *ptr != 0; ptr++)
1514    
1515    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1516    
1517    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1518      {      {
1519      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1520      if (*ptr == 0) return -1;      while (*ptr != 0)
1521          {
1522          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1523          ptr++;
1524    #ifdef SUPPORT_UTF
1525          if (utf) FORWARDCHAR(ptr);
1526    #endif
1527          }
1528        if (*ptr == 0) goto FAIL_EXIT;
1529      continue;      continue;
1530      }      }
1531    
1532    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1533    
1534    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1535      {      {
1536      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1537      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1538      continue;      if (*ptr == 0) goto FAIL_EXIT;
1539        }
1540    
1541      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1542        {
1543        if (dup_parens && *count < hwm_count) *count = hwm_count;
1544        goto FAIL_EXIT;
1545      }      }
1546    
1547    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1548    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1549        if (*count > hwm_count) hwm_count = *count;
1550        *count = start_count;
1551        }
1552      }
1553    
1554    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1555    *ptrptr = ptr;
1556    return -1;
1557    }
1558    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1559    
   count++;  
1560    
1561    if (name == NULL && count == lorn) return count;  
1562    term = *ptr++;  /*************************************************
1563    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1564    thisname = ptr;  *************************************************/
1565    while (*ptr != term) ptr++;  
1566    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1567        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1568      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1569    returns when it reaches a given numbered subpattern. This is used for forward
1570    references to subpatterns. We used to be able to start this scan from the
1571    current compiling point, using the current count value from cd->bracount, and
1572    do it all in a single loop, but the addition of the possibility of duplicate
1573    subpattern numbers means that we have to scan from the very start, in order to
1574    take account of such duplicates, and to use a recursive function to keep track
1575    of the different types of group.
1576    
1577    Arguments:
1578      cd           compile background data
1579      name         name to seek, or NULL if seeking a numbered subpattern
1580      lorn         name length, or subpattern number if name is NULL
1581      xmode        TRUE if we are in /x mode
1582      utf          TRUE if we are in UTF-8 / UTF-16 mode
1583    
1584    Returns:       the number of the found subpattern, or -1 if not found
1585    */
1586    
1587    static int
1588    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1589      BOOL utf)
1590    {
1591    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1592    int count = 0;
1593    int rc;
1594    
1595    /* If the pattern does not start with an opening parenthesis, the first call
1596    to find_parens_sub() will scan right to the end (if necessary). However, if it
1597    does start with a parenthesis, find_parens_sub() will return when it hits the
1598    matching closing parens. That is why we have to have a loop. */
1599    
1600    for (;;)
1601      {
1602      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1603      if (rc > 0 || *ptr++ == 0) break;
1604    }    }
1605    
1606  return -1;  return rc;
1607  }  }
1608    
1609    
1610    
1611    
1612  /*************************************************  /*************************************************
1613  *      Find first significant op code            *  *      Find first significant op code            *
1614  *************************************************/  *************************************************/
1615    
1616  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1617  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
1618  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
1619  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
1620  assertions, and also the \b assertion; for others it does not.  does not.
1621    
1622  Arguments:  Arguments:
1623    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  
1624    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1625    
1626  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1627  */  */
1628    
1629  static const uschar*  static const pcre_uchar*
1630  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1631  {  {
1632  for (;;)  for (;;)
1633    {    {
1634    switch ((int)*code)    switch ((int)*code)
1635      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1636      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1637      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1638      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1639      if (!skipassert) return code;      if (!skipassert) return code;
1640      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1641      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1642      break;      break;
1643    
1644      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1060  for (;;) Line 1648  for (;;)
1648    
1649      case OP_CALLOUT:      case OP_CALLOUT:
1650      case OP_CREF:      case OP_CREF:
1651        case OP_NCREF:
1652      case OP_RREF:      case OP_RREF:
1653        case OP_NRREF:
1654      case OP_DEF:      case OP_DEF:
1655      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1656      break;      break;
1657    
1658      default:      default:
# Line 1076  for (;;) Line 1666  for (;;)
1666    
1667    
1668  /*************************************************  /*************************************************
1669  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1670  *************************************************/  *************************************************/
1671    
1672  /* 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,
1673  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.
1674  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
1675    temporarily terminated with OP_END when this function is called.
1676    
1677    This function is called when a backward assertion is encountered, so that if it
1678    fails, the error message can point to the correct place in the pattern.
1679    However, we cannot do this when the assertion contains subroutine calls,
1680    because they can be forward references. We solve this by remembering this case
1681    and doing the check at the end; a flag specifies which mode we are running in.
1682    
1683  Arguments:  Arguments:
1684    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1685    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1686      atend    TRUE if called when the pattern is complete
1687  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1688               or -2 if \C was encountered  
1689    Returns:   the fixed length,
1690                 or -1 if there is no fixed length,
1691                 or -2 if \C was encountered (in UTF-8 mode only)
1692                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1693                 or -4 if an unknown opcode was encountered (internal error)
1694  */  */
1695    
1696  static int  static int
1697  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1698  {  {
1699  int length = -1;  int length = -1;
1700    
1701  register int branchlength = 0;  register int branchlength = 0;
1702  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1703    
1704  /* 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
1705  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 1707  branch, check the length against that of
1707  for (;;)  for (;;)
1708    {    {
1709    int d;    int d;
1710      pcre_uchar *ce, *cs;
1711    register int op = *cc;    register int op = *cc;
1712    
1713    switch (op)    switch (op)
1714      {      {
1715        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1716        OP_BRA (normal non-capturing bracket) because the other variants of these
1717        opcodes are all concerned with unlimited repeated groups, which of course
1718        are not of fixed length. */
1719    
1720      case OP_CBRA:      case OP_CBRA:
1721      case OP_BRA:      case OP_BRA:
1722      case OP_ONCE:      case OP_ONCE:
1723        case OP_ONCE_NC:
1724      case OP_COND:      case OP_COND:
1725      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1726      if (d < 0) return d;      if (d < 0) return d;
1727      branchlength += d;      branchlength += d;
1728      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1729      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1730      break;      break;
1731    
1732      /* 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.
1733      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
1734      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
1735        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1736        because they all imply an unlimited repeat. */
1737    
1738      case OP_ALT:      case OP_ALT:
1739      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1740      case OP_END:      case OP_END:
1741        case OP_ACCEPT:
1742        case OP_ASSERT_ACCEPT:
1743      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1744        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1745      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1135  for (;;) Line 1747  for (;;)
1747      branchlength = 0;      branchlength = 0;
1748      break;      break;
1749    
1750        /* A true recursion implies not fixed length, but a subroutine call may
1751        be OK. If the subroutine is a forward reference, we can't deal with
1752        it until the end of the pattern, so return -3. */
1753    
1754        case OP_RECURSE:
1755        if (!atend) return -3;
1756        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1757        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1758        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1759        d = find_fixedlength(cs + 2, utf, atend, cd);
1760        if (d < 0) return d;
1761        branchlength += d;
1762        cc += 1 + LINK_SIZE;
1763        break;
1764    
1765      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1766    
1767      case OP_ASSERT:      case OP_ASSERT:
# Line 1142  for (;;) Line 1769  for (;;)
1769      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1770      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1771      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1772      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1773        break;
1774    
1775      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1776    
1777      case OP_REVERSE:      case OP_MARK:
1778        case OP_PRUNE_ARG:
1779        case OP_SKIP_ARG:
1780        case OP_THEN_ARG:
1781        cc += cc[1] + PRIV(OP_lengths)[*cc];
1782        break;
1783    
1784        case OP_CALLOUT:
1785        case OP_CIRC:
1786        case OP_CIRCM:
1787        case OP_CLOSE:
1788        case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1790      case OP_DEF:      case OP_DEF:
1791      case OP_OPT:      case OP_DOLL:
1792      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1793      case OP_EOD:      case OP_EOD:
1794      case OP_EODN:      case OP_EODN:
1795      case OP_CIRC:      case OP_FAIL:
1796      case OP_DOLL:      case OP_NCREF:
1797        case OP_NRREF:
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799        case OP_PRUNE:
1800        case OP_REVERSE:
1801        case OP_RREF:
1802        case OP_SET_SOM:
1803        case OP_SKIP:
1804        case OP_SOD:
1805        case OP_SOM:
1806        case OP_THEN:
1807      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1808      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1809      break;      break;
1810    
1811      /* Handle literal characters */      /* Handle literal characters */
1812    
1813      case OP_CHAR:      case OP_CHAR:
1814      case OP_CHARNC:      case OP_CHARI:
1815      case OP_NOT:      case OP_NOT:
1816        case OP_NOTI:
1817      branchlength++;      branchlength++;
1818      cc += 2;      cc += 2;
1819  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1820      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1821  #endif  #endif
1822      break;      break;
1823    
# Line 1182  for (;;) Line 1825  for (;;)
1825      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1826    
1827      case OP_EXACT:      case OP_EXACT:
1828        case OP_EXACTI:
1829        case OP_NOTEXACT:
1830        case OP_NOTEXACTI:
1831      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1832      cc += 4;      cc += 2 + IMM2_SIZE;
1833  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1834      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1835  #endif  #endif
1836      break;      break;
1837    
1838      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1839      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1840      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;
1841      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1842      break;      break;
1843    
1844      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1205  for (;;) Line 1848  for (;;)
1848      cc += 2;      cc += 2;
1849      /* Fall through */      /* Fall through */
1850    
1851        case OP_HSPACE:
1852        case OP_VSPACE:
1853        case OP_NOT_HSPACE:
1854        case OP_NOT_VSPACE:
1855      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1856      case OP_DIGIT:      case OP_DIGIT:
1857      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1212  for (;;) Line 1859  for (;;)
1859      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1860      case OP_WORDCHAR:      case OP_WORDCHAR:
1861      case OP_ANY:      case OP_ANY:
1862        case OP_ALLANY:
1863      branchlength++;      branchlength++;
1864      cc++;      cc++;
1865      break;      break;
1866    
1867      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1868        otherwise \C is coded as OP_ALLANY. */
1869    
1870      case OP_ANYBYTE:      case OP_ANYBYTE:
1871      return -2;      return -2;
1872    
1873      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1874    
1875  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1876      case OP_XCLASS:      case OP_XCLASS:
1877      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1878      /* Fall through */      /* Fall through */
1879  #endif  #endif
1880    
1881      case OP_CLASS:      case OP_CLASS:
1882      case OP_NCLASS:      case OP_NCLASS:
1883      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1884    
1885      switch (*cc)      switch (*cc)
1886        {        {
1887          case OP_CRPLUS:
1888          case OP_CRMINPLUS:
1889        case OP_CRSTAR:        case OP_CRSTAR:
1890        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1891        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1243  for (;;) Line 1894  for (;;)
1894    
1895        case OP_CRRANGE:        case OP_CRRANGE:
1896        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1897        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1898        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1899        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1900        break;        break;
1901    
1902        default:        default:
# Line 1255  for (;;) Line 1906  for (;;)
1906    
1907      /* Anything else is variable length */      /* Anything else is variable length */
1908    
1909      default:      case OP_ANYNL:
1910        case OP_BRAMINZERO:
1911        case OP_BRAPOS:
1912        case OP_BRAPOSZERO:
1913        case OP_BRAZERO:
1914        case OP_CBRAPOS:
1915        case OP_EXTUNI:
1916        case OP_KETRMAX:
1917        case OP_KETRMIN:
1918        case OP_KETRPOS:
1919        case OP_MINPLUS:
1920        case OP_MINPLUSI:
1921        case OP_MINQUERY:
1922        case OP_MINQUERYI:
1923        case OP_MINSTAR:
1924        case OP_MINSTARI:
1925        case OP_MINUPTO:
1926        case OP_MINUPTOI:
1927        case OP_NOTMINPLUS:
1928        case OP_NOTMINPLUSI:
1929        case OP_NOTMINQUERY:
1930        case OP_NOTMINQUERYI:
1931        case OP_NOTMINSTAR:
1932        case OP_NOTMINSTARI:
1933        case OP_NOTMINUPTO:
1934        case OP_NOTMINUPTOI:
1935        case OP_NOTPLUS:
1936        case OP_NOTPLUSI:
1937        case OP_NOTPOSPLUS:
1938        case OP_NOTPOSPLUSI:
1939        case OP_NOTPOSQUERY:
1940        case OP_NOTPOSQUERYI:
1941        case OP_NOTPOSSTAR:
1942        case OP_NOTPOSSTARI:
1943        case OP_NOTPOSUPTO:
1944        case OP_NOTPOSUPTOI:
1945        case OP_NOTQUERY:
1946        case OP_NOTQUERYI:
1947        case OP_NOTSTAR:
1948        case OP_NOTSTARI:
1949        case OP_NOTUPTO:
1950        case OP_NOTUPTOI:
1951        case OP_PLUS:
1952        case OP_PLUSI:
1953        case OP_POSPLUS:
1954        case OP_POSPLUSI:
1955        case OP_POSQUERY:
1956        case OP_POSQUERYI:
1957        case OP_POSSTAR:
1958        case OP_POSSTARI:
1959        case OP_POSUPTO:
1960        case OP_POSUPTOI:
1961        case OP_QUERY:
1962        case OP_QUERYI:
1963        case OP_REF:
1964        case OP_REFI:
1965        case OP_SBRA:
1966        case OP_SBRAPOS:
1967        case OP_SCBRA:
1968        case OP_SCBRAPOS:
1969        case OP_SCOND:
1970        case OP_SKIPZERO:
1971        case OP_STAR:
1972        case OP_STARI:
1973        case OP_TYPEMINPLUS:
1974        case OP_TYPEMINQUERY:
1975        case OP_TYPEMINSTAR:
1976        case OP_TYPEMINUPTO:
1977        case OP_TYPEPLUS:
1978        case OP_TYPEPOSPLUS:
1979        case OP_TYPEPOSQUERY:
1980        case OP_TYPEPOSSTAR:
1981        case OP_TYPEPOSUPTO:
1982        case OP_TYPEQUERY:
1983        case OP_TYPESTAR:
1984        case OP_TYPEUPTO:
1985        case OP_UPTO:
1986        case OP_UPTOI:
1987      return -1;      return -1;
1988    
1989        /* Catch unrecognized opcodes so that when new ones are added they
1990        are not forgotten, as has happened in the past. */
1991    
1992        default:
1993        return -4;
1994      }      }
1995    }    }
1996  /* Control never gets here */  /* Control never gets here */
# Line 1266  for (;;) Line 2000  for (;;)
2000    
2001    
2002  /*************************************************  /*************************************************
2003  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2004  *************************************************/  *************************************************/
2005    
2006  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2007  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2008    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2009    so that it can be called from pcre_study() when finding the minimum matching
2010    length.
2011    
2012  Arguments:  Arguments:
2013    code        points to start of expression    code        points to start of expression
2014    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2015    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2016    
2017  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
2018  */  */
2019    
2020  static const uschar *  const pcre_uchar *
2021  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2022  {  {
2023  for (;;)  for (;;)
2024    {    {
2025    register int c = *code;    register int c = *code;
2026    
2027    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2028    
2029    /* 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 2032  for (;;)
2032    
2033    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2034    
2035      /* Handle recursion */
2036    
2037      else if (c == OP_REVERSE)
2038        {
2039        if (number < 0) return (pcre_uchar *)code;
2040        code += PRIV(OP_lengths)[c];
2041        }
2042    
2043    /* Handle capturing bracket */    /* Handle capturing bracket */
2044    
2045    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2046               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2047      {      {
2048      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2049      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2050      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2051      }      }
2052    
2053    /* 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
2054    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
2055    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2056      must add in its length. */
2057    
2058    else    else
2059      {      {
# Line 1317  for (;;) Line 2065  for (;;)
2065        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
2066        case OP_TYPEQUERY:        case OP_TYPEQUERY:
2067        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
2068        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
2069        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
2070        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
2071        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2072        break;        break;
2073        }  
2074          case OP_TYPEUPTO:
2075          case OP_TYPEMINUPTO:
2076          case OP_TYPEEXACT:
2077          case OP_TYPEPOSUPTO:
2078          if (code[1 + IMM2_SIZE] == OP_PROP
2079            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2080          break;
2081    
2082          case OP_MARK:
2083          case OP_PRUNE_ARG:
2084          case OP_SKIP_ARG:
2085          code += code[1];
2086          break;
2087    
2088          case OP_THEN_ARG:
2089          code += code[1];
2090          break;
2091          }
2092    
2093      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2094    
2095      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2096    
2097    /* 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
2098    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
2099    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2100    
2101  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2102      if (utf8) switch(c)      if (utf) switch(c)
2103        {        {
2104        case OP_CHAR:        case OP_CHAR:
2105        case OP_CHARNC:        case OP_CHARI:
2106        case OP_EXACT:        case OP_EXACT:
2107          case OP_EXACTI:
2108        case OP_UPTO:        case OP_UPTO:
2109          case OP_UPTOI:
2110        case OP_MINUPTO:        case OP_MINUPTO:
2111          case OP_MINUPTOI:
2112        case OP_POSUPTO:        case OP_POSUPTO:
2113          case OP_POSUPTOI:
2114        case OP_STAR:        case OP_STAR:
2115          case OP_STARI:
2116        case OP_MINSTAR:        case OP_MINSTAR:
2117          case OP_MINSTARI:
2118        case OP_POSSTAR:        case OP_POSSTAR:
2119          case OP_POSSTARI:
2120        case OP_PLUS:        case OP_PLUS:
2121          case OP_PLUSI:
2122        case OP_MINPLUS:        case OP_MINPLUS:
2123          case OP_MINPLUSI:
2124        case OP_POSPLUS:        case OP_POSPLUS:
2125          case OP_POSPLUSI:
2126        case OP_QUERY:        case OP_QUERY:
2127          case OP_QUERYI:
2128        case OP_MINQUERY:        case OP_MINQUERY:
2129          case OP_MINQUERYI:
2130        case OP_POSQUERY:        case OP_POSQUERY:
2131        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2132          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2133        break;        break;
2134        }        }
2135    #else
2136        (void)(utf);  /* Keep compiler happy by referencing function argument */
2137  #endif  #endif
2138      }      }
2139    }    }
# Line 1373  instance of OP_RECURSE. Line 2150  instance of OP_RECURSE.
2150    
2151  Arguments:  Arguments:
2152    code        points to start of expression    code        points to start of expression
2153    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2154    
2155  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
2156  */  */
2157    
2158  static const uschar *  static const pcre_uchar *
2159  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2160  {  {
2161  for (;;)  for (;;)
2162    {    {
2163    register int c = *code;    register int c = *code;
2164    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2165    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2166    
2167    /* 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
2168    map. This includes negated single high-valued characters. The length in    map. This includes negated single high-valued characters. The length in
2169    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
2170    
2171    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2172    
2173    /* 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
2174    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
2175    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2176      must add in its length. */
2177    
2178    else    else
2179      {      {
# Line 1407  for (;;) Line 2185  for (;;)
2185        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
2186        case OP_TYPEQUERY:        case OP_TYPEQUERY:
2187        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
2188        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
2189        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
2190        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
2191        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2192        break;        break;
2193        }  
2194          case OP_TYPEPOSUPTO:
2195          case OP_TYPEUPTO:
2196          case OP_TYPEMINUPTO:
2197          case OP_TYPEEXACT:
2198          if (code[1 + IMM2_SIZE] == OP_PROP
2199            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2200          break;
2201    
2202          case OP_MARK:
2203          case OP_PRUNE_ARG:
2204          case OP_SKIP_ARG:
2205          code += code[1];
2206          break;
2207    
2208          case OP_THEN_ARG:
2209          code += code[1];
2210          break;
2211          }
2212    
2213      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2214    
2215      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2216    
2217      /* 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
2218      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
2219      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2220    
2221  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2222      if (utf8) switch(c)      if (utf) switch(c)
2223        {        {
2224        case OP_CHAR:        case OP_CHAR:
2225        case OP_CHARNC:        case OP_CHARI:
2226        case OP_EXACT:        case OP_EXACT:
2227          case OP_EXACTI:
2228        case OP_UPTO:        case OP_UPTO:
2229          case OP_UPTOI:
2230        case OP_MINUPTO:        case OP_MINUPTO:
2231          case OP_MINUPTOI:
2232        case OP_POSUPTO:        case OP_POSUPTO:
2233          case OP_POSUPTOI:
2234        case OP_STAR:        case OP_STAR:
2235          case OP_STARI:
2236        case OP_MINSTAR:        case OP_MINSTAR:
2237          case OP_MINSTARI:
2238        case OP_POSSTAR:        case OP_POSSTAR:
2239          case OP_POSSTARI:
2240        case OP_PLUS:        case OP_PLUS:
2241          case OP_PLUSI:
2242        case OP_MINPLUS:        case OP_MINPLUS:
2243          case OP_MINPLUSI:
2244        case OP_POSPLUS:        case OP_POSPLUS:
2245          case OP_POSPLUSI:
2246        case OP_QUERY:        case OP_QUERY:
2247          case OP_QUERYI:
2248        case OP_MINQUERY:        case OP_MINQUERY:
2249          case OP_MINQUERYI:
2250        case OP_POSQUERY:        case OP_POSQUERY:
2251        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2252          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2253        break;        break;
2254        }        }
2255    #else
2256        (void)(utf);  /* Keep compiler happy by referencing function argument */
2257  #endif  #endif
2258      }      }
2259    }    }
# Line 1462  for (;;) Line 2269  for (;;)
2269  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()
2270  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
2271  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2272  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
2273  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
2274    bracket whose current branch will already have been scanned.
2275    
2276  Arguments:  Arguments:
2277    code        points to start of search    code        points to start of search
2278    endcode     points to where to stop    endcode     points to where to stop
2279    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2280      cd          contains pointers to tables etc.
2281    
2282  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2283  */  */
2284    
2285  static BOOL  static BOOL
2286  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2287      BOOL utf, compile_data *cd)
2288  {  {
2289  register int c;  register int c;
2290  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2291       code < endcode;       code < endcode;
2292       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2293    {    {
2294    const uschar *ccode;    const pcre_uchar *ccode;
2295    
2296    c = *code;    c = *code;
2297    
2298    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2299      first_significant_code() with a TRUE final argument. */
2300    
2301    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2302      {      {
     code += _pcre_OP_lengths[c];  
2303      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2304      c = *code;      c = *code;
2305      continue;      continue;
2306      }      }
2307    
2308    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2309      implies a backward reference subroutine call, we can scan it. If it's a
2310      forward reference subroutine call, we can't. To detect forward reference
2311      we have to scan up the list that is kept in the workspace. This function is
2312      called only when doing the real compile, not during the pre-compile that
2313      measures the size of the compiled pattern. */
2314    
2315    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2316      {      {
2317        const pcre_uchar *scode;
2318      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2319    
2320      /* Scan a closed bracket */      /* Test for forward reference */
2321    
2322        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2323          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2324    
2325        /* Not a forward reference, test for completed backward reference */
2326    
2327      empty_branch = FALSE;      empty_branch = FALSE;
2328        scode = cd->start_code + GET(code, 1);
2329        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2330    
2331        /* Completed backwards reference */
2332    
2333      do      do
2334        {        {
2335        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2336            {
2337          empty_branch = TRUE;          empty_branch = TRUE;
2338            break;
2339            }
2340          scode += GET(scode, 1);
2341          }
2342        while (*scode == OP_ALT);
2343    
2344        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2345        continue;
2346        }
2347    
2348      /* Groups with zero repeats can of course be empty; skip them. */
2349    
2350      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2351          c == OP_BRAPOSZERO)
2352        {
2353        code += PRIV(OP_lengths)[c];
2354        do code += GET(code, 1); while (*code == OP_ALT);
2355        c = *code;
2356        continue;
2357        }
2358    
2359      /* A nested group that is already marked as "could be empty" can just be
2360      skipped. */
2361    
2362      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2363          c == OP_SCBRA || c == OP_SCBRAPOS)
2364        {
2365        do code += GET(code, 1); while (*code == OP_ALT);
2366        c = *code;
2367        continue;
2368        }
2369    
2370      /* For other groups, scan the branches. */
2371    
2372      if (c == OP_BRA  || c == OP_BRAPOS ||
2373          c == OP_CBRA || c == OP_CBRAPOS ||
2374          c == OP_ONCE || c == OP_ONCE_NC ||
2375          c == OP_COND)
2376        {
2377        BOOL empty_branch;
2378        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2379    
2380        /* If a conditional group has only one branch, there is a second, implied,
2381        empty branch, so just skip over the conditional, because it could be empty.
2382        Otherwise, scan the individual branches of the group. */
2383    
2384        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2385        code += GET(code, 1);        code += GET(code, 1);
2386        else
2387          {
2388          empty_branch = FALSE;
2389          do
2390            {
2391            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2392              empty_branch = TRUE;
2393            code += GET(code, 1);
2394            }
2395          while (*code == OP_ALT);
2396          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2397        }        }
2398      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2399      c = *code;      c = *code;
2400      continue;      continue;
2401      }      }
# Line 1523  for (code = first_significant_code(code Line 2406  for (code = first_significant_code(code
2406      {      {
2407      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2408      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2409      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2410      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"
2411      here. */      here. */
2412    
2413  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2414      case OP_XCLASS:      case OP_XCLASS:
2415      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2416      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1535  for (code = first_significant_code(code Line 2418  for (code = first_significant_code(code
2418    
2419      case OP_CLASS:      case OP_CLASS:
2420      case OP_NCLASS:      case OP_NCLASS:
2421      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2422    
2423  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2424      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2425  #endif  #endif
2426    
# Line 1573  for (code = first_significant_code(code Line 2456  for (code = first_significant_code(code
2456      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2457      case OP_WORDCHAR:      case OP_WORDCHAR:
2458      case OP_ANY:      case OP_ANY:
2459        case OP_ALLANY:
2460      case OP_ANYBYTE:      case OP_ANYBYTE:
2461      case OP_CHAR:      case OP_CHAR:
2462      case OP_CHARNC:      case OP_CHARI:
2463      case OP_NOT:      case OP_NOT:
2464        case OP_NOTI:
2465      case OP_PLUS:      case OP_PLUS:
2466      case OP_MINPLUS:      case OP_MINPLUS:
2467      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1591  for (code = first_significant_code(code Line 2476  for (code = first_significant_code(code
2476      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2477      return FALSE;      return FALSE;
2478    
2479        /* These are going to continue, as they may be empty, but we have to
2480        fudge the length for the \p and \P cases. */
2481    
2482        case OP_TYPESTAR:
2483        case OP_TYPEMINSTAR:
2484        case OP_TYPEPOSSTAR:
2485        case OP_TYPEQUERY:
2486        case OP_TYPEMINQUERY:
2487        case OP_TYPEPOSQUERY:
2488        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2489        break;
2490    
2491        /* Same for these */
2492    
2493        case OP_TYPEUPTO:
2494        case OP_TYPEMINUPTO:
2495        case OP_TYPEPOSUPTO:
2496        if (code[1 + IMM2_SIZE] == OP_PROP
2497          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2498        break;
2499    
2500      /* End of branch */      /* End of branch */
2501    
2502      case OP_KET:      case OP_KET:
2503      case OP_KETRMAX:      case OP_KETRMAX:
2504      case OP_KETRMIN:      case OP_KETRMIN:
2505        case OP_KETRPOS:
2506      case OP_ALT:      case OP_ALT:
2507      return TRUE;      return TRUE;
2508    
2509      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2510      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2511    
2512  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2513      case OP_STAR:      case OP_STAR:
2514        case OP_STARI:
2515      case OP_MINSTAR:      case OP_MINSTAR:
2516        case OP_MINSTARI:
2517      case OP_POSSTAR:      case OP_POSSTAR:
2518        case OP_POSSTARI:
2519      case OP_QUERY:      case OP_QUERY:
2520        case OP_QUERYI:
2521      case OP_MINQUERY:      case OP_MINQUERY:
2522        case OP_MINQUERYI:
2523      case OP_POSQUERY:      case OP_POSQUERY:
2524        case OP_POSQUERYI:
2525        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2526        break;
2527    
2528      case OP_UPTO:      case OP_UPTO:
2529        case OP_UPTOI:
2530      case OP_MINUPTO:      case OP_MINUPTO:
2531        case OP_MINUPTOI:
2532      case OP_POSUPTO:      case OP_POSUPTO:
2533      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2534        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2535      break;      break;
2536  #endif  #endif
2537    
2538        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2539        string. */
2540    
2541        case OP_MARK:
2542        case OP_PRUNE_ARG:
2543        case OP_SKIP_ARG:
2544        code += code[1];
2545        break;
2546    
2547        case OP_THEN_ARG:
2548        code += code[1];
2549        break;
2550    
2551        /* None of the remaining opcodes are required to match a character. */
2552    
2553        default:
2554        break;
2555      }      }
2556    }    }
2557    
# Line 1631  return TRUE; Line 2568  return TRUE;
2568  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
2569  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,
2570  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.
2571    This function is called only during the real compile, not during the
2572    pre-compile.
2573    
2574  Arguments:  Arguments:
2575    code        points to start of the recursion    code        points to start of the recursion
2576    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2577    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2578    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2579      cd          pointers to tables etc
2580    
2581  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2582  */  */
2583    
2584  static BOOL  static BOOL
2585  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2586    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2587  {  {
2588  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2589    {    {
2590    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2591        return FALSE;
2592    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2593    }    }
2594  return TRUE;  return TRUE;
# Line 1660  return TRUE; Line 2601  return TRUE;
2601  *************************************************/  *************************************************/
2602    
2603  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2604  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
2605  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2606  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2607    
2608    Originally, this function only recognized a sequence of letters between the
2609    terminators, but it seems that Perl recognizes any sequence of characters,
2610    though of course unknown POSIX names are subsequently rejected. Perl gives an
2611    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2612    didn't consider this to be a POSIX class. Likewise for [:1234:].
2613    
2614    The problem in trying to be exactly like Perl is in the handling of escapes. We
2615    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2616    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2617    below handles the special case of \], but does not try to do any other escape
2618    processing. This makes it different from Perl for cases such as [:l\ower:]
2619    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2620    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2621    I think.
2622    
2623    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2624    It seems that the appearance of a nested POSIX class supersedes an apparent
2625    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2626    a digit.
2627    
2628    In Perl, unescaped square brackets may also appear as part of class names. For
2629    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2630    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2631    seem right at all. PCRE does not allow closing square brackets in POSIX class
2632    names.
2633    
2634  Argument:  Arguments:
2635    ptr      pointer to the initial [    ptr      pointer to the initial [
2636    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2637    
2638  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2639  */  */
2640    
2641  static BOOL  static BOOL
2642  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2643  {  {
2644  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2645  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2646  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2647    {    {
2648    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2649    return TRUE;      ptr++;
2650      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2651      else
2652        {
2653        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2654          {
2655          *endptr = ptr;
2656          return TRUE;
2657          }
2658        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2659             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2660              ptr[1] == CHAR_EQUALS_SIGN) &&
2661            check_posix_syntax(ptr, endptr))
2662          return FALSE;
2663        }
2664    }    }
2665  return FALSE;  return FALSE;
2666  }  }
# Line 1705  Returns:     a value representing the na Line 2683  Returns:     a value representing the na
2683  */  */
2684    
2685  static int  static int
2686  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2687  {  {
2688    const char *pn = posix_names;
2689  register int yield = 0;  register int yield = 0;
2690  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2691    {    {
2692    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2693      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2694      pn += posix_name_lengths[yield] + 1;
2695    yield++;    yield++;
2696    }    }
2697  return -1;  return -1;
# Line 1726  return -1; Line 2706  return -1;
2706  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2707  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2708  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
2709  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
2710  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
2711  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
2712  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
2713  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2714    OP_END.
2715    
2716  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2717  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 1741  value in the reference (which is a group Line 2722  value in the reference (which is a group
2722  Arguments:  Arguments:
2723    group      points to the start of the group    group      points to the start of the group
2724    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2725    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2726    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2727    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
2728    
# Line 1749  Returns:     nothing Line 2730  Returns:     nothing
2730  */  */
2731    
2732  static void  static void
2733  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2734    uschar *save_hwm)    pcre_uchar *save_hwm)
2735  {  {
2736  uschar *ptr = group;  pcre_uchar *ptr = group;
2737  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2738    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2739    {    {
2740    int offset;    int offset;
2741    uschar *hc;    pcre_uchar *hc;
2742    
2743    /* 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
2744    reference. */    reference. */
# Line 1801  Arguments: Line 2783  Arguments:
2783  Returns:         new code pointer  Returns:         new code pointer
2784  */  */
2785    
2786  static uschar *  static pcre_uchar *
2787  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2788  {  {
2789  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2790  *code++ = 255;  *code++ = 255;
2791  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2792  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2793  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2794  }  }
2795    
2796    
# Line 1830  Returns:             nothing Line 2812  Returns:             nothing
2812  */  */
2813    
2814  static void  static void
2815  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2816  {  {
2817  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2818  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2819  }  }
2820    
# Line 1864  get_othercase_range(unsigned int *cptr, Line 2846  get_othercase_range(unsigned int *cptr,
2846  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2847    
2848  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2849    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2850    
2851  if (c > d) return FALSE;  if (c > d) return FALSE;
2852    
# Line 1873  next = othercase + 1; Line 2855  next = othercase + 1;
2855    
2856  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2857    {    {
2858    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2859    next++;    next++;
2860    }    }
2861    
# Line 1882  for (++c; c <= d; c++) Line 2864  for (++c; c <= d; c++)
2864    
2865  return TRUE;  return TRUE;
2866  }  }
2867    
2868    
2869    
2870    /*************************************************
2871    *        Check a character and a property        *
2872    *************************************************/
2873    
2874    /* This function is called by check_auto_possessive() when a property item
2875    is adjacent to a fixed character.
2876    
2877    Arguments:
2878      c            the character
2879      ptype        the property type
2880      pdata        the data for the type
2881      negated      TRUE if it's a negated property (\P or \p{^)
2882    
2883    Returns:       TRUE if auto-possessifying is OK
2884    */
2885    
2886    static BOOL
2887    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2888    {
2889    const ucd_record *prop = GET_UCD(c);
2890    switch(ptype)
2891      {
2892      case PT_LAMP:
2893      return (prop->chartype == ucp_Lu ||
2894              prop->chartype == ucp_Ll ||
2895              prop->chartype == ucp_Lt) == negated;
2896    
2897      case PT_GC:
2898      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2899    
2900      case PT_PC:
2901      return (pdata == prop->chartype) == negated;
2902    
2903      case PT_SC:
2904      return (pdata == prop->script) == negated;
2905    
2906      /* These are specials */
2907    
2908      case PT_ALNUM:
2909      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2910              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2911    
2912      case PT_SPACE:    /* Perl space */
2913      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2914              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2915              == negated;
2916    
2917      case PT_PXSPACE:  /* POSIX space */
2918      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2919              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2920              c == CHAR_FF || c == CHAR_CR)
2921              == negated;
2922    
2923      case PT_WORD:
2924      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2925              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2926              c == CHAR_UNDERSCORE) == negated;
2927      }
2928    return FALSE;
2929    }
2930  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2931    
2932    
# Line 1895  whether the next thing could possibly ma Line 2940  whether the next thing could possibly ma
2940  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2941    
2942  Arguments:  Arguments:
2943    op_code       the repeated op code    previous      pointer to the repeated opcode
2944    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  
2945    ptr           next character in pattern    ptr           next character in pattern
2946    options       options bits    options       options bits
2947    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1907  Returns:        TRUE if possessifying is Line 2950  Returns:        TRUE if possessifying is
2950  */  */
2951    
2952  static BOOL  static BOOL
2953  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2954    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2955  {  {
2956  int next;  pcre_int32 c, next;
2957    int op_code = *previous++;
2958    
2959  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2960    
# Line 1918  if ((options & PCRE_EXTENDED) != 0) Line 2962  if ((options & PCRE_EXTENDED) != 0)
2962    {    {
2963    for (;;)    for (;;)
2964      {      {
2965      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2966      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2967        {        {
2968        while (*(++ptr) != 0)        ptr++;
2969          while (*ptr != 0)
2970            {
2971          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2972            ptr++;
2973    #ifdef SUPPORT_UTF
2974            if (utf) FORWARDCHAR(ptr);
2975    #endif
2976            }
2977        }        }
2978      else break;      else break;
2979      }      }
# Line 1931  if ((options & PCRE_EXTENDED) != 0) Line 2982  if ((options & PCRE_EXTENDED) != 0)
2982  /* 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
2983  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2984    
2985  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2986    {    {
2987    int temperrorcode = 0;    int temperrorcode = 0;
2988    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2989    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2990    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2991    }    }
2992    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2993    {    {
2994  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2995    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2996  #endif  #endif
2997    next = *ptr++;    next = *ptr++;
2998    }    }
   
2999  else return FALSE;  else return FALSE;
3000    
3001  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1955  if ((options & PCRE_EXTENDED) != 0) Line 3004  if ((options & PCRE_EXTENDED) != 0)
3004    {    {
3005    for (;;)    for (;;)
3006      {      {
3007      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3008      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3009        {        {
3010        while (*(++ptr) != 0)        ptr++;
3011          while (*ptr != 0)
3012            {
3013          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3014            ptr++;
3015    #ifdef SUPPORT_UTF
3016            if (utf) FORWARDCHAR(ptr);
3017    #endif
3018            }
3019        }        }
3020      else break;      else break;
3021      }      }
# Line 1967  if ((options & PCRE_EXTENDED) != 0) Line 3023  if ((options & PCRE_EXTENDED) != 0)
3023    
3024  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3025    
3026  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3027    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3028        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. */  
   
3029    
3030  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3031    the next item is a character. */
3032    
3033  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3034    {    {
3035    case OP_CHAR:    case OP_CHAR:
3036  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3037    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3038    #else
3039      c = *previous;
3040  #endif  #endif
3041    return item != next;    return c != next;
3042    
3043    /* 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
3044    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
3045    high-valued characters. */    high-valued characters. */
3046    
3047    case OP_CHARNC:    case OP_CHARI:
3048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3049    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3050  #endif  #else
3051    if (item == next) return FALSE;    c = *previous;
3052  #ifdef SUPPORT_UTF8  #endif
3053    if (utf8)    if (c == next) return FALSE;
3054    #ifdef SUPPORT_UTF
3055      if (utf)
3056      {      {
3057      unsigned int othercase;      unsigned int othercase;
3058      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3059  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3060      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3061  #else  #else
3062      othercase = NOTACHAR;      othercase = NOTACHAR;
3063  #endif  #endif
3064      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3065      }      }
3066    else    else
3067  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3068    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3069    
3070    /* 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
3071      opcodes are not used for multi-byte characters, because they are coded using
3072      an XCLASS instead. */
3073    
3074    case OP_NOT:    case OP_NOT:
3075    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
3076    if (item == next) return TRUE;  
3077    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
3078  #ifdef SUPPORT_UTF8    if ((c = *previous) == next) return TRUE;
3079    if (utf8)  #ifdef SUPPORT_UTF
3080      if (utf)
3081      {      {
3082      unsigned int othercase;      unsigned int othercase;
3083      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3084  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3085      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3086  #else  #else
3087      othercase = NOTACHAR;      othercase = NOTACHAR;
3088  #endif  #endif
3089      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3090      }      }
3091    else    else
3092  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3093    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3094    
3095      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3096      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3097    
3098    case OP_DIGIT:    case OP_DIGIT:
3099    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2074  if (next >= 0) switch(op_code) Line 3136  if (next >= 0) switch(op_code)
3136      case 0x202f:      case 0x202f:
3137      case 0x205f:      case 0x205f:
3138      case 0x3000:      case 0x3000:
3139      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3140      default:      default:
3141      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3142      }      }
3143    
3144      case OP_ANYNL:
3145    case OP_VSPACE:    case OP_VSPACE:
3146    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3147    switch(next)    switch(next)
# Line 2090  if (next >= 0) switch(op_code) Line 3153  if (next >= 0) switch(op_code)
3153      case 0x85:      case 0x85:
3154      case 0x2028:      case 0x2028:
3155      case 0x2029:      case 0x2029:
3156      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3157      default:      default:
3158      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3159      }      }
3160    
3161    #ifdef SUPPORT_UCP
3162      case OP_PROP:
3163      return check_char_prop(next, previous[0], previous[1], FALSE);
3164    
3165      case OP_NOTPROP:
3166      return check_char_prop(next, previous[0], previous[1], TRUE);
3167    #endif
3168    
3169    default:    default:
3170    return FALSE;    return FALSE;
3171    }    }
3172    
3173    
3174  /* 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
3175    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3176    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3177    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3178    replaced by OP_PROP codes when PCRE_UCP is set. */
3179    
3180  switch(op_code)  switch(op_code)
3181    {    {
3182    case OP_CHAR:    case OP_CHAR:
3183    case OP_CHARNC:    case OP_CHARI:
3184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3185    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3186    #else
3187      c = *previous;
3188  #endif  #endif
3189    switch(-next)    switch(-next)
3190      {      {
3191      case ESC_d:      case ESC_d:
3192      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3193    
3194      case ESC_D:      case ESC_D:
3195      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3196    
3197      case ESC_s:      case ESC_s:
3198      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3199    
3200      case ESC_S:      case ESC_S:
3201      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3202    
3203      case ESC_w:      case ESC_w:
3204      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3205    
3206      case ESC_W:      case ESC_W:
3207      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3208    
3209      case ESC_h:      case ESC_h:
3210      case ESC_H:      case ESC_H:
3211      switch(item)      switch(c)
3212        {        {
3213        case 0x09:        case 0x09:
3214        case 0x20:        case 0x20:
# Line 2159  switch(op_code) Line 3236  switch(op_code)
3236    
3237      case ESC_v:      case ESC_v:
3238      case ESC_V:      case ESC_V:
3239      switch(item)      switch(c)
3240        {        {
3241        case 0x0a:        case 0x0a:
3242        case 0x0b:        case 0x0b:
# Line 2173  switch(op_code) Line 3250  switch(op_code)
3250        return -next == ESC_v;        return -next == ESC_v;
3251        }        }
3252    
3253        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3254        their substitutions and process them. The result will always be either
3255        -ESC_p or -ESC_P. Then fall through to process those values. */
3256    
3257    #ifdef SUPPORT_UCP
3258        case ESC_du:
3259        case ESC_DU:
3260        case ESC_wu:
3261        case ESC_WU:
3262        case ESC_su:
3263        case ESC_SU:
3264          {
3265          int temperrorcode = 0;
3266          ptr = substitutes[-next - ESC_DU];
3267          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3268          if (temperrorcode != 0) return FALSE;
3269          ptr++;    /* For compatibility */
3270          }
3271        /* Fall through */
3272    
3273        case ESC_p:
3274        case ESC_P:
3275          {
3276          int ptype, pdata, errorcodeptr;
3277          BOOL negated;
3278    
3279          ptr--;      /* Make ptr point at the p or P */
3280          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3281          if (ptype < 0) return FALSE;
3282          ptr++;      /* Point past the final curly ket */
3283    
3284          /* If the property item is optional, we have to give up. (When generated
3285          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3286          to the original \d etc. At this point, ptr will point to a zero byte. */
3287    
3288          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3289            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3290              return FALSE;
3291    
3292          /* Do the property check. */
3293    
3294          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3295          }
3296    #endif
3297    
3298      default:      default:
3299      return FALSE;      return FALSE;
3300      }      }
3301    
3302      /* In principle, support for Unicode properties should be integrated here as
3303      well. It means re-organizing the above code so as to get hold of the property
3304      values before switching on the op-code. However, I wonder how many patterns
3305      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3306      these op-codes are never generated.) */
3307    
3308    case OP_DIGIT:    case OP_DIGIT:
3309    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3310           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3311    
3312    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3313    return next == -ESC_d;    return next == -ESC_d;
3314    
3315    case OP_WHITESPACE:    case OP_WHITESPACE:
3316    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3317    
3318    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3319    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3320    
3321    case OP_HSPACE:    case OP_HSPACE:
3322    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3323             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3324    
3325    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3326    return next == -ESC_h;    return next == -ESC_h;
3327    
3328    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3329      case OP_ANYNL:
3330    case OP_VSPACE:    case OP_VSPACE:
3331    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3332    
3333    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3334    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3335    
3336    case OP_WORDCHAR:    case OP_WORDCHAR:
3337    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3338             next == -ESC_v || next == -ESC_R;
3339    
3340    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3341    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2233  Arguments: Line 3364  Arguments:
3364    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3365    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3366    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3367    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3368    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3369    bcptr          points to current branch chain    bcptr          points to current branch chain
3370      cond_depth     conditional nesting depth
3371    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3372    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3373                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2245  Returns:         TRUE on success Line 3377  Returns:         TRUE on success
3377  */  */
3378    
3379  static BOOL  static BOOL
3380  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3381    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3382      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3383    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3384  {  {
3385  int repeat_type, op_type;  int repeat_type, op_type;
3386  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3387  int bravalue = 0;  int bravalue = 0;
3388  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3389  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3390  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3391  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3392  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3393  int after_manual_callout = 0;  int after_manual_callout = 0;
3394  int length_prevgroup = 0;  int length_prevgroup = 0;
3395  register int c;  register int c;
3396  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3397  uschar *last_code = code;  pcre_uchar *last_code = code;
3398  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3399  uschar *tempcode;  pcre_uchar *tempcode;
3400  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3401  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3402  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3403  const uschar *tempptr;  const pcre_uchar *tempptr;
3404  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3405  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3406  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3407  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3408    pcre_uint8 classbits[32];
3409  #ifdef SUPPORT_UTF8  
3410  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3411  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3412  uschar *class_utf8data;  dynamically as we process the pattern. */
3413  uschar utf8_char[6];  
3414    #ifdef SUPPORT_UTF
3415    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3416    BOOL utf = (options & PCRE_UTF8) != 0;
3417    pcre_uchar utf_chars[6];
3418  #else  #else
3419  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3420  uschar *utf8_char = NULL;  #endif
3421    
3422    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3423    
3424    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3425    BOOL xclass;
3426    pcre_uchar *class_uchardata;
3427    pcre_uchar *class_uchardata_base;
3428  #endif  #endif
3429    
3430  #ifdef DEBUG  #ifdef PCRE_DEBUG
3431  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3432  #endif  #endif
3433    
# Line 2294  greedy_non_default = greedy_default ^ 1; Line 3438  greedy_non_default = greedy_default ^ 1;
3438    
3439  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3440  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
3441  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
3442  find one.  find one.
3443    
3444  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
3445  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
3446  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3447  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3448    
3449  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3450    
3451  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3452  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
3453  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
3454  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3455    value. This is used only for ASCII characters. */
3456    
3457  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3458    
3459  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3460    
3461  for (;; ptr++)  for (;; ptr++)
3462    {    {
3463    BOOL negate_class;    BOOL negate_class;
3464      BOOL should_flip_negation;
3465    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3466    BOOL is_quantifier;    BOOL is_quantifier;
3467    BOOL is_recurse;    BOOL is_recurse;
3468    BOOL reset_bracount;    BOOL reset_bracount;
3469    int class_charcount;    int class_has_8bitchar;
3470    int class_lastchar;    int class_single_char;
3471    int newoptions;    int newoptions;
3472    int recno;    int recno;
3473    int refsign;    int refsign;
3474    int skipbytes;    int skipbytes;
3475    int subreqbyte;    int subreqchar;
3476    int subfirstbyte;    int subfirstchar;
3477    int terminator;    int terminator;
3478    int mclength;    int mclength;
3479    uschar mcbuffer[8];    int tempbracount;
3480      pcre_uchar mcbuffer[8];
3481    
3482    /* Get next byte in the pattern */    /* Get next character in the pattern */
3483    
3484    c = *ptr;    c = *ptr;
3485    
3486      /* If we are at the end of a nested substitution, revert to the outer level
3487      string. Nesting only happens one level deep. */
3488    
3489      if (c == 0 && nestptr != NULL)
3490        {
3491        ptr = nestptr;
3492        nestptr = NULL;
3493        c = *ptr;
3494        }
3495    
3496    /* 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
3497    previous cycle of this loop. */    previous cycle of this loop. */
3498    
3499    if (lengthptr != NULL)    if (lengthptr != NULL)
3500      {      {
3501  #ifdef DEBUG  #ifdef PCRE_DEBUG
3502      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3503  #endif  #endif
3504      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3505            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3506        {        {
3507        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3508        goto FAILED;        goto FAILED;
# Line 2366  for (;; ptr++) Line 3524  for (;; ptr++)
3524        goto FAILED;        goto FAILED;
3525        }        }
3526    
3527      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3528      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3529          (int)(code - last_code), c, c));
3530    
3531      /* 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
3532      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,
3533      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 2377  for (;; ptr++) Line 3536  for (;; ptr++)
3536        {        {
3537        if (previous > orig_code)        if (previous > orig_code)
3538          {          {
3539          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3540          code -= previous - orig_code;          code -= previous - orig_code;
3541          previous = orig_code;          previous = orig_code;
3542          }          }
# Line 2393  for (;; ptr++) Line 3552  for (;; ptr++)
3552    /* 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
3553    reference list. */    reference list. */
3554    
3555    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3556               WORK_SIZE_SAFETY_MARGIN)
3557      {      {
3558      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3559      goto FAILED;      goto FAILED;
# Line 2403  for (;; ptr++) Line 3563  for (;; ptr++)
3563    
3564    if (inescq && c != 0)    if (inescq && c != 0)
3565      {      {
3566      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3567        {        {
3568        inescq = FALSE;        inescq = FALSE;
3569        ptr++;        ptr++;
# Line 2429  for (;; ptr++) Line 3589  for (;; ptr++)
3589    /* 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
3590    a quantifier. */    a quantifier. */
3591    
3592    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3593      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3594        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3595    
3596    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3597         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2440  for (;; ptr++) Line 3601  for (;; ptr++)
3601      previous_callout = NULL;      previous_callout = NULL;
3602      }      }
3603    
3604    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3605    
3606    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3607      {      {
3608      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3609      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3610        {        {
3611        while (*(++ptr) != 0)        ptr++;
3612          while (*ptr != 0)
3613          {          {
3614          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3615            ptr++;
3616    #ifdef SUPPORT_UTF
3617            if (utf) FORWARDCHAR(ptr);
3618    #endif
3619          }          }
3620        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3621    
# Line 2470  for (;; ptr++) Line 3636  for (;; ptr++)
3636      {      {
3637      /* ===================================================================*/      /* ===================================================================*/
3638      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3639      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3640      case ')':      case CHAR_RIGHT_PARENTHESIS:
3641      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3642      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3643      *codeptr = code;      *codeptr = code;
3644      *ptrptr = ptr;      *ptrptr = ptr;
3645      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2483  for (;; ptr++) Line 3649  for (;; ptr++)
3649          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3650          goto FAILED;          goto FAILED;
3651          }          }
3652        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3653        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3654        }        }
3655      return TRUE;      return TRUE;
# Line 2493  for (;; ptr++) Line 3659  for (;; ptr++)
3659      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3660      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3661    
3662      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3663        previous = NULL;
3664      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3665        {        {
3666        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3667          *code++ = OP_CIRCM;
3668        }        }
3669      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3670      break;      break;
3671    
3672      case '$':      case CHAR_DOLLAR_SIGN:
3673      previous = NULL;      previous = NULL;
3674      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3675      break;      break;
3676    
3677      /* 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
3678      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3679    
3680      case '.':      case CHAR_DOT:
3681      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3682      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3683      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3684      previous = code;      previous = code;
3685      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3686      break;      break;
3687    
3688    
# Line 2530  for (;; ptr++) Line 3697  for (;; ptr++)
3697      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,
3698      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3699      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.
     */  
3700    
3701      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3702        default (Perl) mode, it is treated as a data character. */
3703    
3704        case CHAR_RIGHT_SQUARE_BRACKET:
3705        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3706          {
3707          *errorcodeptr = ERR64;
3708          goto FAILED;
3709          }
3710        goto NORMAL_CHAR;
3711    
3712        case CHAR_LEFT_SQUARE_BRACKET:
3713      previous = code;      previous = code;
3714    
3715      /* 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
3716      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. */
3717    
3718      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3719          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3720            check_posix_syntax(ptr, &tempptr))
3721        {        {
3722        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3723        goto FAILED;        goto FAILED;
3724        }        }
3725    
# Line 2553  for (;; ptr++) Line 3731  for (;; ptr++)
3731      for (;;)      for (;;)
3732        {        {
3733        c = *(++ptr);        c = *(++ptr);
3734        if (c == '\\')        if (c == CHAR_BACKSLASH)
3735          {          {
3736          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3737            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3738              else break;          else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
3739              ptr += 3;
3740            else
3741              break;
3742          }          }
3743        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3744          negate_class = TRUE;          negate_class = TRUE;
3745        else break;        else break;
3746        }        }
3747    
3748      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3749      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
3750      valued UTF-8 characters, we don't yet do any optimization. */      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3751        [^] must match any character, so generate OP_ALLANY. */
3752    
3753        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3754            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3755          {
3756          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3757          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3758          zerofirstchar = firstchar;
3759          break;
3760          }
3761    
3762        /* If a class contains a negative special such as \S, we need to flip the
3763        negation flag at the end, so that support for characters > 255 works
3764        correctly (they are all included in the class). */
3765    
3766        should_flip_negation = FALSE;
3767    
3768        /* For optimization purposes, we track some properties of the class.
3769        class_has_8bitchar will be non-zero, if the class contains at least one
3770        < 256 character. class_single_char will be 1 if the class contains only
3771        a single character. */
3772    
3773      class_charcount = 0;      class_has_8bitchar = 0;
3774      class_lastchar = -1;      class_single_char = 0;
3775    
3776      /* 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
3777      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains only 1 character (less
3778      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.
3779      */      */
3780    
3781      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3782    
3783  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3784      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3785      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3786        class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3787  #endif  #endif
3788    
3789      /* 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 2589  for (;; ptr++) Line 3792  for (;; ptr++)
3792    
3793      if (c != 0) do      if (c != 0) do
3794        {        {
3795        const uschar *oldptr;        const pcre_uchar *oldptr;
3796    
3797  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3798        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3799          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3800          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3801          }          }
3802  #endif  #endif
3803    
3804    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3805          /* In the pre-compile phase, accumulate the length of any extra
3806          data and reset the pointer. This is so that very large classes that
3807          contain a zillion > 255 characters no longer overwrite the work space
3808          (which is on the stack). */
3809    
3810          if (lengthptr != NULL)
3811            {
3812            *lengthptr += class_uchardata - class_uchardata_base;
3813            class_uchardata = class_uchardata_base;
3814            }
3815    #endif
3816    
3817        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3818    
3819        if (inescq)        if (inescq)
3820          {          {
3821          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3822            {            {
3823            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3824            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2617  for (;; ptr++) Line 3833  for (;; ptr++)
3833        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3834        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3835    
3836        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3837            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3838            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3839          {          {
3840          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3841          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3842          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3843          uschar pbits[32];          pcre_uint8 pbits[32];
3844    
3845          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3846            {            {
3847            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3848            goto FAILED;            goto FAILED;
3849            }            }
3850    
3851          ptr += 2;          ptr += 2;
3852          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3853            {            {
3854            local_negate = TRUE;            local_negate = TRUE;
3855              should_flip_negation = TRUE;  /* Note negative special */
3856            ptr++;            ptr++;
3857            }            }
3858