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

Diff of /code/trunk/pcre_compile.c

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

revision 182 by ph10, Wed Jun 13 15:09:54 2007 UTC revision 904 by ph10, Mon Jan 23 17:30:49 2012 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) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* 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 62  used by pcretest. DEBUG is not defined w Line 70  used by pcretest. DEBUG is not defined w
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
72    
73    /* Maximum length value to check against when making sure that the integer that
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    
81  /*************************************************  /*************************************************
82  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 77  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    /* Private flags added to firstchar and reqchar. */
112    
113  #define COMPILE_WORK_SIZE (4096)  #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 120  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,      0,       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 130  static const short int escapes[] = { Line 200  static const short int escapes[] = {
200  #endif  #endif
201    
202    
203  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
204  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
205  as this is assumed for handling case independence. */  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  static const char *const posix_names[] = {  platforms. */
208    "alpha", "lower", "upper",  
209    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
210    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
211      int   op;                  /* Op when no arg, or -1 if arg mandatory */
212      int   op_arg;              /* Op when arg present, or -1 if not allowed */
213    } verbitem;
214    
215    static const char verbnames[] =
216      "\0"                       /* Empty name is a shorthand for MARK */
217      STRING_MARK0
218      STRING_ACCEPT0
219      STRING_COMMIT0
220      STRING_F0
221      STRING_FAIL0
222      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 const int verbcount = sizeof(verbs)/sizeof(verbitem);
239    
240    
241  static const uschar posix_name_lengths[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
242    now all in a single string, to reduce the number of relocations when a shared
243    library is dynamically loaded. The list of lengths is terminated by a zero
244    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 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 169  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 176  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 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 compiled without UTF 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",    "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 in 8-bit non-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  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
474      /* 60 */
475      "(*VERB) not recognized\0"
476      "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 Unicode property 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 Unicode code point (>= 0xd800 && <= 0xdfff)\0"
491      "invalid UTF-16 string\0"
492      ;
493    
494  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
495  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 268  For convenience, we use the same bit def Line 507  For convenience, we use the same bit def
507    
508  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
509    
510  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
511  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
512    into a subtraction and unsigned comparison). */
513    
514    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
515    
516    #ifndef EBCDIC
517    
518    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
519    UTF-8 mode. */
520    
521    static const pcre_uint8 digitab[] =
522    {    {
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 304  static const unsigned char digitab[] = Line 553  static const unsigned char digitab[] =
553    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
554    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
555    
556  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
557  static const unsigned char digitab[] =  
558    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
559    
560    static const pcre_uint8 digitab[] =
561    {    {
562    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
563    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 340  static const unsigned char digitab[] = Line 592  static const unsigned char digitab[] =
592    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
593    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
594    
595  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
596    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
597    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
598    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 379  static const unsigned char ebcdic_charta Line 631  static const unsigned char ebcdic_charta
631  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
632    
633  static BOOL  static BOOL
634    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
635      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
636    
637    
638    
639  /*************************************************  /*************************************************
640    *            Find an error text                  *
641    *************************************************/
642    
643    /* The error texts are now all in one long string, to save on relocations. As
644    some of the text is of unknown length, we can't use a table of offsets.
645    Instead, just count through the strings. This is not a performance issue
646    because it happens only when there has been a compilation error.
647    
648    Argument:   the error number
649    Returns:    pointer to the error string
650    */
651    
652    static const char *
653    find_error_text(int n)
654    {
655    const char *s = error_texts;
656    for (; n > 0; n--)
657      {
658      while (*s++ != 0) {};
659      if (*s == 0) return "Error text not found (please report)";
660      }
661    return s;
662    }
663    
664    
665    /*************************************************
666    *           Expand the workspace                 *
667    *************************************************/
668    
669    /* This function is called during the second compiling phase, if the number of
670    forward references fills the existing workspace, which is originally a block on
671    the stack. A larger block is obtained from malloc() unless the ultimate limit
672    has been reached or the increase will be rather small.
673    
674    Argument: pointer to the compile data block
675    Returns:  0 if all went well, else an error number
676    */
677    
678    static int
679    expand_workspace(compile_data *cd)
680    {
681    pcre_uchar *newspace;
682    int newsize = cd->workspace_size * 2;
683    
684    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
685    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
686        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
687     return ERR72;
688    
689    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
690    if (newspace == NULL) return ERR21;
691    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
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 405  Arguments: Line 755  Arguments:
755    
756  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
757                   negative => a special escape sequence                   negative => a special escape sequence
758                   on error, errorptr is set                   on error, errorcodeptr is set
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 423  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 440  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 448  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 479  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            {
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;
908          break;
909          }
910    
911      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
912        {        {
913        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
914        return 0;        break;
915          }
916    
917        if (c == 0)
918          {
919          *errorcodeptr = ERR58;
920          break;
921        }        }
922    
923      if (negated)      if (negated)
# Line 501  else Line 925  else
925        if (c > bracount)        if (c > bracount)
926          {          {
927          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
928          return 0;          break;
929          }          }
930        c = bracount - (c - 1);        c = bracount - (c - 1);
931        }        }
# Line 521  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            {
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;
970            break;
971            }
972        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
973          {          {
974          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 542  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 552  else Line 991  else
991      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
992      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
993      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
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 and 16-bit mode,
995      than 3 octal digits. */      but no more 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;  #ifdef COMPILE_PCRE8
1002        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1003    #endif
1004      break;      break;
1005    
1006      /* \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
1007      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.
1008      treated as a data character. */      If not, { is treated as a data character. */
1009    
1010        case CHAR_x:
1011        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1012          {
1013          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1014          Otherwise it is a lowercase x letter. */
1015          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1016            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1017            {
1018            c = 0;
1019            for (i = 0; i < 2; ++i)
1020              {
1021              register int cc = *(++ptr);
1022    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1023              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1024              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1025    #else           /* EBCDIC coding */
1026              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1027              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1028    #endif
1029              }
1030            }
1031          break;
1032          }
1033    
1034      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1035        {        {
1036        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1037    
1038        c = 0;        c = 0;
1039        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1040          {          {
1041          register int cc = *pt++;          register int cc = *pt++;
1042          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1043    
1044  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1046          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1047  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1048          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1049          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050    #endif
1051    
1052    #ifdef COMPILE_PCRE8
1053            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1054    #else
1055    #ifdef COMPILE_PCRE16
1056            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1057    #endif
1058  #endif  #endif
1059          }          }
1060    
1061        if (*pt == '}')        if (c < 0)
1062          {          {
1063          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1064            *errorcodeptr = ERR34;
1065            }
1066    
1067          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1068            {
1069            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1070          ptr = pt;          ptr = pt;
1071          break;          break;
1072          }          }
# Line 602  else Line 1078  else
1078      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1079    
1080      c = 0;      c = 0;
1081      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1082        {        {
1083        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1084        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1085  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1086        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1087        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1088  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1089        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1090        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1091  #endif  #endif
1092        }        }
1093      break;      break;
1094    
1095      /* 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.
1096      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
1097        coding is ASCII-specific, but then the whole concept of \cx is
1098      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1099    
1100      case 'c':      case CHAR_c:
1101      c = *(++ptr);      c = *(++ptr);
1102      if (c == 0)      if (c == 0)
1103        {        {
1104        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1105        return 0;        break;
1106        }        }
1107    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1108  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1109      if (c >= 'a' && c <= 'z') c -= 32;        {
1110          *errorcodeptr = ERR68;
1111          break;
1112          }
1113        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1114      c ^= 0x40;      c ^= 0x40;
1115  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1116      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1117      c ^= 0xC0;      c ^= 0xC0;
1118  #endif  #endif
1119      break;      break;
1120    
1121      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1122      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1123      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
1124      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
1125      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1126    
1127      default:      default:
1128      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 654  else Line 1135  else
1135      }      }
1136    }    }
1137    
1138    /* Perl supports \N{name} for character names, as well as plain \N for "not
1139    newline". PCRE does not support \N{name}. However, it does support
1140    quantification such as \N{2,3}. */
1141    
1142    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1143         !is_counted_repeat(ptr+2))
1144      *errorcodeptr = ERR37;
1145    
1146    /* If PCRE_UCP is set, we change the values for \d etc. */
1147    
1148    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1149      c -= (ESC_DU - ESC_D);
1150    
1151    /* Set the pointer to the final character before returning. */
1152    
1153  *ptrptr = ptr;  *ptrptr = ptr;
1154  return c;  return c;
1155  }  }
# Line 680  Returns:         type value from ucp_typ Line 1176  Returns:         type value from ucp_typ
1176  */  */
1177    
1178  static int  static int
1179  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1180  {  {
1181  int c, i, bot, top;  int c, i, bot, top;
1182  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1183  char name[32];  pcre_uchar name[32];
1184    
1185  c = *(++ptr);  c = *(++ptr);
1186  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 694  if (c == 0) goto ERROR_RETURN; Line 1190  if (c == 0) goto ERROR_RETURN;
1190  /* \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
1191  negation. */  negation. */
1192    
1193  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1194    {    {
1195    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1196      {      {
1197      *negptr = TRUE;      *negptr = TRUE;
1198      ptr++;      ptr++;
1199      }      }
1200    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1201      {      {
1202      c = *(++ptr);      c = *(++ptr);
1203      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1204      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1205      name[i] = c;      name[i] = c;
1206      }      }
1207    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1208    name[i] = 0;    name[i] = 0;
1209    }    }
1210    
# Line 725  else Line 1221  else
1221  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1222    
1223  bot = 0;  bot = 0;
1224  top = _pcre_utt_size;  top = PRIV(utt_size);
1225    
1226  while (bot < top)  while (bot < top)
1227    {    {
1228    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1229    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1230    if (c == 0)    if (c == 0)
1231      {      {
1232      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1233      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1234      }      }
1235    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1236    }    }
# Line 754  return -1; Line 1250  return -1;
1250    
1251    
1252  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1253  *         Read repeat counts                     *  *         Read repeat counts                     *
1254  *************************************************/  *************************************************/
1255    
# Line 805  Returns:         pointer to '}' on succe Line 1268  Returns:         pointer to '}' on succe
1268                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1269  */  */
1270    
1271  static const uschar *  static const pcre_uchar *
1272  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)
1273  {  {
1274  int min = 0;  int min = 0;
1275  int max = -1;  int max = -1;
# Line 814  int max = -1; Line 1277  int max = -1;
1277  /* 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
1278  an integer overflow. */  an integer overflow. */
1279    
1280  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1281  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1282    {    {
1283    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 824  if (min < 0 || min > 65535) Line 1287  if (min < 0 || min > 65535)
1287  /* 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.
1288  Also, max must not be less than min. */  Also, max must not be less than min. */
1289    
1290  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1291    {    {
1292    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1293      {      {
1294      max = 0;      max = 0;
1295      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1296      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1297        {        {
1298        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 854  return p; Line 1317  return p;
1317    
1318    
1319  /*************************************************  /*************************************************
1320  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1321  *************************************************/  *************************************************/
1322    
1323  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1324    top-level call starts at the beginning of the pattern. All other calls must
1325    start at a parenthesis. It scans along a pattern's text looking for capturing
1326  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
1327  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
1328  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
1329  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1330  be terminated by '>' because that is checked in the first pass.  
1331    This function was originally called only from the second pass, in which we know
1332    that if (?< or (?' or (?P< is encountered, the name will be correctly
1333    terminated because that is checked in the first pass. There is now one call to
1334    this function in the first pass, to check for a recursive back reference by
1335    name (so that we can make the whole group atomic). In this case, we need check
1336    only up to the current position in the pattern, and that is still OK because
1337    and previous occurrences will have been checked. To make this work, the test
1338    for "end of pattern" is a check against cd->end_pattern in the main loop,
1339    instead of looking for a binary zero. This means that the special first-pass
1340    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1341    processing items within the loop are OK, because afterwards the main loop will
1342    terminate.)
1343    
1344  Arguments:  Arguments:
1345    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1346    count        current count of capturing parens so far encountered    cd           compile background data
1347    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1348    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1349    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1350      utf          TRUE if we are in UTF-8 / UTF-16 mode
1351      count        pointer to the current capturing subpattern number (updated)
1352    
1353  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1354  */  */
1355    
1356  static int  static int
1357  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,
1358    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1359  {  {
1360  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1361    int start_count = *count;
1362    int hwm_count = start_count;
1363    BOOL dup_parens = FALSE;
1364    
1365    /* If the first character is a parenthesis, check on the type of group we are
1366    dealing with. The very first call may not start with a parenthesis. */
1367    
1368  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1369    {    {
1370    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1371    
1372      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1373    
1374      /* Handle a normal, unnamed capturing parenthesis. */
1375    
1376      else if (ptr[1] != CHAR_QUESTION_MARK)
1377        {
1378        *count += 1;
1379        if (name == NULL && *count == lorn) return *count;
1380        ptr++;
1381        }
1382    
1383      /* All cases now have (? at the start. Remember when we are in a group
1384      where the parenthesis numbers are duplicated. */
1385    
1386      else if (ptr[2] == CHAR_VERTICAL_LINE)
1387        {
1388        ptr += 3;
1389        dup_parens = TRUE;
1390        }
1391    
1392      /* Handle comments; all characters are allowed until a ket is reached. */
1393    
1394      else if (ptr[2] == CHAR_NUMBER_SIGN)
1395        {
1396        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1397        goto FAIL_EXIT;
1398        }
1399    
1400      /* Handle a condition. If it is an assertion, just carry on so that it
1401      is processed as normal. If not, skip to the closing parenthesis of the
1402      condition (there can't be any nested parens). */
1403    
1404      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1405        {
1406        ptr += 2;
1407        if (ptr[1] != CHAR_QUESTION_MARK)
1408          {
1409          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1410          if (*ptr != 0) ptr++;
1411          }
1412        }
1413    
1414      /* Start with (? but not a condition. */
1415    
1416      else
1417        {
1418        ptr += 2;
1419        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1420    
1421        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1422    
1423        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1424            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1425          {
1426          int term;
1427          const pcre_uchar *thisname;
1428          *count += 1;
1429          if (name == NULL && *count == lorn) return *count;
1430          term = *ptr++;
1431          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1432          thisname = ptr;
1433          while (*ptr != term) ptr++;
1434          if (name != NULL && lorn == ptr - thisname &&
1435              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1436            return *count;
1437          term++;
1438          }
1439        }
1440      }
1441    
1442    /* Past any initial parenthesis handling, scan for parentheses or vertical
1443    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1444    first-pass call when this value is temporarily adjusted to stop at the current
1445    position. So DO NOT change this to a test for binary zero. */
1446    
1447    for (; ptr < cd->end_pattern; ptr++)
1448      {
1449    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1450    
1451    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1452      {      {
1453      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1454      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1455        {        {
1456        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1457        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1458        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1459        }        }
1460      continue;      continue;
1461      }      }
1462    
1463    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1464      are handled for real. If the first character is '^', skip it. Also, if the
1465      first few characters (either before or after ^) are \Q\E or \E we skip them
1466      too. This makes for compatibility with Perl. Note the use of STR macros to
1467      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1468    
1469      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1470        {
1471        BOOL negate_class = FALSE;
1472        for (;;)
1473          {
1474          if (ptr[1] == CHAR_BACKSLASH)
1475            {
1476            if (ptr[2] == CHAR_E)
1477              ptr+= 2;
1478            else if (STRNCMP_UC_C8(ptr + 2,
1479                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1480              ptr += 4;
1481            else
1482              break;
1483            }
1484          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1485            {
1486            negate_class = TRUE;
1487            ptr++;
1488            }
1489          else break;
1490          }
1491    
1492        /* If the next character is ']', it is a data character that must be
1493        skipped, except in JavaScript compatibility mode. */
1494    
1495    if (*ptr == '[')      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1496      {          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1497      while (*(++ptr) != ']')        ptr++;
1498    
1499        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1500        {        {
1501        if (*ptr == '\\')        if (*ptr == 0) return -1;
1502          if (*ptr == CHAR_BACKSLASH)
1503          {          {
1504          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1505          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1506            {            {
1507            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1508            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1509            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1510            }            }
1511          continue;          continue;
1512          }          }
# Line 921  for (; *ptr != 0; ptr++) Line 1516  for (; *ptr != 0; ptr++)
1516    
1517    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1518    
1519    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1520      {      {
1521      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1522      if (*ptr == 0) return -1;      while (*ptr != 0)
1523          {
1524          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1525          ptr++;
1526    #ifdef SUPPORT_UTF
1527          if (utf) FORWARDCHAR(ptr);
1528    #endif
1529          }
1530        if (*ptr == 0) goto FAIL_EXIT;
1531      continue;      continue;
1532      }      }
1533    
1534    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1535    
1536    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1537      {      {
1538      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1539      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1540      continue;      if (*ptr == 0) goto FAIL_EXIT;
1541      }      }
1542    
1543    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1544    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1545        if (dup_parens && *count < hwm_count) *count = hwm_count;
1546        goto FAIL_EXIT;
1547        }
1548    
1549    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1550        {
1551        if (*count > hwm_count) hwm_count = *count;
1552        *count = start_count;
1553        }
1554      }
1555    
1556    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1557         *ptr != '\'')  *ptrptr = ptr;
1558      continue;  return -1;
1559    }
1560    
1561    
1562    
1563    
1564    /*************************************************
1565    *       Find forward referenced subpattern       *
1566    *************************************************/
1567    
1568    count++;  /* This function scans along a pattern's text looking for capturing
1569    subpatterns, and counting them. If it finds a named pattern that matches the
1570    name it is given, it returns its number. Alternatively, if the name is NULL, it
1571    returns when it reaches a given numbered subpattern. This is used for forward
1572    references to subpatterns. We used to be able to start this scan from the
1573    current compiling point, using the current count value from cd->bracount, and
1574    do it all in a single loop, but the addition of the possibility of duplicate
1575    subpattern numbers means that we have to scan from the very start, in order to
1576    take account of such duplicates, and to use a recursive function to keep track
1577    of the different types of group.
1578    
1579    if (name == NULL && count == lorn) return count;  Arguments:
1580    term = *ptr++;    cd           compile background data
1581    if (term == '<') term = '>';    name         name to seek, or NULL if seeking a numbered subpattern
1582    thisname = ptr;    lorn         name length, or subpattern number if name is NULL
1583    while (*ptr != term) ptr++;    xmode        TRUE if we are in /x mode
1584    if (name != NULL && lorn == ptr - thisname &&    utf          TRUE if we are in UTF-8 / UTF-16 mode
1585        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
1586      return count;  Returns:       the number of the found subpattern, or -1 if not found
1587    */
1588    
1589    static int
1590    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1591      BOOL utf)
1592    {
1593    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1594    int count = 0;
1595    int rc;
1596    
1597    /* If the pattern does not start with an opening parenthesis, the first call
1598    to find_parens_sub() will scan right to the end (if necessary). However, if it
1599    does start with a parenthesis, find_parens_sub() will return when it hits the
1600    matching closing parens. That is why we have to have a loop. */
1601    
1602    for (;;)
1603      {
1604      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1605      if (rc > 0 || *ptr++ == 0) break;
1606    }    }
1607    
1608  return -1;  return rc;
1609  }  }
1610    
1611    
1612    
1613    
1614  /*************************************************  /*************************************************
1615  *      Find first significant op code            *  *      Find first significant op code            *
1616  *************************************************/  *************************************************/
1617    
1618  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1619  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
1620  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
1621  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
1622  assertions, and also the \b assertion; for others it does not.  does not.
1623    
1624  Arguments:  Arguments:
1625    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  
1626    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1627    
1628  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1629  */  */
1630    
1631  static const uschar*  static const pcre_uchar*
1632  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1633  {  {
1634  for (;;)  for (;;)
1635    {    {
1636    switch ((int)*code)    switch ((int)*code)
1637      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1638      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1639      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1640      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1641      if (!skipassert) return code;      if (!skipassert) return code;
1642      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1643      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1644      break;      break;
1645    
1646      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1013  for (;;) Line 1650  for (;;)
1650    
1651      case OP_CALLOUT:      case OP_CALLOUT:
1652      case OP_CREF:      case OP_CREF:
1653        case OP_NCREF:
1654      case OP_RREF:      case OP_RREF:
1655        case OP_NRREF:
1656      case OP_DEF:      case OP_DEF:
1657      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1658      break;      break;
1659    
1660      default:      default:
# Line 1029  for (;;) Line 1668  for (;;)
1668    
1669    
1670  /*************************************************  /*************************************************
1671  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1672  *************************************************/  *************************************************/
1673    
1674  /* 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,
1675  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.
1676  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
1677    temporarily terminated with OP_END when this function is called.
1678    
1679    This function is called when a backward assertion is encountered, so that if it
1680    fails, the error message can point to the correct place in the pattern.
1681    However, we cannot do this when the assertion contains subroutine calls,
1682    because they can be forward references. We solve this by remembering this case
1683    and doing the check at the end; a flag specifies which mode we are running in.
1684    
1685  Arguments:  Arguments:
1686    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1687    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1688      atend    TRUE if called when the pattern is complete
1689  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1690               or -2 if \C was encountered  
1691    Returns:   the fixed length,
1692                 or -1 if there is no fixed length,
1693                 or -2 if \C was encountered (in UTF-8 mode only)
1694                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1695                 or -4 if an unknown opcode was encountered (internal error)
1696  */  */
1697    
1698  static int  static int
1699  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1700  {  {
1701  int length = -1;  int length = -1;
1702    
1703  register int branchlength = 0;  register int branchlength = 0;
1704  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1705    
1706  /* 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
1707  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1058  branch, check the length against that of Line 1709  branch, check the length against that of
1709  for (;;)  for (;;)
1710    {    {
1711    int d;    int d;
1712      pcre_uchar *ce, *cs;
1713    register int op = *cc;    register int op = *cc;
1714    
1715    switch (op)    switch (op)
1716      {      {
1717        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1718        OP_BRA (normal non-capturing bracket) because the other variants of these
1719        opcodes are all concerned with unlimited repeated groups, which of course
1720        are not of fixed length. */
1721    
1722      case OP_CBRA:      case OP_CBRA:
1723      case OP_BRA:      case OP_BRA:
1724      case OP_ONCE:      case OP_ONCE:
1725        case OP_ONCE_NC:
1726      case OP_COND:      case OP_COND:
1727      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1728      if (d < 0) return d;      if (d < 0) return d;
1729      branchlength += d;      branchlength += d;
1730      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1731      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1732      break;      break;
1733    
1734      /* 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.
1735      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
1736      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
1737        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1738        because they all imply an unlimited repeat. */
1739    
1740      case OP_ALT:      case OP_ALT:
1741      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1742      case OP_END:      case OP_END:
1743        case OP_ACCEPT:
1744        case OP_ASSERT_ACCEPT:
1745      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1746        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1747      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1089  for (;;) Line 1749  for (;;)
1749      branchlength = 0;      branchlength = 0;
1750      break;      break;
1751    
1752        /* A true recursion implies not fixed length, but a subroutine call may
1753        be OK. If the subroutine is a forward reference, we can't deal with
1754        it until the end of the pattern, so return -3. */
1755    
1756        case OP_RECURSE:
1757        if (!atend) return -3;
1758        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1759        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1760        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1761        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1762        if (d < 0) return d;
1763        branchlength += d;
1764        cc += 1 + LINK_SIZE;
1765        break;
1766    
1767      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1768    
1769      case OP_ASSERT:      case OP_ASSERT:
# Line 1096  for (;;) Line 1771  for (;;)
1771      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1772      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1773      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1774      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1775        break;
1776    
1777      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1778    
1779      case OP_REVERSE:      case OP_MARK:
1780        case OP_PRUNE_ARG:
1781        case OP_SKIP_ARG:
1782        case OP_THEN_ARG:
1783        cc += cc[1] + PRIV(OP_lengths)[*cc];
1784        break;
1785    
1786        case OP_CALLOUT:
1787        case OP_CIRC:
1788        case OP_CIRCM:
1789        case OP_CLOSE:
1790        case OP_COMMIT:
1791      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1792      case OP_DEF:      case OP_DEF:
1793      case OP_OPT:      case OP_DOLL:
1794      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_CIRC:      case OP_FAIL:
1798      case OP_DOLL:      case OP_NCREF:
1799        case OP_NRREF:
1800      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1801        case OP_PRUNE:
1802        case OP_REVERSE:
1803        case OP_RREF:
1804        case OP_SET_SOM:
1805        case OP_SKIP:
1806        case OP_SOD:
1807        case OP_SOM:
1808        case OP_THEN:
1809      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1810      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1811      break;      break;
1812    
1813      /* Handle literal characters */      /* Handle literal characters */
1814    
1815      case OP_CHAR:      case OP_CHAR:
1816      case OP_CHARNC:      case OP_CHARI:
1817      case OP_NOT:      case OP_NOT:
1818        case OP_NOTI:
1819      branchlength++;      branchlength++;
1820      cc += 2;      cc += 2;
1821  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1822      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1823  #endif  #endif
1824      break;      break;
1825    
# Line 1136  for (;;) Line 1827  for (;;)
1827      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1828    
1829      case OP_EXACT:      case OP_EXACT:
1830        case OP_EXACTI:
1831        case OP_NOTEXACT:
1832        case OP_NOTEXACTI:
1833      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1834      cc += 4;      cc += 2 + IMM2_SIZE;
1835  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1836      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1837  #endif  #endif
1838      break;      break;
1839    
1840      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1841      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1842      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1843        cc += 1 + IMM2_SIZE + 1;
1844      break;      break;
1845    
1846      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1158  for (;;) Line 1850  for (;;)
1850      cc += 2;      cc += 2;
1851      /* Fall through */      /* Fall through */
1852    
1853        case OP_HSPACE:
1854        case OP_VSPACE:
1855        case OP_NOT_HSPACE:
1856        case OP_NOT_VSPACE:
1857      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1858      case OP_DIGIT:      case OP_DIGIT:
1859      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1165  for (;;) Line 1861  for (;;)
1861      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1862      case OP_WORDCHAR:      case OP_WORDCHAR:
1863      case OP_ANY:      case OP_ANY:
1864        case OP_ALLANY:
1865      branchlength++;      branchlength++;
1866      cc++;      cc++;
1867      break;      break;
1868    
1869      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1870        otherwise \C is coded as OP_ALLANY. */
1871    
1872      case OP_ANYBYTE:      case OP_ANYBYTE:
1873      return -2;      return -2;
1874    
1875      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1876    
1877  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1878      case OP_XCLASS:      case OP_XCLASS:
1879      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1880      /* Fall through */      /* Fall through */
1881  #endif  #endif
1882    
1883      case OP_CLASS:      case OP_CLASS:
1884      case OP_NCLASS:      case OP_NCLASS:
1885      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1886    
1887      switch (*cc)      switch (*cc)
1888        {        {
1889          case OP_CRPLUS:
1890          case OP_CRMINPLUS:
1891        case OP_CRSTAR:        case OP_CRSTAR:
1892        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1893        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1196  for (;;) Line 1896  for (;;)
1896    
1897        case OP_CRRANGE:        case OP_CRRANGE:
1898        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1899        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1900        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1901        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1902        break;        break;
1903    
1904        default:        default:
# Line 1208  for (;;) Line 1908  for (;;)
1908    
1909      /* Anything else is variable length */      /* Anything else is variable length */
1910    
1911      default:      case OP_ANYNL:
1912        case OP_BRAMINZERO:
1913        case OP_BRAPOS:
1914        case OP_BRAPOSZERO:
1915        case OP_BRAZERO:
1916        case OP_CBRAPOS:
1917        case OP_EXTUNI:
1918        case OP_KETRMAX:
1919        case OP_KETRMIN:
1920        case OP_KETRPOS:
1921        case OP_MINPLUS:
1922        case OP_MINPLUSI:
1923        case OP_MINQUERY:
1924        case OP_MINQUERYI:
1925        case OP_MINSTAR:
1926        case OP_MINSTARI:
1927        case OP_MINUPTO:
1928        case OP_MINUPTOI:
1929        case OP_NOTMINPLUS:
1930        case OP_NOTMINPLUSI:
1931        case OP_NOTMINQUERY:
1932        case OP_NOTMINQUERYI:
1933        case OP_NOTMINSTAR:
1934        case OP_NOTMINSTARI:
1935        case OP_NOTMINUPTO:
1936        case OP_NOTMINUPTOI:
1937        case OP_NOTPLUS:
1938        case OP_NOTPLUSI:
1939        case OP_NOTPOSPLUS:
1940        case OP_NOTPOSPLUSI:
1941        case OP_NOTPOSQUERY:
1942        case OP_NOTPOSQUERYI:
1943        case OP_NOTPOSSTAR:
1944        case OP_NOTPOSSTARI:
1945        case OP_NOTPOSUPTO:
1946        case OP_NOTPOSUPTOI:
1947        case OP_NOTQUERY:
1948        case OP_NOTQUERYI:
1949        case OP_NOTSTAR:
1950        case OP_NOTSTARI:
1951        case OP_NOTUPTO:
1952        case OP_NOTUPTOI:
1953        case OP_PLUS:
1954        case OP_PLUSI:
1955        case OP_POSPLUS:
1956        case OP_POSPLUSI:
1957        case OP_POSQUERY:
1958        case OP_POSQUERYI:
1959        case OP_POSSTAR:
1960        case OP_POSSTARI:
1961        case OP_POSUPTO:
1962        case OP_POSUPTOI:
1963        case OP_QUERY:
1964        case OP_QUERYI:
1965        case OP_REF:
1966        case OP_REFI:
1967        case OP_SBRA:
1968        case OP_SBRAPOS:
1969        case OP_SCBRA:
1970        case OP_SCBRAPOS:
1971        case OP_SCOND:
1972        case OP_SKIPZERO:
1973        case OP_STAR:
1974        case OP_STARI:
1975        case OP_TYPEMINPLUS:
1976        case OP_TYPEMINQUERY:
1977        case OP_TYPEMINSTAR:
1978        case OP_TYPEMINUPTO:
1979        case OP_TYPEPLUS:
1980        case OP_TYPEPOSPLUS:
1981        case OP_TYPEPOSQUERY:
1982        case OP_TYPEPOSSTAR:
1983        case OP_TYPEPOSUPTO:
1984        case OP_TYPEQUERY:
1985        case OP_TYPESTAR:
1986        case OP_TYPEUPTO:
1987        case OP_UPTO:
1988        case OP_UPTOI:
1989      return -1;      return -1;
1990    
1991        /* Catch unrecognized opcodes so that when new ones are added they
1992        are not forgotten, as has happened in the past. */
1993    
1994        default:
1995        return -4;
1996      }      }
1997    }    }
1998  /* Control never gets here */  /* Control never gets here */
# Line 1219  for (;;) Line 2002  for (;;)
2002    
2003    
2004  /*************************************************  /*************************************************
2005  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2006  *************************************************/  *************************************************/
2007    
2008  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2009  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2010    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2011    so that it can be called from pcre_study() when finding the minimum matching
2012    length.
2013    
2014  Arguments:  Arguments:
2015    code        points to start of expression    code        points to start of expression
2016    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2017    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2018    
2019  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
2020  */  */
2021    
2022  static const uschar *  const pcre_uchar *
2023  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2024  {  {
2025  for (;;)  for (;;)
2026    {    {
2027    register int c = *code;    register int c = *code;
2028    
2029    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2030    
2031    /* 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 1247  for (;;) Line 2034  for (;;)
2034    
2035    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2036    
2037      /* Handle recursion */
2038    
2039      else if (c == OP_REVERSE)
2040        {
2041        if (number < 0) return (pcre_uchar *)code;
2042        code += PRIV(OP_lengths)[c];
2043        }
2044    
2045    /* Handle capturing bracket */    /* Handle capturing bracket */
2046    
2047    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2048               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2049      {      {
2050      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2051      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2052      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2053      }      }
2054    
2055      /* Otherwise, we can get the item's length from the table, except that for
2056      repeated character types, we have to test for \p and \P, which have an extra
2057      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2058      must add in its length. */
2059    
2060      else
2061        {
2062        switch(c)
2063          {
2064          case OP_TYPESTAR:
2065          case OP_TYPEMINSTAR:
2066          case OP_TYPEPLUS:
2067          case OP_TYPEMINPLUS:
2068          case OP_TYPEQUERY:
2069          case OP_TYPEMINQUERY:
2070          case OP_TYPEPOSSTAR:
2071          case OP_TYPEPOSPLUS:
2072          case OP_TYPEPOSQUERY:
2073          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2074          break;
2075    
2076          case OP_TYPEUPTO:
2077          case OP_TYPEMINUPTO:
2078          case OP_TYPEEXACT:
2079          case OP_TYPEPOSUPTO:
2080          if (code[1 + IMM2_SIZE] == OP_PROP
2081            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2082          break;
2083    
2084          case OP_MARK:
2085          case OP_PRUNE_ARG:
2086          case OP_SKIP_ARG:
2087          code += code[1];
2088          break;
2089    
2090          case OP_THEN_ARG:
2091          code += code[1];
2092          break;
2093          }
2094    
2095        /* Add in the fixed length from the table */
2096    
2097        code += PRIV(OP_lengths)[c];
2098    
2099    /* 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
2100    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
2101    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2102    
2103    else  #ifdef SUPPORT_UTF
2104      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2105        {        {
2106        case OP_CHAR:        case OP_CHAR:
2107        case OP_CHARNC:        case OP_CHARI:
2108        case OP_EXACT:        case OP_EXACT:
2109          case OP_EXACTI:
2110        case OP_UPTO:        case OP_UPTO:
2111          case OP_UPTOI:
2112        case OP_MINUPTO:        case OP_MINUPTO:
2113          case OP_MINUPTOI:
2114        case OP_POSUPTO:        case OP_POSUPTO:
2115          case OP_POSUPTOI:
2116        case OP_STAR:        case OP_STAR:
2117          case OP_STARI:
2118        case OP_MINSTAR:        case OP_MINSTAR:
2119          case OP_MINSTARI:
2120        case OP_POSSTAR:        case OP_POSSTAR:
2121          case OP_POSSTARI:
2122        case OP_PLUS:        case OP_PLUS:
2123          case OP_PLUSI:
2124        case OP_MINPLUS:        case OP_MINPLUS:
2125          case OP_MINPLUSI:
2126        case OP_POSPLUS:        case OP_POSPLUS:
2127          case OP_POSPLUSI:
2128        case OP_QUERY:        case OP_QUERY:
2129          case OP_QUERYI:
2130        case OP_MINQUERY:        case OP_MINQUERY:
2131          case OP_MINQUERYI:
2132        case OP_POSQUERY:        case OP_POSQUERY:
2133        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2134          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2135        break;        break;
2136        }        }
2137    #else
2138        (void)(utf);  /* Keep compiler happy by referencing function argument */
2139  #endif  #endif
2140      }      }
2141    }    }
# Line 1300  instance of OP_RECURSE. Line 2152  instance of OP_RECURSE.
2152    
2153  Arguments:  Arguments:
2154    code        points to start of expression    code        points to start of expression
2155    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2156    
2157  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
2158  */  */
2159    
2160  static const uschar *  static const pcre_uchar *
2161  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2162  {  {
2163  for (;;)  for (;;)
2164    {    {
# Line 1320  for (;;) Line 2172  for (;;)
2172    
2173    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2174    
2175    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
2176    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
2177    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2178    bytes. */    must add in its length. */
2179    
2180    else    else
2181      {      {
2182      code += _pcre_OP_lengths[c];      switch(c)
2183  #ifdef SUPPORT_UTF8        {
2184      if (utf8) switch(c)        case OP_TYPESTAR:
2185          case OP_TYPEMINSTAR:
2186          case OP_TYPEPLUS:
2187          case OP_TYPEMINPLUS:
2188          case OP_TYPEQUERY:
2189          case OP_TYPEMINQUERY:
2190          case OP_TYPEPOSSTAR:
2191          case OP_TYPEPOSPLUS:
2192          case OP_TYPEPOSQUERY:
2193          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2194          break;
2195    
2196          case OP_TYPEPOSUPTO:
2197          case OP_TYPEUPTO:
2198          case OP_TYPEMINUPTO:
2199          case OP_TYPEEXACT:
2200          if (code[1 + IMM2_SIZE] == OP_PROP
2201            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2202          break;
2203    
2204          case OP_MARK:
2205          case OP_PRUNE_ARG:
2206          case OP_SKIP_ARG:
2207          code += code[1];
2208          break;
2209    
2210          case OP_THEN_ARG:
2211          code += code[1];
2212          break;
2213          }
2214    
2215        /* Add in the fixed length from the table */
2216    
2217        code += PRIV(OP_lengths)[c];
2218    
2219        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2220        by a multi-byte character. The length in the table is a minimum, so we have
2221        to arrange to skip the extra bytes. */
2222    
2223    #ifdef SUPPORT_UTF
2224        if (utf) switch(c)
2225        {        {
2226        case OP_CHAR:        case OP_CHAR:
2227        case OP_CHARNC:        case OP_CHARI:
2228        case OP_EXACT:        case OP_EXACT:
2229          case OP_EXACTI:
2230        case OP_UPTO:        case OP_UPTO:
2231          case OP_UPTOI:
2232        case OP_MINUPTO:        case OP_MINUPTO:
2233          case OP_MINUPTOI:
2234        case OP_POSUPTO:        case OP_POSUPTO:
2235          case OP_POSUPTOI:
2236        case OP_STAR:        case OP_STAR:
2237          case OP_STARI:
2238        case OP_MINSTAR:        case OP_MINSTAR:
2239          case OP_MINSTARI:
2240        case OP_POSSTAR:        case OP_POSSTAR:
2241          case OP_POSSTARI:
2242        case OP_PLUS:        case OP_PLUS:
2243          case OP_PLUSI:
2244        case OP_MINPLUS:        case OP_MINPLUS:
2245          case OP_MINPLUSI:
2246        case OP_POSPLUS:        case OP_POSPLUS:
2247          case OP_POSPLUSI:
2248        case OP_QUERY:        case OP_QUERY:
2249          case OP_QUERYI:
2250        case OP_MINQUERY:        case OP_MINQUERY:
2251          case OP_MINQUERYI:
2252        case OP_POSQUERY:        case OP_POSQUERY:
2253        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2254          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2255        break;        break;
2256        }        }
2257    #else
2258        (void)(utf);  /* Keep compiler happy by referencing function argument */
2259  #endif  #endif
2260      }      }
2261    }    }
# Line 1364  for (;;) Line 2271  for (;;)
2271  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()
2272  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
2273  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2274  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
2275  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
2276    bracket whose current branch will already have been scanned.
2277    
2278  Arguments:  Arguments:
2279    code        points to start of search    code        points to start of search
2280    endcode     points to where to stop    endcode     points to where to stop
2281    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2282      cd          contains pointers to tables etc.
2283    
2284  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2285  */  */
2286    
2287  static BOOL  static BOOL
2288  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2289      BOOL utf, compile_data *cd)
2290  {  {
2291  register int c;  register int c;
2292  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2293       code < endcode;       code < endcode;
2294       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2295    {    {
2296    const uschar *ccode;    const pcre_uchar *ccode;
2297    
2298    c = *code;    c = *code;
2299    
2300    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2301      first_significant_code() with a TRUE final argument. */
2302    
2303    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2304      {      {
     code += _pcre_OP_lengths[c];  
2305      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2306      c = *code;      c = *code;
2307      continue;      continue;
2308      }      }
2309    
2310    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2311      implies a backward reference subroutine call, we can scan it. If it's a
2312      forward reference subroutine call, we can't. To detect forward reference
2313      we have to scan up the list that is kept in the workspace. This function is
2314      called only when doing the real compile, not during the pre-compile that
2315      measures the size of the compiled pattern. */
2316    
2317    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_RECURSE)
2318      {      {
2319        const pcre_uchar *scode;
2320      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2321    
2322      /* Scan a closed bracket */      /* Test for forward reference */
2323    
2324        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2325          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2326    
2327        /* Not a forward reference, test for completed backward reference */
2328    
2329      empty_branch = FALSE;      empty_branch = FALSE;
2330        scode = cd->start_code + GET(code, 1);
2331        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2332    
2333        /* Completed backwards reference */
2334    
2335      do      do
2336        {        {
2337        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2338            {
2339          empty_branch = TRUE;          empty_branch = TRUE;
2340            break;
2341            }
2342          scode += GET(scode, 1);
2343          }
2344        while (*scode == OP_ALT);
2345    
2346        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2347        continue;
2348        }
2349    
2350      /* Groups with zero repeats can of course be empty; skip them. */
2351    
2352      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2353          c == OP_BRAPOSZERO)
2354        {
2355        code += PRIV(OP_lengths)[c];
2356        do code += GET(code, 1); while (*code == OP_ALT);
2357        c = *code;
2358        continue;
2359        }
2360    
2361      /* A nested group that is already marked as "could be empty" can just be
2362      skipped. */
2363    
2364      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2365          c == OP_SCBRA || c == OP_SCBRAPOS)
2366        {
2367        do code += GET(code, 1); while (*code == OP_ALT);
2368        c = *code;
2369        continue;
2370        }
2371    
2372      /* For other groups, scan the branches. */
2373    
2374      if (c == OP_BRA  || c == OP_BRAPOS ||
2375          c == OP_CBRA || c == OP_CBRAPOS ||
2376          c == OP_ONCE || c == OP_ONCE_NC ||
2377          c == OP_COND)
2378        {
2379        BOOL empty_branch;
2380        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2381    
2382        /* If a conditional group has only one branch, there is a second, implied,
2383        empty branch, so just skip over the conditional, because it could be empty.
2384        Otherwise, scan the individual branches of the group. */
2385    
2386        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2387        code += GET(code, 1);        code += GET(code, 1);
2388        else
2389          {
2390          empty_branch = FALSE;
2391          do
2392            {
2393            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2394              empty_branch = TRUE;
2395            code += GET(code, 1);
2396            }
2397          while (*code == OP_ALT);
2398          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2399        }        }
2400      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2401      c = *code;      c = *code;
2402      continue;      continue;
2403      }      }
# Line 1423  for (code = first_significant_code(code Line 2406  for (code = first_significant_code(code
2406    
2407    switch (c)    switch (c)
2408      {      {
2409      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2410        cannot be represented just by a bit map. This includes negated single
2411        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2412        actual length is stored in the compiled code, so we must update "code"
2413        here. */
2414    
2415  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2416      case OP_XCLASS:      case OP_XCLASS:
2417      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2418      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2419  #endif  #endif
2420    
2421      case OP_CLASS:      case OP_CLASS:
2422      case OP_NCLASS:      case OP_NCLASS:
2423      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2424    
2425  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2426      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2427  #endif  #endif
2428    
# Line 1471  for (code = first_significant_code(code Line 2458  for (code = first_significant_code(code
2458      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2459      case OP_WORDCHAR:      case OP_WORDCHAR:
2460      case OP_ANY:      case OP_ANY:
2461        case OP_ALLANY:
2462      case OP_ANYBYTE:      case OP_ANYBYTE:
2463      case OP_CHAR:      case OP_CHAR:
2464      case OP_CHARNC:      case OP_CHARI:
2465      case OP_NOT:      case OP_NOT:
2466        case OP_NOTI:
2467      case OP_PLUS:      case OP_PLUS:
2468      case OP_MINPLUS:      case OP_MINPLUS:
2469      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1489  for (code = first_significant_code(code Line 2478  for (code = first_significant_code(code
2478      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2479      return FALSE;      return FALSE;
2480    
2481        /* These are going to continue, as they may be empty, but we have to
2482        fudge the length for the \p and \P cases. */
2483    
2484        case OP_TYPESTAR:
2485        case OP_TYPEMINSTAR:
2486        case OP_TYPEPOSSTAR:
2487        case OP_TYPEQUERY:
2488        case OP_TYPEMINQUERY:
2489        case OP_TYPEPOSQUERY:
2490        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2491        break;
2492    
2493        /* Same for these */
2494    
2495        case OP_TYPEUPTO:
2496        case OP_TYPEMINUPTO:
2497        case OP_TYPEPOSUPTO:
2498        if (code[1 + IMM2_SIZE] == OP_PROP
2499          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2500        break;
2501    
2502      /* End of branch */      /* End of branch */
2503    
2504      case OP_KET:      case OP_KET:
2505      case OP_KETRMAX:      case OP_KETRMAX:
2506      case OP_KETRMIN:      case OP_KETRMIN:
2507        case OP_KETRPOS:
2508      case OP_ALT:      case OP_ALT:
2509      return TRUE;      return TRUE;
2510    
2511      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2512      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2513    
2514  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2515      case OP_STAR:      case OP_STAR:
2516        case OP_STARI:
2517      case OP_MINSTAR:      case OP_MINSTAR:
2518        case OP_MINSTARI:
2519      case OP_POSSTAR:      case OP_POSSTAR:
2520        case OP_POSSTARI:
2521      case OP_QUERY:      case OP_QUERY:
2522        case OP_QUERYI:
2523      case OP_MINQUERY:      case OP_MINQUERY:
2524        case OP_MINQUERYI:
2525      case OP_POSQUERY:      case OP_POSQUERY:
2526        case OP_POSQUERYI:
2527        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2528        break;
2529    
2530      case OP_UPTO:      case OP_UPTO:
2531        case OP_UPTOI:
2532      case OP_MINUPTO:      case OP_MINUPTO:
2533        case OP_MINUPTOI:
2534      case OP_POSUPTO:      case OP_POSUPTO:
2535      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2536        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2537      break;      break;
2538  #endif  #endif
2539    
2540        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2541        string. */
2542    
2543        case OP_MARK:
2544        case OP_PRUNE_ARG:
2545        case OP_SKIP_ARG:
2546        code += code[1];
2547        break;
2548    
2549        case OP_THEN_ARG:
2550        code += code[1];
2551        break;
2552    
2553        /* None of the remaining opcodes are required to match a character. */
2554    
2555        default:
2556        break;
2557      }      }
2558    }    }
2559    
# Line 1529  return TRUE; Line 2570  return TRUE;
2570  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
2571  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,
2572  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.
2573    This function is called only during the real compile, not during the
2574    pre-compile.
2575    
2576  Arguments:  Arguments:
2577    code        points to start of the recursion    code        points to start of the recursion
2578    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2579    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2580    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2581      cd          pointers to tables etc
2582    
2583  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2584  */  */
2585    
2586  static BOOL  static BOOL
2587  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2588    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2589  {  {
2590  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2591    {    {
2592    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2593        return FALSE;
2594    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2595    }    }
2596  return TRUE;  return TRUE;
# Line 1558  return TRUE; Line 2603  return TRUE;
2603  *************************************************/  *************************************************/
2604    
2605  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2606  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
2607  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2608  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2609    
2610    Originally, this function only recognized a sequence of letters between the
2611    terminators, but it seems that Perl recognizes any sequence of characters,
2612    though of course unknown POSIX names are subsequently rejected. Perl gives an
2613    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2614    didn't consider this to be a POSIX class. Likewise for [:1234:].
2615    
2616    The problem in trying to be exactly like Perl is in the handling of escapes. We
2617    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2618    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2619    below handles the special case of \], but does not try to do any other escape
2620    processing. This makes it different from Perl for cases such as [:l\ower:]
2621    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2622    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2623    I think.
2624    
2625    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2626    It seems that the appearance of a nested POSIX class supersedes an apparent
2627    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2628    a digit.
2629    
2630    In Perl, unescaped square brackets may also appear as part of class names. For
2631    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2632    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2633    seem right at all. PCRE does not allow closing square brackets in POSIX class
2634    names.
2635    
2636  Argument:  Arguments:
2637    ptr      pointer to the initial [    ptr      pointer to the initial [
2638    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2639    
2640  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2641  */  */
2642    
2643  static BOOL  static BOOL
2644  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2645  {  {
2646  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2647  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2648  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2649    {    {
2650    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2651    return TRUE;      ptr++;
2652      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2653      else
2654        {
2655        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2656          {
2657          *endptr = ptr;
2658          return TRUE;
2659          }
2660        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2661             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2662              ptr[1] == CHAR_EQUALS_SIGN) &&
2663            check_posix_syntax(ptr, endptr))
2664          return FALSE;
2665        }
2666    }    }
2667  return FALSE;  return FALSE;
2668  }  }
# Line 1603  Returns:     a value representing the na Line 2685  Returns:     a value representing the na
2685  */  */
2686    
2687  static int  static int
2688  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2689  {  {
2690    const char *pn = posix_names;
2691  register int yield = 0;  register int yield = 0;
2692  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2693    {    {
2694    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2695      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2696      pn += posix_name_lengths[yield] + 1;
2697    yield++;    yield++;
2698    }    }
2699  return -1;  return -1;
# Line 1624  return -1; Line 2708  return -1;
2708  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2709  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2710  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
2711  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
2712  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
2713  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
2714  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
2715  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2716    OP_END.
2717    
2718  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2719  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 1639  value in the reference (which is a group Line 2724  value in the reference (which is a group
2724  Arguments:  Arguments:
2725    group      points to the start of the group    group      points to the start of the group
2726    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2727    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2728    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2729    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
2730    
# Line 1647  Returns:     nothing Line 2732  Returns:     nothing
2732  */  */
2733    
2734  static void  static void
2735  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2736    uschar *save_hwm)    pcre_uchar *save_hwm)
2737  {  {
2738  uschar *ptr = group;  pcre_uchar *ptr = group;
2739  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2740    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2741    {    {
2742    int offset;    int offset;
2743    uschar *hc;    pcre_uchar *hc;
2744    
2745    /* 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
2746    reference. */    reference. */
# Line 1699  Arguments: Line 2785  Arguments:
2785  Returns:         new code pointer  Returns:         new code pointer
2786  */  */
2787    
2788  static uschar *  static pcre_uchar *
2789  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2790  {  {
2791  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2792  *code++ = 255;  *code++ = 255;
2793  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2794  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2795  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2796  }  }
2797    
2798    
# Line 1728  Returns:             nothing Line 2814  Returns:             nothing
2814  */  */
2815    
2816  static void  static void
2817  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2818  {  {
2819  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2820  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2821  }  }
2822    
# Line 1762  get_othercase_range(unsigned int *cptr, Line 2848  get_othercase_range(unsigned int *cptr,
2848  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2849    
2850  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2851    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2852    
2853  if (c > d) return FALSE;  if (c > d) return FALSE;
2854    
# Line 1771  next = othercase + 1; Line 2857  next = othercase + 1;
2857    
2858  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2859    {    {
2860    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2861    next++;    next++;
2862    }    }
2863    
# Line 1780  for (++c; c <= d; c++) Line 2866  for (++c; c <= d; c++)
2866    
2867  return TRUE;  return TRUE;
2868  }  }
2869    
2870    
2871    
2872    /*************************************************
2873    *        Check a character and a property        *
2874    *************************************************/
2875    
2876    /* This function is called by check_auto_possessive() when a property item
2877    is adjacent to a fixed character.
2878    
2879    Arguments:
2880      c            the character
2881      ptype        the property type
2882      pdata        the data for the type
2883      negated      TRUE if it's a negated property (\P or \p{^)
2884    
2885    Returns:       TRUE if auto-possessifying is OK
2886    */
2887    
2888    static BOOL
2889    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2890    {
2891    const ucd_record *prop = GET_UCD(c);
2892    switch(ptype)
2893      {
2894      case PT_LAMP:
2895      return (prop->chartype == ucp_Lu ||
2896              prop->chartype == ucp_Ll ||
2897              prop->chartype == ucp_Lt) == negated;
2898    
2899      case PT_GC:
2900      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2901    
2902      case PT_PC:
2903      return (pdata == prop->chartype) == negated;
2904    
2905      case PT_SC:
2906      return (pdata == prop->script) == negated;
2907    
2908      /* These are specials */
2909    
2910      case PT_ALNUM:
2911      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2912              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2913    
2914      case PT_SPACE:    /* Perl space */
2915      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2916              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2917              == negated;
2918    
2919      case PT_PXSPACE:  /* POSIX space */
2920      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2921              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2922              c == CHAR_FF || c == CHAR_CR)
2923              == negated;
2924    
2925      case PT_WORD:
2926      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2927              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2928              c == CHAR_UNDERSCORE) == negated;
2929      }
2930    return FALSE;
2931    }
2932  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2933    
2934    
# Line 1793  whether the next thing could possibly ma Line 2942  whether the next thing could possibly ma
2942  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2943    
2944  Arguments:  Arguments:
2945    op_code       the repeated op code    previous      pointer to the repeated opcode
2946    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  
2947    ptr           next character in pattern    ptr           next character in pattern
2948    options       options bits    options       options bits
2949    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1805  Returns:        TRUE if possessifying is Line 2952  Returns:        TRUE if possessifying is
2952  */  */
2953    
2954  static BOOL  static BOOL
2955  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2956    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2957  {  {
2958  int next;  pcre_int32 c, next;
2959    int op_code = *previous++;
2960    
2961  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2962    
# Line 1816  if ((options & PCRE_EXTENDED) != 0) Line 2964  if ((options & PCRE_EXTENDED) != 0)
2964    {    {
2965    for (;;)    for (;;)
2966      {      {
2967      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2968      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2969        {        {
2970        while (*(++ptr) != 0)        ptr++;
2971          while (*ptr != 0)
2972            {
2973          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2974            ptr++;
2975    #ifdef SUPPORT_UTF
2976            if (utf) FORWARDCHAR(ptr);
2977    #endif
2978            }
2979        }        }
2980      else break;      else break;
2981      }      }
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 2984  if ((options & PCRE_EXTENDED) != 0)
2984  /* 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
2985  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2986    
2987  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2988    {    {
2989    int temperrorcode = 0;    int temperrorcode = 0;
2990    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2991    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2992    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2993    }    }
2994    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2995    {    {
2996  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2997    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2998  #endif  #endif
2999    next = *ptr++;    next = *ptr++;
3000    }    }
   
3001  else return FALSE;  else return FALSE;
3002    
3003  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1853  if ((options & PCRE_EXTENDED) != 0) Line 3006  if ((options & PCRE_EXTENDED) != 0)
3006    {    {
3007    for (;;)    for (;;)
3008      {      {
3009      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3010      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3011        {        {
3012        while (*(++ptr) != 0)        ptr++;
3013          while (*ptr != 0)
3014            {
3015          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3016            ptr++;
3017    #ifdef SUPPORT_UTF
3018            if (utf) FORWARDCHAR(ptr);
3019    #endif
3020            }
3021        }        }
3022      else break;      else break;
3023      }      }
# Line 1865  if ((options & PCRE_EXTENDED) != 0) Line 3025  if ((options & PCRE_EXTENDED) != 0)
3025    
3026  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3027    
3028  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3029    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3030        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. */  
   
3031    
3032  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3033    the next item is a character. */
3034    
3035  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3036    {    {
3037    case OP_CHAR:    case OP_CHAR:
3038  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3039    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3040    #else
3041      c = *previous;
3042  #endif  #endif
3043    return item != next;    return c != next;
3044    
3045    /* 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
3046    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
3047    high-valued characters. */    high-valued characters. */
3048    
3049    case OP_CHARNC:    case OP_CHARI:
3050  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3051    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3052  #endif  #else
3053    if (item == next) return FALSE;    c = *previous;
3054  #ifdef SUPPORT_UTF8  #endif
3055    if (utf8)    if (c == next) return FALSE;
3056    #ifdef SUPPORT_UTF
3057      if (utf)
3058      {      {
3059      unsigned int othercase;      unsigned int othercase;
3060      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3061  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3062      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3063  #else  #else
3064      othercase = NOTACHAR;      othercase = NOTACHAR;
3065  #endif  #endif
3066      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3067      }      }
3068    else    else
3069  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3070    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */
3071    
3072    /* 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
3073      opcodes are not used for multi-byte characters, because they are coded using
3074      an XCLASS instead. */
3075    
3076    case OP_NOT:    case OP_NOT:
3077    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
3078    if (item == next) return TRUE;  
3079    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
3080  #ifdef SUPPORT_UTF8    if ((c = *previous) == next) return TRUE;
3081    if (utf8)  #ifdef SUPPORT_UTF
3082      if (utf)
3083      {      {
3084      unsigned int othercase;      unsigned int othercase;
3085      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3086  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3087      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3088  #else  #else
3089      othercase = NOTACHAR;      othercase = NOTACHAR;
3090  #endif  #endif
3091      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3092      }      }
3093    else    else
3094  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3095    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == (int)(TABLE_GET((unsigned int)next, cd->fcc, next)));  /* Non-UTF-8 mode */
3096    
3097      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3098      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3099    
3100    case OP_DIGIT:    case OP_DIGIT:
3101    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1972  if (next >= 0) switch(op_code) Line 3138  if (next >= 0) switch(op_code)
3138      case 0x202f:      case 0x202f:
3139      case 0x205f:      case 0x205f:
3140      case 0x3000:      case 0x3000:
3141      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3142      default:      default:
3143      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3144      }      }
3145    
3146      case OP_ANYNL:
3147    case OP_VSPACE:    case OP_VSPACE:
3148    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3149    switch(next)    switch(next)
# Line 1988  if (next >= 0) switch(op_code) Line 3155  if (next >= 0) switch(op_code)
3155      case 0x85:      case 0x85:
3156      case 0x2028:      case 0x2028:
3157      case 0x2029:      case 0x2029:
3158      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3159      default:      default:
3160      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3161      }      }
3162    
3163    #ifdef SUPPORT_UCP
3164      case OP_PROP:
3165      return check_char_prop(next, previous[0], previous[1], FALSE);
3166    
3167      case OP_NOTPROP:
3168      return check_char_prop(next, previous[0], previous[1], TRUE);
3169    #endif
3170    
3171    default:    default:
3172    return FALSE;    return FALSE;
3173    }    }
3174    
3175    
3176  /* 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
3177    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3178    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3179    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3180    replaced by OP_PROP codes when PCRE_UCP is set. */
3181    
3182  switch(op_code)  switch(op_code)
3183    {    {
3184    case OP_CHAR:    case OP_CHAR:
3185    case OP_CHARNC:    case OP_CHARI:
3186  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3187    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3188    #else
3189      c = *previous;
3190  #endif  #endif
3191    switch(-next)    switch(-next)
3192      {      {
3193      case ESC_d:      case ESC_d:
3194      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3195    
3196      case ESC_D:      case ESC_D:
3197      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3198    
3199      case ESC_s:      case ESC_s:
3200      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3201    
3202      case ESC_S:      case ESC_S:
3203      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3204    
3205      case ESC_w:      case ESC_w:
3206      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3207    
3208      case ESC_W:      case ESC_W:
3209      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3210    
3211      case ESC_h:      case ESC_h:
3212      case ESC_H:      case ESC_H:
3213      switch(item)      switch(c)
3214        {        {
3215        case 0x09:        case 0x09:
3216        case 0x20:        case 0x20:
# Line 2057  switch(op_code) Line 3238  switch(op_code)
3238    
3239      case ESC_v:      case ESC_v:
3240      case ESC_V:      case ESC_V:
3241      switch(item)      switch(c)
3242        {        {
3243        case 0x0a:        case 0x0a:
3244        case 0x0b:        case 0x0b:
# Line 2071  switch(op_code) Line 3252  switch(op_code)
3252        return -next == ESC_v;        return -next == ESC_v;
3253        }        }
3254    
3255        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3256        their substitutions and process them. The result will always be either
3257        -ESC_p or -ESC_P. Then fall through to process those values. */
3258    
3259    #ifdef SUPPORT_UCP
3260        case ESC_du:
3261        case ESC_DU:
3262        case ESC_wu:
3263        case ESC_WU:
3264        case ESC_su:
3265        case ESC_SU:
3266          {
3267          int temperrorcode = 0;
3268          ptr = substitutes[-next - ESC_DU];
3269          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3270          if (temperrorcode != 0) return FALSE;
3271          ptr++;    /* For compatibility */
3272          }
3273        /* Fall through */
3274    
3275        case ESC_p:
3276        case ESC_P:
3277          {
3278          int ptype, pdata, errorcodeptr;
3279          BOOL negated;
3280    
3281          ptr--;      /* Make ptr point at the p or P */
3282          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3283          if (ptype < 0) return FALSE;
3284          ptr++;      /* Point past the final curly ket */
3285    
3286          /* If the property item is optional, we have to give up. (When generated
3287          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3288          to the original \d etc. At this point, ptr will point to a zero byte. */
3289    
3290          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3291            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3292              return FALSE;
3293    
3294          /* Do the property check. */
3295    
3296          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3297          }
3298    #endif
3299    
3300      default:      default:
3301      return FALSE;      return FALSE;
3302      }      }
3303    
3304      /* In principle, support for Unicode properties should be integrated here as
3305      well. It means re-organizing the above code so as to get hold of the property
3306      values before switching on the op-code. However, I wonder how many patterns
3307      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3308      these op-codes are never generated.) */
3309    
3310    case OP_DIGIT:    case OP_DIGIT:
3311    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3312           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3313    
3314    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3315    return next == -ESC_d;    return next == -ESC_d;
3316    
3317    case OP_WHITESPACE:    case OP_WHITESPACE:
3318    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3319    
3320    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3321    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3322    
3323    case OP_HSPACE:    case OP_HSPACE:
3324    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3325             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3326    
3327    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3328    return next == -ESC_h;    return next == -ESC_h;
3329    
3330    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3331      case OP_ANYNL:
3332    case OP_VSPACE:    case OP_VSPACE:
3333    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3334    
3335    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3336    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3337    
3338    case OP_WORDCHAR:    case OP_WORDCHAR:
3339    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3340             next == -ESC_v || next == -ESC_R;
3341    
3342    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3343    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2131  Arguments: Line 3366  Arguments:
3366    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3367    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3368    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3369    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3370    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3371    bcptr          points to current branch chain    bcptr          points to current branch chain
3372      cond_depth     conditional nesting depth
3373    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3374    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3375                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2143  Returns:         TRUE on success Line 3379  Returns:         TRUE on success
3379  */  */
3380    
3381  static BOOL  static BOOL
3382  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3383    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3384      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3385    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3386  {  {
3387  int repeat_type, op_type;  int repeat_type, op_type;
3388  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3389  int bravalue = 0;  int bravalue = 0;
3390  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3391  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3392  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3393  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3394  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3395  int after_manual_callout = 0;  int after_manual_callout = 0;
3396  int length_prevgroup = 0;  int length_prevgroup = 0;
3397  register int c;  register int c;
3398  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3399  uschar *last_code = code;  pcre_uchar *last_code = code;
3400  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3401  uschar *tempcode;  pcre_uchar *tempcode;
3402  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3403  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3404  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3405  const uschar *tempptr;  const pcre_uchar *tempptr;
3406  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3407  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3408  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3409  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3410    pcre_uint8 classbits[32];
3411  #ifdef SUPPORT_UTF8  
3412  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3413  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3414  uschar *class_utf8data;  dynamically as we process the pattern. */
3415  uschar utf8_char[6];  
3416    #ifdef SUPPORT_UTF
3417    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3418    BOOL utf = (options & PCRE_UTF8) != 0;
3419    pcre_uchar utf_chars[6];
3420  #else  #else
3421  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3422  uschar *utf8_char = NULL;  #endif
3423    
3424    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3425    
3426    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3427    BOOL xclass;
3428    pcre_uchar *class_uchardata;
3429    pcre_uchar *class_uchardata_base;
3430  #endif  #endif
3431    
3432  #ifdef DEBUG  #ifdef PCRE_DEBUG
3433  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3434  #endif  #endif
3435    
# Line 2192  greedy_non_default = greedy_default ^ 1; Line 3440  greedy_non_default = greedy_default ^ 1;
3440    
3441  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3442  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
3443  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
3444  find one.  find one.
3445    
3446  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
3447  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
3448  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3449  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3450    
3451  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3452    
3453  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3454  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
3455  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
3456  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3457    value. This is used only for ASCII characters. */
3458    
3459  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3460    
3461  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3462    
3463  for (;; ptr++)  for (;; ptr++)
3464    {    {
3465    BOOL negate_class;    BOOL negate_class;
3466      BOOL should_flip_negation;
3467    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3468    BOOL is_quantifier;    BOOL is_quantifier;
3469    BOOL is_recurse;    BOOL is_recurse;
3470    BOOL reset_bracount;    BOOL reset_bracount;
3471    int class_charcount;    int class_has_8bitchar;
3472    int class_lastchar;    int class_single_char;
3473    int newoptions;    int newoptions;
3474    int recno;    int recno;
3475    int refsign;    int refsign;
3476    int skipbytes;    int skipbytes;
3477    int subreqbyte;    int subreqchar;
3478    int subfirstbyte;    int subfirstchar;
3479    int terminator;    int terminator;
3480    int mclength;    int mclength;
3481    uschar mcbuffer[8];    int tempbracount;
3482      pcre_uchar mcbuffer[8];
3483    
3484    /* Get next byte in the pattern */    /* Get next character in the pattern */
3485    
3486    c = *ptr;    c = *ptr;
3487    
3488      /* If we are at the end of a nested substitution, revert to the outer level
3489      string. Nesting only happens one level deep. */
3490    
3491      if (c == 0 && nestptr != NULL)
3492        {
3493        ptr = nestptr;
3494        nestptr = NULL;
3495        c = *ptr;
3496        }
3497    
3498    /* 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
3499    previous cycle of this loop. */    previous cycle of this loop. */
3500    
3501    if (lengthptr != NULL)    if (lengthptr != NULL)
3502      {      {
3503  #ifdef DEBUG  #ifdef PCRE_DEBUG
3504      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3505  #endif  #endif
3506      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3507            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3508        {        {
3509        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3510        goto FAILED;        goto FAILED;
# Line 2255  for (;; ptr++) Line 3517  for (;; ptr++)
3517      */      */
3518    
3519      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3520      *lengthptr += code - last_code;  
3521      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      /* Paranoid check for integer overflow */
3522    
3523        if (OFLOW_MAX - *lengthptr < code - last_code)
3524          {
3525          *errorcodeptr = ERR20;
3526          goto FAILED;
3527          }
3528    
3529        *lengthptr += (int)(code - last_code);
3530        DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3531          (int)(code - last_code), c, c));
3532    
3533      /* 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
3534      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,
# Line 2266  for (;; ptr++) Line 3538  for (;; ptr++)
3538        {        {
3539        if (previous > orig_code)        if (previous > orig_code)
3540          {          {
3541          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3542          code -= previous - orig_code;          code -= previous - orig_code;
3543          previous = orig_code;          previous = orig_code;
3544          }          }
# Line 2282  for (;; ptr++) Line 3554  for (;; ptr++)
3554    /* 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
3555    reference list. */    reference list. */
3556    
3557    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3558               WORK_SIZE_SAFETY_MARGIN)
3559      {      {
3560      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3561      goto FAILED;      goto FAILED;
# Line 2292  for (;; ptr++) Line 3565  for (;; ptr++)
3565    
3566    if (inescq && c != 0)    if (inescq && c != 0)
3567      {      {
3568      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3569        {        {
3570        inescq = FALSE;        inescq = FALSE;
3571        ptr++;        ptr++;
# Line 2318  for (;; ptr++) Line 3591  for (;; ptr++)
3591    /* 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
3592    a quantifier. */    a quantifier. */
3593    
3594    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3595      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3596        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3597    
3598    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3599         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2329  for (;; ptr++) Line 3603  for (;; ptr++)
3603      previous_callout = NULL;      previous_callout = NULL;
3604      }      }
3605    
3606    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3607    
3608    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3609      {      {
3610      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3611      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3612        {        {
3613        while (*(++ptr) != 0)        ptr++;
3614          while (*ptr != 0)
3615          {          {
3616          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3617            ptr++;
3618    #ifdef SUPPORT_UTF
3619            if (utf) FORWARDCHAR(ptr);
3620    #endif
3621          }          }
3622        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3623    
# Line 2359  for (;; ptr++) Line 3638  for (;; ptr++)
3638      {      {
3639      /* ===================================================================*/      /* ===================================================================*/
3640      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3641      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3642      case ')':      case CHAR_RIGHT_PARENTHESIS:
3643      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3644      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3645      *codeptr = code;      *codeptr = code;
3646      *ptrptr = ptr;      *ptrptr = ptr;
3647      if (lengthptr != NULL)      if (lengthptr != NULL)
3648        {        {
3649        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3650            {
3651            *errorcodeptr = ERR20;
3652            goto FAILED;
3653            }
3654          *lengthptr += (int)(code - last_code);   /* To include callout length */
3655        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3656        }        }
3657      return TRUE;      return TRUE;
# Line 2377  for (;; ptr++) Line 3661  for (;; ptr++)
3661      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3662      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3663    
3664      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3665        previous = NULL;
3666      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3667        {        {
3668        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3669          *code++ = OP_CIRCM;
3670        }        }
3671      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3672      break;      break;
3673    
3674      case '$':      case CHAR_DOLLAR_SIGN:
3675      previous = NULL;      previous = NULL;
3676      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3677      break;      break;
3678    
3679      /* 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
3680      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3681    
3682      case '.':      case CHAR_DOT:
3683      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3684      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3685      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3686      previous = code;      previous = code;
3687      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3688      break;      break;
3689    
3690    
# Line 2414  for (;; ptr++) Line 3699  for (;; ptr++)
3699      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,
3700      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3701      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.
     */  
3702    
3703      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3704        default (Perl) mode, it is treated as a data character. */
3705    
3706        case CHAR_RIGHT_SQUARE_BRACKET:
3707        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3708          {
3709          *errorcodeptr = ERR64;
3710          goto FAILED;
3711          }
3712        goto NORMAL_CHAR;
3713    
3714        case CHAR_LEFT_SQUARE_BRACKET:
3715      previous = code;      previous = code;
3716    
3717      /* 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
3718      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. */
3719    
3720      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3721          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3722            check_posix_syntax(ptr, &tempptr))
3723        {        {
3724        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3725        goto FAILED;        goto FAILED;
3726        }        }
3727    
3728      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3729        if the first few characters (either before or after ^) are \Q\E or \E we
3730        skip them too. This makes for compatibility with Perl. */
3731    
3732      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3733        for (;;)
3734        {        {
       negate_class = TRUE;  
3735        c = *(++ptr);        c = *(++ptr);
3736        }        if (c == CHAR_BACKSLASH)
3737      else          {
3738        {          if (ptr[1] == CHAR_E)
3739        negate_class = FALSE;            ptr++;
3740            else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
3741              ptr += 3;
3742            else
3743              break;
3744            }
3745          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3746            negate_class = TRUE;
3747          else break;
3748          }
3749    
3750        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3751        an initial ']' is taken as a data character -- the code below handles
3752        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3753        [^] must match any character, so generate OP_ALLANY. */
3754    
3755        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3756            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3757          {
3758          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3759          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3760          zerofirstchar = firstchar;
3761          break;
3762        }        }
3763    
3764      /* Keep a count of chars with values < 256 so that we can optimize the case      /* If a class contains a negative special such as \S, we need to flip the
3765      of just a single character (as long as it's < 256). However, For higher      negation flag at the end, so that support for characters > 255 works
3766      valued UTF-8 characters, we don't yet do any optimization. */      correctly (they are all included in the class). */
3767    
3768        should_flip_negation = FALSE;
3769    
3770        /* For optimization purposes, we track some properties of the class.
3771        class_has_8bitchar will be non-zero, if the class contains at least one
3772        < 256 character. class_single_char will be 1 if the class contains only
3773        a single character. */
3774    
3775      class_charcount = 0;      class_has_8bitchar = 0;
3776      class_lastchar = -1;      class_single_char = 0;
3777    
3778      /* 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
3779      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
3780      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.
3781      */      */
3782    
3783      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3784    
3785  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3786      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3787      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3788        class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3789  #endif  #endif
3790    
3791      /* 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 2466  for (;; ptr++) Line 3794  for (;; ptr++)
3794    
3795      if (c != 0) do      if (c != 0) do
3796        {        {
3797        const uschar *oldptr;        const pcre_uchar *oldptr;
3798    
3799  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3800        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3801          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3802          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3803          }          }
3804  #endif  #endif
3805    
3806    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3807          /* In the pre-compile phase, accumulate the length of any extra
3808          data and reset the pointer. This is so that very large classes that
3809          contain a zillion > 255 characters no longer overwrite the work space
3810          (which is on the stack). */
3811    
3812          if (lengthptr != NULL)
3813            {
3814            *lengthptr += class_uchardata - class_uchardata_base;
3815            class_uchardata = class_uchardata_base;
3816            }
3817    #endif
3818    
3819        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3820    
3821        if (inescq)        if (inescq)
3822          {          {
3823          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3824            {            {
3825            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3826            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2494  for (;; ptr++) Line 3835  for (;; ptr++)
3835        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3836        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3837    
3838        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3839            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3840            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3841          {          {
3842          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3843          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3844          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3845          uschar pbits[32];          pcre_uint8 pbits[32];
3846    
3847          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3848            {            {
3849            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3850            goto FAILED;            goto FAILED;
3851            }            }
3852    
3853          ptr += 2;          ptr += 2;
3854          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3855            {            {
3856            local_negate = TRUE;            local_negate = TRUE;
3857              should_flip_negation = TRUE;  /* Note negative special */
3858            ptr++;            ptr++;
3859            }            }
3860    
3861          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3862          if (posix_class < 0)          if (posix_class < 0)
3863            {            {
3864            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2530  for (;; ptr++) Line 3872  for (;; ptr++)
3872          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3873            posix_class = 0;            posix_class = 0;
3874    
3875          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3876          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3877          subtract bits that may be in the main map already. At the end we or the  
3878          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3879            if ((options & PCRE_UCP) != 0)
3880              {
3881              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3882              if (posix_substitutes[pc] != NULL)
3883                {
3884                nestptr = tempptr + 1;
3885                ptr = posix_substitutes[pc] - 1;
3886                continue;
3887                }
3888              }
3889    #endif
3890            /* In the non-UCP case, we build the bit map for the POSIX class in a
3891            chunk of local store because we may be adding and subtracting from it,
3892            and we don't want to subtract bits that may be in the main map already.
3893            At the end we or the result into the bit map that is being built. */
3894    
3895          posix_class *= 3;          posix_class *= 3;
3896    
3897          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3898    
3899          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3900            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3901    
3902          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3903    
# Line 2571  for (;; ptr++) Line 3928  for (;; ptr++)
3928            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3929    
3930          ptr = tempptr + 1;          ptr = tempptr + 1;
3931          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3932            class_has_8bitchar = 1;
3933            /* Every class contains at least two characters. */
3934            class_single_char = 2;
3935          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3936          }          }
3937    
3938        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3939        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3940        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3941        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so