/[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 1060 by chpe, Tue Oct 16 15:53:57 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_uint32 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 pcre_uint32 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 pcre_uint32 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;
1061        int count = 0;        BOOL overflow;
1062    
1063        c = 0;        c = 0;
1064        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1065          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1066          {          {
1067          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1068          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1069          count++;  
1070    #ifdef COMPILE_PCRE32
1071  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1072          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1073          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1074    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1075            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1076            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1077  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1078          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1079          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1080    #endif
1081    
1082    #if defined COMPILE_PCRE8
1083            if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }
1084    #elif defined COMPILE_PCRE16
1085            if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }
1086    #elif defined COMPILE_PCRE32
1087            if (utf && c > 0x10ffff) { overflow = TRUE; break; }
1088  #endif  #endif
1089          }          }
1090    
1091        if (*pt == '}')        if (overflow)
1092            {
1093            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1094            *errorcodeptr = ERR34;
1095            }
1096    
1097          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1098          {          {
1099          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1100          ptr = pt;          ptr = pt;
1101          break;          break;
1102          }          }
# Line 612  else Line 1108  else
1108      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1109    
1110      c = 0;      c = 0;
1111      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1112        {        {
1113        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1114        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1115  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1116        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1117        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1118  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1119        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1120        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1121  #endif  #endif
1122        }        }
1123      break;      break;
1124    
1125      /* 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.
1126      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
1127        coding is ASCII-specific, but then the whole concept of \cx is
1128      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1129    
1130      case 'c':      case CHAR_c:
1131      c = *(++ptr);      c = *(++ptr);
1132      if (c == 0)      if (c == 0)
1133        {        {
1134        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1135        return 0;        break;
1136        }        }
1137    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1138  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1139      if (c >= 'a' && c <= 'z') c -= 32;        {
1140          *errorcodeptr = ERR68;
1141          break;
1142          }
1143        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1144      c ^= 0x40;      c ^= 0x40;
1145  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1146      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1147      c ^= 0xC0;      c ^= 0xC0;
1148  #endif  #endif
1149      break;      break;
1150    
1151      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1152      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1153      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
1154      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
1155      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1156    
1157      default:      default:
1158      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 664  else Line 1165  else
1165      }      }
1166    }    }
1167    
1168  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1169  return c;  newline". PCRE does not support \N{name}. However, it does support
1170  }  quantification such as \N{2,3}. */
1171    
1172    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1173         !is_counted_repeat(ptr+2))
1174      *errorcodeptr = ERR37;
1175    
1176    /* If PCRE_UCP is set, we change the values for \d etc. */
1177    
1178    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1179      escape += (ESC_DU - ESC_D);
1180    
1181    /* Set the pointer to the final character before returning. */
1182    
1183    *ptrptr = ptr;
1184    *chptr = c;
1185    return escape;
1186    }
1187    
1188  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1189  /*************************************************  /*************************************************
# Line 690  Returns:         type value from ucp_typ Line 1205  Returns:         type value from ucp_typ
1205  */  */
1206    
1207  static int  static int
1208  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1209  {  {
1210  int c, i, bot, top;  int c, i, bot, top;
1211  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1212  char name[32];  pcre_uchar name[32];
1213    
1214  c = *(++ptr);  c = *(++ptr);
1215  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 704  if (c == 0) goto ERROR_RETURN; Line 1219  if (c == 0) goto ERROR_RETURN;
1219  /* \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
1220  negation. */  negation. */
1221    
1222  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1223    {    {
1224    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1225      {      {
1226      *negptr = TRUE;      *negptr = TRUE;
1227      ptr++;      ptr++;
1228      }      }
1229    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1230      {      {
1231      c = *(++ptr);      c = *(++ptr);
1232      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1233      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1234      name[i] = c;      name[i] = c;
1235      }      }
1236    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1237    name[i] = 0;    name[i] = 0;
1238    }    }
1239    
# Line 735  else Line 1250  else
1250  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1251    
1252  bot = 0;  bot = 0;
1253  top = _pcre_utt_size;  top = PRIV(utt_size);
1254    
1255  while (bot < top)  while (bot < top)
1256    {    {
1257    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1258    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1259    if (c == 0)    if (c == 0)
1260      {      {
1261      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1262      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1263      }      }
1264    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1265    }    }
# Line 764  return -1; Line 1279  return -1;
1279    
1280    
1281  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1282  *         Read repeat counts                     *  *         Read repeat counts                     *
1283  *************************************************/  *************************************************/
1284    
# Line 815  Returns:         pointer to '}' on succe Line 1297  Returns:         pointer to '}' on succe
1297                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1298  */  */
1299    
1300  static const uschar *  static const pcre_uchar *
1301  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)
1302  {  {
1303  int min = 0;  int min = 0;
1304  int max = -1;  int max = -1;
# Line 824  int max = -1; Line 1306  int max = -1;
1306  /* 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
1307  an integer overflow. */  an integer overflow. */
1308    
1309  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1310  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1311    {    {
1312    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 834  if (min < 0 || min > 65535) Line 1316  if (min < 0 || min > 65535)
1316  /* 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.
1317  Also, max must not be less than min. */  Also, max must not be less than min. */
1318    
1319  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1320    {    {
1321    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1322      {      {
1323      max = 0;      max = 0;
1324      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1325      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1326        {        {
1327        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 864  return p; Line 1346  return p;
1346    
1347    
1348  /*************************************************  /*************************************************
1349  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1350  *************************************************/  *************************************************/
1351    
1352  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1353    top-level call starts at the beginning of the pattern. All other calls must
1354    start at a parenthesis. It scans along a pattern's text looking for capturing
1355  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
1356  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
1357  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
1358  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1359  be terminated by '>' because that is checked in the first pass.  
1360    This function was originally called only from the second pass, in which we know
1361    that if (?< or (?' or (?P< is encountered, the name will be correctly
1362    terminated because that is checked in the first pass. There is now one call to
1363    this function in the first pass, to check for a recursive back reference by
1364    name (so that we can make the whole group atomic). In this case, we need check
1365    only up to the current position in the pattern, and that is still OK because
1366    and previous occurrences will have been checked. To make this work, the test
1367    for "end of pattern" is a check against cd->end_pattern in the main loop,
1368    instead of looking for a binary zero. This means that the special first-pass
1369    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1370    processing items within the loop are OK, because afterwards the main loop will
1371    terminate.)
1372    
1373  Arguments:  Arguments:
1374    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1375    count        current count of capturing parens so far encountered    cd           compile background data
1376    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1377    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1378    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1379      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1380      count        pointer to the current capturing subpattern number (updated)
1381    
1382  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1383  */  */
1384    
1385  static int  static int
1386  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,
1387    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1388  {  {
1389  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1390    int start_count = *count;
1391    int hwm_count = start_count;
1392    BOOL dup_parens = FALSE;
1393    
1394  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1395    dealing with. The very first call may not start with a parenthesis. */
1396    
1397    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1398    {    {
1399    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1400    
1401      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1402    
1403      /* Handle a normal, unnamed capturing parenthesis. */
1404    
1405      else if (ptr[1] != CHAR_QUESTION_MARK)
1406        {
1407        *count += 1;
1408        if (name == NULL && *count == lorn) return *count;
1409        ptr++;
1410        }
1411    
1412      /* All cases now have (? at the start. Remember when we are in a group
1413      where the parenthesis numbers are duplicated. */
1414    
1415      else if (ptr[2] == CHAR_VERTICAL_LINE)
1416        {
1417        ptr += 3;
1418        dup_parens = TRUE;
1419        }
1420    
1421      /* Handle comments; all characters are allowed until a ket is reached. */
1422    
1423      else if (ptr[2] == CHAR_NUMBER_SIGN)
1424        {
1425        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1426        goto FAIL_EXIT;
1427        }
1428    
1429      /* Handle a condition. If it is an assertion, just carry on so that it
1430      is processed as normal. If not, skip to the closing parenthesis of the
1431      condition (there can't be any nested parens). */
1432    
1433      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1434        {
1435        ptr += 2;
1436        if (ptr[1] != CHAR_QUESTION_MARK)
1437          {
1438          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1439          if (*ptr != 0) ptr++;
1440          }
1441        }
1442    
1443      /* Start with (? but not a condition. */
1444    
1445      else
1446        {
1447        ptr += 2;
1448        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1449    
1450        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1451    
1452        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1453            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1454          {
1455          int term;
1456          const pcre_uchar *thisname;
1457          *count += 1;
1458          if (name == NULL && *count == lorn) return *count;
1459          term = *ptr++;
1460          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1461          thisname = ptr;
1462          while (*ptr != term) ptr++;
1463          if (name != NULL && lorn == ptr - thisname &&
1464              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1465            return *count;
1466          term++;
1467          }
1468        }
1469      }
1470    
1471    /* Past any initial parenthesis handling, scan for parentheses or vertical
1472    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1473    first-pass call when this value is temporarily adjusted to stop at the current
1474    position. So DO NOT change this to a test for binary zero. */
1475    
1476    for (; ptr < cd->end_pattern; ptr++)
1477      {
1478    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1479    
1480    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1481      {      {
1482      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1483      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1484        {        {
1485        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1486        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1487        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1488        }        }
1489      continue;      continue;
1490      }      }
1491    
1492    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1493      are handled for real. If the first character is '^', skip it. Also, if the
1494      first few characters (either before or after ^) are \Q\E or \E we skip them
1495      too. This makes for compatibility with Perl. Note the use of STR macros to
1496      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1497    
1498    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1499      {      {
1500      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1501        for (;;)
1502          {
1503          if (ptr[1] == CHAR_BACKSLASH)
1504            {
1505            if (ptr[2] == CHAR_E)
1506              ptr+= 2;
1507            else if (STRNCMP_UC_C8(ptr + 2,
1508                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1509              ptr += 4;
1510            else
1511              break;
1512            }
1513          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1514            {
1515            negate_class = TRUE;
1516            ptr++;
1517            }
1518          else break;
1519          }
1520    
1521        /* If the next character is ']', it is a data character that must be
1522        skipped, except in JavaScript compatibility mode. */
1523    
1524        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1525            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1526          ptr++;
1527    
1528        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1529        {        {
1530        if (*ptr == '\\')        if (*ptr == 0) return -1;
1531          if (*ptr == CHAR_BACKSLASH)
1532          {          {
1533          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1534          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1535            {            {
1536            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1537            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1538            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1539            }            }
1540          continue;          continue;
1541          }          }
# Line 931  for (; *ptr != 0; ptr++) Line 1545  for (; *ptr != 0; ptr++)
1545    
1546    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1547    
1548    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1549      {      {
1550      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1551      if (*ptr == 0) return -1;      while (*ptr != 0)
1552          {
1553          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1554          ptr++;
1555    #ifdef SUPPORT_UTF
1556          if (utf) FORWARDCHAR(ptr);
1557    #endif
1558          }
1559        if (*ptr == 0) goto FAIL_EXIT;
1560      continue;      continue;
1561      }      }
1562    
1563    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1564    
1565    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1566      {      {
1567      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1568      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1569      continue;      if (*ptr == 0) goto FAIL_EXIT;
1570      }      }
1571    
1572    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1573    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1574        if (dup_parens && *count < hwm_count) *count = hwm_count;
1575    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1576        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1577    
1578    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1579    term = *ptr++;      {
1580    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1581    thisname = ptr;      *count = start_count;
1582    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1583    }    }
1584    
1585    FAIL_EXIT:
1586    *ptrptr = ptr;
1587  return -1;  return -1;
1588  }  }
1589    
1590    
1591    
1592    
1593  /*************************************************  /*************************************************
1594  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1595    *************************************************/
1596    
1597    /* This function scans along a pattern's text looking for capturing
1598    subpatterns, and counting them. If it finds a named pattern that matches the
1599    name it is given, it returns its number. Alternatively, if the name is NULL, it
1600    returns when it reaches a given numbered subpattern. This is used for forward
1601    references to subpatterns. We used to be able to start this scan from the
1602    current compiling point, using the current count value from cd->bracount, and
1603    do it all in a single loop, but the addition of the possibility of duplicate
1604    subpattern numbers means that we have to scan from the very start, in order to
1605    take account of such duplicates, and to use a recursive function to keep track
1606    of the different types of group.
1607    
1608    Arguments:
1609      cd           compile background data
1610      name         name to seek, or NULL if seeking a numbered subpattern
1611      lorn         name length, or subpattern number if name is NULL
1612      xmode        TRUE if we are in /x mode
1613      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1614    
1615    Returns:       the number of the found subpattern, or -1 if not found
1616    */
1617    
1618    static int
1619    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1620      BOOL utf)
1621    {
1622    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1623    int count = 0;
1624    int rc;
1625    
1626    /* If the pattern does not start with an opening parenthesis, the first call
1627    to find_parens_sub() will scan right to the end (if necessary). However, if it
1628    does start with a parenthesis, find_parens_sub() will return when it hits the
1629    matching closing parens. That is why we have to have a loop. */
1630    
1631    for (;;)
1632      {
1633      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1634      if (rc > 0 || *ptr++ == 0) break;
1635      }
1636    
1637    return rc;
1638    }
1639    
1640    
1641    
1642    
1643    /*************************************************
1644    *      Find first significant op code            *
1645  *************************************************/  *************************************************/
1646    
1647  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1648  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1649  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1650  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1651  assertions, and also the \b assertion; for others it does not.  does not.
1652    
1653  Arguments:  Arguments:
1654    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  
1655    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1656    
1657  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1658  */  */
1659    
1660  static const uschar*  static const pcre_uchar*
1661  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1662  {  {
1663  for (;;)  for (;;)
1664    {    {
1665    switch ((int)*code)    switch ((int)*code)
1666      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1667      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1668      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1669      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1670      if (!skipassert) return code;      if (!skipassert) return code;
1671      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1672      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1673      break;      break;
1674    
1675      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1023  for (;;) Line 1679  for (;;)
1679    
1680      case OP_CALLOUT:      case OP_CALLOUT:
1681      case OP_CREF:      case OP_CREF:
1682        case OP_NCREF:
1683      case OP_RREF:      case OP_RREF:
1684        case OP_NRREF:
1685      case OP_DEF:      case OP_DEF:
1686      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1687      break;      break;
1688    
1689      default:      default:
# Line 1039  for (;;) Line 1697  for (;;)
1697    
1698    
1699  /*************************************************  /*************************************************
1700  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1701  *************************************************/  *************************************************/
1702    
1703  /* 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,
1704  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.
1705  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
1706    temporarily terminated with OP_END when this function is called.
1707    
1708    This function is called when a backward assertion is encountered, so that if it
1709    fails, the error message can point to the correct place in the pattern.
1710    However, we cannot do this when the assertion contains subroutine calls,
1711    because they can be forward references. We solve this by remembering this case
1712    and doing the check at the end; a flag specifies which mode we are running in.
1713    
1714  Arguments:  Arguments:
1715    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1716    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1717      atend    TRUE if called when the pattern is complete
1718  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1719               or -2 if \C was encountered  
1720    Returns:   the fixed length,
1721                 or -1 if there is no fixed length,
1722                 or -2 if \C was encountered (in UTF-8 mode only)
1723                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1724                 or -4 if an unknown opcode was encountered (internal error)
1725  */  */
1726    
1727  static int  static int
1728  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1729  {  {
1730  int length = -1;  int length = -1;
1731    
1732  register int branchlength = 0;  register int branchlength = 0;
1733  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1734    
1735  /* 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
1736  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 1738  branch, check the length against that of
1738  for (;;)  for (;;)
1739    {    {
1740    int d;    int d;
1741      pcre_uchar *ce, *cs;
1742    register int op = *cc;    register int op = *cc;
1743    
1744    switch (op)    switch (op)
1745      {      {
1746        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1747        OP_BRA (normal non-capturing bracket) because the other variants of these
1748        opcodes are all concerned with unlimited repeated groups, which of course
1749        are not of fixed length. */
1750    
1751      case OP_CBRA:      case OP_CBRA:
1752      case OP_BRA:      case OP_BRA:
1753      case OP_ONCE:      case OP_ONCE:
1754        case OP_ONCE_NC:
1755      case OP_COND:      case OP_COND:
1756      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1757      if (d < 0) return d;      if (d < 0) return d;
1758      branchlength += d;      branchlength += d;
1759      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1760      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1761      break;      break;
1762    
1763      /* 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.
1764      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
1765      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
1766        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1767        because they all imply an unlimited repeat. */
1768    
1769      case OP_ALT:      case OP_ALT:
1770      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1771      case OP_END:      case OP_END:
1772        case OP_ACCEPT:
1773        case OP_ASSERT_ACCEPT:
1774      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1775        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1776      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1099  for (;;) Line 1778  for (;;)
1778      branchlength = 0;      branchlength = 0;
1779      break;      break;
1780    
1781        /* A true recursion implies not fixed length, but a subroutine call may
1782        be OK. If the subroutine is a forward reference, we can't deal with
1783        it until the end of the pattern, so return -3. */
1784    
1785        case OP_RECURSE:
1786        if (!atend) return -3;
1787        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1788        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1789        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1790        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1791        if (d < 0) return d;
1792        branchlength += d;
1793        cc += 1 + LINK_SIZE;
1794        break;
1795    
1796      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1797    
1798      case OP_ASSERT:      case OP_ASSERT:
# Line 1106  for (;;) Line 1800  for (;;)
1800      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1801      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1802      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1803      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1804        break;
1805    
1806      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1807    
1808      case OP_REVERSE:      case OP_MARK:
1809        case OP_PRUNE_ARG:
1810        case OP_SKIP_ARG:
1811        case OP_THEN_ARG:
1812        cc += cc[1] + PRIV(OP_lengths)[*cc];
1813        break;
1814    
1815        case OP_CALLOUT:
1816        case OP_CIRC:
1817        case OP_CIRCM:
1818        case OP_CLOSE:
1819        case OP_COMMIT:
1820      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1821      case OP_DEF:      case OP_DEF:
1822      case OP_OPT:      case OP_DOLL:
1823      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1824      case OP_EOD:      case OP_EOD:
1825      case OP_EODN:      case OP_EODN:
1826      case OP_CIRC:      case OP_FAIL:
1827      case OP_DOLL:      case OP_NCREF:
1828        case OP_NRREF:
1829      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1830        case OP_PRUNE:
1831        case OP_REVERSE:
1832        case OP_RREF:
1833        case OP_SET_SOM:
1834        case OP_SKIP:
1835        case OP_SOD:
1836        case OP_SOM:
1837        case OP_THEN:
1838      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1839      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1840      break;      break;
1841    
1842      /* Handle literal characters */      /* Handle literal characters */
1843    
1844      case OP_CHAR:      case OP_CHAR:
1845      case OP_CHARNC:      case OP_CHARI:
1846      case OP_NOT:      case OP_NOT:
1847        case OP_NOTI:
1848      branchlength++;      branchlength++;
1849      cc += 2;      cc += 2;
1850  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1851      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1852  #endif  #endif
1853      break;      break;
1854    
# Line 1146  for (;;) Line 1856  for (;;)
1856      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1857    
1858      case OP_EXACT:      case OP_EXACT:
1859        case OP_EXACTI:
1860        case OP_NOTEXACT:
1861        case OP_NOTEXACTI:
1862      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1863      cc += 4;      cc += 2 + IMM2_SIZE;
1864  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1865      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1866  #endif  #endif
1867      break;      break;
1868    
1869      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1870      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1871      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1872          cc += 2;
1873        cc += 1 + IMM2_SIZE + 1;
1874      break;      break;
1875    
1876      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1168  for (;;) Line 1880  for (;;)
1880      cc += 2;      cc += 2;
1881      /* Fall through */      /* Fall through */
1882    
1883        case OP_HSPACE:
1884        case OP_VSPACE:
1885        case OP_NOT_HSPACE:
1886        case OP_NOT_VSPACE:
1887      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1888      case OP_DIGIT:      case OP_DIGIT:
1889      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1175  for (;;) Line 1891  for (;;)
1891      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1892      case OP_WORDCHAR:      case OP_WORDCHAR:
1893      case OP_ANY:      case OP_ANY:
1894        case OP_ALLANY:
1895      branchlength++;      branchlength++;
1896      cc++;      cc++;
1897      break;      break;
1898    
1899      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1900        otherwise \C is coded as OP_ALLANY. */
1901    
1902      case OP_ANYBYTE:      case OP_ANYBYTE:
1903      return -2;      return -2;
1904    
1905      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1906    
1907  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1908      case OP_XCLASS:      case OP_XCLASS:
1909      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1910      /* Fall through */      /* Fall through */
1911  #endif  #endif
1912    
1913      case OP_CLASS:      case OP_CLASS:
1914      case OP_NCLASS:      case OP_NCLASS:
1915      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1916    
1917      switch (*cc)      switch (*cc)
1918        {        {
1919          case OP_CRPLUS:
1920          case OP_CRMINPLUS:
1921        case OP_CRSTAR:        case OP_CRSTAR:
1922        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1923        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1206  for (;;) Line 1926  for (;;)
1926    
1927        case OP_CRRANGE:        case OP_CRRANGE:
1928        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1929        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1930        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1931        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1932        break;        break;
1933    
1934        default:        default:
# Line 1218  for (;;) Line 1938  for (;;)
1938    
1939      /* Anything else is variable length */      /* Anything else is variable length */
1940    
1941      default:      case OP_ANYNL:
1942        case OP_BRAMINZERO:
1943        case OP_BRAPOS:
1944        case OP_BRAPOSZERO:
1945        case OP_BRAZERO:
1946        case OP_CBRAPOS:
1947        case OP_EXTUNI:
1948        case OP_KETRMAX:
1949        case OP_KETRMIN:
1950        case OP_KETRPOS:
1951        case OP_MINPLUS:
1952        case OP_MINPLUSI:
1953        case OP_MINQUERY:
1954        case OP_MINQUERYI:
1955        case OP_MINSTAR:
1956        case OP_MINSTARI:
1957        case OP_MINUPTO:
1958        case OP_MINUPTOI:
1959        case OP_NOTMINPLUS:
1960        case OP_NOTMINPLUSI:
1961        case OP_NOTMINQUERY:
1962        case OP_NOTMINQUERYI:
1963        case OP_NOTMINSTAR:
1964        case OP_NOTMINSTARI:
1965        case OP_NOTMINUPTO:
1966        case OP_NOTMINUPTOI:
1967        case OP_NOTPLUS:
1968        case OP_NOTPLUSI:
1969        case OP_NOTPOSPLUS:
1970        case OP_NOTPOSPLUSI:
1971        case OP_NOTPOSQUERY:
1972        case OP_NOTPOSQUERYI:
1973        case OP_NOTPOSSTAR:
1974        case OP_NOTPOSSTARI:
1975        case OP_NOTPOSUPTO:
1976        case OP_NOTPOSUPTOI:
1977        case OP_NOTQUERY:
1978        case OP_NOTQUERYI:
1979        case OP_NOTSTAR:
1980        case OP_NOTSTARI:
1981        case OP_NOTUPTO:
1982        case OP_NOTUPTOI:
1983        case OP_PLUS:
1984        case OP_PLUSI:
1985        case OP_POSPLUS:
1986        case OP_POSPLUSI:
1987        case OP_POSQUERY:
1988        case OP_POSQUERYI:
1989        case OP_POSSTAR:
1990        case OP_POSSTARI:
1991        case OP_POSUPTO:
1992        case OP_POSUPTOI:
1993        case OP_QUERY:
1994        case OP_QUERYI:
1995        case OP_REF:
1996        case OP_REFI:
1997        case OP_SBRA:
1998        case OP_SBRAPOS:
1999        case OP_SCBRA:
2000        case OP_SCBRAPOS:
2001        case OP_SCOND:
2002        case OP_SKIPZERO:
2003        case OP_STAR:
2004        case OP_STARI:
2005        case OP_TYPEMINPLUS:
2006        case OP_TYPEMINQUERY:
2007        case OP_TYPEMINSTAR:
2008        case OP_TYPEMINUPTO:
2009        case OP_TYPEPLUS:
2010        case OP_TYPEPOSPLUS:
2011        case OP_TYPEPOSQUERY:
2012        case OP_TYPEPOSSTAR:
2013        case OP_TYPEPOSUPTO:
2014        case OP_TYPEQUERY:
2015        case OP_TYPESTAR:
2016        case OP_TYPEUPTO:
2017        case OP_UPTO:
2018        case OP_UPTOI:
2019      return -1;      return -1;
2020    
2021        /* Catch unrecognized opcodes so that when new ones are added they
2022        are not forgotten, as has happened in the past. */
2023    
2024        default:
2025        return -4;
2026      }      }
2027    }    }
2028  /* Control never gets here */  /* Control never gets here */
# Line 1229  for (;;) Line 2032  for (;;)
2032    
2033    
2034  /*************************************************  /*************************************************
2035  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2036  *************************************************/  *************************************************/
2037    
2038  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2039  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2040    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2041    so that it can be called from pcre_study() when finding the minimum matching
2042    length.
2043    
2044  Arguments:  Arguments:
2045    code        points to start of expression    code        points to start of expression
2046    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2047    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2048    
2049  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
2050  */  */
2051    
2052  static const uschar *  const pcre_uchar *
2053  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2054  {  {
2055  for (;;)  for (;;)
2056    {    {
2057    register int c = *code;    register int c = *code;
2058    
2059    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2060    
2061    /* 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 2064  for (;;)
2064    
2065    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2066    
2067      /* Handle recursion */
2068    
2069      else if (c == OP_REVERSE)
2070        {
2071        if (number < 0) return (pcre_uchar *)code;
2072        code += PRIV(OP_lengths)[c];
2073        }
2074    
2075    /* Handle capturing bracket */    /* Handle capturing bracket */
2076    
2077    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2078               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2079      {      {
2080      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2081      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2082      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2083      }      }
2084    
2085      /* Otherwise, we can get the item's length from the table, except that for
2086      repeated character types, we have to test for \p and \P, which have an extra
2087      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2088      must add in its length. */
2089    
2090      else
2091        {
2092        switch(c)
2093          {
2094          case OP_TYPESTAR:
2095          case OP_TYPEMINSTAR:
2096          case OP_TYPEPLUS:
2097          case OP_TYPEMINPLUS:
2098          case OP_TYPEQUERY:
2099          case OP_TYPEMINQUERY:
2100          case OP_TYPEPOSSTAR:
2101          case OP_TYPEPOSPLUS:
2102          case OP_TYPEPOSQUERY:
2103          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2104          break;
2105    
2106          case OP_TYPEUPTO:
2107          case OP_TYPEMINUPTO:
2108          case OP_TYPEEXACT:
2109          case OP_TYPEPOSUPTO:
2110          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2111            code += 2;
2112          break;
2113    
2114          case OP_MARK:
2115          case OP_PRUNE_ARG:
2116          case OP_SKIP_ARG:
2117          code += code[1];
2118          break;
2119    
2120          case OP_THEN_ARG:
2121          code += code[1];
2122          break;
2123          }
2124    
2125        /* Add in the fixed length from the table */
2126    
2127        code += PRIV(OP_lengths)[c];
2128    
2129    /* 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
2130    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
2131    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2132    
2133    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2134      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2135        {        {
2136        case OP_CHAR:        case OP_CHAR:
2137        case OP_CHARNC:        case OP_CHARI:
2138        case OP_EXACT:        case OP_EXACT:
2139          case OP_EXACTI:
2140        case OP_UPTO:        case OP_UPTO:
2141          case OP_UPTOI:
2142        case OP_MINUPTO:        case OP_MINUPTO:
2143          case OP_MINUPTOI:
2144        case OP_POSUPTO:        case OP_POSUPTO:
2145          case OP_POSUPTOI:
2146        case OP_STAR:        case OP_STAR:
2147          case OP_STARI:
2148        case OP_MINSTAR:        case OP_MINSTAR:
2149          case OP_MINSTARI:
2150        case OP_POSSTAR:        case OP_POSSTAR:
2151          case OP_POSSTARI:
2152        case OP_PLUS:        case OP_PLUS:
2153          case OP_PLUSI:
2154        case OP_MINPLUS:        case OP_MINPLUS:
2155          case OP_MINPLUSI:
2156        case OP_POSPLUS:        case OP_POSPLUS:
2157          case OP_POSPLUSI:
2158        case OP_QUERY:        case OP_QUERY:
2159          case OP_QUERYI:
2160        case OP_MINQUERY:        case OP_MINQUERY:
2161          case OP_MINQUERYI:
2162        case OP_POSQUERY:        case OP_POSQUERY:
2163        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2164          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2165        break;        break;
2166        }        }
2167    #else
2168        (void)(utf);  /* Keep compiler happy by referencing function argument */
2169  #endif  #endif
2170      }      }
2171    }    }
# Line 1310  instance of OP_RECURSE. Line 2182  instance of OP_RECURSE.
2182    
2183  Arguments:  Arguments:
2184    code        points to start of expression    code        points to start of expression
2185    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2186    
2187  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
2188  */  */
2189    
2190  static const uschar *  static const pcre_uchar *
2191  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2192  {  {
2193  for (;;)  for (;;)
2194    {    {
# Line 1330  for (;;) Line 2202  for (;;)
2202    
2203    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2204    
2205    /* 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
2206    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
2207    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
2208    bytes. */    must add in its length. */
2209    
2210    else    else
2211      {      {
2212      code += _pcre_OP_lengths[c];      switch(c)
2213  #ifdef SUPPORT_UTF8        {
2214      if (utf8) switch(c)        case OP_TYPESTAR:
2215          case OP_TYPEMINSTAR:
2216          case OP_TYPEPLUS:
2217          case OP_TYPEMINPLUS:
2218          case OP_TYPEQUERY:
2219          case OP_TYPEMINQUERY:
2220          case OP_TYPEPOSSTAR:
2221          case OP_TYPEPOSPLUS:
2222          case OP_TYPEPOSQUERY:
2223          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2224          break;
2225    
2226          case OP_TYPEPOSUPTO:
2227          case OP_TYPEUPTO:
2228          case OP_TYPEMINUPTO:
2229          case OP_TYPEEXACT:
2230          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2231            code += 2;
2232          break;
2233    
2234          case OP_MARK:
2235          case OP_PRUNE_ARG:
2236          case OP_SKIP_ARG:
2237          code += code[1];
2238          break;
2239    
2240          case OP_THEN_ARG:
2241          code += code[1];
2242          break;
2243          }
2244    
2245        /* Add in the fixed length from the table */
2246    
2247        code += PRIV(OP_lengths)[c];
2248    
2249        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2250        by a multi-byte character. The length in the table is a minimum, so we have
2251        to arrange to skip the extra bytes. */
2252    
2253    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2254        if (utf) switch(c)
2255        {        {
2256        case OP_CHAR:        case OP_CHAR:
2257        case OP_CHARNC:        case OP_CHARI:
2258          case OP_NOT:
2259          case OP_NOTI:
2260        case OP_EXACT:        case OP_EXACT:
2261          case OP_EXACTI:
2262          case OP_NOTEXACT:
2263          case OP_NOTEXACTI:
2264        case OP_UPTO:        case OP_UPTO:
2265          case OP_UPTOI:
2266          case OP_NOTUPTO:
2267          case OP_NOTUPTOI:
2268        case OP_MINUPTO:        case OP_MINUPTO:
2269          case OP_MINUPTOI:
2270          case OP_NOTMINUPTO:
2271          case OP_NOTMINUPTOI:
2272        case OP_POSUPTO:        case OP_POSUPTO:
2273          case OP_POSUPTOI:
2274          case OP_NOTPOSUPTO:
2275          case OP_NOTPOSUPTOI:
2276        case OP_STAR:        case OP_STAR:
2277          case OP_STARI:
2278          case OP_NOTSTAR:
2279          case OP_NOTSTARI:
2280        case OP_MINSTAR:        case OP_MINSTAR:
2281          case OP_MINSTARI:
2282          case OP_NOTMINSTAR:
2283          case OP_NOTMINSTARI:
2284        case OP_POSSTAR:        case OP_POSSTAR:
2285          case OP_POSSTARI:
2286          case OP_NOTPOSSTAR:
2287          case OP_NOTPOSSTARI:
2288        case OP_PLUS:        case OP_PLUS:
2289          case OP_PLUSI:
2290          case OP_NOTPLUS:
2291          case OP_NOTPLUSI:
2292        case OP_MINPLUS:        case OP_MINPLUS:
2293          case OP_MINPLUSI:
2294          case OP_NOTMINPLUS:
2295          case OP_NOTMINPLUSI:
2296        case OP_POSPLUS:        case OP_POSPLUS:
2297          case OP_POSPLUSI:
2298          case OP_NOTPOSPLUS:
2299          case OP_NOTPOSPLUSI:
2300        case OP_QUERY:        case OP_QUERY:
2301          case OP_QUERYI:
2302          case OP_NOTQUERY:
2303          case OP_NOTQUERYI:
2304        case OP_MINQUERY:        case OP_MINQUERY:
2305          case OP_MINQUERYI:
2306          case OP_NOTMINQUERY:
2307          case OP_NOTMINQUERYI:
2308        case OP_POSQUERY:        case OP_POSQUERY:
2309        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2310          case OP_NOTPOSQUERY:
2311          case OP_NOTPOSQUERYI:
2312          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2313        break;        break;
2314        }        }
2315    #else
2316        (void)(utf);  /* Keep compiler happy by referencing function argument */
2317  #endif  #endif
2318      }      }
2319    }    }
# Line 1374  for (;;) Line 2329  for (;;)
2329  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()
2330  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
2331  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2332  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
2333  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
2334    bracket whose current branch will already have been scanned.
2335    
2336  Arguments:  Arguments:
2337    code        points to start of search    code        points to start of search
2338    endcode     points to where to stop    endcode     points to where to stop
2339    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2340      cd          contains pointers to tables etc.
2341    
2342  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2343  */  */
2344    
2345  static BOOL  static BOOL
2346  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2347      BOOL utf, compile_data *cd)
2348  {  {
2349  register int c;  register int c;
2350  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2351       code < endcode;       code < endcode;
2352       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2353    {    {
2354    const uschar *ccode;    const pcre_uchar *ccode;
2355    
2356    c = *code;    c = *code;
2357    
2358    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2359      first_significant_code() with a TRUE final argument. */
2360    
2361    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2362      {      {
     code += _pcre_OP_lengths[c];  
2363      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2364      c = *code;      c = *code;
2365      continue;      continue;
2366      }      }
2367    
2368    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2369      implies a backward reference subroutine call, we can scan it. If it's a
2370      forward reference subroutine call, we can't. To detect forward reference
2371      we have to scan up the list that is kept in the workspace. This function is
2372      called only when doing the real compile, not during the pre-compile that
2373      measures the size of the compiled pattern. */
2374    
2375    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2376      {      {
2377        const pcre_uchar *scode;
2378      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2379    
2380      /* Scan a closed bracket */      /* Test for forward reference */
2381    
2382        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2383          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2384    
2385        /* Not a forward reference, test for completed backward reference */
2386    
2387      empty_branch = FALSE;      empty_branch = FALSE;
2388        scode = cd->start_code + GET(code, 1);
2389        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2390    
2391        /* Completed backwards reference */
2392    
2393      do      do
2394        {        {
2395        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2396            {
2397          empty_branch = TRUE;          empty_branch = TRUE;
2398            break;
2399            }
2400          scode += GET(scode, 1);
2401          }
2402        while (*scode == OP_ALT);
2403    
2404        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2405        continue;
2406        }
2407    
2408      /* Groups with zero repeats can of course be empty; skip them. */
2409    
2410      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2411          c == OP_BRAPOSZERO)
2412        {
2413        code += PRIV(OP_lengths)[c];
2414        do code += GET(code, 1); while (*code == OP_ALT);
2415        c = *code;
2416        continue;
2417        }
2418    
2419      /* A nested group that is already marked as "could be empty" can just be
2420      skipped. */
2421    
2422      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2423          c == OP_SCBRA || c == OP_SCBRAPOS)
2424        {
2425        do code += GET(code, 1); while (*code == OP_ALT);
2426        c = *code;
2427        continue;
2428        }
2429    
2430      /* For other groups, scan the branches. */
2431    
2432      if (c == OP_BRA  || c == OP_BRAPOS ||
2433          c == OP_CBRA || c == OP_CBRAPOS ||
2434          c == OP_ONCE || c == OP_ONCE_NC ||
2435          c == OP_COND)
2436        {
2437        BOOL empty_branch;
2438        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2439    
2440        /* If a conditional group has only one branch, there is a second, implied,
2441        empty branch, so just skip over the conditional, because it could be empty.
2442        Otherwise, scan the individual branches of the group. */
2443    
2444        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2445        code += GET(code, 1);        code += GET(code, 1);
2446        else
2447          {
2448          empty_branch = FALSE;
2449          do
2450            {
2451            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2452              empty_branch = TRUE;
2453            code += GET(code, 1);
2454            }
2455          while (*code == OP_ALT);
2456          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2457        }        }
2458      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2459      c = *code;      c = *code;
2460      continue;      continue;
2461      }      }
# Line 1433  for (code = first_significant_code(code Line 2464  for (code = first_significant_code(code
2464    
2465    switch (c)    switch (c)
2466      {      {
2467      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2468        cannot be represented just by a bit map. This includes negated single
2469        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2470        actual length is stored in the compiled code, so we must update "code"
2471        here. */
2472    
2473  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2474      case OP_XCLASS:      case OP_XCLASS:
2475      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2476      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2477  #endif  #endif
2478    
2479      case OP_CLASS:      case OP_CLASS:
2480      case OP_NCLASS:      case OP_NCLASS:
2481      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2482    
2483  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2484      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2485  #endif  #endif
2486    
# Line 1481  for (code = first_significant_code(code Line 2516  for (code = first_significant_code(code
2516      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2517      case OP_WORDCHAR:      case OP_WORDCHAR:
2518      case OP_ANY:      case OP_ANY:
2519        case OP_ALLANY:
2520      case OP_ANYBYTE:      case OP_ANYBYTE:
2521      case OP_CHAR:      case OP_CHAR:
2522      case OP_CHARNC:      case OP_CHARI:
2523      case OP_NOT:      case OP_NOT:
2524        case OP_NOTI:
2525      case OP_PLUS:      case OP_PLUS:
2526      case OP_MINPLUS:      case OP_MINPLUS:
2527      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1499  for (code = first_significant_code(code Line 2536  for (code = first_significant_code(code
2536      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2537      return FALSE;      return FALSE;
2538    
2539        /* These are going to continue, as they may be empty, but we have to
2540        fudge the length for the \p and \P cases. */
2541    
2542        case OP_TYPESTAR:
2543        case OP_TYPEMINSTAR:
2544        case OP_TYPEPOSSTAR:
2545        case OP_TYPEQUERY:
2546        case OP_TYPEMINQUERY:
2547        case OP_TYPEPOSQUERY:
2548        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2549        break;
2550    
2551        /* Same for these */
2552    
2553        case OP_TYPEUPTO:
2554        case OP_TYPEMINUPTO:
2555        case OP_TYPEPOSUPTO:
2556        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2557          code += 2;
2558        break;
2559    
2560      /* End of branch */      /* End of branch */
2561    
2562      case OP_KET:      case OP_KET:
2563      case OP_KETRMAX:      case OP_KETRMAX:
2564      case OP_KETRMIN:      case OP_KETRMIN:
2565        case OP_KETRPOS:
2566      case OP_ALT:      case OP_ALT:
2567      return TRUE;      return TRUE;
2568    
2569      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2570      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2571    
2572  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2573      case OP_STAR:      case OP_STAR:
2574        case OP_STARI:
2575      case OP_MINSTAR:      case OP_MINSTAR:
2576        case OP_MINSTARI:
2577      case OP_POSSTAR:      case OP_POSSTAR:
2578        case OP_POSSTARI:
2579      case OP_QUERY:      case OP_QUERY:
2580        case OP_QUERYI:
2581      case OP_MINQUERY:      case OP_MINQUERY:
2582        case OP_MINQUERYI:
2583      case OP_POSQUERY:      case OP_POSQUERY:
2584        case OP_POSQUERYI:
2585        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2586        break;
2587    
2588      case OP_UPTO:      case OP_UPTO:
2589        case OP_UPTOI:
2590      case OP_MINUPTO:      case OP_MINUPTO:
2591        case OP_MINUPTOI:
2592      case OP_POSUPTO:      case OP_POSUPTO:
2593      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2594        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2595      break;      break;
2596  #endif  #endif
2597    
2598        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2599        string. */
2600    
2601        case OP_MARK:
2602        case OP_PRUNE_ARG:
2603        case OP_SKIP_ARG:
2604        code += code[1];
2605        break;
2606    
2607        case OP_THEN_ARG:
2608        code += code[1];
2609        break;
2610    
2611        /* None of the remaining opcodes are required to match a character. */
2612    
2613        default:
2614        break;
2615      }      }
2616    }    }
2617    
# Line 1539  return TRUE; Line 2628  return TRUE;
2628  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
2629  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,
2630  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.
2631    This function is called only during the real compile, not during the
2632    pre-compile.
2633    
2634  Arguments:  Arguments:
2635    code        points to start of the recursion    code        points to start of the recursion
2636    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2637    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2638    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2639      cd          pointers to tables etc
2640    
2641  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2642  */  */
2643    
2644  static BOOL  static BOOL
2645  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2646    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2647  {  {
2648  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2649    {    {
2650    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2651        return FALSE;
2652    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2653    }    }
2654  return TRUE;  return TRUE;
# Line 1568  return TRUE; Line 2661  return TRUE;
2661  *************************************************/  *************************************************/
2662    
2663  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2664  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
2665  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2666  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2667    
2668    Originally, this function only recognized a sequence of letters between the
2669    terminators, but it seems that Perl recognizes any sequence of characters,
2670    though of course unknown POSIX names are subsequently rejected. Perl gives an
2671    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2672    didn't consider this to be a POSIX class. Likewise for [:1234:].
2673    
2674    The problem in trying to be exactly like Perl is in the handling of escapes. We
2675    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2676    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2677    below handles the special case of \], but does not try to do any other escape
2678    processing. This makes it different from Perl for cases such as [:l\ower:]
2679    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2680    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2681    I think.
2682    
2683    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2684    It seems that the appearance of a nested POSIX class supersedes an apparent
2685    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2686    a digit.
2687    
2688    In Perl, unescaped square brackets may also appear as part of class names. For
2689    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2690    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2691    seem right at all. PCRE does not allow closing square brackets in POSIX class
2692    names.
2693    
2694  Argument:  Arguments:
2695    ptr      pointer to the initial [    ptr      pointer to the initial [
2696    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2697    
2698  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2699  */  */
2700    
2701  static BOOL  static BOOL
2702  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2703  {  {
2704  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2705  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2706  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2707    {    {
2708    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2709    return TRUE;      ptr++;
2710      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2711      else
2712        {
2713        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2714          {
2715          *endptr = ptr;
2716          return TRUE;
2717          }
2718        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2719             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2720              ptr[1] == CHAR_EQUALS_SIGN) &&
2721            check_posix_syntax(ptr, endptr))
2722          return FALSE;
2723        }
2724    }    }
2725  return FALSE;  return FALSE;
2726  }  }
# Line 1613  Returns:     a value representing the na Line 2743  Returns:     a value representing the na
2743  */  */
2744    
2745  static int  static int
2746  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2747  {  {
2748    const char *pn = posix_names;
2749  register int yield = 0;  register int yield = 0;
2750  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2751    {    {
2752    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2753      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2754      pn += posix_name_lengths[yield] + 1;
2755    yield++;    yield++;
2756    }    }
2757  return -1;  return -1;
# Line 1634  return -1; Line 2766  return -1;
2766  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2767  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2768  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
2769  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
2770  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
2771  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
2772  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
2773  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2774    OP_END.
2775    
2776  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2777  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 2782  value in the reference (which is a group
2782  Arguments:  Arguments:
2783    group      points to the start of the group    group      points to the start of the group
2784    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2785    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2786    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2787    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
2788    
# Line 1657  Returns:     nothing Line 2790  Returns:     nothing
2790  */  */
2791    
2792  static void  static void
2793  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2794    uschar *save_hwm)    pcre_uchar *save_hwm)
2795  {  {
2796  uschar *ptr = group;  pcre_uchar *ptr = group;
2797  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2798    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2799    {    {
2800    int offset;    int offset;
2801    uschar *hc;    pcre_uchar *hc;
2802    
2803    /* 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
2804    reference. */    reference. */
# Line 1709  Arguments: Line 2843  Arguments:
2843  Returns:         new code pointer  Returns:         new code pointer
2844  */  */
2845    
2846  static uschar *  static pcre_uchar *
2847  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2848  {  {
2849  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2850  *code++ = 255;  *code++ = 255;
2851  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2852  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2853  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2854  }  }
2855    
2856    
# Line 1738  Returns:             nothing Line 2872  Returns:             nothing
2872  */  */
2873    
2874  static void  static void
2875  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2876  {  {
2877  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2878  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2879  }  }
2880    
# Line 1752  PUT(previous_callout, 2 + LINK_SIZE, len Line 2886  PUT(previous_callout, 2 + LINK_SIZE, len
2886  *************************************************/  *************************************************/
2887    
2888  /* 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
2889  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
2890  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
2891  start address.  start address. A character with multiple other cases is returned on its own
2892    with a special return value.
2893    
2894  Arguments:  Arguments:
2895    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1762  Arguments: Line 2897  Arguments:
2897    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2898    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2899    
2900  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2901                   0 when a range is returned
2902                  >0 the CASESET offset for char with multiple other cases
2903                    in this case, ocptr contains the original
2904  */  */
2905    
2906  static BOOL  static int
2907  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2908    unsigned int *odptr)    unsigned int *odptr)
2909  {  {
2910  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2911    int co;
2912    
2913    /* Find the first character that has an other case. If it has multiple other
2914    cases, return its case offset value. */
2915    
2916  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2917    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2918      if ((co = UCD_CASESET(c)) != 0)
2919        {
2920        *ocptr = c++;   /* Character that has the set */
2921        *cptr = c;      /* Rest of input range */
2922        return co;
2923        }
2924      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2925      }
2926    
2927  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2928    
2929  *ocptr = othercase;  *ocptr = othercase;
2930  next = othercase + 1;  next = othercase + 1;
2931    
2932  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2933    {    {
2934    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2935    next++;    next++;
2936    }    }
2937    
2938  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2939  *cptr = c;  *cptr = c;             /* Rest of input range */
2940    return 0;
2941    }
2942    
2943    
2944  return TRUE;  
2945    /*************************************************
2946    *        Check a character and a property        *
2947    *************************************************/
2948    
2949    /* This function is called by check_auto_possessive() when a property item
2950    is adjacent to a fixed character.
2951    
2952    Arguments:
2953      c            the character
2954      ptype        the property type
2955      pdata        the data for the type
2956      negated      TRUE if it's a negated property (\P or \p{^)
2957    
2958    Returns:       TRUE if auto-possessifying is OK
2959    */
2960    
2961    static BOOL
2962    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2963    {
2964    #ifdef SUPPORT_UCP
2965    const pcre_uint32 *p;
2966    #endif
2967    
2968    const ucd_record *prop = GET_UCD(c);
2969    
2970    switch(ptype)
2971      {
2972      case PT_LAMP:
2973      return (prop->chartype == ucp_Lu ||
2974              prop->chartype == ucp_Ll ||
2975              prop->chartype == ucp_Lt) == negated;
2976    
2977      case PT_GC:
2978      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2979    
2980      case PT_PC:
2981      return (pdata == prop->chartype) == negated;
2982    
2983      case PT_SC:
2984      return (pdata == prop->script) == negated;
2985    
2986      /* These are specials */
2987    
2988      case PT_ALNUM:
2989      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2990              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2991    
2992      case PT_SPACE:    /* Perl space */
2993      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2994              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2995              == negated;
2996    
2997      case PT_PXSPACE:  /* POSIX space */
2998      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2999              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3000              c == CHAR_FF || c == CHAR_CR)
3001              == negated;
3002    
3003      case PT_WORD:
3004      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3005              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3006              c == CHAR_UNDERSCORE) == negated;
3007    
3008    #ifdef SUPPORT_UCP
3009      case PT_CLIST:
3010      p = PRIV(ucd_caseless_sets) + prop->caseset;
3011      for (;;)
3012        {
3013        if ((unsigned int)c < *p) return !negated;
3014        if ((unsigned int)c == *p++) return negated;
3015        }
3016      break;  /* Control never reaches here */
3017    #endif
3018      }
3019    
3020    return FALSE;
3021  }  }
3022  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3023    
# Line 1803  whether the next thing could possibly ma Line 3032  whether the next thing could possibly ma
3032  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3033    
3034  Arguments:  Arguments:
3035    op_code       the repeated op code    previous      pointer to the repeated opcode
3036    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  
3037    ptr           next character in pattern    ptr           next character in pattern
3038    options       options bits    options       options bits
3039    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1815  Returns:        TRUE if possessifying is Line 3042  Returns:        TRUE if possessifying is
3042  */  */
3043    
3044  static BOOL  static BOOL
3045  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3046    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3047  {  {
3048  int next;  pcre_int32 c = NOTACHAR; // FIXMEchpe pcre_uint32
3049    pcre_int32 next;
3050    int escape;
3051    int op_code = *previous++;
3052    
3053  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3054    
# Line 1826  if ((options & PCRE_EXTENDED) != 0) Line 3056  if ((options & PCRE_EXTENDED) != 0)
3056    {    {
3057    for (;;)    for (;;)
3058      {      {
3059      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3060      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3061        {        {
3062        while (*(++ptr) != 0)        ptr++;
3063          while (*ptr != 0)
3064            {
3065          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3066            ptr++;
3067    #ifdef SUPPORT_UTF
3068            if (utf) FORWARDCHAR(ptr);
3069    #endif
3070            }
3071        }        }
3072      else break;      else break;
3073      }      }
# Line 1839  if ((options & PCRE_EXTENDED) != 0) Line 3076  if ((options & PCRE_EXTENDED) != 0)
3076  /* 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
3077  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3078    
3079  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3080    {    {
3081    int temperrorcode = 0;    int temperrorcode = 0;
3082    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3083    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3084    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3085    }    }
3086    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3087    {    {
3088  #ifdef SUPPORT_UTF8    escape = 0;
3089    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3090      if (utf) { GETCHARINC(next, ptr); } else
3091  #endif  #endif
3092    next = *ptr++;    next = *ptr++;
3093    }    }
   
3094  else return FALSE;  else return FALSE;
3095    
3096  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1863  if ((options & PCRE_EXTENDED) != 0) Line 3099  if ((options & PCRE_EXTENDED) != 0)
3099    {    {
3100    for (;;)    for (;;)
3101      {      {
3102      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3103      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3104        {        {
3105        while (*(++ptr) != 0)        ptr++;
3106          while (*ptr != 0)
3107            {
3108          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3109            ptr++;
3110    #ifdef SUPPORT_UTF
3111            if (utf) FORWARDCHAR(ptr);
3112    #endif
3113            }
3114        }        }
3115      else break;      else break;
3116      }      }
# Line 1875  if ((options & PCRE_EXTENDED) != 0) Line 3118  if ((options & PCRE_EXTENDED) != 0)
3118    
3119  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3120    
3121  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3122    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3123        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. */  
   
3124    
3125  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3126    
3127  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3128        op_code == OP_NOT || op_code == OP_NOTI)
3129      //if (escape == 0) switch(op_code)
3130    {    {
3131    case OP_CHAR:  #ifdef SUPPORT_UTF
3132  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3133    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3134      c = *previous;
3135  #endif  #endif
3136    return item != next;    }
3137    
3138    /* 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
3139    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3140    
3141    case OP_CHARNC:  if (escape == 0)
3142  #ifdef SUPPORT_UTF8    {
3143    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
3144      case, which maps to the special PT_CLIST property. Check this first. */
3145    
3146    #ifdef SUPPORT_UCP
3147      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3148        {
3149        int ocs = UCD_CASESET(next);
3150        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3151        }
3152  #endif  #endif
3153    if (item == next) return FALSE;  
3154  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3155      {      {
3156      unsigned int othercase;      case OP_CHAR:
3157      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3158    
3159        /* For CHARI (caseless character) we must check the other case. If we have
3160        Unicode property support, we can use it to test the other case of
3161        high-valued characters. We know that next can have only one other case,
3162        because multi-other-case characters are dealt with above. */
3163    
3164        case OP_CHARI:
3165        if (c == next) return FALSE;
3166    #ifdef SUPPORT_UTF
3167        if (utf)
3168          {
3169          unsigned int othercase;
3170          if (next < 128) othercase = cd->fcc[next]; else
3171  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3172      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3173  #else  #else
3174      othercase = NOTACHAR;        othercase = NOTACHAR;
3175  #endif  #endif
3176      return (unsigned int)item != othercase;        return (unsigned int)c != othercase;
3177      }        }
3178    else      else
3179  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3180    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3181    
3182    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3183        return c == next;
3184    case OP_NOT:  
3185    if (next < 0) return FALSE;  /* Not a character */      case OP_NOTI:
3186    if (item == next) return TRUE;      if (c == next) return TRUE;
3187    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3188  #ifdef SUPPORT_UTF8      if (utf)
3189    if (utf8)        {
3190      {        unsigned int othercase;
3191      unsigned int othercase;        if (next < 128) othercase = cd->fcc[next]; else
     if (next < 128) othercase = cd->fcc[next]; else  
3192  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3193      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3194  #else  #else
3195      othercase = NOTACHAR;        othercase = NOTACHAR;
3196  #endif  #endif
3197      return (unsigned int)item == othercase;        return (unsigned int)c == othercase;
3198      }        }
3199    else      else
3200  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3201    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3202    
3203    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3204    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3205    
3206    case OP_NOT_DIGIT:      case OP_DIGIT:
3207    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3208    
3209    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3210    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3211    
3212    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3213    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3214    
3215    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3216    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3217    
3218    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3219    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3220    
3221    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3222    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3223    switch(next)  
3224      {      case OP_HSPACE:
3225      case 0x09:      case OP_NOT_HSPACE:
3226      case 0x20:      switch(next)
3227      case 0xa0:        {
3228      case 0x1680:        HSPACE_CASES:
3229      case 0x180e:        return op_code == OP_NOT_HSPACE;
3230      case 0x2000:  
3231      case 0x2001:        default:
3232      case 0x2002:        return op_code != OP_NOT_HSPACE;
3233      case 0x2003:        }
3234      case 0x2004:  
3235      case 0x2005:      case OP_ANYNL:
3236      case 0x2006:      case OP_VSPACE:
3237      case 0x2007:      case OP_NOT_VSPACE:
3238      case 0x2008:      switch(next)
3239      case 0x2009:        {
3240      case 0x200A:        VSPACE_CASES:
3241      case 0x202f:        return op_code == OP_NOT_VSPACE;
3242      case 0x205f:  
3243      case 0x3000:        default:
3244      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3245      default:        }
3246      return op_code == OP_HSPACE;  
3247      }  #ifdef SUPPORT_UCP
3248        case OP_PROP:
3249        return check_char_prop(next, previous[0], previous[1], FALSE);
3250    
3251        case OP_NOTPROP:
3252        return check_char_prop(next, previous[0], previous[1], TRUE);
3253    #endif
3254    
   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;  
3255      default:      default:
3256      return op_code == OP_VSPACE;      return FALSE;
3257      }      }
   
   default:  
   return FALSE;  
3258    }    }
3259    
3260    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3261  /* 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
3262    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3263    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3264    replaced by OP_PROP codes when PCRE_UCP is set. */
3265    
3266  switch(op_code)  switch(op_code)
3267    {    {
3268    case OP_CHAR:    case OP_CHAR:
3269    case OP_CHARNC:    case OP_CHARI:
3270  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3271      {      {
3272      case ESC_d:      case ESC_d:
3273      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3274    
3275      case ESC_D:      case ESC_D:
3276      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3277    
3278      case ESC_s:      case ESC_s:
3279      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3280    
3281      case ESC_S:      case ESC_S:
3282      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3283    
3284      case ESC_w:      case ESC_w:
3285      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3286    
3287      case ESC_W:      case ESC_W:
3288      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3289    
3290      case ESC_h:      case ESC_h:
3291      case ESC_H:      case ESC_H:
3292      switch(item)      switch(c)
3293        {        {
3294        case 0x09:        HSPACE_CASES:
3295        case 0x20:        return escape != ESC_h;
3296        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;  
3297        default:        default:
3298        return -next == ESC_h;        return escape == ESC_h;
3299        }        }
3300    
3301      case ESC_v:      case ESC_v:
3302      case ESC_V:      case ESC_V:
3303      switch(item)      switch(c)
3304        {        {
3305        case 0x0a:        VSPACE_CASES:
3306        case 0x0b:        return escape != ESC_v;
3307        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3308        default:        default:
3309        return -next == ESC_v;        return escape == ESC_v;
3310        }        }
3311    
3312        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3313        their substitutions and process them. The result will always be either
3314        ESC_p or ESC_P. Then fall through to process those values. */
3315    
3316    #ifdef SUPPORT_UCP
3317        case ESC_du:
3318        case ESC_DU:
3319        case ESC_wu:
3320        case ESC_WU:
3321        case ESC_su:
3322        case ESC_SU:
3323          {
3324          int temperrorcode = 0;
3325          ptr = substitutes[escape - ESC_DU];
3326          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3327          if (temperrorcode != 0) return FALSE;
3328          ptr++;    /* For compatibility */
3329          }
3330        /* Fall through */
3331    
3332        case ESC_p:
3333        case ESC_P:
3334          {
3335          int ptype, pdata, errorcodeptr;
3336          BOOL negated;
3337    
3338          ptr--;      /* Make ptr point at the p or P */
3339          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3340          if (ptype < 0) return FALSE;
3341          ptr++;      /* Point past the final curly ket */
3342    
3343          /* If the property item is optional, we have to give up. (When generated
3344          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3345          to the original \d etc. At this point, ptr will point to a zero byte. */
3346    
3347          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3348            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3349              return FALSE;
3350    
3351          /* Do the property check. */
3352    
3353          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3354          }
3355    #endif
3356    
3357      default:      default:
3358      return FALSE;      return FALSE;
3359      }      }
3360    
3361      /* In principle, support for Unicode properties should be integrated here as
3362      well. It means re-organizing the above code so as to get hold of the property
3363      values before switching on the op-code. However, I wonder how many patterns
3364      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3365      these op-codes are never generated.) */
3366    
3367    case OP_DIGIT:    case OP_DIGIT:
3368    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3369           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3370    
3371    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3372    return next == -ESC_d;    return escape == ESC_d;
3373    
3374    case OP_WHITESPACE:    case OP_WHITESPACE:
3375    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3376    
3377    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3378    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3379    
3380    case OP_HSPACE:    case OP_HSPACE:
3381    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3382             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3383    
3384    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3385    return next == -ESC_h;    return escape == ESC_h;
3386    
3387    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3388      case OP_ANYNL:
3389    case OP_VSPACE:    case OP_VSPACE:
3390    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3391    
3392    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3393    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3394    
3395    case OP_WORDCHAR:    case OP_WORDCHAR:
3396    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3397             escape == ESC_v || escape == ESC_R;
3398    
3399    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3400    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3401    
3402    default:    default:
3403    return FALSE;    return FALSE;
# Line 2127  switch(op_code) Line 3409  switch(op_code)
3409    
3410    
3411  /*************************************************  /*************************************************
3412    *        Add a character or range to a class     *
3413    *************************************************/
3414    
3415    /* This function packages up the logic of adding a character or range of
3416    characters to a class. The character values in the arguments will be within the
3417    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3418    mutually recursive with the function immediately below.
3419    
3420    Arguments:
3421      classbits     the bit map for characters < 256
3422      uchardptr     points to the pointer for extra data
3423      options       the options word
3424      cd            contains pointers to tables etc.
3425      start         start of range character
3426      end           end of range character
3427    
3428    Returns:        the number of < 256 characters added
3429                    the pointer to extra data is updated
3430    */
3431    
3432    static int
3433    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3434      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3435    {
3436    pcre_uint32 c;
3437    int n8 = 0;
3438    
3439    /* If caseless matching is required, scan the range and process alternate
3440    cases. In Unicode, there are 8-bit characters that have alternate cases that
3441    are greater than 255 and vice-versa. Sometimes we can just extend the original
3442    range. */
3443    
3444    if ((options & PCRE_CASELESS) != 0)
3445      {
3446    #ifdef SUPPORT_UCP
3447      if ((options & PCRE_UTF8) != 0)
3448        {
3449        int rc;
3450        pcre_uint32 oc, od;
3451    
3452        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3453        c = start;
3454    
3455        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3456          {
3457          /* Handle a single character that has more than one other case. */
3458    
3459          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3460            PRIV(ucd_caseless_sets) + rc, oc);
3461    
3462          /* Do nothing if the other case range is within the original range. */
3463    
3464          else if (oc >= start && od <= end) continue;
3465    
3466          /* Extend the original range if there is overlap, noting that if oc < c, we
3467          can't have od > end because a subrange is always shorter than the basic
3468          range. Otherwise, use a recursive call to add the additional range. */
3469    
3470          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3471          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3472          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3473          }
3474        }
3475      else
3476    #endif  /* SUPPORT_UCP */
3477    
3478      /* Not UTF-mode, or no UCP */
3479    
3480      for (c = start; c <= end && c < 256; c++)
3481        {
3482        SETBIT(classbits, cd->fcc[c]);
3483        n8++;
3484        }
3485      }
3486    
3487    /* Now handle the original range. Adjust the final value according to the bit
3488    length - this means that the same lists of (e.g.) horizontal spaces can be used
3489    in all cases. */
3490    
3491    #if defined COMPILE_PCRE8
3492    #ifdef SUPPORT_UTF
3493      if ((options & PCRE_UTF8) == 0)
3494    #endif
3495      if (end > 0xff) end = 0xff;
3496    
3497    #elif defined COMPILE_PCRE16
3498    #ifdef SUPPORT_UTF
3499      if ((options & PCRE_UTF16) == 0)
3500    #endif
3501      if (end > 0xffff) end = 0xffff;
3502    
3503    #endif /* COMPILE_PCRE[8|16] */
3504    
3505    /* If all characters are less than 256, use the bit map. Otherwise use extra
3506    data. */
3507    
3508    if (end < 0x100)
3509      {
3510      for (c = start; c <= end; c++)
3511        {
3512        n8++;
3513        SETBIT(classbits, c);
3514        }
3515      }
3516    
3517    else
3518      {
3519      pcre_uchar *uchardata = *uchardptr;
3520    
3521    #ifdef SUPPORT_UTF
3522      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3523        {
3524        if (start < end)
3525          {
3526          *uchardata++ = XCL_RANGE;
3527          uchardata += PRIV(ord2utf)(start, uchardata);
3528          uchardata += PRIV(ord2utf)(end, uchardata);
3529          }
3530        else if (start == end)
3531          {
3532          *uchardata++ = XCL_SINGLE;
3533          uchardata += PRIV(ord2utf)(start, uchardata);
3534          }
3535        }
3536      else
3537    #endif  /* SUPPORT_UTF */
3538    
3539      /* Without UTF support, character values are constrained by the bit length,
3540      and can only be > 256 for 16-bit and 32-bit libraries. */
3541    
3542    #ifdef COMPILE_PCRE8
3543        {}
3544    #else
3545      if (start < end)
3546        {
3547        *uchardata++ = XCL_RANGE;
3548        *uchardata++ = start;
3549        *uchardata++ = end;
3550        }
3551      else if (start == end)
3552        {
3553        *uchardata++ = XCL_SINGLE;
3554        *uchardata++ = start;
3555        }
3556    #endif
3557    
3558      *uchardptr = uchardata;   /* Updata extra data pointer */
3559      }
3560    
3561    return n8;    /* Number of 8-bit characters */
3562    }
3563    
3564    
3565    
3566    
3567    /*************************************************
3568    *        Add a list of characters to a class     *
3569    *************************************************/
3570    
3571    /* This function is used for adding a list of case-equivalent characters to a
3572    class, and also for adding a list of horizontal or vertical whitespace. If the
3573    list is in order (which it should be), ranges of characters are detected and
3574    handled appropriately. This function is mutually recursive with the function
3575    above.
3576    
3577    Arguments:
3578      classbits     the bit map for characters < 256
3579      uchardptr     points to the pointer for extra data
3580      options       the options word
3581      cd            contains pointers to tables etc.
3582      p             points to row of 32-bit values, terminated by NOTACHAR
3583      except        character to omit; this is used when adding lists of
3584                      case-equivalent characters to avoid including the one we
3585                      already know about
3586    
3587    Returns:        the number of < 256 characters added
3588                    the pointer to extra data is updated
3589    */
3590    
3591    static int
3592    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3593      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3594    {
3595    int n8 = 0;
3596    while (p[0] < NOTACHAR)
3597      {
3598      int n = 0;
3599      if (p[0] != except)
3600        {
3601        while(p[n+1] == p[0] + n + 1) n++;
3602        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3603        }
3604      p += n + 1;
3605      }
3606    return n8;
3607    }
3608    
3609    
3610    
3611    /*************************************************
3612    *    Add characters not in a list to a class     *
3613    *************************************************/
3614    
3615    /* This function is used for adding the complement of a list of horizontal or
3616    vertical whitespace to a class. The list must be in order.
3617    
3618    Arguments:
3619      classbits     the bit map for characters < 256
3620      uchardptr     points to the pointer for extra data
3621      options       the options word
3622      cd            contains pointers to tables etc.
3623      p             points to row of 32-bit values, terminated by NOTACHAR
3624    
3625    Returns:        the number of < 256 characters added
3626                    the pointer to extra data is updated
3627    */
3628    
3629    static int
3630    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3631      int options, compile_data *cd, const pcre_uint32 *p)
3632    {
3633    BOOL utf = (options & PCRE_UTF8) != 0;
3634    int n8 = 0;
3635    if (p[0] > 0)
3636      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3637    while (p[0] < NOTACHAR)
3638      {
3639      while (p[1] == p[0] + 1) p++;
3640      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3641        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3642      p++;
3643      }
3644    return n8;
3645    }
3646    
3647    
3648    
3649    /*************************************************
3650  *           Compile one branch                   *  *           Compile one branch                   *
3651  *************************************************/  *************************************************/
3652    
# Line 2141  Arguments: Line 3661  Arguments:
3661    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3662    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3663    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3664    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3665    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3666    bcptr          points to current branch chain    bcptr          points to current branch chain
3667      cond_depth     conditional nesting depth
3668    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3669    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3670                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2153  Returns:         TRUE on success Line 3674  Returns:         TRUE on success
3674  */  */
3675    
3676  static BOOL  static BOOL
3677  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3678    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3679      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3680    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3681  {  {
3682  int repeat_type, op_type;  int repeat_type, op_type;
3683  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3684  int bravalue = 0;  int bravalue = 0;
3685  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3686  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3687  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3688  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3689  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3690  int after_manual_callout = 0;  int after_manual_callout = 0;
3691  int length_prevgroup = 0;  int length_prevgroup = 0;
3692  register int c;  register int c;
3693  register uschar *code = *codeptr;  int escape;
3694  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3695  uschar *orig_code = code;  pcre_uchar *last_code = code;
3696  uschar *tempcode;  pcre_uchar *orig_code = code;
3697    pcre_uchar *tempcode;
3698  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3699  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3700  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3701  const uschar *tempptr;  const pcre_uchar *tempptr;
3702  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3703  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3704  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3705  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3706    pcre_uint8 classbits[32];
3707  #ifdef SUPPORT_UTF8  
3708  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3709  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3710  uschar *class_utf8data;  dynamically as we process the pattern. */
3711  uschar utf8_char[6];  
3712    #ifdef SUPPORT_UTF
3713    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3714    BOOL utf = (options & PCRE_UTF8) != 0;
3715    pcre_uchar utf_chars[6];
3716  #else  #else
3717  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3718  uschar *utf8_char = NULL;  #endif
3719    
3720    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3721    class_uchardata always so that it can be passed to add_to_class() always,
3722    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3723    alternative calls for the different cases. */
3724    
3725    pcre_uchar *class_uchardata;
3726    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3727