/[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 206 by ph10, Fri Aug 3 14:53:04 2007 UTC revision 1059 by chpe, Tue Oct 16 15:53:53 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 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_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    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* 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  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,
88        int, int, int *, int *, branch_chain *, compile_data *, int *);
89    
90    
91    
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 103  so this number is very generous.
103  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
104  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
105  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
106  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
107    filled up by repetitions of forward references, for example patterns like
108    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
109    that the workspace is expanded using malloc() in this situation. The value
110    below is therefore a minimum, and we put a maximum on it for safety. The
111    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
112    kicks in at the same number of forward references in all cases. */
113    
114    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
115    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
116    
117  #define COMPILE_WORK_SIZE (4096)  /* The overrun tests check for a slightly smaller size so that they detect the
118    overrun before it actually does run off the end of the data block. */
119    
120    #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    #define REQ_MASK       (REQ_CASELESS | REQ_VARY)
127    
128    /* Repeated character flags. */
129    
130    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
131    
132  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
133  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
134  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
135  is invalid. */  is invalid. */
136    
137  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
138    
139    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
140    in UTF-8 mode. */
141    
142  static const short int escapes[] = {  static const short int escapes[] = {
143       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
144       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
145     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
146  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
147  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
148  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
149     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
150  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
151  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
152       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
153         -ESC_D,                  -ESC_E,
154         0,                       -ESC_G,
155         -ESC_H,                  0,
156         0,                       -ESC_K,
157         0,                       0,
158         -ESC_N,                  0,
159         -ESC_P,                  -ESC_Q,
160         -ESC_R,                  -ESC_S,
161         0,                       0,
162         -ESC_V,                  -ESC_W,
163         -ESC_X,                  0,
164         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
165         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
166         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
167         CHAR_GRAVE_ACCENT,       7,
168         -ESC_b,                  0,
169         -ESC_d,                  ESC_e,
170         ESC_f,                   0,
171         -ESC_h,                  0,
172         0,                       -ESC_k,
173         0,                       0,
174         ESC_n,                   0,
175         -ESC_p,                  0,
176         ESC_r,                   -ESC_s,
177         ESC_tee,                 0,
178         -ESC_v,                  -ESC_w,
179         0,                       0,
180         -ESC_z
181  };  };
182    
183  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
184    
185    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
186    
187  static const short int escapes[] = {  static const short int escapes[] = {
188  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
189  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 202  static const short int escapes[] = {
202  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
203  /*  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,
204  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
205  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
206  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
207  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
208  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 212  static const short int escapes[] = {
212  #endif  #endif
213    
214    
215  /* 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
216  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
217  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
218    string is built from string macros so that it works in UTF-8 mode on EBCDIC
219  static const char *const posix_names[] = {  platforms. */
220    "alpha", "lower", "upper",  
221    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
222    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
223      int   op;                  /* Op when no arg, or -1 if arg mandatory */
224      int   op_arg;              /* Op when arg present, or -1 if not allowed */
225    } verbitem;
226    
227    static const char verbnames[] =
228      "\0"                       /* Empty name is a shorthand for MARK */
229      STRING_MARK0
230      STRING_ACCEPT0
231      STRING_COMMIT0
232      STRING_F0
233      STRING_FAIL0
234      STRING_PRUNE0
235      STRING_SKIP0
236      STRING_THEN;
237    
238    static const verbitem verbs[] = {
239      { 0, -1,        OP_MARK },
240      { 4, -1,        OP_MARK },
241      { 6, OP_ACCEPT, -1 },
242      { 6, OP_COMMIT, -1 },
243      { 1, OP_FAIL,   -1 },
244      { 4, OP_FAIL,   -1 },
245      { 5, OP_PRUNE,  OP_PRUNE_ARG },
246      { 4, OP_SKIP,   OP_SKIP_ARG  },
247      { 4, OP_THEN,   OP_THEN_ARG  }
248    };
249    
250    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
251    
252    
253  static const uschar posix_name_lengths[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
254    now all in a single string, to reduce the number of relocations when a shared
255    library is dynamically loaded. The list of lengths is terminated by a zero
256    length entry. The first three must be alpha, lower, upper, as this is assumed
257    for handling case independence. */
258    
259    static const char posix_names[] =
260      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
261      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
262      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
263      STRING_word0  STRING_xdigit;
264    
265    static const pcre_uint8 posix_name_lengths[] = {
266    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 };
267    
268  /* 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 179  static const int posix_class_maps[] = { Line 292  static const int posix_class_maps[] = {
292    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
293  };  };
294    
295    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
296    substitutes must be in the order of the names, defined above, and there are
297    both positive and negative cases. NULL means no substitute. */
298    
299    #ifdef SUPPORT_UCP
300    static const pcre_uchar string_PNd[]  = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pNd[]  = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_PXsp[] = {
307      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_pXsp[] = {
310      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_PXwd[] = {
313      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_pXwd[] = {
316      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    
319    static const pcre_uchar *substitutes[] = {
320      string_PNd,           /* \D */
321      string_pNd,           /* \d */
322      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
323      string_pXsp,          /* \s */
324      string_PXwd,          /* \W */
325      string_pXwd           /* \w */
326    };
327    
328    static const pcre_uchar string_pL[] =   {
329      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pLl[] =  {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLu[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXan[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_h[] =    {
341      CHAR_BACKSLASH, CHAR_h, '\0' };
342    static const pcre_uchar string_pXps[] = {
343      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_PL[] =   {
346      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PLl[] =  {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLu[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PXan[] = {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_H[] =    {
358      CHAR_BACKSLASH, CHAR_H, '\0' };
359    static const pcre_uchar string_PXps[] = {
360      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
361      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
362    
363    static const pcre_uchar *posix_substitutes[] = {
364      string_pL,            /* alpha */
365      string_pLl,           /* lower */
366      string_pLu,           /* upper */
367      string_pXan,          /* alnum */
368      NULL,                 /* ascii */
369      string_h,             /* blank */
370      NULL,                 /* cntrl */
371      string_pNd,           /* digit */
372      NULL,                 /* graph */
373      NULL,                 /* print */
374      NULL,                 /* punct */
375      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
376      string_pXwd,          /* word */
377      NULL,                 /* xdigit */
378      /* Negated cases */
379      string_PL,            /* ^alpha */
380      string_PLl,           /* ^lower */
381      string_PLu,           /* ^upper */
382      string_PXan,          /* ^alnum */
383      NULL,                 /* ^ascii */
384      string_H,             /* ^blank */
385      NULL,                 /* ^cntrl */
386      string_PNd,           /* ^digit */
387      NULL,                 /* ^graph */
388      NULL,                 /* ^print */
389      NULL,                 /* ^punct */
390      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
391      string_PXwd,          /* ^word */
392      NULL                  /* ^xdigit */
393    };
394    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
395    #endif
396    
397  #define STRING(a)  # a  #define STRING(a)  # a
398  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 186  static const int posix_class_maps[] = { Line 400  static const int posix_class_maps[] = {
400  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
401  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
402  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
403  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
404    the number of relocations needed when a shared library is loaded dynamically,
405  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
406    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
407    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
408    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
409    "unrecognized character follows \\",  
410    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
411    substring, so that the whole string ends with \0\0, which can be detected when
412    counting through. */
413    
414    static const char error_texts[] =
415      "no error\0"
416      "\\ at end of pattern\0"
417      "\\c at end of pattern\0"
418      "unrecognized character follows \\\0"
419      "numbers out of order in {} quantifier\0"
420    /* 5 */    /* 5 */
421    "number too big in {} quantifier",    "number too big in {} quantifier\0"
422    "missing terminating ] for character class",    "missing terminating ] for character class\0"
423    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
424    "range out of order in character class",    "range out of order in character class\0"
425    "nothing to repeat",    "nothing to repeat\0"
426    /* 10 */    /* 10 */
427    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
428    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
429    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
430    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
431    "missing )",    "missing )\0"
432    /* 15 */    /* 15 */
433    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
434    "erroffset passed as NULL",    "erroffset passed as NULL\0"
435    "unknown option bit(s) set",    "unknown option bit(s) set\0"
436    "missing ) after comment",    "missing ) after comment\0"
437    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
438    /* 20 */    /* 20 */
439    "regular expression is too large",    "regular expression is too large\0"
440    "failed to get memory",    "failed to get memory\0"
441    "unmatched parentheses",    "unmatched parentheses\0"
442    "internal error: code overflow",    "internal error: code overflow\0"
443    "unrecognized character after (?<",    "unrecognized character after (?<\0"
444    /* 25 */    /* 25 */
445    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
446    "malformed number or name after (?(",    "malformed number or name after (?(\0"
447    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
448    "assertion expected after (?(",    "assertion expected after (?(\0"
449    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
450    /* 30 */    /* 30 */
451    "unknown POSIX class name",    "unknown POSIX class name\0"
452    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
453    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
454    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
455    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
456    /* 35 */    /* 35 */
457    "invalid condition (?(0)",    "invalid condition (?(0)\0"
458    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
459    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
460    "number after (?C is > 255",    "number after (?C is > 255\0"
461    "closing ) for (?C expected",    "closing ) for (?C expected\0"
462    /* 40 */    /* 40 */
463    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
464    "unrecognized character after (?P",    "unrecognized character after (?P\0"
465    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
466    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
467    "invalid UTF-8 string",    "invalid UTF-8 string\0"
468    /* 45 */    /* 45 */
469    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
470    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
471    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
472    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
473    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
474    /* 50 */    /* 50 */
475    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
476    "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"
477    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
478    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
479    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
480    /* 55 */    /* 55 */
481    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
482    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
483    "\\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"
484    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
485  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
486      /* 60 */
487      "(*VERB) not recognized\0"
488      "number is too big\0"
489      "subpattern name expected\0"
490      "digit expected after (?+\0"
491      "] is an invalid data character in JavaScript compatibility mode\0"
492      /* 65 */
493      "different names for subpatterns of the same number are not allowed\0"
494      "(*MARK) must have an argument\0"
495      "this version of PCRE is not compiled with Unicode property support\0"
496      "\\c must be followed by an ASCII character\0"
497      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
498      /* 70 */
499      "internal error: unknown opcode in find_fixedlength()\0"
500      "\\N is not supported in a class\0"
501      "too many forward references\0"
502      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
503      "invalid UTF-16 string\0"
504      /* 75 */
505      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
506      "character value in \\u.... sequence is too large\0"
507      "invalid UTF-32 string\0"
508      ;
509    
510  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
511  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 278  For convenience, we use the same bit def Line 523  For convenience, we use the same bit def
523    
524  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
525    
526  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
527  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
528    into a subtraction and unsigned comparison). */
529    
530    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
531    
532    #ifndef EBCDIC
533    
534    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
535    UTF-8 mode. */
536    
537    static const pcre_uint8 digitab[] =
538    {    {
539    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
540    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 314  static const unsigned char digitab[] = Line 569  static const unsigned char digitab[] =
569    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
570    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
571    
572  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
573  static const unsigned char digitab[] =  
574    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
575    
576    static const pcre_uint8 digitab[] =
577    {    {
578    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
579    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 350  static const unsigned char digitab[] = Line 608  static const unsigned char digitab[] =
608    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
609    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
610    
611  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
612    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
613    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
614    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 386  static const unsigned char ebcdic_charta Line 644  static const unsigned char ebcdic_charta
644  #endif  #endif
645    
646    
647  /* Definition to allow mutual recursion */  
648    
649    /*************************************************
650    *            Find an error text                  *
651    *************************************************/
652    
653    /* The error texts are now all in one long string, to save on relocations. As
654    some of the text is of unknown length, we can't use a table of offsets.
655    Instead, just count through the strings. This is not a performance issue
656    because it happens only when there has been a compilation error.
657    
658    Argument:   the error number
659    Returns:    pointer to the error string
660    */
661    
662    static const char *
663    find_error_text(int n)
664    {
665    const char *s = error_texts;
666    for (; n > 0; n--)
667      {
668      while (*s++ != 0) {};
669      if (*s == 0) return "Error text not found (please report)";
670      }
671    return s;
672    }
673    
674    
675    /*************************************************
676    *           Expand the workspace                 *
677    *************************************************/
678    
679    /* This function is called during the second compiling phase, if the number of
680    forward references fills the existing workspace, which is originally a block on
681    the stack. A larger block is obtained from malloc() unless the ultimate limit
682    has been reached or the increase will be rather small.
683    
684    Argument: pointer to the compile data block
685    Returns:  0 if all went well, else an error number
686    */
687    
688    static int
689    expand_workspace(compile_data *cd)
690    {
691    pcre_uchar *newspace;
692    int newsize = cd->workspace_size * 2;
693    
694    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
695    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
696        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
697     return ERR72;
698    
699    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
700    if (newspace == NULL) return ERR21;
701    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
702    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
703    if (cd->workspace_size > COMPILE_WORK_SIZE)
704      (PUBL(free))((void *)cd->start_workspace);
705    cd->start_workspace = newspace;
706    cd->workspace_size = newsize;
707    return 0;
708    }
709    
710    
711    
712    /*************************************************
713    *            Check for counted repeat            *
714    *************************************************/
715    
716    /* This function is called when a '{' is encountered in a place where it might
717    start a quantifier. It looks ahead to see if it really is a quantifier or not.
718    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
719    where the ddds are digits.
720    
721    Arguments:
722      p         pointer to the first char after '{'
723    
724    Returns:    TRUE or FALSE
725    */
726    
727  static BOOL  static BOOL
728    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  is_counted_repeat(const pcre_uchar *p)
729      int *, int *, branch_chain *, compile_data *, int *);  {
730    if (!IS_DIGIT(*p)) return FALSE;
731    p++;
732    while (IS_DIGIT(*p)) p++;
733    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
734    
735    if (*p++ != CHAR_COMMA) return FALSE;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (!IS_DIGIT(*p)) return FALSE;
739    p++;
740    while (IS_DIGIT(*p)) p++;
741    
742    return (*p == CHAR_RIGHT_CURLY_BRACKET);
743    }
744    
745    
746    
# Line 399  static BOOL Line 749  static BOOL
749  *************************************************/  *************************************************/
750    
751  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
752  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character
753  encodes one of the more complicated things such as \d. A backreference to group  which will be placed in chptr. A backreference to group
754  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  n is returned as ESC_REF + n; ESC_REF is the highest ESC_xxx macro. When
755  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  UTF-8 is enabled, a positive value greater than 255 may be returned in chptr.
756  ptr is pointing at the \. On exit, it is on the final character of the escape  On entry,ptr is pointing at the \. On exit, it is on the final character of the
757  sequence.  escape sequence.
758    
759  Arguments:  Arguments:
760    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
761      chptr          points to the data character
762    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
763    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
764    options        the options bits    options        the options bits
765    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
766    
767  Returns:         zero or positive => a data character  Returns:         zero => a data character
768                   negative => a special escape sequence                   positive => a special escape sequence
769                   on error, errorptr is set                   on error, errorcodeptr is set
770  */  */
771    
772  static int  static int
773  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *chptr, int *errorcodeptr,
774    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
775  {  {
776  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
777  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
778  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
779    pcre_int32 c;
780    int escape = 0;
781    int i;
782    
783  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
784  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 433  ptr--;                            /* Set Line 787  ptr--;                            /* Set
787    
788  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
789    
790  /* 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
791  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.
792  Otherwise further processing may be required. */  Otherwise further processing may be required. */
793    
794  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
795  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
796  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
797    else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = i; else escape = -i; }
798    
799  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
800  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
801  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
802    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = i; else escape = -i; }
803  #endif  #endif
804    
805  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
806    
807  else  else
808    {    {
809    const uschar *oldptr;    const pcre_uchar *oldptr;
810    BOOL braced, negated;    BOOL braced, negated;
811    
812    switch (c)    switch (c)
# Line 458  else Line 814  else
814      /* 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
815      error. */      error. */
816    
817      case 'l':      case CHAR_l:
818      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
819      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
820      break;      break;
821    
822      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
823      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
824      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
825      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
826      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
827      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
828            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
829      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
830      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
831        {          {
832        const uschar *p;          c = 0;
833        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
834          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
835        if (*p != 0 && *p != '}')            register int cc = *(++ptr);
836    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
837              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
838              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
839    #else           /* EBCDIC coding */
840              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
841              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
842    #endif
843              }
844    
845    #if defined COMPILE_PCRE8
846            if (c > (utf ? 0x10ffff : 0xff))
847    #elif defined COMPILE_PCRE16
848            if (c > (utf ? 0x10ffff : 0xffff))
849    #elif defined COMPILE_PCRE32
850            if (utf && c > 0x10ffff)
851    #endif
852              {
853              *errorcodeptr = ERR76;
854              }
855            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
856            }
857          }
858        else
859          *errorcodeptr = ERR37;
860        break;
861    
862        case CHAR_U:
863        /* In JavaScript, \U is an uppercase U letter. */
864        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
865        break;
866    
867        /* In a character class, \g is just a literal "g". Outside a character
868        class, \g must be followed by one of a number of specific things:
869    
870        (1) A number, either plain or braced. If positive, it is an absolute
871        backreference. If negative, it is a relative backreference. This is a Perl
872        5.10 feature.
873    
874        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
875        is part of Perl's movement towards a unified syntax for back references. As
876        this is synonymous with \k{name}, we fudge it up by pretending it really
877        was \k.
878    
879        (3) For Oniguruma compatibility we also support \g followed by a name or a
880        number either in angle brackets or in single quotes. However, these are
881        (possibly recursive) subroutine calls, _not_ backreferences. Just return
882        the ESC_g code (cf \k). */
883    
884        case CHAR_g:
885        if (isclass) break;
886        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
887          {
888          escape = ESC_g;
889          break;
890          }
891    
892        /* Handle the Perl-compatible cases */
893    
894        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
895          {
896          const pcre_uchar *p;
897          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
898            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
899          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
900          {          {
901          c = -ESC_k;          escape = ESC_k;
902          break;          break;
903          }          }
904        braced = TRUE;        braced = TRUE;
# Line 489  else Line 906  else
906        }        }
907      else braced = FALSE;      else braced = FALSE;
908    
909      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
910        {        {
911        negated = TRUE;        negated = TRUE;
912        ptr++;        ptr++;
913        }        }
914      else negated = FALSE;      else negated = FALSE;
915    
916        /* The integer range is limited by the machine's int representation. */
917      c = 0;      c = 0;
918      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
919        c = c * 10 + *(++ptr) - '0';        {
920          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
921            {
922            c = -1;
923            break;
924            }
925          c = c * 10 + *(++ptr) - CHAR_0;
926          }
927        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
928          {
929          while (IS_DIGIT(ptr[1]))
930            ptr++;
931          *errorcodeptr = ERR61;
932          break;
933          }
934    
935      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
936        {        {
937        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
938        return 0;        break;
939          }
940    
941        if (c == 0)
942          {
943          *errorcodeptr = ERR58;
944          break;
945        }        }
946    
947      if (negated)      if (negated)
# Line 511  else Line 949  else
949        if (c > bracount)        if (c > bracount)
950          {          {
951          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
952          return 0;          break;
953          }          }
954        c = bracount - (c - 1);        c = bracount - (c - 1);
955        }        }
956    
957      c = -(ESC_REF + c);      escape = ESC_REF + c;
958      break;      break;
959    
960      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 531  else Line 969  else
969      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
970      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
971    
972      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:
973      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
974    
975      if (!isclass)      if (!isclass)
976        {        {
977        oldptr = ptr;        oldptr = ptr;
978        c -= '0';        /* The integer range is limited by the machine's int representation. */
979        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
980          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
981            {
982            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
983              {
984              c = -1;
985              break;
986              }
987            c = c * 10 + *(++ptr) - CHAR_0;
988            }
989          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
990            {
991            while (IS_DIGIT(ptr[1]))
992              ptr++;
993            *errorcodeptr = ERR61;
994            break;
995            }
996        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
997          {          {
998          c = -(ESC_REF + c);          escape = ESC_REF + c;
999          break;          break;
1000          }          }
1001        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 552  else Line 1005  else
1005      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.
1006      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1007    
1008      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1009        {        {
1010        ptr--;        ptr--;
1011        c = 0;        c = 0;
# Line 562  else Line 1015  else
1015      /* \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
1016      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
1017      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
1018      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,
1019      than 3 octal digits. */      but no more than 3 octal digits. */
1020    
1021      case '0':      case CHAR_0:
1022      c -= '0';      c -= CHAR_0;
1023      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1024          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1025      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1026        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1027    #endif
1028      break;      break;
1029    
1030      /* \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
1031      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.
1032      treated as a data character. */      If not, { is treated as a data character. */
1033    
1034        case CHAR_x:
1035        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1036          {
1037          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1038          Otherwise it is a lowercase x letter. */
1039          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1040            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1041            {
1042            c = 0;
1043            for (i = 0; i < 2; ++i)
1044              {
1045              register int cc = *(++ptr);
1046    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1047              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1048              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1049    #else           /* EBCDIC coding */
1050              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1051              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1052    #endif
1053              }
1054            }
1055          break;
1056          }
1057    
1058      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1059        {        {
1060        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1061    
1062        c = 0;        c = 0;
1063        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1064          {          {
1065          register int cc = *pt++;          register int cc = *pt++;
1066          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1067    
1068  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1069          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1070          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1071  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1072          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1073          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1074    #endif
1075    
1076    #if defined COMPILE_PCRE8
1077            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1078    #elif defined COMPILE_PCRE16
1079            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1080    #elif defined COMPILE_PCRE32
1081            if (utf && c > 0x10ffff) { c = -1; break; }
1082  #endif  #endif
1083          }          }
1084    
1085        if (*pt == '}')        if (c < 0)
1086          {          {
1087          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1088            *errorcodeptr = ERR34;
1089            }
1090    
1091          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1092            {
1093            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1094          ptr = pt;          ptr = pt;
1095          break;          break;
1096          }          }
# Line 612  else Line 1102  else
1102      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1103    
1104      c = 0;      c = 0;
1105      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1106        {        {
1107        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1108        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1109  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1110        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1111        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1112  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1113        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1114        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1115  #endif  #endif
1116        }        }
1117      break;      break;
1118    
1119      /* 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.
1120      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
1121        coding is ASCII-specific, but then the whole concept of \cx is
1122      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1123    
1124      case 'c':      case CHAR_c:
1125      c = *(++ptr);      c = *(++ptr);
1126      if (c == 0)      if (c == 0)
1127        {        {
1128        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1129        return 0;        break;
1130        }        }
1131    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1132  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1133      if (c >= 'a' && c <= 'z') c -= 32;        {
1134          *errorcodeptr = ERR68;
1135          break;
1136          }
1137        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1138      c ^= 0x40;      c ^= 0x40;
1139  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1140      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1141      c ^= 0xC0;      c ^= 0xC0;
1142  #endif  #endif
1143      break;      break;
1144    
1145      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1146      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1147      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
1148      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
1149      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1150    
1151      default:      default:
1152      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 664  else Line 1159  else
1159      }      }
1160    }    }
1161    
1162  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1163  return c;  newline". PCRE does not support \N{name}. However, it does support
1164  }  quantification such as \N{2,3}. */
1165    
1166    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1167         !is_counted_repeat(ptr+2))
1168      *errorcodeptr = ERR37;
1169    
1170    /* If PCRE_UCP is set, we change the values for \d etc. */
1171    
1172    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1173      escape += (ESC_DU - ESC_D);
1174    
1175    /* Set the pointer to the final character before returning. */
1176    
1177    *ptrptr = ptr;
1178    *chptr = c;
1179    return escape;
1180    }
1181    
1182  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1183  /*************************************************  /*************************************************
# Line 690  Returns:         type value from ucp_typ Line 1199  Returns:         type value from ucp_typ
1199  */  */
1200    
1201  static int  static int
1202  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1203  {  {
1204  int c, i, bot, top;  int c, i, bot, top;
1205  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1206  char name[32];  pcre_uchar name[32];
1207    
1208  c = *(++ptr);  c = *(++ptr);
1209  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 704  if (c == 0) goto ERROR_RETURN; Line 1213  if (c == 0) goto ERROR_RETURN;
1213  /* \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
1214  negation. */  negation. */
1215    
1216  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1217    {    {
1218    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1219      {      {
1220      *negptr = TRUE;      *negptr = TRUE;
1221      ptr++;      ptr++;
1222      }      }
1223    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1224      {      {
1225      c = *(++ptr);      c = *(++ptr);
1226      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1227      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1228      name[i] = c;      name[i] = c;
1229      }      }
1230    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1231    name[i] = 0;    name[i] = 0;
1232    }    }
1233    
# Line 735  else Line 1244  else
1244  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1245    
1246  bot = 0;  bot = 0;
1247  top = _pcre_utt_size;  top = PRIV(utt_size);
1248    
1249  while (bot < top)  while (bot < top)
1250    {    {
1251    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1252    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1253    if (c == 0)    if (c == 0)
1254      {      {
1255      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1256      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1257      }      }
1258    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1259    }    }
# Line 764  return -1; Line 1273  return -1;
1273    
1274    
1275  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1276  *         Read repeat counts                     *  *         Read repeat counts                     *
1277  *************************************************/  *************************************************/
1278    
# Line 815  Returns:         pointer to '}' on succe Line 1291  Returns:         pointer to '}' on succe
1291                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1292  */  */
1293    
1294  static const uschar *  static const pcre_uchar *
1295  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)
1296  {  {
1297  int min = 0;  int min = 0;
1298  int max = -1;  int max = -1;
# Line 824  int max = -1; Line 1300  int max = -1;
1300  /* 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
1301  an integer overflow. */  an integer overflow. */
1302    
1303  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1304  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1305    {    {
1306    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 834  if (min < 0 || min > 65535) Line 1310  if (min < 0 || min > 65535)
1310  /* 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.
1311  Also, max must not be less than min. */  Also, max must not be less than min. */
1312    
1313  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1314    {    {
1315    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1316      {      {
1317      max = 0;      max = 0;
1318      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1319      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1320        {        {
1321        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 864  return p; Line 1340  return p;
1340    
1341    
1342  /*************************************************  /*************************************************
1343  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1344  *************************************************/  *************************************************/
1345    
1346  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1347    top-level call starts at the beginning of the pattern. All other calls must
1348    start at a parenthesis. It scans along a pattern's text looking for capturing
1349  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
1350  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
1351  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
1352  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1353  be terminated by '>' because that is checked in the first pass.  
1354    This function was originally called only from the second pass, in which we know
1355    that if (?< or (?' or (?P< is encountered, the name will be correctly
1356    terminated because that is checked in the first pass. There is now one call to
1357    this function in the first pass, to check for a recursive back reference by
1358    name (so that we can make the whole group atomic). In this case, we need check
1359    only up to the current position in the pattern, and that is still OK because
1360    and previous occurrences will have been checked. To make this work, the test
1361    for "end of pattern" is a check against cd->end_pattern in the main loop,
1362    instead of looking for a binary zero. This means that the special first-pass
1363    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1364    processing items within the loop are OK, because afterwards the main loop will
1365    terminate.)
1366    
1367  Arguments:  Arguments:
1368    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1369    count        current count of capturing parens so far encountered    cd           compile background data
1370    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1371    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1372    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1373      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1374      count        pointer to the current capturing subpattern number (updated)
1375    
1376  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1377  */  */
1378    
1379  static int  static int
1380  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,
1381    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1382  {  {
1383  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1384    int start_count = *count;
1385    int hwm_count = start_count;
1386    BOOL dup_parens = FALSE;
1387    
1388    /* If the first character is a parenthesis, check on the type of group we are
1389    dealing with. The very first call may not start with a parenthesis. */
1390    
1391  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1392    {    {
1393    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1394    
1395      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1396    
1397      /* Handle a normal, unnamed capturing parenthesis. */
1398    
1399      else if (ptr[1] != CHAR_QUESTION_MARK)
1400        {
1401        *count += 1;
1402        if (name == NULL && *count == lorn) return *count;
1403        ptr++;
1404        }
1405    
1406      /* All cases now have (? at the start. Remember when we are in a group
1407      where the parenthesis numbers are duplicated. */
1408    
1409      else if (ptr[2] == CHAR_VERTICAL_LINE)
1410        {
1411        ptr += 3;
1412        dup_parens = TRUE;
1413        }
1414    
1415      /* Handle comments; all characters are allowed until a ket is reached. */
1416    
1417      else if (ptr[2] == CHAR_NUMBER_SIGN)
1418        {
1419        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1420        goto FAIL_EXIT;
1421        }
1422    
1423      /* Handle a condition. If it is an assertion, just carry on so that it
1424      is processed as normal. If not, skip to the closing parenthesis of the
1425      condition (there can't be any nested parens). */
1426    
1427      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1428        {
1429        ptr += 2;
1430        if (ptr[1] != CHAR_QUESTION_MARK)
1431          {
1432          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1433          if (*ptr != 0) ptr++;
1434          }
1435        }
1436    
1437      /* Start with (? but not a condition. */
1438    
1439      else
1440        {
1441        ptr += 2;
1442        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1443    
1444        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1445    
1446        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1447            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1448          {
1449          int term;
1450          const pcre_uchar *thisname;
1451          *count += 1;
1452          if (name == NULL && *count == lorn) return *count;
1453          term = *ptr++;
1454          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1455          thisname = ptr;
1456          while (*ptr != term) ptr++;
1457          if (name != NULL && lorn == ptr - thisname &&
1458              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1459            return *count;
1460          term++;
1461          }
1462        }
1463      }
1464    
1465    /* Past any initial parenthesis handling, scan for parentheses or vertical
1466    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1467    first-pass call when this value is temporarily adjusted to stop at the current
1468    position. So DO NOT change this to a test for binary zero. */
1469    
1470    for (; ptr < cd->end_pattern; ptr++)
1471      {
1472    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1473    
1474    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1475      {      {
1476      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1477      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1478        {        {
1479        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1480        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1481        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1482        }        }
1483      continue;      continue;
1484      }      }
1485    
1486    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1487      are handled for real. If the first character is '^', skip it. Also, if the
1488      first few characters (either before or after ^) are \Q\E or \E we skip them
1489      too. This makes for compatibility with Perl. Note the use of STR macros to
1490      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1491    
1492    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1493      {      {
1494      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1495        for (;;)
1496          {
1497          if (ptr[1] == CHAR_BACKSLASH)
1498            {
1499            if (ptr[2] == CHAR_E)
1500              ptr+= 2;
1501            else if (STRNCMP_UC_C8(ptr + 2,
1502                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1503              ptr += 4;
1504            else
1505              break;
1506            }
1507          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1508            {
1509            negate_class = TRUE;
1510            ptr++;
1511            }
1512          else break;
1513          }
1514    
1515        /* If the next character is ']', it is a data character that must be
1516        skipped, except in JavaScript compatibility mode. */
1517    
1518        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1519            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1520          ptr++;
1521    
1522        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1523        {        {
1524        if (*ptr == '\\')        if (*ptr == 0) return -1;
1525          if (*ptr == CHAR_BACKSLASH)
1526          {          {
1527          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1528          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1529            {            {
1530            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1531            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1532            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1533            }            }
1534          continue;          continue;
1535          }          }
# Line 931  for (; *ptr != 0; ptr++) Line 1539  for (; *ptr != 0; ptr++)
1539    
1540    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1541    
1542    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1543      {      {
1544      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1545      if (*ptr == 0) return -1;      while (*ptr != 0)
1546          {
1547          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1548          ptr++;
1549    #ifdef SUPPORT_UTF
1550          if (utf) FORWARDCHAR(ptr);
1551    #endif
1552          }
1553        if (*ptr == 0) goto FAIL_EXIT;
1554      continue;      continue;
1555      }      }
1556    
1557    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1558    
1559    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1560      {      {
1561      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1562      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1563      continue;      if (*ptr == 0) goto FAIL_EXIT;
1564      }      }
1565    
1566    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1567    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1568        if (dup_parens && *count < hwm_count) *count = hwm_count;
1569    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1570        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1571    
1572    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1573    term = *ptr++;      {
1574    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1575    thisname = ptr;      *count = start_count;
1576    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1577    }    }
1578    
1579    FAIL_EXIT:
1580    *ptrptr = ptr;
1581  return -1;  return -1;
1582  }  }
1583    
1584    
1585    
1586    
1587  /*************************************************  /*************************************************
1588  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1589  *************************************************/  *************************************************/
1590    
1591  /* This is called by several functions that scan a compiled expression looking  /* This function scans along a pattern's text looking for capturing
1592  for a fixed first character, or an anchoring op code etc. It skips over things  subpatterns, and counting them. If it finds a named pattern that matches the
1593  that do not influence this. For some calls, a change of option is important.  name it is given, it returns its number. Alternatively, if the name is NULL, it
1594  For some calls, it makes sense to skip negative forward and all backward  returns when it reaches a given numbered subpattern. This is used for forward
1595  assertions, and also the \b assertion; for others it does not.  references to subpatterns. We used to be able to start this scan from the
1596    current compiling point, using the current count value from cd->bracount, and
1597    do it all in a single loop, but the addition of the possibility of duplicate
1598    subpattern numbers means that we have to scan from the very start, in order to
1599    take account of such duplicates, and to use a recursive function to keep track
1600    of the different types of group.
1601    
1602    Arguments:
1603      cd           compile background data
1604      name         name to seek, or NULL if seeking a numbered subpattern
1605      lorn         name length, or subpattern number if name is NULL
1606      xmode        TRUE if we are in /x mode
1607      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1608    
1609    Returns:       the number of the found subpattern, or -1 if not found
1610    */
1611    
1612    static int
1613    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1614      BOOL utf)
1615    {
1616    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1617    int count = 0;
1618    int rc;
1619    
1620    /* If the pattern does not start with an opening parenthesis, the first call
1621    to find_parens_sub() will scan right to the end (if necessary). However, if it
1622    does start with a parenthesis, find_parens_sub() will return when it hits the
1623    matching closing parens. That is why we have to have a loop. */
1624    
1625    for (;;)
1626      {
1627      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1628      if (rc > 0 || *ptr++ == 0) break;
1629      }
1630    
1631    return rc;
1632    }
1633    
1634    
1635    
1636    
1637    /*************************************************
1638    *      Find first significant op code            *
1639    *************************************************/
1640    
1641    /* This is called by several functions that scan a compiled expression looking
1642    for a fixed first character, or an anchoring op code etc. It skips over things
1643    that do not influence this. For some calls, it makes sense to skip negative
1644    forward and all backward assertions, and also the \b assertion; for others it
1645    does not.
1646    
1647  Arguments:  Arguments:
1648    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  
1649    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1650    
1651  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1652  */  */
1653    
1654  static const uschar*  static const pcre_uchar*
1655  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1656  {  {
1657  for (;;)  for (;;)
1658    {    {
1659    switch ((int)*code)    switch ((int)*code)
1660      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1661      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1662      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1663      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1664      if (!skipassert) return code;      if (!skipassert) return code;
1665      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1666      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1667      break;      break;
1668    
1669      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1023  for (;;) Line 1673  for (;;)
1673    
1674      case OP_CALLOUT:      case OP_CALLOUT:
1675      case OP_CREF:      case OP_CREF:
1676        case OP_NCREF:
1677      case OP_RREF:      case OP_RREF:
1678        case OP_NRREF:
1679      case OP_DEF:      case OP_DEF:
1680      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1681      break;      break;
1682    
1683      default:      default:
# Line 1039  for (;;) Line 1691  for (;;)
1691    
1692    
1693  /*************************************************  /*************************************************
1694  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1695  *************************************************/  *************************************************/
1696    
1697  /* 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,
1698  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.
1699  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
1700    temporarily terminated with OP_END when this function is called.
1701    
1702    This function is called when a backward assertion is encountered, so that if it
1703    fails, the error message can point to the correct place in the pattern.
1704    However, we cannot do this when the assertion contains subroutine calls,
1705    because they can be forward references. We solve this by remembering this case
1706    and doing the check at the end; a flag specifies which mode we are running in.
1707    
1708  Arguments:  Arguments:
1709    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1710    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1711      atend    TRUE if called when the pattern is complete
1712  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1713               or -2 if \C was encountered  
1714    Returns:   the fixed length,
1715                 or -1 if there is no fixed length,
1716                 or -2 if \C was encountered (in UTF-8 mode only)
1717                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1718                 or -4 if an unknown opcode was encountered (internal error)
1719  */  */
1720    
1721  static int  static int
1722  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1723  {  {
1724  int length = -1;  int length = -1;
1725    
1726  register int branchlength = 0;  register int branchlength = 0;
1727  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1728    
1729  /* 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
1730  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1068  branch, check the length against that of Line 1732  branch, check the length against that of
1732  for (;;)  for (;;)
1733    {    {
1734    int d;    int d;
1735      pcre_uchar *ce, *cs;
1736    register int op = *cc;    register int op = *cc;
1737    
1738    switch (op)    switch (op)
1739      {      {
1740        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1741        OP_BRA (normal non-capturing bracket) because the other variants of these
1742        opcodes are all concerned with unlimited repeated groups, which of course
1743        are not of fixed length. */
1744    
1745      case OP_CBRA:      case OP_CBRA:
1746      case OP_BRA:      case OP_BRA:
1747      case OP_ONCE:      case OP_ONCE:
1748        case OP_ONCE_NC:
1749      case OP_COND:      case OP_COND:
1750      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1751      if (d < 0) return d;      if (d < 0) return d;
1752      branchlength += d;      branchlength += d;
1753      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1754      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1755      break;      break;
1756    
1757      /* 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.
1758      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
1759      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
1760        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1761        because they all imply an unlimited repeat. */
1762    
1763      case OP_ALT:      case OP_ALT:
1764      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1765      case OP_END:      case OP_END:
1766        case OP_ACCEPT:
1767        case OP_ASSERT_ACCEPT:
1768      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1769        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1770      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1099  for (;;) Line 1772  for (;;)
1772      branchlength = 0;      branchlength = 0;
1773      break;      break;
1774    
1775        /* A true recursion implies not fixed length, but a subroutine call may
1776        be OK. If the subroutine is a forward reference, we can't deal with
1777        it until the end of the pattern, so return -3. */
1778    
1779        case OP_RECURSE:
1780        if (!atend) return -3;
1781        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1782        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1783        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1784        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1785        if (d < 0) return d;
1786        branchlength += d;
1787        cc += 1 + LINK_SIZE;
1788        break;
1789    
1790      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1791    
1792      case OP_ASSERT:      case OP_ASSERT:
# Line 1106  for (;;) Line 1794  for (;;)
1794      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1795      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1796      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1797      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1798        break;
1799    
1800      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1801    
1802      case OP_REVERSE:      case OP_MARK:
1803        case OP_PRUNE_ARG:
1804        case OP_SKIP_ARG:
1805        case OP_THEN_ARG:
1806        cc += cc[1] + PRIV(OP_lengths)[*cc];
1807        break;
1808    
1809        case OP_CALLOUT:
1810        case OP_CIRC:
1811        case OP_CIRCM:
1812        case OP_CLOSE:
1813        case OP_COMMIT:
1814      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1815      case OP_DEF:      case OP_DEF:
1816      case OP_OPT:      case OP_DOLL:
1817      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1818      case OP_EOD:      case OP_EOD:
1819      case OP_EODN:      case OP_EODN:
1820      case OP_CIRC:      case OP_FAIL:
1821      case OP_DOLL:      case OP_NCREF:
1822        case OP_NRREF:
1823      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1824        case OP_PRUNE:
1825        case OP_REVERSE:
1826        case OP_RREF:
1827        case OP_SET_SOM:
1828        case OP_SKIP:
1829        case OP_SOD:
1830        case OP_SOM:
1831        case OP_THEN:
1832      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1833      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1834      break;      break;
1835    
1836      /* Handle literal characters */      /* Handle literal characters */
1837    
1838      case OP_CHAR:      case OP_CHAR:
1839      case OP_CHARNC:      case OP_CHARI:
1840      case OP_NOT:      case OP_NOT:
1841        case OP_NOTI:
1842      branchlength++;      branchlength++;
1843      cc += 2;      cc += 2;
1844  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1845      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1846  #endif  #endif
1847      break;      break;
1848    
# Line 1146  for (;;) Line 1850  for (;;)
1850      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1851    
1852      case OP_EXACT:      case OP_EXACT:
1853        case OP_EXACTI:
1854        case OP_NOTEXACT:
1855        case OP_NOTEXACTI:
1856      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1857      cc += 4;      cc += 2 + IMM2_SIZE;
1858  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1859      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1860  #endif  #endif
1861      break;      break;
1862    
1863      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1864      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1865      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1866          cc += 2;
1867        cc += 1 + IMM2_SIZE + 1;
1868      break;      break;
1869    
1870      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1168  for (;;) Line 1874  for (;;)
1874      cc += 2;      cc += 2;
1875      /* Fall through */      /* Fall through */
1876    
1877        case OP_HSPACE:
1878        case OP_VSPACE:
1879        case OP_NOT_HSPACE:
1880        case OP_NOT_VSPACE:
1881      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1882      case OP_DIGIT:      case OP_DIGIT:
1883      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1175  for (;;) Line 1885  for (;;)
1885      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1886      case OP_WORDCHAR:      case OP_WORDCHAR:
1887      case OP_ANY:      case OP_ANY:
1888        case OP_ALLANY:
1889      branchlength++;      branchlength++;
1890      cc++;      cc++;
1891      break;      break;
1892    
1893      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1894        otherwise \C is coded as OP_ALLANY. */
1895    
1896      case OP_ANYBYTE:      case OP_ANYBYTE:
1897      return -2;      return -2;
1898    
1899      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1900    
1901  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1902      case OP_XCLASS:      case OP_XCLASS:
1903      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1904      /* Fall through */      /* Fall through */
1905  #endif  #endif
1906    
1907      case OP_CLASS:      case OP_CLASS:
1908      case OP_NCLASS:      case OP_NCLASS:
1909      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1910    
1911      switch (*cc)      switch (*cc)
1912        {        {
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRSTAR:        case OP_CRSTAR:
1916        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1917        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1206  for (;;) Line 1920  for (;;)
1920    
1921        case OP_CRRANGE:        case OP_CRRANGE:
1922        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1923        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1924        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1925        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1926        break;        break;
1927    
1928        default:        default:
# Line 1218  for (;;) Line 1932  for (;;)
1932    
1933      /* Anything else is variable length */      /* Anything else is variable length */
1934    
1935      default:      case OP_ANYNL:
1936        case OP_BRAMINZERO:
1937        case OP_BRAPOS:
1938        case OP_BRAPOSZERO:
1939        case OP_BRAZERO:
1940        case OP_CBRAPOS:
1941        case OP_EXTUNI:
1942        case OP_KETRMAX:
1943        case OP_KETRMIN:
1944        case OP_KETRPOS:
1945        case OP_MINPLUS:
1946        case OP_MINPLUSI:
1947        case OP_MINQUERY:
1948        case OP_MINQUERYI:
1949        case OP_MINSTAR:
1950        case OP_MINSTARI:
1951        case OP_MINUPTO:
1952        case OP_MINUPTOI:
1953        case OP_NOTMINPLUS:
1954        case OP_NOTMINPLUSI:
1955        case OP_NOTMINQUERY:
1956        case OP_NOTMINQUERYI:
1957        case OP_NOTMINSTAR:
1958        case OP_NOTMINSTARI:
1959        case OP_NOTMINUPTO:
1960        case OP_NOTMINUPTOI:
1961        case OP_NOTPLUS:
1962        case OP_NOTPLUSI:
1963        case OP_NOTPOSPLUS:
1964        case OP_NOTPOSPLUSI:
1965        case OP_NOTPOSQUERY:
1966        case OP_NOTPOSQUERYI:
1967        case OP_NOTPOSSTAR:
1968        case OP_NOTPOSSTARI:
1969        case OP_NOTPOSUPTO:
1970        case OP_NOTPOSUPTOI:
1971        case OP_NOTQUERY:
1972        case OP_NOTQUERYI:
1973        case OP_NOTSTAR:
1974        case OP_NOTSTARI:
1975        case OP_NOTUPTO:
1976        case OP_NOTUPTOI:
1977        case OP_PLUS:
1978        case OP_PLUSI:
1979        case OP_POSPLUS:
1980        case OP_POSPLUSI:
1981        case OP_POSQUERY:
1982        case OP_POSQUERYI:
1983        case OP_POSSTAR:
1984        case OP_POSSTARI:
1985        case OP_POSUPTO:
1986        case OP_POSUPTOI:
1987        case OP_QUERY:
1988        case OP_QUERYI:
1989        case OP_REF:
1990        case OP_REFI:
1991        case OP_SBRA:
1992        case OP_SBRAPOS:
1993        case OP_SCBRA:
1994        case OP_SCBRAPOS:
1995        case OP_SCOND:
1996        case OP_SKIPZERO:
1997        case OP_STAR:
1998        case OP_STARI:
1999        case OP_TYPEMINPLUS:
2000        case OP_TYPEMINQUERY:
2001        case OP_TYPEMINSTAR:
2002        case OP_TYPEMINUPTO:
2003        case OP_TYPEPLUS:
2004        case OP_TYPEPOSPLUS:
2005        case OP_TYPEPOSQUERY:
2006        case OP_TYPEPOSSTAR:
2007        case OP_TYPEPOSUPTO:
2008        case OP_TYPEQUERY:
2009        case OP_TYPESTAR:
2010        case OP_TYPEUPTO:
2011        case OP_UPTO:
2012        case OP_UPTOI:
2013      return -1;      return -1;
2014    
2015        /* Catch unrecognized opcodes so that when new ones are added they
2016        are not forgotten, as has happened in the past. */
2017    
2018        default:
2019        return -4;
2020      }      }
2021    }    }
2022  /* Control never gets here */  /* Control never gets here */
# Line 1229  for (;;) Line 2026  for (;;)
2026    
2027    
2028  /*************************************************  /*************************************************
2029  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2030  *************************************************/  *************************************************/
2031    
2032  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2033  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2034    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2035    so that it can be called from pcre_study() when finding the minimum matching
2036    length.
2037    
2038  Arguments:  Arguments:
2039    code        points to start of expression    code        points to start of expression
2040    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2041    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2042    
2043  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
2044  */  */
2045    
2046  static const uschar *  const pcre_uchar *
2047  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2048  {  {
2049  for (;;)  for (;;)
2050    {    {
2051    register int c = *code;    register int c = *code;
2052    
2053    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2054    
2055    /* 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 1257  for (;;) Line 2058  for (;;)
2058    
2059    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2060    
2061      /* Handle recursion */
2062    
2063      else if (c == OP_REVERSE)
2064        {
2065        if (number < 0) return (pcre_uchar *)code;
2066        code += PRIV(OP_lengths)[c];
2067        }
2068    
2069    /* Handle capturing bracket */    /* Handle capturing bracket */
2070    
2071    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2072               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2073      {      {
2074      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2075      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2076      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2077      }      }
2078    
2079      /* Otherwise, we can get the item's length from the table, except that for
2080      repeated character types, we have to test for \p and \P, which have an extra
2081      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2082      must add in its length. */
2083    
2084      else
2085        {
2086        switch(c)
2087          {
2088          case OP_TYPESTAR:
2089          case OP_TYPEMINSTAR:
2090          case OP_TYPEPLUS:
2091          case OP_TYPEMINPLUS:
2092          case OP_TYPEQUERY:
2093          case OP_TYPEMINQUERY:
2094          case OP_TYPEPOSSTAR:
2095          case OP_TYPEPOSPLUS:
2096          case OP_TYPEPOSQUERY:
2097          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2098          break;
2099    
2100          case OP_TYPEUPTO:
2101          case OP_TYPEMINUPTO:
2102          case OP_TYPEEXACT:
2103          case OP_TYPEPOSUPTO:
2104          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2105            code += 2;
2106          break;
2107    
2108          case OP_MARK:
2109          case OP_PRUNE_ARG:
2110          case OP_SKIP_ARG:
2111          code += code[1];
2112          break;
2113    
2114          case OP_THEN_ARG:
2115          code += code[1];
2116          break;
2117          }
2118    
2119        /* Add in the fixed length from the table */
2120    
2121        code += PRIV(OP_lengths)[c];
2122    
2123    /* 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
2124    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
2125    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2126    
2127    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2128      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2129        {        {
2130        case OP_CHAR:        case OP_CHAR:
2131        case OP_CHARNC:        case OP_CHARI:
2132        case OP_EXACT:        case OP_EXACT:
2133          case OP_EXACTI:
2134        case OP_UPTO:        case OP_UPTO:
2135          case OP_UPTOI:
2136        case OP_MINUPTO:        case OP_MINUPTO:
2137          case OP_MINUPTOI:
2138        case OP_POSUPTO:        case OP_POSUPTO:
2139          case OP_POSUPTOI:
2140        case OP_STAR:        case OP_STAR:
2141          case OP_STARI:
2142        case OP_MINSTAR:        case OP_MINSTAR:
2143          case OP_MINSTARI:
2144        case OP_POSSTAR:        case OP_POSSTAR:
2145          case OP_POSSTARI:
2146        case OP_PLUS:        case OP_PLUS:
2147          case OP_PLUSI:
2148        case OP_MINPLUS:        case OP_MINPLUS:
2149          case OP_MINPLUSI:
2150        case OP_POSPLUS:        case OP_POSPLUS:
2151          case OP_POSPLUSI:
2152        case OP_QUERY:        case OP_QUERY:
2153          case OP_QUERYI:
2154        case OP_MINQUERY:        case OP_MINQUERY:
2155          case OP_MINQUERYI:
2156        case OP_POSQUERY:        case OP_POSQUERY:
2157        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2158          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2159        break;        break;
2160        }        }
2161    #else
2162        (void)(utf);  /* Keep compiler happy by referencing function argument */
2163  #endif  #endif
2164      }      }
2165    }    }
# Line 1310  instance of OP_RECURSE. Line 2176  instance of OP_RECURSE.
2176    
2177  Arguments:  Arguments:
2178    code        points to start of expression    code        points to start of expression
2179    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2180    
2181  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
2182  */  */
2183    
2184  static const uschar *  static const pcre_uchar *
2185  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2186  {  {
2187  for (;;)  for (;;)
2188    {    {
# Line 1330  for (;;) Line 2196  for (;;)
2196    
2197    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2198    
2199    /* 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
2200    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
2201    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
2202    bytes. */    must add in its length. */
2203    
2204    else    else
2205      {      {
2206      code += _pcre_OP_lengths[c];      switch(c)
2207  #ifdef SUPPORT_UTF8        {
2208      if (utf8) switch(c)        case OP_TYPESTAR:
2209          case OP_TYPEMINSTAR:
2210          case OP_TYPEPLUS:
2211          case OP_TYPEMINPLUS:
2212          case OP_TYPEQUERY:
2213          case OP_TYPEMINQUERY:
2214          case OP_TYPEPOSSTAR:
2215          case OP_TYPEPOSPLUS:
2216          case OP_TYPEPOSQUERY:
2217          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2218          break;
2219    
2220          case OP_TYPEPOSUPTO:
2221          case OP_TYPEUPTO:
2222          case OP_TYPEMINUPTO:
2223          case OP_TYPEEXACT:
2224          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2225            code += 2;
2226          break;
2227    
2228          case OP_MARK:
2229          case OP_PRUNE_ARG:
2230          case OP_SKIP_ARG:
2231          code += code[1];
2232          break;
2233    
2234          case OP_THEN_ARG:
2235          code += code[1];
2236          break;
2237          }
2238    
2239        /* Add in the fixed length from the table */
2240    
2241        code += PRIV(OP_lengths)[c];
2242    
2243        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2244        by a multi-byte character. The length in the table is a minimum, so we have
2245        to arrange to skip the extra bytes. */
2246    
2247    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2248        if (utf) switch(c)
2249        {        {
2250        case OP_CHAR:        case OP_CHAR:
2251        case OP_CHARNC:        case OP_CHARI:
2252          case OP_NOT:
2253          case OP_NOTI:
2254        case OP_EXACT:        case OP_EXACT:
2255          case OP_EXACTI:
2256          case OP_NOTEXACT:
2257          case OP_NOTEXACTI:
2258        case OP_UPTO:        case OP_UPTO:
2259          case OP_UPTOI:
2260          case OP_NOTUPTO:
2261          case OP_NOTUPTOI:
2262        case OP_MINUPTO:        case OP_MINUPTO:
2263          case OP_MINUPTOI:
2264          case OP_NOTMINUPTO:
2265          case OP_NOTMINUPTOI:
2266        case OP_POSUPTO:        case OP_POSUPTO:
2267          case OP_POSUPTOI:
2268          case OP_NOTPOSUPTO:
2269          case OP_NOTPOSUPTOI:
2270        case OP_STAR:        case OP_STAR:
2271          case OP_STARI:
2272          case OP_NOTSTAR:
2273          case OP_NOTSTARI:
2274        case OP_MINSTAR:        case OP_MINSTAR:
2275          case OP_MINSTARI:
2276          case OP_NOTMINSTAR:
2277          case OP_NOTMINSTARI:
2278        case OP_POSSTAR:        case OP_POSSTAR:
2279          case OP_POSSTARI:
2280          case OP_NOTPOSSTAR:
2281          case OP_NOTPOSSTARI:
2282        case OP_PLUS:        case OP_PLUS:
2283          case OP_PLUSI:
2284          case OP_NOTPLUS:
2285          case OP_NOTPLUSI:
2286        case OP_MINPLUS:        case OP_MINPLUS:
2287          case OP_MINPLUSI:
2288          case OP_NOTMINPLUS:
2289          case OP_NOTMINPLUSI:
2290        case OP_POSPLUS:        case OP_POSPLUS:
2291          case OP_POSPLUSI:
2292          case OP_NOTPOSPLUS:
2293          case OP_NOTPOSPLUSI:
2294        case OP_QUERY:        case OP_QUERY:
2295          case OP_QUERYI:
2296          case OP_NOTQUERY:
2297          case OP_NOTQUERYI:
2298        case OP_MINQUERY:        case OP_MINQUERY:
2299          case OP_MINQUERYI:
2300          case OP_NOTMINQUERY:
2301          case OP_NOTMINQUERYI:
2302        case OP_POSQUERY:        case OP_POSQUERY:
2303        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2304          case OP_NOTPOSQUERY:
2305          case OP_NOTPOSQUERYI:
2306          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2307        break;        break;
2308        }        }
2309    #else
2310        (void)(utf);  /* Keep compiler happy by referencing function argument */
2311  #endif  #endif
2312      }      }
2313    }    }
# Line 1374  for (;;) Line 2323  for (;;)
2323  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()
2324  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
2325  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2326  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
2327  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
2328    bracket whose current branch will already have been scanned.
2329    
2330  Arguments:  Arguments:
2331    code        points to start of search    code        points to start of search
2332    endcode     points to where to stop    endcode     points to where to stop
2333    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2334      cd          contains pointers to tables etc.
2335    
2336  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2337  */  */
2338    
2339  static BOOL  static BOOL
2340  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2341      BOOL utf, compile_data *cd)
2342  {  {
2343  register int c;  register int c;
2344  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2345       code < endcode;       code < endcode;
2346       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2347    {    {
2348    const uschar *ccode;    const pcre_uchar *ccode;
2349    
2350    c = *code;    c = *code;
2351    
2352    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2353      first_significant_code() with a TRUE final argument. */
2354    
2355    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2356      {      {
     code += _pcre_OP_lengths[c];  
2357      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2358      c = *code;      c = *code;
2359      continue;      continue;
2360      }      }
2361    
2362    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2363      implies a backward reference subroutine call, we can scan it. If it's a
2364      forward reference subroutine call, we can't. To detect forward reference
2365      we have to scan up the list that is kept in the workspace. This function is
2366      called only when doing the real compile, not during the pre-compile that
2367      measures the size of the compiled pattern. */
2368    
2369    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2370      {      {
2371        const pcre_uchar *scode;
2372      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2373    
2374      /* Scan a closed bracket */      /* Test for forward reference */
2375    
2376        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2377          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2378    
2379        /* Not a forward reference, test for completed backward reference */
2380    
2381      empty_branch = FALSE;      empty_branch = FALSE;
2382        scode = cd->start_code + GET(code, 1);
2383        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2384    
2385        /* Completed backwards reference */
2386    
2387      do      do
2388        {        {
2389        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2390            {
2391          empty_branch = TRUE;          empty_branch = TRUE;
2392            break;
2393            }
2394          scode += GET(scode, 1);
2395          }
2396        while (*scode == OP_ALT);
2397    
2398        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2399        continue;
2400        }
2401    
2402      /* Groups with zero repeats can of course be empty; skip them. */
2403    
2404      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2405          c == OP_BRAPOSZERO)
2406        {
2407        code += PRIV(OP_lengths)[c];
2408        do code += GET(code, 1); while (*code == OP_ALT);
2409        c = *code;
2410        continue;
2411        }
2412    
2413      /* A nested group that is already marked as "could be empty" can just be
2414      skipped. */
2415    
2416      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2417          c == OP_SCBRA || c == OP_SCBRAPOS)
2418        {
2419        do code += GET(code, 1); while (*code == OP_ALT);
2420        c = *code;
2421        continue;
2422        }
2423    
2424      /* For other groups, scan the branches. */
2425    
2426      if (c == OP_BRA  || c == OP_BRAPOS ||
2427          c == OP_CBRA || c == OP_CBRAPOS ||
2428          c == OP_ONCE || c == OP_ONCE_NC ||
2429          c == OP_COND)
2430        {
2431        BOOL empty_branch;
2432        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2433    
2434        /* If a conditional group has only one branch, there is a second, implied,
2435        empty branch, so just skip over the conditional, because it could be empty.
2436        Otherwise, scan the individual branches of the group. */
2437    
2438        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2439        code += GET(code, 1);        code += GET(code, 1);
2440        else
2441          {
2442          empty_branch = FALSE;
2443          do
2444            {
2445            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2446              empty_branch = TRUE;
2447            code += GET(code, 1);
2448            }
2449          while (*code == OP_ALT);
2450          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2451        }        }
2452      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2453      c = *code;      c = *code;
2454      continue;      continue;
2455      }      }
# Line 1433  for (code = first_significant_code(code Line 2458  for (code = first_significant_code(code
2458    
2459    switch (c)    switch (c)
2460      {      {
2461      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2462        cannot be represented just by a bit map. This includes negated single
2463        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2464        actual length is stored in the compiled code, so we must update "code"
2465        here. */
2466    
2467  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2468      case OP_XCLASS:      case OP_XCLASS:
2469      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2470      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2471  #endif  #endif
2472    
2473      case OP_CLASS:      case OP_CLASS:
2474      case OP_NCLASS:      case OP_NCLASS:
2475      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2476    
2477  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2478      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2479  #endif  #endif
2480    
# Line 1481  for (code = first_significant_code(code Line 2510  for (code = first_significant_code(code
2510      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2511      case OP_WORDCHAR:      case OP_WORDCHAR:
2512      case OP_ANY:      case OP_ANY:
2513        case OP_ALLANY:
2514      case OP_ANYBYTE:      case OP_ANYBYTE:
2515      case OP_CHAR:      case OP_CHAR:
2516      case OP_CHARNC:      case OP_CHARI:
2517      case OP_NOT:      case OP_NOT:
2518        case OP_NOTI:
2519      case OP_PLUS:      case OP_PLUS:
2520      case OP_MINPLUS:      case OP_MINPLUS:
2521      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1499  for (code = first_significant_code(code Line 2530  for (code = first_significant_code(code
2530      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2531      return FALSE;      return FALSE;
2532    
2533        /* These are going to continue, as they may be empty, but we have to
2534        fudge the length for the \p and \P cases. */
2535    
2536        case OP_TYPESTAR:
2537        case OP_TYPEMINSTAR:
2538        case OP_TYPEPOSSTAR:
2539        case OP_TYPEQUERY:
2540        case OP_TYPEMINQUERY:
2541        case OP_TYPEPOSQUERY:
2542        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2543        break;
2544    
2545        /* Same for these */
2546    
2547        case OP_TYPEUPTO:
2548        case OP_TYPEMINUPTO:
2549        case OP_TYPEPOSUPTO:
2550        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2551          code += 2;
2552        break;
2553    
2554      /* End of branch */      /* End of branch */
2555    
2556      case OP_KET:      case OP_KET:
2557      case OP_KETRMAX:      case OP_KETRMAX:
2558      case OP_KETRMIN:      case OP_KETRMIN:
2559        case OP_KETRPOS:
2560      case OP_ALT:      case OP_ALT:
2561      return TRUE;      return TRUE;
2562    
2563      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2564      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2565    
2566  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2567      case OP_STAR:      case OP_STAR:
2568        case OP_STARI:
2569      case OP_MINSTAR:      case OP_MINSTAR:
2570        case OP_MINSTARI:
2571      case OP_POSSTAR:      case OP_POSSTAR:
2572        case OP_POSSTARI:
2573      case OP_QUERY:      case OP_QUERY:
2574        case OP_QUERYI:
2575      case OP_MINQUERY:      case OP_MINQUERY:
2576        case OP_MINQUERYI:
2577      case OP_POSQUERY:      case OP_POSQUERY:
2578        case OP_POSQUERYI:
2579        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2580        break;
2581    
2582      case OP_UPTO:      case OP_UPTO:
2583        case OP_UPTOI:
2584      case OP_MINUPTO:      case OP_MINUPTO:
2585        case OP_MINUPTOI:
2586      case OP_POSUPTO:      case OP_POSUPTO:
2587      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2588        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2589      break;      break;
2590  #endif  #endif
2591    
2592        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2593        string. */
2594    
2595        case OP_MARK:
2596        case OP_PRUNE_ARG:
2597        case OP_SKIP_ARG:
2598        code += code[1];
2599        break;
2600    
2601        case OP_THEN_ARG:
2602        code += code[1];
2603        break;
2604    
2605        /* None of the remaining opcodes are required to match a character. */
2606    
2607        default:
2608        break;
2609      }      }
2610    }    }
2611    
# Line 1539  return TRUE; Line 2622  return TRUE;
2622  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
2623  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,
2624  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.
2625    This function is called only during the real compile, not during the
2626    pre-compile.
2627    
2628  Arguments:  Arguments:
2629    code        points to start of the recursion    code        points to start of the recursion
2630    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2631    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2632    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2633      cd          pointers to tables etc
2634    
2635  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2636  */  */
2637    
2638  static BOOL  static BOOL
2639  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2640    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2641  {  {
2642  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2643    {    {
2644    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2645        return FALSE;
2646    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2647    }    }
2648  return TRUE;  return TRUE;
# Line 1568  return TRUE; Line 2655  return TRUE;
2655  *************************************************/  *************************************************/
2656    
2657  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2658  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
2659  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2660  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2661    
2662    Originally, this function only recognized a sequence of letters between the
2663    terminators, but it seems that Perl recognizes any sequence of characters,
2664    though of course unknown POSIX names are subsequently rejected. Perl gives an
2665    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2666    didn't consider this to be a POSIX class. Likewise for [:1234:].
2667    
2668    The problem in trying to be exactly like Perl is in the handling of escapes. We
2669    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2670    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2671    below handles the special case of \], but does not try to do any other escape
2672    processing. This makes it different from Perl for cases such as [:l\ower:]
2673    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2674    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2675    I think.
2676    
2677    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2678    It seems that the appearance of a nested POSIX class supersedes an apparent
2679    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2680    a digit.
2681    
2682    In Perl, unescaped square brackets may also appear as part of class names. For
2683    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2684    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2685    seem right at all. PCRE does not allow closing square brackets in POSIX class
2686    names.
2687    
2688  Argument:  Arguments:
2689    ptr      pointer to the initial [    ptr      pointer to the initial [
2690    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2691    
2692  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2693  */  */
2694    
2695  static BOOL  static BOOL
2696  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2697  {  {
2698  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2699  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2700  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2701    {    {
2702    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2703    return TRUE;      ptr++;
2704      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2705      else
2706        {
2707        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2708          {
2709          *endptr = ptr;
2710          return TRUE;
2711          }
2712        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2713             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2714              ptr[1] == CHAR_EQUALS_SIGN) &&
2715            check_posix_syntax(ptr, endptr))
2716          return FALSE;
2717        }
2718    }    }
2719  return FALSE;  return FALSE;
2720  }  }
# Line 1613  Returns:     a value representing the na Line 2737  Returns:     a value representing the na
2737  */  */
2738    
2739  static int  static int
2740  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2741  {  {
2742    const char *pn = posix_names;
2743  register int yield = 0;  register int yield = 0;
2744  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2745    {    {
2746    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2747      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2748      pn += posix_name_lengths[yield] + 1;
2749    yield++;    yield++;
2750    }    }
2751  return -1;  return -1;
# Line 1634  return -1; Line 2760  return -1;
2760  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2761  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2762  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
2763  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
2764  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
2765  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
2766  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
2767  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2768    OP_END.
2769    
2770  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2771  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 1649  value in the reference (which is a group Line 2776  value in the reference (which is a group
2776  Arguments:  Arguments:
2777    group      points to the start of the group    group      points to the start of the group
2778    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2779    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2780    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2781    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
2782    
# Line 1657  Returns:     nothing Line 2784  Returns:     nothing
2784  */  */
2785    
2786  static void  static void
2787  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2788    uschar *save_hwm)    pcre_uchar *save_hwm)
2789  {  {
2790  uschar *ptr = group;  pcre_uchar *ptr = group;
2791  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2792    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2793    {    {
2794    int offset;    int offset;
2795    uschar *hc;    pcre_uchar *hc;
2796    
2797    /* 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
2798    reference. */    reference. */
# Line 1709  Arguments: Line 2837  Arguments:
2837  Returns:         new code pointer  Returns:         new code pointer
2838  */  */
2839    
2840  static uschar *  static pcre_uchar *
2841  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2842  {  {
2843  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2844  *code++ = 255;  *code++ = 255;
2845  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2846  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2847  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2848  }  }
2849    
2850    
# Line 1738  Returns:             nothing Line 2866  Returns:             nothing
2866  */  */
2867    
2868  static void  static void
2869  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2870  {  {
2871  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2872  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2873  }  }
2874    
# Line 1752  PUT(previous_callout, 2 + LINK_SIZE, len Line 2880  PUT(previous_callout, 2 + LINK_SIZE, len
2880  *************************************************/  *************************************************/
2881    
2882  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
2883  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
2884  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
2885  start address.  start address. A character with multiple other cases is returned on its own
2886    with a special return value.
2887    
2888  Arguments:  Arguments:
2889    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1762  Arguments: Line 2891  Arguments:
2891    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2892    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2893    
2894  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2895                   0 when a range is returned
2896                  >0 the CASESET offset for char with multiple other cases
2897                    in this case, ocptr contains the original
2898  */  */
2899    
2900  static BOOL  static int
2901  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2902    unsigned int *odptr)    unsigned int *odptr)
2903  {  {
2904  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2905    int co;
2906    
2907    /* Find the first character that has an other case. If it has multiple other
2908    cases, return its case offset value. */
2909    
2910  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2911    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2912      if ((co = UCD_CASESET(c)) != 0)
2913        {
2914        *ocptr = c++;   /* Character that has the set */
2915        *cptr = c;      /* Rest of input range */
2916        return co;
2917        }
2918      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2919      }
2920    
2921  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2922    
2923  *ocptr = othercase;  *ocptr = othercase;
2924  next = othercase + 1;  next = othercase + 1;
2925    
2926  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2927    {    {
2928    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2929    next++;    next++;
2930    }    }
2931    
2932  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2933  *cptr = c;  *cptr = c;             /* Rest of input range */
2934    return 0;
2935    }
2936    
2937    
2938  return TRUE;  
2939    /*************************************************
2940    *        Check a character and a property        *
2941    *************************************************/
2942    
2943    /* This function is called by check_auto_possessive() when a property item
2944    is adjacent to a fixed character.
2945    
2946    Arguments:
2947      c            the character
2948      ptype        the property type
2949      pdata        the data for the type
2950      negated      TRUE if it's a negated property (\P or \p{^)
2951    
2952    Returns:       TRUE if auto-possessifying is OK
2953    */
2954    
2955    static BOOL
2956    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2957    {
2958    #ifdef SUPPORT_UCP
2959    const pcre_uint32 *p;
2960    #endif
2961    
2962    const ucd_record *prop = GET_UCD(c);
2963    
2964    switch(ptype)
2965      {
2966      case PT_LAMP:
2967      return (prop->chartype == ucp_Lu ||
2968              prop->chartype == ucp_Ll ||
2969              prop->chartype == ucp_Lt) == negated;
2970    
2971      case PT_GC:
2972      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2973    
2974      case PT_PC:
2975      return (pdata == prop->chartype) == negated;
2976    
2977      case PT_SC:
2978      return (pdata == prop->script) == negated;
2979    
2980      /* These are specials */
2981    
2982      case PT_ALNUM:
2983      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2984              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2985    
2986      case PT_SPACE:    /* Perl space */
2987      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2988              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2989              == negated;
2990    
2991      case PT_PXSPACE:  /* POSIX space */
2992      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2993              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2994              c == CHAR_FF || c == CHAR_CR)
2995              == negated;
2996    
2997      case PT_WORD:
2998      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2999              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3000              c == CHAR_UNDERSCORE) == negated;
3001    
3002    #ifdef SUPPORT_UCP
3003      case PT_CLIST:
3004      p = PRIV(ucd_caseless_sets) + prop->caseset;
3005      for (;;)
3006        {
3007        if ((unsigned int)c < *p) return !negated;
3008        if ((unsigned int)c == *p++) return negated;
3009        }
3010      break;  /* Control never reaches here */
3011    #endif
3012      }
3013    
3014    return FALSE;
3015  }  }
3016  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3017    
# Line 1803  whether the next thing could possibly ma Line 3026  whether the next thing could possibly ma
3026  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3027    
3028  Arguments:  Arguments:
3029    op_code       the repeated op code    previous      pointer to the repeated opcode
3030    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
3031    ptr           next character in pattern    ptr           next character in pattern
3032    options       options bits    options       options bits
3033    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1815  Returns:        TRUE if possessifying is Line 3036  Returns:        TRUE if possessifying is
3036  */  */
3037    
3038  static BOOL  static BOOL
3039  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3040    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3041  {  {
3042  int next;  pcre_int32 c = NOTACHAR; // FIXMEchpe pcre_uint32
3043    pcre_int32 next;
3044    int escape;
3045    int op_code = *previous++;
3046    
3047  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3048    
# Line 1826  if ((options & PCRE_EXTENDED) != 0) Line 3050  if ((options & PCRE_EXTENDED) != 0)
3050    {    {
3051    for (;;)    for (;;)
3052      {      {
3053      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3054      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3055        {        {
3056        while (*(++ptr) != 0)        ptr++;
3057          while (*ptr != 0)
3058            {
3059          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3060            ptr++;
3061    #ifdef SUPPORT_UTF
3062            if (utf) FORWARDCHAR(ptr);
3063    #endif
3064            }
3065        }        }
3066      else break;      else break;
3067      }      }
# Line 1839  if ((options & PCRE_EXTENDED) != 0) Line 3070  if ((options & PCRE_EXTENDED) != 0)
3070  /* 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
3071  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3072    
3073  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3074    {    {
3075    int temperrorcode = 0;    int temperrorcode = 0;
3076    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3077    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3078    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3079    }    }
3080    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3081    {    {
3082  #ifdef SUPPORT_UTF8    escape = 0;
3083    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3084      if (utf) { GETCHARINC(next, ptr); } else
3085  #endif  #endif
3086    next = *ptr++;    next = *ptr++;
3087    }    }
   
3088  else return FALSE;  else return FALSE;
3089    
3090  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1863  if ((options & PCRE_EXTENDED) != 0) Line 3093  if ((options & PCRE_EXTENDED) != 0)
3093    {    {
3094    for (;;)    for (;;)
3095      {      {
3096      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3097      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3098        {        {
3099        while (*(++ptr) != 0)        ptr++;
3100          while (*ptr != 0)
3101            {
3102          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3103            ptr++;
3104    #ifdef SUPPORT_UTF
3105            if (utf) FORWARDCHAR(ptr);
3106    #endif
3107            }
3108        }        }
3109      else break;      else break;
3110      }      }
# Line 1875  if ((options & PCRE_EXTENDED) != 0) Line 3112  if ((options & PCRE_EXTENDED) != 0)
3112    
3113  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3114    
3115  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3116    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3117        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. */  
   
3118    
3119  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3120    
3121  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3122        op_code == OP_NOT || op_code == OP_NOTI)
3123      //if (escape == 0) switch(op_code)
3124    {    {
3125    case OP_CHAR:  #ifdef SUPPORT_UTF
3126  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3127    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3128      c = *previous;
3129  #endif  #endif
3130    return item != next;    }
3131    
3132    /* For CHARNC (caseless character) we must check the other case. If we have  /* Now compare the next item with the previous opcode. First, handle cases when
3133    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3134    
3135    case OP_CHARNC:  if (escape == 0)
3136  #ifdef SUPPORT_UTF8    {
3137    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
3138      case, which maps to the special PT_CLIST property. Check this first. */
3139    
3140    #ifdef SUPPORT_UCP
3141      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3142        {
3143        int ocs = UCD_CASESET(next);
3144        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3145        }
3146  #endif  #endif
3147    if (item == next) return FALSE;  
3148  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3149      {      {
3150      unsigned int othercase;      case OP_CHAR:
3151      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3152    
3153        /* For CHARI (caseless character) we must check the other case. If we have
3154        Unicode property support, we can use it to test the other case of
3155        high-valued characters. We know that next can have only one other case,
3156        because multi-other-case characters are dealt with above. */
3157    
3158        case OP_CHARI:
3159        if (c == next) return FALSE;
3160    #ifdef SUPPORT_UTF
3161        if (utf)
3162          {
3163          unsigned int othercase;
3164          if (next < 128) othercase = cd->fcc[next]; else
3165  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3166      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3167  #else  #else
3168      othercase = NOTACHAR;        othercase = NOTACHAR;
3169  #endif  #endif
3170      return (unsigned int)item != othercase;        return (unsigned int)c != othercase;
3171      }        }
3172    else      else
3173  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3174    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3175    
3176    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3177        return c == next;
3178    case OP_NOT:  
3179    if (next < 0) return FALSE;  /* Not a character */      case OP_NOTI:
3180    if (item == next) return TRUE;      if (c == next) return TRUE;
3181    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3182  #ifdef SUPPORT_UTF8      if (utf)
3183    if (utf8)        {
3184      {        unsigned int othercase;
3185      unsigned int othercase;        if (next < 128) othercase = cd->fcc[next]; else
     if (next < 128) othercase = cd->fcc[next]; else  
3186  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3187      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3188  #else  #else
3189      othercase = NOTACHAR;        othercase = NOTACHAR;
3190  #endif  #endif
3191      return (unsigned int)item == othercase;        return (unsigned int)c == othercase;
3192      }        }
3193    else      else
3194  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3195    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3196    
3197    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3198    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3199    
3200    case OP_NOT_DIGIT:      case OP_DIGIT:
3201    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3202    
3203    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3204    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3205    
3206    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3207    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3208    
3209    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3210    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3211    
3212    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3213    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3214    
3215    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3216    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3217    switch(next)  
3218      {      case OP_HSPACE:
3219      case 0x09:      case OP_NOT_HSPACE:
3220      case 0x20:      switch(next)
3221      case 0xa0:        {
3222      case 0x1680:        HSPACE_CASES:
3223      case 0x180e:        return op_code == OP_NOT_HSPACE;
3224      case 0x2000:  
3225      case 0x2001:        default:
3226      case 0x2002:        return op_code != OP_NOT_HSPACE;
3227      case 0x2003:        }
3228      case 0x2004:  
3229      case 0x2005:      case OP_ANYNL:
3230      case 0x2006:      case OP_VSPACE:
3231      case 0x2007:      case OP_NOT_VSPACE:
3232      case 0x2008:      switch(next)
3233      case 0x2009:        {
3234      case 0x200A:        VSPACE_CASES:
3235      case 0x202f:        return op_code == OP_NOT_VSPACE;
3236      case 0x205f:  
3237      case 0x3000:        default:
3238      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3239      default:        }
3240      return op_code == OP_HSPACE;  
3241      }  #ifdef SUPPORT_UCP
3242        case OP_PROP:
3243        return check_char_prop(next, previous[0], previous[1], FALSE);
3244    
3245        case OP_NOTPROP:
3246        return check_char_prop(next, previous[0], previous[1], TRUE);
3247    #endif
3248    
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
     {  
     case 0x0a:  
     case 0x0b:  
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code != OP_VSPACE;  
3249      default:      default:
3250      return op_code == OP_VSPACE;      return FALSE;
3251      }      }
   
   default:  
   return FALSE;  
3252    }    }
3253    
3254    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3255  /* Handle the case when the next item is \d, \s, etc. */  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3256    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3257    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3258    replaced by OP_PROP codes when PCRE_UCP is set. */
3259    
3260  switch(op_code)  switch(op_code)
3261    {    {
3262    case OP_CHAR:    case OP_CHAR:
3263    case OP_CHARNC:    case OP_CHARI:
3264  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3265      {      {
3266      case ESC_d:      case ESC_d:
3267      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3268    
3269      case ESC_D:      case ESC_D:
3270      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3271    
3272      case ESC_s:      case ESC_s:
3273      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3274    
3275      case ESC_S:      case ESC_S:
3276      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3277    
3278      case ESC_w:      case ESC_w:
3279      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3280    
3281      case ESC_W:      case ESC_W:
3282      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3283    
3284      case ESC_h:      case ESC_h:
3285      case ESC_H:      case ESC_H:
3286      switch(item)      switch(c)
3287        {        {
3288        case 0x09:        HSPACE_CASES:
3289        case 0x20:        return escape != ESC_h;
3290        case 0xa0:  
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
3291        default:        default:
3292        return -next == ESC_h;        return escape == ESC_h;
3293        }        }
3294    
3295      case ESC_v:      case ESC_v:
3296      case ESC_V:      case ESC_V:
3297      switch(item)      switch(c)
3298        {        {
3299        case 0x0a:        VSPACE_CASES:
3300        case 0x0b:        return escape != ESC_v;
3301        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3302        default:        default:
3303        return -next == ESC_v;        return escape == ESC_v;
3304        }        }
3305    
3306        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3307        their substitutions and process them. The result will always be either
3308        ESC_p or ESC_P. Then fall through to process those values. */
3309    
3310    #ifdef SUPPORT_UCP
3311        case ESC_du:
3312        case ESC_DU:
3313        case ESC_wu:
3314        case ESC_WU:
3315        case ESC_su:
3316        case ESC_SU:
3317          {
3318          int temperrorcode = 0;
3319          ptr = substitutes[escape - ESC_DU];
3320          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3321          if (temperrorcode != 0) return FALSE;
3322          ptr++;    /* For compatibility */
3323          }
3324        /* Fall through */
3325    
3326        case ESC_p:
3327        case ESC_P:
3328          {
3329          int ptype, pdata, errorcodeptr;
3330          BOOL negated;
3331    
3332          ptr--;      /* Make ptr point at the p or P */
3333          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3334          if (ptype < 0) return FALSE;
3335          ptr++;      /* Point past the final curly ket */
3336    
3337          /* If the property item is optional, we have to give up. (When generated
3338          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3339          to the original \d etc. At this point, ptr will point to a zero byte. */
3340    
3341          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3342            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3343              return FALSE;
3344    
3345          /* Do the property check. */
3346    
3347          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3348          }
3349    #endif
3350    
3351      default:      default:
3352      return FALSE;      return FALSE;
3353      }      }
3354    
3355      /* In principle, support for Unicode properties should be integrated here as
3356      well. It means re-organizing the above code so as to get hold of the property
3357      values before switching on the op-code. However, I wonder how many patterns
3358      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3359      these op-codes are never generated.) */
3360    
3361    case OP_DIGIT:    case OP_DIGIT:
3362    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3363           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3364    
3365    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3366    return next == -ESC_d;    return escape == ESC_d;
3367    
3368    case OP_WHITESPACE:    case OP_WHITESPACE:
3369    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3370    
3371    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3372    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3373    
3374    case OP_HSPACE:    case OP_HSPACE:
3375    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3376             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3377    
3378    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3379    return next == -ESC_h;    return escape == ESC_h;
3380    
3381    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3382      case OP_ANYNL:
3383    case OP_VSPACE:    case OP_VSPACE:
3384    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3385    
3386    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3387    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3388    
3389    case OP_WORDCHAR:    case OP_WORDCHAR:
3390    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3391             escape == ESC_v || escape == ESC_R;
3392    
3393    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3394    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3395    
3396    default:    default:
3397    return FALSE;    return FALSE;
# Line 2127  switch(op_code) Line 3403  switch(op_code)
3403    
3404    
3405  /*************************************************  /*************************************************
3406    *        Add a character or range to a class     *
3407    *************************************************/
3408    
3409    /* This function packages up the logic of adding a character or range of
3410    characters to a class. The character values in the arguments will be within the
3411    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3412    mutually recursive with the function immediately below.
3413    
3414    Arguments:
3415      classbits     the bit map for characters < 256
3416      uchardptr     points to the pointer for extra data
3417      options       the options word
3418      cd            contains pointers to tables etc.
3419      start         start of range character
3420      end           end of range character
3421    
3422    Returns:        the number of < 256 characters added
3423                    the pointer to extra data is updated
3424    */
3425    
3426    static int
3427    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3428      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3429    {
3430    pcre_uint32 c;
3431    int n8 = 0;
3432    
3433    /* If caseless matching is required, scan the range and process alternate
3434    cases. In Unicode, there are 8-bit characters that have alternate cases that
3435    are greater than 255 and vice-versa. Sometimes we can just extend the original
3436    range. */
3437    
3438    if ((options & PCRE_CASELESS) != 0)
3439      {
3440    #ifdef SUPPORT_UCP
3441      if ((options & PCRE_UTF8) != 0)
3442        {
3443        int rc;
3444        pcre_uint32 oc, od;
3445    
3446        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3447        c = start;
3448    
3449        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3450          {
3451          /* Handle a single character that has more than one other case. */
3452    
3453          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3454            PRIV(ucd_caseless_sets) + rc, oc);
3455    
3456          /* Do nothing if the other case range is within the original range. */
3457    
3458          else if (oc >= start && od <= end) continue;
3459    
3460          /* Extend the original range if there is overlap, noting that if oc < c, we
3461          can't have od > end because a subrange is always shorter than the basic
3462          range. Otherwise, use a recursive call to add the additional range. */
3463    
3464          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3465          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3466          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3467          }
3468        }
3469      else
3470    #endif  /* SUPPORT_UCP */
3471    
3472      /* Not UTF-mode, or no UCP */
3473    
3474      for (c = start; c <= end && c < 256; c++)
3475        {
3476        SETBIT(classbits, cd->fcc[c]);
3477        n8++;
3478        }
3479      }
3480    
3481    /* Now handle the original range. Adjust the final value according to the bit
3482    length - this means that the same lists of (e.g.) horizontal spaces can be used
3483    in all cases. */
3484    
3485    #if defined COMPILE_PCRE8
3486    #ifdef SUPPORT_UTF
3487      if ((options & PCRE_UTF8) == 0)
3488    #endif
3489      if (end > 0xff) end = 0xff;
3490    
3491    #elif defined COMPILE_PCRE16
3492    #ifdef SUPPORT_UTF
3493      if ((options & PCRE_UTF16) == 0)
3494    #endif
3495      if (end > 0xffff) end = 0xffff;
3496    
3497    #endif /* COMPILE_PCRE[8|16] */
3498    
3499    /* If all characters are less than 256, use the bit map. Otherwise use extra
3500    data. */
3501    
3502    if (end < 0x100)
3503      {
3504      for (c = start; c <= end; c++)
3505        {
3506        n8++;
3507        SETBIT(classbits, c);
3508        }
3509      }
3510    
3511    else
3512      {
3513      pcre_uchar *uchardata = *uchardptr;
3514    
3515    #ifdef SUPPORT_UTF
3516      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3517        {
3518        if (start < end)
3519          {
3520          *uchardata++ = XCL_RANGE;
3521          uchardata += PRIV(ord2utf)(start, uchardata);
3522          uchardata += PRIV(ord2utf)(end, uchardata);
3523          }
3524        else if (start == end)
3525          {
3526          *uchardata++ = XCL_SINGLE;
3527          uchardata += PRIV(ord2utf)(start, uchardata);
3528          }
3529        }
3530      else
3531    #endif  /* SUPPORT_UTF */
3532    
3533      /* Without UTF support, character values are constrained by the bit length,
3534      and can only be > 256 for 16-bit and 32-bit libraries. */
3535    
3536    #ifdef COMPILE_PCRE8
3537        {}
3538    #else
3539      if (start < end)
3540        {
3541        *uchardata++ = XCL_RANGE;
3542        *uchardata++ = start;
3543        *uchardata++ = end;
3544        }
3545      else if (start == end)
3546        {
3547        *uchardata++ = XCL_SINGLE;
3548        *uchardata++ = start;
3549        }
3550    #endif
3551    
3552      *uchardptr = uchardata;   /* Updata extra data pointer */
3553      }
3554    
3555    return n8;    /* Number of 8-bit characters */
3556    }
3557    
3558    
3559    
3560    
3561    /*************************************************
3562    *        Add a list of characters to a class     *
3563    *************************************************/
3564    
3565    /* This function is used for adding a list of case-equivalent characters to a
3566    class, and also for adding a list of horizontal or vertical whitespace. If the
3567    list is in order (which it should be), ranges of characters are detected and
3568    handled appropriately. This function is mutually recursive with the function
3569    above.
3570    
3571    Arguments:
3572      classbits     the bit map for characters < 256
3573      uchardptr     points to the pointer for extra data
3574      options       the options word
3575      cd            contains pointers to tables etc.
3576      p             points to row of 32-bit values, terminated by NOTACHAR
3577      except        character to omit; this is used when adding lists of
3578                      case-equivalent characters to avoid including the one we
3579                      already know about
3580    
3581    Returns:        the number of < 256 characters added
3582                    the pointer to extra data is updated
3583    */
3584    
3585    static int
3586    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3587      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3588    {
3589    int n8 = 0;
3590    while (p[0] < NOTACHAR)
3591      {
3592      int n = 0;
3593      if (p[0] != except)
3594        {
3595        while(p[n+1] == p[0] + n + 1) n++;
3596        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3597        }
3598      p += n + 1;
3599      }
3600    return n8;
3601    }
3602    
3603    
3604    
3605    /*************************************************
3606    *    Add characters not in a list to a class     *
3607    *************************************************/
3608    
3609    /* This function is used for adding the complement of a list of horizontal or
3610    vertical whitespace to a class. The list must be in order.
3611    
3612    Arguments:
3613      classbits     the bit map for characters < 256
3614      uchardptr     points to the pointer for extra data
3615      options       the options word
3616      cd            contains pointers to tables etc.
3617      p             points to row of 32-bit values, terminated by NOTACHAR
3618    
3619    Returns:        the number of < 256 characters added
3620                    the pointer to extra data is updated
3621    */
3622    
3623    static int
3624    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3625      int options, compile_data *cd, const pcre_uint32 *p)
3626    {
3627    BOOL utf = (options & PCRE_UTF8) != 0;
3628    int n8 = 0;
3629    if (p[0] > 0)
3630      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3631    while (p[0] < NOTACHAR)
3632      {
3633      while (p[1] == p[0] + 1) p++;
3634      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3635        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3636      p++;
3637      }
3638    return n8;
3639    }
3640    
3641    
3642    
3643    /*************************************************
3644  *           Compile one branch                   *  *           Compile one branch                   *
3645  *************************************************/  *************************************************/
3646    
# Line 2141  Arguments: Line 3655  Arguments:
3655    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3656    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3657    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3658    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3659    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3660    bcptr          points to current branch chain    bcptr          points to current branch chain
3661      cond_depth     conditional nesting depth
3662    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3663    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3664                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2153  Returns:         TRUE on success Line 3668  Returns:         TRUE on success
3668  */  */
3669    
3670  static BOOL  static BOOL
3671  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3672    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3673      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3674    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3675  {  {
3676  int repeat_type, op_type;  int repeat_type, op_type;
3677  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3678  int bravalue = 0;  int bravalue = 0;
3679  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3680  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3681  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3682  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3683  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3684  int after_manual_callout = 0;  int after_manual_callout = 0;
3685  int length_prevgroup = 0;  int length_prevgroup = 0;
3686  register int c;  register int c;
3687  register uschar *code = *codeptr;  int escape;
3688  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3689  uschar *orig_code = code;  pcre_uchar *last_code = code;
3690  uschar *tempcode;  pcre_uchar *orig_code = code;
3691    pcre_uchar *tempcode;
3692  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3693  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3694  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3695  const uschar *tempptr;  const pcre_uchar *tempptr;
3696  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3697  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3698  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3699  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3700    pcre_uint8 classbits[32];
3701  #ifdef SUPPORT_UTF8  
3702  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3703  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3704  uschar *class_utf8data;  dynamically as we process the pattern. */
3705  uschar utf8_char[6];  
3706    #ifdef SUPPORT_UTF
3707    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3708    BOOL utf = (options & PCRE_UTF8) != 0;
3709    pcre_uchar utf_chars[6];
3710  #else  #else
3711  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3712  uschar *utf8_char = NULL;  #endif
3713    
3714    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3715    class_uchardata always so that it can be passed to add_to_class() always,
3716    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3717    alternative calls for the different cases. */
3718    
3719    pcre_uchar *class_uchardata;
3720    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3721    BOOL xclass;
3722    pcre_uchar *class_uchardata_base;
3723  #endif  #endif
3724    
3725  #ifdef DEBUG  #ifdef PCRE_DEBUG
3726  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3727  #endif  #endif
3728    
# Line 2202  greedy_non_default = greedy_default ^ 1; Line 3733  greedy_non_default = greedy_default ^ 1;
3733