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

Diff of /code/trunk/pcre_compile.c

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

revision 182 by ph10, Wed Jun 13 15:09:54 2007 UTC revision 1061 by chpe, Tue Oct 16 15:54:02 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|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
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* 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  /*************************************************  /*************************************************
# Line 77  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    /* 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 COMPILE_WORK_SIZE (4096)  #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 120  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,      0,       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 130  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    /* 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 uschar posix_name_lengths[] = {  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 169  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 176  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 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",    "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 268  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 304  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 340  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 376  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 389  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 n is returned as
754  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  negative n. When UTF-8 is enabled, a positive value greater than 255 may
755  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  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                   negative => a back reference
770                     on error, errorcodeptr is set
771  */  */
772    
773  static int  static int
774  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *chptr, int *errorcodeptr,
775    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
776  {  {
777  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
778  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
779  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
780    pcre_uint32 c;
781    int escape = 0;
782    int i;
783    
784  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
785  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 423  ptr--;                            /* Set Line 788  ptr--;                            /* Set
788    
789  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
790    
791  /* 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
792  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.
793  Otherwise further processing may be required. */  Otherwise further processing may be required. */
794    
795  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
796  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
797  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
798    else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = i; else escape = -i; }
799    
800  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
801  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
802  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
803    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = i; else escape = -i; }
804  #endif  #endif
805    
806  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
807    
808  else  else
809    {    {
810    const uschar *oldptr;    const pcre_uchar *oldptr;
811    BOOL braced, negated;    BOOL braced, negated;
812    
813    switch (c)    switch (c)
# Line 448  else Line 815  else
815      /* 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
816      error. */      error. */
817    
818      case 'l':      case CHAR_l:
819      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
820      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
821      break;      break;
822    
823      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
824      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
825      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
826      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
827      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
828      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
829            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
830      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
831      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
832        {          {
833        const uschar *p;          c = 0;
834        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
835          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
836        if (*p != 0 && *p != '}')            register pcre_uint32 cc = *(++ptr);
837    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
838              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
839              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
840    #else           /* EBCDIC coding */
841              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
842              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
843    #endif
844              }
845    
846    #if defined COMPILE_PCRE8
847            if (c > (utf ? 0x10ffff : 0xff))
848    #elif defined COMPILE_PCRE16
849            if (c > (utf ? 0x10ffff : 0xffff))
850    #elif defined COMPILE_PCRE32
851            if (utf && c > 0x10ffff)
852    #endif
853              {
854              *errorcodeptr = ERR76;
855              }
856            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
857            }
858          }
859        else
860          *errorcodeptr = ERR37;
861        break;
862    
863        case CHAR_U:
864        /* In JavaScript, \U is an uppercase U letter. */
865        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
866        break;
867    
868        /* In a character class, \g is just a literal "g". Outside a character
869        class, \g must be followed by one of a number of specific things:
870    
871        (1) A number, either plain or braced. If positive, it is an absolute
872        backreference. If negative, it is a relative backreference. This is a Perl
873        5.10 feature.
874    
875        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
876        is part of Perl's movement towards a unified syntax for back references. As
877        this is synonymous with \k{name}, we fudge it up by pretending it really
878        was \k.
879    
880        (3) For Oniguruma compatibility we also support \g followed by a name or a
881        number either in angle brackets or in single quotes. However, these are
882        (possibly recursive) subroutine calls, _not_ backreferences. Just return
883        the ESC_g code (cf \k). */
884    
885        case CHAR_g:
886        if (isclass) break;
887        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
888          {
889          escape = ESC_g;
890          break;
891          }
892    
893        /* Handle the Perl-compatible cases */
894    
895        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
896          {
897          const pcre_uchar *p;
898          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
899            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
900          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
901          {          {
902          c = -ESC_k;          escape = ESC_k;
903          break;          break;
904          }          }
905        braced = TRUE;        braced = TRUE;
# Line 479  else Line 907  else
907        }        }
908      else braced = FALSE;      else braced = FALSE;
909    
910      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
911        {        {
912        negated = TRUE;        negated = TRUE;
913        ptr++;        ptr++;
914        }        }
915      else negated = FALSE;      else negated = FALSE;
916    
917        /* The integer range is limited by the machine's int representation. */
918      c = 0;      c = 0;
919      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
920        c = c * 10 + *(++ptr) - '0';        {
921          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
922            {
923            c = -1;
924            break;
925            }
926          c = c * 10 + *(++ptr) - CHAR_0;
927          }
928        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
929          {
930          while (IS_DIGIT(ptr[1]))
931            ptr++;
932          *errorcodeptr = ERR61;
933          break;
934          }
935    
936      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
937        {        {
938        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
939        return 0;        break;
940          }
941    
942        if (c == 0)
943          {
944          *errorcodeptr = ERR58;
945          break;
946        }        }
947    
948      if (negated)      if (negated)
# Line 501  else Line 950  else
950        if (c > bracount)        if (c > bracount)
951          {          {
952          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
953          return 0;          break;
954          }          }
955        c = bracount - (c - 1);        c = bracount - (c - 1);
956        }        }
957    
958      c = -(ESC_REF + c);      escape = -c;
959      break;      break;
960    
961      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 521  else Line 970  else
970      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
971      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
972    
973      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:
974      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
975    
976      if (!isclass)      if (!isclass)
977        {        {
978        oldptr = ptr;        oldptr = ptr;
979        c -= '0';        /* The integer range is limited by the machine's int representation. */
980        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
981          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
982            {
983            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
984              {
985              c = -1;
986              break;
987              }
988            c = c * 10 + *(++ptr) - CHAR_0;
989            }
990          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
991            {
992            while (IS_DIGIT(ptr[1]))
993              ptr++;
994            *errorcodeptr = ERR61;
995            break;
996            }
997        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
998          {          {
999          c = -(ESC_REF + c);          escape = -c;
1000          break;          break;
1001          }          }
1002        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 542  else Line 1006  else
1006      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.
1007      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1008    
1009      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1010        {        {
1011        ptr--;        ptr--;
1012        c = 0;        c = 0;
# Line 552  else Line 1016  else
1016      /* \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
1017      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
1018      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
1019      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,
1020      than 3 octal digits. */      but no more than 3 octal digits. */
1021    
1022      case '0':      case CHAR_0:
1023      c -= '0';      c -= CHAR_0;
1024      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1025          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1026      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1027        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1028    #endif
1029      break;      break;
1030    
1031      /* \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
1032      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.
1033      treated as a data character. */      If not, { is treated as a data character. */
1034    
1035        case CHAR_x:
1036        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1037          {
1038          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1039          Otherwise it is a lowercase x letter. */
1040          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1041            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1042            {
1043            c = 0;
1044            for (i = 0; i < 2; ++i)
1045              {
1046              register pcre_uint32 cc = *(++ptr);
1047    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1048              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1049              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1050    #else           /* EBCDIC coding */
1051              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1052              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1053    #endif
1054              }
1055            }
1056          break;
1057          }
1058    
1059      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1060        {        {
1061        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
1062        int count = 0;        BOOL overflow;
1063    
1064        c = 0;        c = 0;
1065        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1066          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1067          {          {
1068          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1069          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1070          count++;  
1071    #ifdef COMPILE_PCRE32
1072  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1073          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1074          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1075    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1076            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1077            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1078  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1079          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1080          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1081    #endif
1082    
1083    #if defined COMPILE_PCRE8
1084            if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }
1085    #elif defined COMPILE_PCRE16
1086            if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }
1087    #elif defined COMPILE_PCRE32
1088            if (utf && c > 0x10ffff) { overflow = TRUE; break; }
1089  #endif  #endif
1090          }          }
1091    
1092        if (*pt == '}')        if (overflow)
1093          {          {
1094          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1095            *errorcodeptr = ERR34;
1096            }
1097    
1098          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1099            {
1100            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1101          ptr = pt;          ptr = pt;
1102          break;          break;
1103          }          }
# Line 602  else Line 1109  else
1109      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1110    
1111      c = 0;      c = 0;
1112      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1113        {        {
1114        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1115        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1116  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1117        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1118        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1119  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1120        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1121        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1122  #endif  #endif
1123        }        }
1124      break;      break;
1125    
1126      /* 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.
1127      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
1128        coding is ASCII-specific, but then the whole concept of \cx is
1129      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1130    
1131      case 'c':      case CHAR_c:
1132      c = *(++ptr);      c = *(++ptr);
1133      if (c == 0)      if (c == 0)
1134        {        {
1135        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1136        return 0;        break;
1137        }        }
1138    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1139  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1140      if (c >= 'a' && c <= 'z') c -= 32;        {
1141          *errorcodeptr = ERR68;
1142          break;
1143          }
1144        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1145      c ^= 0x40;      c ^= 0x40;
1146  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1147      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1148      c ^= 0xC0;      c ^= 0xC0;
1149  #endif  #endif
1150      break;      break;
1151    
1152      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1153      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1154      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
1155      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
1156      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1157    
1158      default:      default:
1159      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 654  else Line 1166  else
1166      }      }
1167    }    }
1168    
1169  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1170  return c;  newline". PCRE does not support \N{name}. However, it does support
1171  }  quantification such as \N{2,3}. */
1172    
1173    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1174         !is_counted_repeat(ptr+2))
1175      *errorcodeptr = ERR37;
1176    
1177    /* If PCRE_UCP is set, we change the values for \d etc. */
1178    
1179    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1180      escape += (ESC_DU - ESC_D);
1181    
1182    /* Set the pointer to the final character before returning. */
1183    
1184    *ptrptr = ptr;
1185    *chptr = c;
1186    return escape;
1187    }
1188    
1189  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1190  /*************************************************  /*************************************************
# Line 680  Returns:         type value from ucp_typ Line 1206  Returns:         type value from ucp_typ
1206  */  */
1207    
1208  static int  static int
1209  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1210  {  {
1211  int c, i, bot, top;  int c, i, bot, top;
1212  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1213  char name[32];  pcre_uchar name[32];
1214    
1215  c = *(++ptr);  c = *(++ptr);
1216  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 694  if (c == 0) goto ERROR_RETURN; Line 1220  if (c == 0) goto ERROR_RETURN;
1220  /* \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
1221  negation. */  negation. */
1222    
1223  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1224    {    {
1225    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1226      {      {
1227      *negptr = TRUE;      *negptr = TRUE;
1228      ptr++;      ptr++;
1229      }      }
1230    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1231      {      {
1232      c = *(++ptr);      c = *(++ptr);
1233      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1234      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1235      name[i] = c;      name[i] = c;
1236      }      }
1237    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1238    name[i] = 0;    name[i] = 0;
1239    }    }
1240    
# Line 725  else Line 1251  else
1251  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1252    
1253  bot = 0;  bot = 0;
1254  top = _pcre_utt_size;  top = PRIV(utt_size);
1255    
1256  while (bot < top)  while (bot < top)
1257    {    {
1258    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1259    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1260    if (c == 0)    if (c == 0)
1261      {      {
1262      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1263      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1264      }      }
1265    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1266    }    }
# Line 754  return -1; Line 1280  return -1;
1280    
1281    
1282  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1283  *         Read repeat counts                     *  *         Read repeat counts                     *
1284  *************************************************/  *************************************************/
1285    
# Line 805  Returns:         pointer to '}' on succe Line 1298  Returns:         pointer to '}' on succe
1298                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1299  */  */
1300    
1301  static const uschar *  static const pcre_uchar *
1302  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)
1303  {  {
1304  int min = 0;  int min = 0;
1305  int max = -1;  int max = -1;
# Line 814  int max = -1; Line 1307  int max = -1;
1307  /* 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
1308  an integer overflow. */  an integer overflow. */
1309    
1310  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1311  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1312    {    {
1313    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 824  if (min < 0 || min > 65535) Line 1317  if (min < 0 || min > 65535)
1317  /* 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.
1318  Also, max must not be less than min. */  Also, max must not be less than min. */
1319    
1320  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1321    {    {
1322    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1323      {      {
1324      max = 0;      max = 0;
1325      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1326      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1327        {        {
1328        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 854  return p; Line 1347  return p;
1347    
1348    
1349  /*************************************************  /*************************************************
1350  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1351  *************************************************/  *************************************************/
1352    
1353  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1354    top-level call starts at the beginning of the pattern. All other calls must
1355    start at a parenthesis. It scans along a pattern's text looking for capturing
1356  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
1357  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
1358  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
1359  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1360  be terminated by '>' because that is checked in the first pass.  
1361    This function was originally called only from the second pass, in which we know
1362    that if (?< or (?' or (?P< is encountered, the name will be correctly
1363    terminated because that is checked in the first pass. There is now one call to
1364    this function in the first pass, to check for a recursive back reference by
1365    name (so that we can make the whole group atomic). In this case, we need check
1366    only up to the current position in the pattern, and that is still OK because
1367    and previous occurrences will have been checked. To make this work, the test
1368    for "end of pattern" is a check against cd->end_pattern in the main loop,
1369    instead of looking for a binary zero. This means that the special first-pass
1370    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1371    processing items within the loop are OK, because afterwards the main loop will
1372    terminate.)
1373    
1374  Arguments:  Arguments:
1375    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1376    count        current count of capturing parens so far encountered    cd           compile background data
1377    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1378    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1379    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1380      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1381      count        pointer to the current capturing subpattern number (updated)
1382    
1383  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1384  */  */
1385    
1386  static int  static int
1387  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,
1388    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1389  {  {
1390  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1391    int start_count = *count;
1392    int hwm_count = start_count;
1393    BOOL dup_parens = FALSE;
1394    
1395    /* If the first character is a parenthesis, check on the type of group we are
1396    dealing with. The very first call may not start with a parenthesis. */
1397    
1398  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1399    {    {
1400    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1401    
1402      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1403    
1404      /* Handle a normal, unnamed capturing parenthesis. */
1405    
1406      else if (ptr[1] != CHAR_QUESTION_MARK)
1407        {
1408        *count += 1;
1409        if (name == NULL && *count == lorn) return *count;
1410        ptr++;
1411        }
1412    
1413      /* All cases now have (? at the start. Remember when we are in a group
1414      where the parenthesis numbers are duplicated. */
1415    
1416      else if (ptr[2] == CHAR_VERTICAL_LINE)
1417        {
1418        ptr += 3;
1419        dup_parens = TRUE;
1420        }
1421    
1422      /* Handle comments; all characters are allowed until a ket is reached. */
1423    
1424      else if (ptr[2] == CHAR_NUMBER_SIGN)
1425        {
1426        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1427        goto FAIL_EXIT;
1428        }
1429    
1430      /* Handle a condition. If it is an assertion, just carry on so that it
1431      is processed as normal. If not, skip to the closing parenthesis of the
1432      condition (there can't be any nested parens). */
1433    
1434      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1435        {
1436        ptr += 2;
1437        if (ptr[1] != CHAR_QUESTION_MARK)
1438          {
1439          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1440          if (*ptr != 0) ptr++;
1441          }
1442        }
1443    
1444      /* Start with (? but not a condition. */
1445    
1446      else
1447        {
1448        ptr += 2;
1449        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1450    
1451        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1452    
1453        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1454            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1455          {
1456          int term;
1457          const pcre_uchar *thisname;
1458          *count += 1;
1459          if (name == NULL && *count == lorn) return *count;
1460          term = *ptr++;
1461          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1462          thisname = ptr;
1463          while (*ptr != term) ptr++;
1464          if (name != NULL && lorn == ptr - thisname &&
1465              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1466            return *count;
1467          term++;
1468          }
1469        }
1470      }
1471    
1472    /* Past any initial parenthesis handling, scan for parentheses or vertical
1473    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1474    first-pass call when this value is temporarily adjusted to stop at the current
1475    position. So DO NOT change this to a test for binary zero. */
1476    
1477    for (; ptr < cd->end_pattern; ptr++)
1478      {
1479    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1480    
1481    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1482      {      {
1483      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1484      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1485        {        {
1486        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1487        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1488        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1489        }        }
1490      continue;      continue;
1491      }      }
1492    
1493    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1494      are handled for real. If the first character is '^', skip it. Also, if the
1495      first few characters (either before or after ^) are \Q\E or \E we skip them
1496      too. This makes for compatibility with Perl. Note the use of STR macros to
1497      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1498    
1499      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1500        {
1501        BOOL negate_class = FALSE;
1502        for (;;)
1503          {
1504          if (ptr[1] == CHAR_BACKSLASH)
1505            {
1506            if (ptr[2] == CHAR_E)
1507              ptr+= 2;
1508            else if (STRNCMP_UC_C8(ptr + 2,
1509                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1510              ptr += 4;
1511            else
1512              break;
1513            }
1514          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1515            {
1516            negate_class = TRUE;
1517            ptr++;
1518            }
1519          else break;
1520          }
1521    
1522        /* If the next character is ']', it is a data character that must be
1523        skipped, except in JavaScript compatibility mode. */
1524    
1525    if (*ptr == '[')      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1526      {          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1527      while (*(++ptr) != ']')        ptr++;
1528    
1529        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1530        {        {
1531        if (*ptr == '\\')        if (*ptr == 0) return -1;
1532          if (*ptr == CHAR_BACKSLASH)
1533          {          {
1534          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1535          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1536            {            {
1537            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1538            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1539            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1540            }            }
1541          continue;          continue;
1542          }          }
# Line 921  for (; *ptr != 0; ptr++) Line 1546  for (; *ptr != 0; ptr++)
1546    
1547    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1548    
1549    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1550      {      {
1551      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1552      if (*ptr == 0) return -1;      while (*ptr != 0)
1553          {
1554          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1555          ptr++;
1556    #ifdef SUPPORT_UTF
1557          if (utf) FORWARDCHAR(ptr);
1558    #endif
1559          }
1560        if (*ptr == 0) goto FAIL_EXIT;
1561      continue;      continue;
1562      }      }
1563    
1564    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1565    
1566    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1567      {      {
1568      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1569      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1570      continue;      if (*ptr == 0) goto FAIL_EXIT;
1571      }      }
1572    
1573    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1574    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1575        if (dup_parens && *count < hwm_count) *count = hwm_count;
1576    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1577        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1578    
1579    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1580    term = *ptr++;      {
1581    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1582    thisname = ptr;      *count = start_count;
1583    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1584    }    }
1585    
1586    FAIL_EXIT:
1587    *ptrptr = ptr;
1588  return -1;  return -1;
1589  }  }
1590    
1591    
1592    
1593    
1594  /*************************************************  /*************************************************
1595  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1596    *************************************************/
1597    
1598    /* This function scans along a pattern's text looking for capturing
1599    subpatterns, and counting them. If it finds a named pattern that matches the
1600    name it is given, it returns its number. Alternatively, if the name is NULL, it
1601    returns when it reaches a given numbered subpattern. This is used for forward
1602    references to subpatterns. We used to be able to start this scan from the
1603    current compiling point, using the current count value from cd->bracount, and
1604    do it all in a single loop, but the addition of the possibility of duplicate
1605    subpattern numbers means that we have to scan from the very start, in order to
1606    take account of such duplicates, and to use a recursive function to keep track
1607    of the different types of group.
1608    
1609    Arguments:
1610      cd           compile background data
1611      name         name to seek, or NULL if seeking a numbered subpattern
1612      lorn         name length, or subpattern number if name is NULL
1613      xmode        TRUE if we are in /x mode
1614      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1615    
1616    Returns:       the number of the found subpattern, or -1 if not found
1617    */
1618    
1619    static int
1620    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1621      BOOL utf)
1622    {
1623    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1624    int count = 0;
1625    int rc;
1626    
1627    /* If the pattern does not start with an opening parenthesis, the first call
1628    to find_parens_sub() will scan right to the end (if necessary). However, if it
1629    does start with a parenthesis, find_parens_sub() will return when it hits the
1630    matching closing parens. That is why we have to have a loop. */
1631    
1632    for (;;)
1633      {
1634      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1635      if (rc > 0 || *ptr++ == 0) break;
1636      }
1637    
1638    return rc;
1639    }
1640    
1641    
1642    
1643    
1644    /*************************************************
1645    *      Find first significant op code            *
1646  *************************************************/  *************************************************/
1647    
1648  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1649  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
1650  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
1651  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
1652  assertions, and also the \b assertion; for others it does not.  does not.
1653    
1654  Arguments:  Arguments:
1655    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  
1656    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1657    
1658  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1659  */  */
1660    
1661  static const uschar*  static const pcre_uchar*
1662  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1663  {  {
1664  for (;;)  for (;;)
1665    {    {
1666    switch ((int)*code)    switch ((int)*code)
1667      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1668      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1669      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1670      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1671      if (!skipassert) return code;      if (!skipassert) return code;
1672      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1673      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1674      break;      break;
1675    
1676      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1013  for (;;) Line 1680  for (;;)
1680    
1681      case OP_CALLOUT:      case OP_CALLOUT:
1682      case OP_CREF:      case OP_CREF:
1683        case OP_NCREF:
1684      case OP_RREF:      case OP_RREF:
1685        case OP_NRREF:
1686      case OP_DEF:      case OP_DEF:
1687      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1688      break;      break;
1689    
1690      default:      default:
# Line 1029  for (;;) Line 1698  for (;;)
1698    
1699    
1700  /*************************************************  /*************************************************
1701  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1702  *************************************************/  *************************************************/
1703    
1704  /* 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,
1705  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.
1706  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
1707    temporarily terminated with OP_END when this function is called.
1708    
1709    This function is called when a backward assertion is encountered, so that if it
1710    fails, the error message can point to the correct place in the pattern.
1711    However, we cannot do this when the assertion contains subroutine calls,
1712    because they can be forward references. We solve this by remembering this case
1713    and doing the check at the end; a flag specifies which mode we are running in.
1714    
1715  Arguments:  Arguments:
1716    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1717    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1718      atend    TRUE if called when the pattern is complete
1719  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1720               or -2 if \C was encountered  
1721    Returns:   the fixed length,
1722                 or -1 if there is no fixed length,
1723                 or -2 if \C was encountered (in UTF-8 mode only)
1724                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1725                 or -4 if an unknown opcode was encountered (internal error)
1726  */  */
1727    
1728  static int  static int
1729  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1730  {  {
1731  int length = -1;  int length = -1;
1732    
1733  register int branchlength = 0;  register int branchlength = 0;
1734  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1735    
1736  /* 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
1737  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1058  branch, check the length against that of Line 1739  branch, check the length against that of
1739  for (;;)  for (;;)
1740    {    {
1741    int d;    int d;
1742      pcre_uchar *ce, *cs;
1743    register int op = *cc;    register int op = *cc;
1744    
1745    switch (op)    switch (op)
1746      {      {
1747        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1748        OP_BRA (normal non-capturing bracket) because the other variants of these
1749        opcodes are all concerned with unlimited repeated groups, which of course
1750        are not of fixed length. */
1751    
1752      case OP_CBRA:      case OP_CBRA:
1753      case OP_BRA:      case OP_BRA:
1754      case OP_ONCE:      case OP_ONCE:
1755        case OP_ONCE_NC:
1756      case OP_COND:      case OP_COND:
1757      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1758      if (d < 0) return d;      if (d < 0) return d;
1759      branchlength += d;      branchlength += d;
1760      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1761      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1762      break;      break;
1763    
1764      /* 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.
1765      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
1766      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
1767        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1768        because they all imply an unlimited repeat. */
1769    
1770      case OP_ALT:      case OP_ALT:
1771      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1772      case OP_END:      case OP_END:
1773        case OP_ACCEPT:
1774        case OP_ASSERT_ACCEPT:
1775      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1776        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1777      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1089  for (;;) Line 1779  for (;;)
1779      branchlength = 0;      branchlength = 0;
1780      break;      break;
1781    
1782        /* A true recursion implies not fixed length, but a subroutine call may
1783        be OK. If the subroutine is a forward reference, we can't deal with
1784        it until the end of the pattern, so return -3. */
1785    
1786        case OP_RECURSE:
1787        if (!atend) return -3;
1788        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1789        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1790        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1791        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1792        if (d < 0) return d;
1793        branchlength += d;
1794        cc += 1 + LINK_SIZE;
1795        break;
1796    
1797      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1798    
1799      case OP_ASSERT:      case OP_ASSERT:
# Line 1096  for (;;) Line 1801  for (;;)
1801      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1802      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1803      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1804      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1805        break;
1806    
1807      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1808    
1809      case OP_REVERSE:      case OP_MARK:
1810        case OP_PRUNE_ARG:
1811        case OP_SKIP_ARG:
1812        case OP_THEN_ARG:
1813        cc += cc[1] + PRIV(OP_lengths)[*cc];
1814        break;
1815    
1816        case OP_CALLOUT:
1817        case OP_CIRC:
1818        case OP_CIRCM:
1819        case OP_CLOSE:
1820        case OP_COMMIT:
1821      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1822      case OP_DEF:      case OP_DEF:
1823      case OP_OPT:      case OP_DOLL:
1824      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1825      case OP_EOD:      case OP_EOD:
1826      case OP_EODN:      case OP_EODN:
1827      case OP_CIRC:      case OP_FAIL:
1828      case OP_DOLL:      case OP_NCREF:
1829        case OP_NRREF:
1830      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1831        case OP_PRUNE:
1832        case OP_REVERSE:
1833        case OP_RREF:
1834        case OP_SET_SOM:
1835        case OP_SKIP:
1836        case OP_SOD:
1837        case OP_SOM:
1838        case OP_THEN:
1839      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1840      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1841      break;      break;
1842    
1843      /* Handle literal characters */      /* Handle literal characters */
1844    
1845      case OP_CHAR:      case OP_CHAR:
1846      case OP_CHARNC:      case OP_CHARI:
1847      case OP_NOT:      case OP_NOT:
1848        case OP_NOTI:
1849      branchlength++;      branchlength++;
1850      cc += 2;      cc += 2;
1851  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1852      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1853  #endif  #endif
1854      break;      break;
1855    
# Line 1136  for (;;) Line 1857  for (;;)
1857      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1858    
1859      case OP_EXACT:      case OP_EXACT:
1860        case OP_EXACTI:
1861        case OP_NOTEXACT:
1862        case OP_NOTEXACTI:
1863      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1864      cc += 4;      cc += 2 + IMM2_SIZE;
1865  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1866      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1867  #endif  #endif
1868      break;      break;
1869    
1870      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1871      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1872      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1873          cc += 2;
1874        cc += 1 + IMM2_SIZE + 1;
1875      break;      break;
1876    
1877      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1158  for (;;) Line 1881  for (;;)
1881      cc += 2;      cc += 2;
1882      /* Fall through */      /* Fall through */
1883    
1884        case OP_HSPACE:
1885        case OP_VSPACE:
1886        case OP_NOT_HSPACE:
1887        case OP_NOT_VSPACE:
1888      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1889      case OP_DIGIT:      case OP_DIGIT:
1890      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1165  for (;;) Line 1892  for (;;)
1892      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1893      case OP_WORDCHAR:      case OP_WORDCHAR:
1894      case OP_ANY:      case OP_ANY:
1895        case OP_ALLANY:
1896      branchlength++;      branchlength++;
1897      cc++;      cc++;
1898      break;      break;
1899    
1900      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1901        otherwise \C is coded as OP_ALLANY. */
1902    
1903      case OP_ANYBYTE:      case OP_ANYBYTE:
1904      return -2;      return -2;
1905    
1906      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1907    
1908  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1909      case OP_XCLASS:      case OP_XCLASS:
1910      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1911      /* Fall through */      /* Fall through */
1912  #endif  #endif
1913    
1914      case OP_CLASS:      case OP_CLASS:
1915      case OP_NCLASS:      case OP_NCLASS:
1916      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1917    
1918      switch (*cc)      switch (*cc)
1919        {        {
1920          case OP_CRPLUS:
1921          case OP_CRMINPLUS:
1922        case OP_CRSTAR:        case OP_CRSTAR:
1923        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1924        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1196  for (;;) Line 1927  for (;;)
1927    
1928        case OP_CRRANGE:        case OP_CRRANGE:
1929        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1930        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1931        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1932        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1933        break;        break;
1934    
1935        default:        default:
# Line 1208  for (;;) Line 1939  for (;;)
1939    
1940      /* Anything else is variable length */      /* Anything else is variable length */
1941    
1942      default:      case OP_ANYNL:
1943        case OP_BRAMINZERO:
1944        case OP_BRAPOS:
1945        case OP_BRAPOSZERO:
1946        case OP_BRAZERO:
1947        case OP_CBRAPOS:
1948        case OP_EXTUNI:
1949        case OP_KETRMAX:
1950        case OP_KETRMIN:
1951        case OP_KETRPOS:
1952        case OP_MINPLUS:
1953        case OP_MINPLUSI:
1954        case OP_MINQUERY:
1955        case OP_MINQUERYI:
1956        case OP_MINSTAR:
1957        case OP_MINSTARI:
1958        case OP_MINUPTO:
1959        case OP_MINUPTOI:
1960        case OP_NOTMINPLUS:
1961        case OP_NOTMINPLUSI:
1962        case OP_NOTMINQUERY:
1963        case OP_NOTMINQUERYI:
1964        case OP_NOTMINSTAR:
1965        case OP_NOTMINSTARI:
1966        case OP_NOTMINUPTO:
1967        case OP_NOTMINUPTOI:
1968        case OP_NOTPLUS:
1969        case OP_NOTPLUSI:
1970        case OP_NOTPOSPLUS:
1971        case OP_NOTPOSPLUSI:
1972        case OP_NOTPOSQUERY:
1973        case OP_NOTPOSQUERYI:
1974        case OP_NOTPOSSTAR:
1975        case OP_NOTPOSSTARI:
1976        case OP_NOTPOSUPTO:
1977        case OP_NOTPOSUPTOI:
1978        case OP_NOTQUERY:
1979        case OP_NOTQUERYI:
1980        case OP_NOTSTAR:
1981        case OP_NOTSTARI:
1982        case OP_NOTUPTO:
1983        case OP_NOTUPTOI:
1984        case OP_PLUS:
1985        case OP_PLUSI:
1986        case OP_POSPLUS:
1987        case OP_POSPLUSI:
1988        case OP_POSQUERY:
1989        case OP_POSQUERYI:
1990        case OP_POSSTAR:
1991        case OP_POSSTARI:
1992        case OP_POSUPTO:
1993        case OP_POSUPTOI:
1994        case OP_QUERY:
1995        case OP_QUERYI:
1996        case OP_REF:
1997        case OP_REFI:
1998        case OP_SBRA:
1999        case OP_SBRAPOS:
2000        case OP_SCBRA:
2001        case OP_SCBRAPOS:
2002        case OP_SCOND:
2003        case OP_SKIPZERO:
2004        case OP_STAR:
2005        case OP_STARI:
2006        case OP_TYPEMINPLUS:
2007        case OP_TYPEMINQUERY:
2008        case OP_TYPEMINSTAR:
2009        case OP_TYPEMINUPTO:
2010        case OP_TYPEPLUS:
2011        case OP_TYPEPOSPLUS:
2012        case OP_TYPEPOSQUERY:
2013        case OP_TYPEPOSSTAR:
2014        case OP_TYPEPOSUPTO:
2015        case OP_TYPEQUERY:
2016        case OP_TYPESTAR:
2017        case OP_TYPEUPTO:
2018        case OP_UPTO:
2019        case OP_UPTOI:
2020      return -1;      return -1;
2021    
2022        /* Catch unrecognized opcodes so that when new ones are added they
2023        are not forgotten, as has happened in the past. */
2024    
2025        default:
2026        return -4;
2027      }      }
2028    }    }
2029  /* Control never gets here */  /* Control never gets here */
# Line 1219  for (;;) Line 2033  for (;;)
2033    
2034    
2035  /*************************************************  /*************************************************
2036  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2037  *************************************************/  *************************************************/
2038    
2039  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2040  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2041    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2042    so that it can be called from pcre_study() when finding the minimum matching
2043    length.
2044    
2045  Arguments:  Arguments:
2046    code        points to start of expression    code        points to start of expression
2047    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2048    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2049    
2050  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
2051  */  */
2052    
2053  static const uschar *  const pcre_uchar *
2054  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2055  {  {
2056  for (;;)  for (;;)
2057    {    {
2058    register int c = *code;    register int c = *code;
2059    
2060    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2061    
2062    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1247  for (;;) Line 2065  for (;;)
2065    
2066    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2067    
2068      /* Handle recursion */
2069    
2070      else if (c == OP_REVERSE)
2071        {
2072        if (number < 0) return (pcre_uchar *)code;
2073        code += PRIV(OP_lengths)[c];
2074        }
2075    
2076    /* Handle capturing bracket */    /* Handle capturing bracket */
2077    
2078    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2079               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2080      {      {
2081      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2082      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2083      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2084      }      }
2085    
2086      /* Otherwise, we can get the item's length from the table, except that for
2087      repeated character types, we have to test for \p and \P, which have an extra
2088      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2089      must add in its length. */
2090    
2091      else
2092        {
2093        switch(c)
2094          {
2095          case OP_TYPESTAR:
2096          case OP_TYPEMINSTAR:
2097          case OP_TYPEPLUS:
2098          case OP_TYPEMINPLUS:
2099          case OP_TYPEQUERY:
2100          case OP_TYPEMINQUERY:
2101          case OP_TYPEPOSSTAR:
2102          case OP_TYPEPOSPLUS:
2103          case OP_TYPEPOSQUERY:
2104          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2105          break;
2106    
2107          case OP_TYPEUPTO:
2108          case OP_TYPEMINUPTO:
2109          case OP_TYPEEXACT:
2110          case OP_TYPEPOSUPTO:
2111          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2112            code += 2;
2113          break;
2114    
2115          case OP_MARK:
2116          case OP_PRUNE_ARG:
2117          case OP_SKIP_ARG:
2118          code += code[1];
2119          break;
2120    
2121          case OP_THEN_ARG:
2122          code += code[1];
2123          break;
2124          }
2125    
2126        /* Add in the fixed length from the table */
2127    
2128        code += PRIV(OP_lengths)[c];
2129    
2130    /* 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
2131    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
2132    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2133    
2134    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2135      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2136        {        {
2137        case OP_CHAR:        case OP_CHAR:
2138        case OP_CHARNC:        case OP_CHARI:
2139        case OP_EXACT:        case OP_EXACT:
2140          case OP_EXACTI:
2141        case OP_UPTO:        case OP_UPTO:
2142          case OP_UPTOI:
2143        case OP_MINUPTO:        case OP_MINUPTO:
2144          case OP_MINUPTOI:
2145        case OP_POSUPTO:        case OP_POSUPTO:
2146          case OP_POSUPTOI:
2147        case OP_STAR:        case OP_STAR:
2148          case OP_STARI:
2149        case OP_MINSTAR:        case OP_MINSTAR:
2150          case OP_MINSTARI:
2151        case OP_POSSTAR:        case OP_POSSTAR:
2152          case OP_POSSTARI:
2153        case OP_PLUS:        case OP_PLUS:
2154          case OP_PLUSI:
2155        case OP_MINPLUS:        case OP_MINPLUS:
2156          case OP_MINPLUSI:
2157        case OP_POSPLUS:        case OP_POSPLUS:
2158          case OP_POSPLUSI:
2159        case OP_QUERY:        case OP_QUERY:
2160          case OP_QUERYI:
2161        case OP_MINQUERY:        case OP_MINQUERY:
2162          case OP_MINQUERYI:
2163        case OP_POSQUERY:        case OP_POSQUERY:
2164        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2165          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2166        break;        break;
2167        }        }
2168    #else
2169        (void)(utf);  /* Keep compiler happy by referencing function argument */
2170  #endif  #endif
2171      }      }
2172    }    }
# Line 1300  instance of OP_RECURSE. Line 2183  instance of OP_RECURSE.
2183    
2184  Arguments:  Arguments:
2185    code        points to start of expression    code        points to start of expression
2186    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2187    
2188  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
2189  */  */
2190    
2191  static const uschar *  static const pcre_uchar *
2192  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2193  {  {
2194  for (;;)  for (;;)
2195    {    {
# Line 1320  for (;;) Line 2203  for (;;)
2203    
2204    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2205    
2206    /* 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
2207    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
2208    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
2209    bytes. */    must add in its length. */
2210    
2211    else    else
2212      {      {
2213      code += _pcre_OP_lengths[c];      switch(c)
2214  #ifdef SUPPORT_UTF8        {
2215      if (utf8) switch(c)        case OP_TYPESTAR:
2216          case OP_TYPEMINSTAR:
2217          case OP_TYPEPLUS:
2218          case OP_TYPEMINPLUS:
2219          case OP_TYPEQUERY:
2220          case OP_TYPEMINQUERY:
2221          case OP_TYPEPOSSTAR:
2222          case OP_TYPEPOSPLUS:
2223          case OP_TYPEPOSQUERY:
2224          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2225          break;
2226    
2227          case OP_TYPEPOSUPTO:
2228          case OP_TYPEUPTO:
2229          case OP_TYPEMINUPTO:
2230          case OP_TYPEEXACT:
2231          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2232            code += 2;
2233          break;
2234    
2235          case OP_MARK:
2236          case OP_PRUNE_ARG:
2237          case OP_SKIP_ARG:
2238          code += code[1];
2239          break;
2240    
2241          case OP_THEN_ARG:
2242          code += code[1];
2243          break;
2244          }
2245    
2246        /* Add in the fixed length from the table */
2247    
2248        code += PRIV(OP_lengths)[c];
2249    
2250        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2251        by a multi-byte character. The length in the table is a minimum, so we have
2252        to arrange to skip the extra bytes. */
2253    
2254    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2255        if (utf) switch(c)
2256        {        {
2257        case OP_CHAR:        case OP_CHAR:
2258        case OP_CHARNC:        case OP_CHARI:
2259          case OP_NOT:
2260          case OP_NOTI:
2261        case OP_EXACT:        case OP_EXACT:
2262          case OP_EXACTI:
2263          case OP_NOTEXACT:
2264          case OP_NOTEXACTI:
2265        case OP_UPTO:        case OP_UPTO:
2266          case OP_UPTOI:
2267          case OP_NOTUPTO:
2268          case OP_NOTUPTOI:
2269        case OP_MINUPTO:        case OP_MINUPTO:
2270          case OP_MINUPTOI:
2271          case OP_NOTMINUPTO:
2272          case OP_NOTMINUPTOI:
2273        case OP_POSUPTO:        case OP_POSUPTO:
2274          case OP_POSUPTOI:
2275          case OP_NOTPOSUPTO:
2276          case OP_NOTPOSUPTOI:
2277        case OP_STAR:        case OP_STAR:
2278          case OP_STARI:
2279          case OP_NOTSTAR:
2280          case OP_NOTSTARI:
2281        case OP_MINSTAR:        case OP_MINSTAR:
2282          case OP_MINSTARI:
2283          case OP_NOTMINSTAR:
2284          case OP_NOTMINSTARI:
2285        case OP_POSSTAR:        case OP_POSSTAR:
2286          case OP_POSSTARI:
2287          case OP_NOTPOSSTAR:
2288          case OP_NOTPOSSTARI:
2289        case OP_PLUS:        case OP_PLUS:
2290          case OP_PLUSI:
2291          case OP_NOTPLUS:
2292          case OP_NOTPLUSI:
2293        case OP_MINPLUS:        case OP_MINPLUS:
2294          case OP_MINPLUSI:
2295          case OP_NOTMINPLUS:
2296          case OP_NOTMINPLUSI:
2297        case OP_POSPLUS:        case OP_POSPLUS:
2298          case OP_POSPLUSI:
2299          case OP_NOTPOSPLUS:
2300          case OP_NOTPOSPLUSI:
2301        case OP_QUERY:        case OP_QUERY:
2302          case OP_QUERYI:
2303          case OP_NOTQUERY:
2304          case OP_NOTQUERYI:
2305        case OP_MINQUERY:        case OP_MINQUERY:
2306          case OP_MINQUERYI:
2307          case OP_NOTMINQUERY:
2308          case OP_NOTMINQUERYI:
2309        case OP_POSQUERY:        case OP_POSQUERY:
2310        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2311          case OP_NOTPOSQUERY:
2312          case OP_NOTPOSQUERYI:
2313          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2314        break;        break;
2315        }        }
2316    #else
2317        (void)(utf);  /* Keep compiler happy by referencing function argument */
2318  #endif  #endif
2319      }      }
2320    }    }
# Line 1364  for (;;) Line 2330  for (;;)
2330  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()
2331  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
2332  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2333  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
2334  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
2335    bracket whose current branch will already have been scanned.
2336    
2337  Arguments:  Arguments:
2338    code        points to start of search    code        points to start of search
2339    endcode     points to where to stop    endcode     points to where to stop
2340    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2341      cd          contains pointers to tables etc.
2342    
2343  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2344  */  */
2345    
2346  static BOOL  static BOOL
2347  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2348      BOOL utf, compile_data *cd)
2349  {  {
2350  register int c;  register int c;
2351  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2352       code < endcode;       code < endcode;
2353       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2354    {    {
2355    const uschar *ccode;    const pcre_uchar *ccode;
2356    
2357    c = *code;    c = *code;
2358    
2359    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2360      first_significant_code() with a TRUE final argument. */
2361    
2362    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2363      {      {
     code += _pcre_OP_lengths[c];  
2364      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2365      c = *code;      c = *code;
2366      continue;      continue;
2367      }      }
2368    
2369    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2370      implies a backward reference subroutine call, we can scan it. If it's a
2371      forward reference subroutine call, we can't. To detect forward reference
2372      we have to scan up the list that is kept in the workspace. This function is
2373      called only when doing the real compile, not during the pre-compile that
2374      measures the size of the compiled pattern. */
2375    
2376    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_RECURSE)
2377      {      {
2378        const pcre_uchar *scode;
2379      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2380    
2381      /* Scan a closed bracket */      /* Test for forward reference */
2382    
2383        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2384          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2385    
2386        /* Not a forward reference, test for completed backward reference */
2387    
2388      empty_branch = FALSE;      empty_branch = FALSE;
2389        scode = cd->start_code + GET(code, 1);
2390        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2391    
2392        /* Completed backwards reference */
2393    
2394      do      do
2395        {        {
2396        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2397            {
2398          empty_branch = TRUE;          empty_branch = TRUE;
2399            break;
2400            }
2401          scode += GET(scode, 1);
2402          }
2403        while (*scode == OP_ALT);
2404    
2405        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2406        continue;
2407        }
2408    
2409      /* Groups with zero repeats can of course be empty; skip them. */
2410    
2411      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2412          c == OP_BRAPOSZERO)
2413        {
2414        code += PRIV(OP_lengths)[c];
2415        do code += GET(code, 1); while (*code == OP_ALT);
2416        c = *code;
2417        continue;
2418        }
2419    
2420      /* A nested group that is already marked as "could be empty" can just be
2421      skipped. */
2422    
2423      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2424          c == OP_SCBRA || c == OP_SCBRAPOS)
2425        {
2426        do code += GET(code, 1); while (*code == OP_ALT);
2427        c = *code;
2428        continue;
2429        }
2430    
2431      /* For other groups, scan the branches. */
2432    
2433      if (c == OP_BRA  || c == OP_BRAPOS ||
2434          c == OP_CBRA || c == OP_CBRAPOS ||
2435          c == OP_ONCE || c == OP_ONCE_NC ||
2436          c == OP_COND)
2437        {
2438        BOOL empty_branch;
2439        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2440    
2441        /* If a conditional group has only one branch, there is a second, implied,
2442        empty branch, so just skip over the conditional, because it could be empty.
2443        Otherwise, scan the individual branches of the group. */
2444    
2445        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2446        code += GET(code, 1);        code += GET(code, 1);
2447        else
2448          {
2449          empty_branch = FALSE;
2450          do
2451            {
2452            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2453              empty_branch = TRUE;
2454            code += GET(code, 1);
2455            }
2456          while (*code == OP_ALT);
2457          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2458        }        }
2459      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2460      c = *code;      c = *code;
2461      continue;      continue;
2462      }      }
# Line 1423  for (code = first_significant_code(code Line 2465  for (code = first_significant_code(code
2465    
2466    switch (c)    switch (c)
2467      {      {
2468      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2469        cannot be represented just by a bit map. This includes negated single
2470        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2471        actual length is stored in the compiled code, so we must update "code"
2472        here. */
2473    
2474  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2475      case OP_XCLASS:      case OP_XCLASS:
2476      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2477      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2478  #endif  #endif
2479    
2480      case OP_CLASS:      case OP_CLASS:
2481      case OP_NCLASS:      case OP_NCLASS:
2482      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2483    
2484  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2485      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2486  #endif  #endif
2487    
# Line 1471  for (code = first_significant_code(code Line 2517  for (code = first_significant_code(code
2517      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2518      case OP_WORDCHAR:      case OP_WORDCHAR:
2519      case OP_ANY:      case OP_ANY:
2520        case OP_ALLANY:
2521      case OP_ANYBYTE:      case OP_ANYBYTE:
2522      case OP_CHAR:      case OP_CHAR:
2523      case OP_CHARNC:      case OP_CHARI:
2524      case OP_NOT:      case OP_NOT:
2525        case OP_NOTI:
2526      case OP_PLUS:      case OP_PLUS:
2527      case OP_MINPLUS:      case OP_MINPLUS:
2528      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1489  for (code = first_significant_code(code Line 2537  for (code = first_significant_code(code
2537      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2538      return FALSE;      return FALSE;
2539    
2540        /* These are going to continue, as they may be empty, but we have to
2541        fudge the length for the \p and \P cases. */
2542    
2543        case OP_TYPESTAR:
2544        case OP_TYPEMINSTAR:
2545        case OP_TYPEPOSSTAR:
2546        case OP_TYPEQUERY:
2547        case OP_TYPEMINQUERY:
2548        case OP_TYPEPOSQUERY:
2549        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2550        break;
2551    
2552        /* Same for these */
2553    
2554        case OP_TYPEUPTO:
2555        case OP_TYPEMINUPTO:
2556        case OP_TYPEPOSUPTO:
2557        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2558          code += 2;
2559        break;
2560    
2561      /* End of branch */      /* End of branch */
2562    
2563      case OP_KET:      case OP_KET:
2564      case OP_KETRMAX:      case OP_KETRMAX:
2565      case OP_KETRMIN:      case OP_KETRMIN:
2566        case OP_KETRPOS:
2567      case OP_ALT:      case OP_ALT:
2568      return TRUE;      return TRUE;
2569    
2570      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2571      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2572    
2573  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2574      case OP_STAR:      case OP_STAR:
2575        case OP_STARI:
2576      case OP_MINSTAR:      case OP_MINSTAR:
2577        case OP_MINSTARI:
2578      case OP_POSSTAR:      case OP_POSSTAR:
2579        case OP_POSSTARI:
2580      case OP_QUERY:      case OP_QUERY:
2581        case OP_QUERYI:
2582      case OP_MINQUERY:      case OP_MINQUERY:
2583        case OP_MINQUERYI:
2584      case OP_POSQUERY:      case OP_POSQUERY:
2585        case OP_POSQUERYI:
2586        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2587        break;
2588    
2589      case OP_UPTO:      case OP_UPTO:
2590        case OP_UPTOI:
2591      case OP_MINUPTO:      case OP_MINUPTO:
2592        case OP_MINUPTOI:
2593      case OP_POSUPTO:      case OP_POSUPTO:
2594      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2595        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2596      break;      break;
2597  #endif  #endif
2598    
2599        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2600        string. */
2601    
2602        case OP_MARK:
2603        case OP_PRUNE_ARG:
2604        case OP_SKIP_ARG:
2605        code += code[1];
2606        break;
2607    
2608        case OP_THEN_ARG:
2609        code += code[1];
2610        break;
2611    
2612        /* None of the remaining opcodes are required to match a character. */
2613    
2614        default:
2615        break;
2616      }      }
2617    }    }
2618    
# Line 1529  return TRUE; Line 2629  return TRUE;
2629  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
2630  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,
2631  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.
2632    This function is called only during the real compile, not during the
2633    pre-compile.
2634    
2635  Arguments:  Arguments:
2636    code        points to start of the recursion    code        points to start of the recursion
2637    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2638    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2639    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2640      cd          pointers to tables etc
2641    
2642  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2643  */  */
2644    
2645  static BOOL  static BOOL
2646  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2647    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2648  {  {
2649  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2650    {    {
2651    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2652        return FALSE;
2653    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2654    }    }
2655  return TRUE;  return TRUE;
# Line 1558  return TRUE; Line 2662  return TRUE;
2662  *************************************************/  *************************************************/
2663    
2664  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2665  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
2666  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2667  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2668    
2669    Originally, this function only recognized a sequence of letters between the
2670    terminators, but it seems that Perl recognizes any sequence of characters,
2671    though of course unknown POSIX names are subsequently rejected. Perl gives an
2672    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2673    didn't consider this to be a POSIX class. Likewise for [:1234:].
2674    
2675    The problem in trying to be exactly like Perl is in the handling of escapes. We
2676    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2677    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2678    below handles the special case of \], but does not try to do any other escape
2679    processing. This makes it different from Perl for cases such as [:l\ower:]
2680    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2681    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2682    I think.
2683    
2684    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2685    It seems that the appearance of a nested POSIX class supersedes an apparent
2686    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2687    a digit.
2688    
2689    In Perl, unescaped square brackets may also appear as part of class names. For
2690    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2691    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2692    seem right at all. PCRE does not allow closing square brackets in POSIX class
2693    names.
2694    
2695  Argument:  Arguments:
2696    ptr      pointer to the initial [    ptr      pointer to the initial [
2697    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2698    
2699  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2700  */  */
2701    
2702  static BOOL  static BOOL
2703  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2704  {  {
2705  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2706  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2707  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2708    {    {
2709    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2710    return TRUE;      ptr++;
2711      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2712      else
2713        {
2714        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2715          {
2716          *endptr = ptr;
2717          return TRUE;
2718          }
2719        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2720             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2721              ptr[1] == CHAR_EQUALS_SIGN) &&
2722            check_posix_syntax(ptr, endptr))
2723          return FALSE;
2724        }
2725    }    }
2726  return FALSE;  return FALSE;
2727  }  }
# Line 1603  Returns:     a value representing the na Line 2744  Returns:     a value representing the na
2744  */  */
2745    
2746  static int  static int
2747  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2748  {  {
2749    const char *pn = posix_names;
2750  register int yield = 0;  register int yield = 0;
2751  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2752    {    {
2753    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2754      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2755      pn += posix_name_lengths[yield] + 1;
2756    yield++;    yield++;
2757    }    }
2758  return -1;  return -1;
# Line 1624  return -1; Line 2767  return -1;
2767  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2768  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2769  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
2770  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
2771  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
2772  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
2773  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
2774  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2775    OP_END.
2776    
2777  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2778  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1639  value in the reference (which is a group Line 2783  value in the reference (which is a group
2783  Arguments:  Arguments:
2784    group      points to the start of the group    group      points to the start of the group
2785    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2786    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2787    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2788    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
2789    
# Line 1647  Returns:     nothing Line 2791  Returns:     nothing
2791  */  */
2792    
2793  static void  static void
2794  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2795    uschar *save_hwm)    pcre_uchar *save_hwm)
2796  {  {
2797  uschar *ptr = group;  pcre_uchar *ptr = group;
2798  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2799    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2800    {    {
2801    int offset;    int offset;
2802    uschar *hc;    pcre_uchar *hc;
2803    
2804    /* 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
2805    reference. */    reference. */
# Line 1699  Arguments: Line 2844  Arguments:
2844  Returns:         new code pointer  Returns:         new code pointer
2845  */  */
2846    
2847  static uschar *  static pcre_uchar *
2848  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2849  {  {
2850  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2851  *code++ = 255;  *code++ = 255;
2852  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2853  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2854  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2855  }  }
2856    
2857    
# Line 1728  Returns:             nothing Line 2873  Returns:             nothing
2873  */  */
2874    
2875  static void  static void
2876  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2877  {  {
2878  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2879  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2880  }  }
2881    
# Line 1742  PUT(previous_callout, 2 + LINK_SIZE, len Line 2887  PUT(previous_callout, 2 + LINK_SIZE, len
2887  *************************************************/  *************************************************/
2888    
2889  /* 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
2890  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
2891  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
2892  start address.  start address. A character with multiple other cases is returned on its own
2893    with a special return value.
2894    
2895  Arguments:  Arguments:
2896    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1752  Arguments: Line 2898  Arguments:
2898    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2899    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2900    
2901  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2902                   0 when a range is returned
2903                  >0 the CASESET offset for char with multiple other cases
2904                    in this case, ocptr contains the original
2905  */  */
2906    
2907  static BOOL  static int
2908  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2909    unsigned int *odptr)    unsigned int *odptr)
2910  {  {
2911  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2912    int co;
2913    
2914    /* Find the first character that has an other case. If it has multiple other
2915    cases, return its case offset value. */
2916    
2917  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2918    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2919      if ((co = UCD_CASESET(c)) != 0)
2920        {
2921        *ocptr = c++;   /* Character that has the set */
2922        *cptr = c;      /* Rest of input range */
2923        return co;
2924        }
2925      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2926      }
2927    
2928  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2929    
2930  *ocptr = othercase;  *ocptr = othercase;
2931  next = othercase + 1;  next = othercase + 1;
2932    
2933  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2934    {    {
2935    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2936    next++;    next++;
2937    }    }
2938    
2939  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2940  *cptr = c;  *cptr = c;             /* Rest of input range */
2941    return 0;
2942    }
2943    
2944    
2945  return TRUE;  
2946    /*************************************************
2947    *        Check a character and a property        *
2948    *************************************************/
2949    
2950    /* This function is called by check_auto_possessive() when a property item
2951    is adjacent to a fixed character.
2952    
2953    Arguments:
2954      c            the character
2955      ptype        the property type
2956      pdata        the data for the type
2957      negated      TRUE if it's a negated property (\P or \p{^)
2958    
2959    Returns:       TRUE if auto-possessifying is OK
2960    */
2961    
2962    static BOOL
2963    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2964    {
2965    #ifdef SUPPORT_UCP
2966    const pcre_uint32 *p;
2967    #endif
2968    
2969    const ucd_record *prop = GET_UCD(c);
2970    
2971    switch(ptype)
2972      {
2973      case PT_LAMP:
2974      return (prop->chartype == ucp_Lu ||
2975              prop->chartype == ucp_Ll ||
2976              prop->chartype == ucp_Lt) == negated;
2977    
2978      case PT_GC:
2979      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2980    
2981      case PT_PC:
2982      return (pdata == prop->chartype) == negated;
2983    
2984      case PT_SC:
2985      return (pdata == prop->script) == negated;
2986    
2987      /* These are specials */
2988    
2989      case PT_ALNUM:
2990      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2991              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2992    
2993      case PT_SPACE:    /* Perl space */
2994      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2995              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2996              == negated;
2997    
2998      case PT_PXSPACE:  /* POSIX space */
2999      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3000              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3001              c == CHAR_FF || c == CHAR_CR)
3002              == negated;
3003    
3004      case PT_WORD:
3005      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3006              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3007              c == CHAR_UNDERSCORE) == negated;
3008    
3009    #ifdef SUPPORT_UCP
3010      case PT_CLIST:
3011      p = PRIV(ucd_caseless_sets) + prop->caseset;
3012      for (;;)
3013        {
3014        if ((unsigned int)c < *p) return !negated;
3015        if ((unsigned int)c == *p++) return negated;
3016        }
3017      break;  /* Control never reaches here */
3018    #endif
3019      }
3020    
3021    return FALSE;
3022  }  }
3023  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3024    
# Line 1793  whether the next thing could possibly ma Line 3033  whether the next thing could possibly ma
3033  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3034    
3035  Arguments:  Arguments:
3036    op_code       the repeated op code    previous      pointer to the repeated opcode
3037    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  
3038    ptr           next character in pattern    ptr           next character in pattern
3039    options       options bits    options       options bits
3040    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1805  Returns:        TRUE if possessifying is Line 3043  Returns:        TRUE if possessifying is
3043  */  */
3044    
3045  static BOOL  static BOOL
3046  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3047    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3048  {  {
3049  int next;  pcre_int32 c = NOTACHAR; // FIXMEchpe pcre_uint32
3050    pcre_int32 next;
3051    int escape;
3052    int op_code = *previous++;
3053    
3054  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3055    
# Line 1816  if ((options & PCRE_EXTENDED) != 0) Line 3057  if ((options & PCRE_EXTENDED) != 0)
3057    {    {
3058    for (;;)    for (;;)
3059      {      {
3060      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3061      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3062        {        {
3063        while (*(++ptr) != 0)        ptr++;
3064          while (*ptr != 0)
3065            {
3066          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3067            ptr++;
3068    #ifdef SUPPORT_UTF
3069            if (utf) FORWARDCHAR(ptr);
3070    #endif
3071            }
3072        }        }
3073      else break;      else break;
3074      }      }
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 3077  if ((options & PCRE_EXTENDED) != 0)
3077  /* 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
3078  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3079    
3080  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3081    {    {
3082    int temperrorcode = 0;    int temperrorcode = 0;
3083    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3084    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3085    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3086    }    }
3087    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3088    {    {
3089  #ifdef SUPPORT_UTF8    escape = 0;
3090    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3091      if (utf) { GETCHARINC(next, ptr); } else
3092  #endif  #endif
3093    next = *ptr++;    next = *ptr++;
3094    }    }
   
3095  else return FALSE;  else return FALSE;
3096    
3097  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1853  if ((options & PCRE_EXTENDED) != 0) Line 3100  if ((options & PCRE_EXTENDED) != 0)
3100    {    {
3101    for (;;)    for (;;)
3102      {      {
3103      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3104      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3105        {        {
3106        while (*(++ptr) != 0)        ptr++;
3107          while (*ptr != 0)
3108            {
3109          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3110            ptr++;
3111    #ifdef SUPPORT_UTF
3112            if (utf) FORWARDCHAR(ptr);
3113    #endif
3114            }
3115        }        }
3116      else break;      else break;
3117      }      }
# Line 1865  if ((options & PCRE_EXTENDED) != 0) Line 3119  if ((options & PCRE_EXTENDED) != 0)
3119    
3120  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3121    
3122  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3123    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3124        return FALSE;
3125    
3126  /* Now compare the next item with the previous opcode. If the previous is a  /* If the previous item is a character, get its value. */
 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. */  
3127    
3128    if (op_code == OP_CHAR || op_code == OP_CHARI ||
3129        op_code == OP_NOT || op_code == OP_NOTI)
3130      //if (escape == 0) switch(op_code)
3131      {
3132    #ifdef SUPPORT_UTF
3133      GETCHARTEST(c, previous);
3134    #else
3135      c = *previous;
3136    #endif
3137      }
3138    
3139  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3140    the next item is a character. */
3141    
3142  if (next >= 0) switch(op_code)  if (escape == 0)
3143    {    {
3144    case OP_CHAR:    /* For a caseless UTF match, the next character may have more than one other
3145  #ifdef SUPPORT_UTF8    case, which maps to the special PT_CLIST property. Check this first. */
3146    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
3147    #ifdef SUPPORT_UCP
3148      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3149        {
3150        int ocs = UCD_CASESET(next);
3151        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3152        }
3153  #endif  #endif
   return item != next;  
3154    
3155    /* For CHARNC (caseless character) we must check the other case. If we have    switch(op_code)
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
   
   case OP_CHARNC:  
 #ifdef SUPPORT_UTF8  
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   if (item == next) return FALSE;  
 #ifdef SUPPORT_UTF8  
   if (utf8)  
3156      {      {
3157      unsigned int othercase;      case OP_CHAR:
3158      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3159    
3160        /* For CHARI (caseless character) we must check the other case. If we have
3161        Unicode property support, we can use it to test the other case of
3162        high-valued characters. We know that next can have only one other case,
3163        because multi-other-case characters are dealt with above. */
3164    
3165        case OP_CHARI:
3166        if (c == next) return FALSE;
3167    #ifdef SUPPORT_UTF
3168        if (utf)
3169          {
3170          unsigned int othercase;
3171          if (next < 128) othercase = cd->fcc[next]; else
3172  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3173      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3174  #else  #else
3175      othercase = NOTACHAR;        othercase = NOTACHAR;
3176  #endif  #endif
3177      return (unsigned int)item != othercase;        return (unsigned int)c != othercase;
3178      }        }
3179    else      else
3180  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3181    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3182    
3183    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3184        return c == next;
3185    case OP_NOT:  
3186    if (next < 0) return FALSE;  /* Not a character */      case OP_NOTI:
3187    if (item == next) return TRUE;      if (c == next) return TRUE;
3188    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3189  #ifdef SUPPORT_UTF8      if (utf)
3190    if (utf8)        {
3191      {        unsigned int othercase;
3192      unsigned int othercase;        if (next < 128) othercase = cd->fcc[next]; else
     if (next < 128) othercase = cd->fcc[next]; else  
3193  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3194      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3195  #else  #else
3196      othercase = NOTACHAR;        othercase = NOTACHAR;
3197  #endif  #endif
3198      return (unsigned int)item == othercase;        return (unsigned int)c == othercase;
3199      }        }
3200    else      else
3201  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3202    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3203    
3204    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3205    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3206    
3207    case OP_NOT_DIGIT:      case OP_DIGIT:
3208    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3209    
3210    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3211    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3212    
3213    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3214    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3215    
3216    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3217    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3218    
3219    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3220    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3221    
3222    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3223    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3224    switch(next)  
3225      {      case OP_HSPACE:
3226      case 0x09:      case OP_NOT_HSPACE:
3227      case 0x20:      switch(next)
3228      case 0xa0:        {
3229      case 0x1680:        HSPACE_CASES:
3230      case 0x180e:        return op_code == OP_NOT_HSPACE;
3231      case 0x2000:  
3232      case 0x2001:        default:
3233      case 0x2002:        return op_code != OP_NOT_HSPACE;
3234      case 0x2003:        }
3235      case 0x2004:  
3236      case 0x2005:      case OP_ANYNL:
3237      case 0x2006:      case OP_VSPACE:
3238      case 0x2007:      case OP_NOT_VSPACE:
3239      case 0x2008:      switch(next)
3240      case 0x2009:        {
3241      case 0x200A:        VSPACE_CASES:
3242      case 0x202f:        return op_code == OP_NOT_VSPACE;
3243      case 0x205f:  
3244      case 0x3000:        default:
3245      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3246      default:        }
3247      return op_code == OP_HSPACE;  
3248      }  #ifdef SUPPORT_UCP
3249        case OP_PROP:
3250        return check_char_prop(next, previous[0], previous[1], FALSE);
3251    
3252        case OP_NOTPROP:
3253        return check_char_prop(next, previous[0], previous[1], TRUE);
3254    #endif
3255    
   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;  
3256      default:      default:
3257      return op_code == OP_VSPACE;      return FALSE;
3258      }      }
   
   default:  
   return FALSE;  
3259    }    }
3260    
3261    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3262  /* 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
3263    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3264    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3265    replaced by OP_PROP codes when PCRE_UCP is set. */
3266    
3267  switch(op_code)  switch(op_code)
3268    {    {
3269    case OP_CHAR:    case OP_CHAR:
3270    case OP_CHARNC:    case OP_CHARI:
3271  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3272      {      {
3273      case ESC_d:      case ESC_d:
3274      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3275    
3276      case ESC_D:      case ESC_D:
3277      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3278    
3279      case ESC_s:      case ESC_s:
3280      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3281    
3282      case ESC_S:      case ESC_S:
3283      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3284    
3285      case ESC_w:      case ESC_w:
3286      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3287    
3288      case ESC_W:      case ESC_W:
3289      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3290    
3291      case ESC_h:      case ESC_h:
3292      case ESC_H:      case ESC_H:
3293      switch(item)      switch(c)
3294        {        {
3295        case 0x09:        HSPACE_CASES:
3296        case 0x20:        return escape != ESC_h;
3297        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;  
3298        default:        default:
3299        return -next == ESC_h;        return escape == ESC_h;
3300        }        }
3301    
3302      case ESC_v:      case ESC_v:
3303      case ESC_V:      case ESC_V:
3304      switch(item)      switch(c)
3305        {        {
3306        case 0x0a:        VSPACE_CASES:
3307        case 0x0b:        return escape != ESC_v;
3308        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3309        default:        default:
3310        return -next == ESC_v;        return escape == ESC_v;
3311          }
3312    
3313        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3314        their substitutions and process them. The result will always be either
3315        ESC_p or ESC_P. Then fall through to process those values. */
3316    
3317    #ifdef SUPPORT_UCP
3318        case ESC_du:
3319        case ESC_DU:
3320        case ESC_wu:
3321        case ESC_WU:
3322        case ESC_su:
3323        case ESC_SU:
3324          {
3325          int temperrorcode = 0;
3326          ptr = substitutes[escape - ESC_DU];
3327          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3328          if (temperrorcode != 0) return FALSE;
3329          ptr++;    /* For compatibility */
3330        }        }
3331        /* Fall through */
3332    
3333        case ESC_p:
3334        case ESC_P:
3335          {
3336          int ptype, pdata, errorcodeptr;
3337          BOOL negated;
3338    
3339          ptr--;      /* Make ptr point at the p or P */
3340          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3341          if (ptype < 0) return FALSE;
3342          ptr++;      /* Point past the final curly ket */
3343    
3344          /* If the property item is optional, we have to give up. (When generated
3345          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3346          to the original \d etc. At this point, ptr will point to a zero byte. */
3347    
3348          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3349            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3350              return FALSE;
3351    
3352          /* Do the property check. */
3353    
3354          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3355          }
3356    #endif
3357    
3358      default:      default:
3359      return FALSE;      return FALSE;
3360      }      }
3361    
3362      /* In principle, support for Unicode properties should be integrated here as
3363      well. It means re-organizing the above code so as to get hold of the property
3364      values before switching on the op-code. However, I wonder how many patterns
3365      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3366      these op-codes are never generated.) */
3367    
3368    case OP_DIGIT:    case OP_DIGIT:
3369    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3370           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3371    
3372    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3373    return next == -ESC_d;    return escape == ESC_d;
3374    
3375    case OP_WHITESPACE:    case OP_WHITESPACE:
3376    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3377    
3378    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3379    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3380    
3381    case OP_HSPACE:    case OP_HSPACE:
3382    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3383             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3384    
3385    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3386    return next == -ESC_h;    return escape == ESC_h;
3387    
3388    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3389      case OP_ANYNL:
3390    case OP_VSPACE:    case OP_VSPACE:
3391    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3392    
3393    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3394    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3395    
3396    case OP_WORDCHAR:    case OP_WORDCHAR:
3397    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3398             escape == ESC_v || escape == ESC_R;
3399    
3400    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3401    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3402    
3403    default:    default:
3404    return FALSE;    return FALSE;
# Line 2117  switch(op_code) Line 3410  switch(op_code)
3410    
3411    
3412  /*************************************************  /*************************************************
3413    *        Add a character or range to a class     *
3414    *************************************************/
3415    
3416    /* This function packages up the logic of adding a character or range of
3417    characters to a class. The character values in the arguments will be within the
3418    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3419    mutually recursive with the function immediately below.
3420    
3421    Arguments:
3422      classbits     the bit map for characters < 256
3423      uchardptr     points to the pointer for extra data
3424      options       the options word
3425      cd            contains pointers to tables etc.
3426      start         start of range character
3427      end           end of range character
3428    
3429    Returns:        the number of < 256 characters added
3430                    the pointer to extra data is updated
3431    */
3432    
3433    static int
3434    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3435      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3436    {
3437    pcre_uint32 c;
3438    int n8 = 0;
3439    
3440    /* If caseless matching is required, scan the range and process alternate
3441    cases. In Unicode, there are 8-bit characters that have alternate cases that
3442    are greater than 255 and vice-versa. Sometimes we can just extend the original
3443    range. */
3444    
3445    if ((options & PCRE_CASELESS) != 0)
3446      {
3447    #ifdef SUPPORT_UCP
3448      if ((options & PCRE_UTF8) != 0)
3449        {
3450        int rc;
3451        pcre_uint32 oc, od;
3452    
3453        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3454        c = start;
3455    
3456        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3457          {
3458          /* Handle a single character that has more than one other case. */
3459    
3460          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3461            PRIV(ucd_caseless_sets) + rc, oc);
3462    
3463          /* Do nothing if the other case range is within the original range. */
3464    
3465          else if (oc >= start && od <= end) continue;
3466    
3467          /* Extend the original range if there is overlap, noting that if oc < c, we
3468          can't have od > end because a subrange is always shorter than the basic
3469          range. Otherwise, use a recursive call to add the additional range. */
3470    
3471          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3472          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3473          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3474          }
3475        }
3476      else
3477    #endif  /* SUPPORT_UCP */
3478    
3479      /* Not UTF-mode, or no UCP */
3480    
3481      for (c = start; c <= end && c < 256; c++)
3482        {
3483        SETBIT(classbits, cd->fcc[c]);
3484        n8++;
3485        }
3486      }
3487    
3488    /* Now handle the original range. Adjust the final value according to the bit
3489    length - this means that the same lists of (e.g.) horizontal spaces can be used
3490    in all cases. */
3491    
3492    #if defined COMPILE_PCRE8
3493    #ifdef SUPPORT_UTF
3494      if ((options & PCRE_UTF8) == 0)
3495    #endif
3496      if (end > 0xff) end = 0xff;
3497    
3498    #elif defined COMPILE_PCRE16
3499    #ifdef SUPPORT_UTF
3500      if ((options & PCRE_UTF16) == 0)
3501    #endif
3502      if (end > 0xffff) end = 0xffff;
3503    
3504    #endif /* COMPILE_PCRE[8|16] */
3505    
3506    /* If all characters are less than 256, use the bit map. Otherwise use extra
3507    data. */
3508    
3509    if (end < 0x100)
3510      {
3511      for (c = start; c <= end; c++)
3512        {
3513        n8++;
3514        SETBIT(classbits, c);
3515        }
3516      }
3517    
3518    else
3519      {
3520      pcre_uchar *uchardata = *uchardptr;
3521    
3522    #ifdef SUPPORT_UTF
3523      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3524        {
3525        if (start < end)
3526          {
3527          *uchardata++ = XCL_RANGE;
3528          uchardata += PRIV(ord2utf)(start, uchardata);
3529          uchardata += PRIV(ord2utf)(end, uchardata);
3530          }
3531        else if (start == end)
3532          {
3533          *uchardata++ = XCL_SINGLE;
3534          uchardata += PRIV(ord2utf)(start, uchardata);
3535          }
3536        }
3537      else
3538    #endif  /* SUPPORT_UTF */
3539    
3540      /* Without UTF support, character values are constrained by the bit length,
3541      and can only be > 256 for 16-bit and 32-bit libraries. */
3542    
3543    #ifdef COMPILE_PCRE8
3544        {}
3545    #else
3546      if (start < end)
3547        {
3548        *uchardata++ = XCL_RANGE;
3549        *uchardata++ = start;
3550        *uchardata++ = end;
3551        }
3552      else if (start == end)
3553        {
3554        *uchardata++ = XCL_SINGLE;
3555        *uchardata++ = start;
3556        }
3557    #endif
3558    
3559      *uchardptr = uchardata;   /* Updata extra data pointer */
3560      }
3561    
3562    return n8;    /* Number of 8-bit characters */
3563    }
3564    
3565    
3566    
3567    
3568    /*************************************************
3569    *        Add a list of characters to a class     *
3570    *************************************************/
3571    
3572    /* This function is used for adding a list of case-equivalent characters to a
3573    class, and also for adding a list of horizontal or vertical whitespace. If the
3574    list is in order (which it should be), ranges of characters are detected and
3575    handled appropriately. This function is mutually recursive with the function
3576    above.
3577    
3578    Arguments:
3579      classbits     the bit map for characters < 256
3580      uchardptr     points to the pointer for extra data
3581      options       the options word
3582      cd            contains pointers to tables etc.
3583      p             points to row of 32-bit values, terminated by NOTACHAR
3584      except        character to omit; this is used when adding lists of
3585                      case-equivalent characters to avoid including the one we
3586                      already know about
3587    
3588    Returns:        the number of < 256 characters added
3589                    the pointer to extra data is updated
3590    */
3591    
3592    static int
3593    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3594      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3595    {
3596    int n8 = 0;
3597    while (p[0] < NOTACHAR)
3598      {
3599      int n = 0;
3600      if (p[0] != except)
3601        {
3602        while(p[n+1] == p[0] + n + 1) n++;
3603        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3604        }
3605      p += n + 1;
3606      }
3607    return n8;
3608    }
3609    
3610    
3611    
3612    /*************************************************
3613    *    Add characters not in a list to a class     *
3614    *************************************************/
3615    
3616    /* This function is used for adding the complement of a list of horizontal or
3617    vertical whitespace to a class. The list must be in order.
3618    
3619    Arguments:
3620      classbits     the bit map for characters < 256
3621      uchardptr     points to the pointer for extra data
3622      options       the options word
3623      cd            contains pointers to tables etc.
3624      p             points to row of 32-bit values, terminated by NOTACHAR
3625    
3626    Returns:        the number of < 256 characters added
3627                    the pointer to extra data is updated
3628    */
3629    
3630    static int
3631    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3632      int options, compile_data *cd, const pcre_uint32 *p)
3633    {
3634    BOOL utf = (options & PCRE_UTF8) != 0;
3635    int n8 = 0;
3636    if (p[0] > 0)
3637      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3638    while (p[0] < NOTACHAR)
3639      {
3640      while (p[1] == p[0] + 1) p++;
3641      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3642        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3643      p++;
3644      }
3645    return n8;
3646    }
3647    
3648    
3649    
3650    /*************************************************
3651  *           Compile one branch                   *  *           Compile one branch                   *
3652  *************************************************/  *************************************************/
3653    
# Line 2131  Arguments: Line 3662  Arguments:
3662    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3663    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3664    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3665    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3666    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3667    bcptr          points to current branch chain    bcptr          points to current branch chain
3668      cond_depth     conditional nesting depth
3669    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3670    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3671                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2143  Returns:         TRUE on success Line 3675  Returns:         TRUE on success
3675  */  */
3676    
3677  static BOOL  static BOOL
3678  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3679    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3680      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3681    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3682  {  {
3683  int repeat_type, op_type;  int repeat_type, op_type;
3684  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3685  int bravalue = 0;  int bravalue = 0;
3686  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3687  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3688  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3689  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3690  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3691  int after_manual_callout = 0;  int after_manual_callout = 0;
3692  int length_prevgroup = 0;  int length_prevgroup = 0;
3693  register int c;  register int c;
3694  register uschar *code = *codeptr;  int escape;
3695  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3696  uschar *orig_code = code;  pcre_uchar *last_code = code;
3697  uschar *tempcode;  pcre_uchar *orig_code = code;
3698    pcre_uchar *tempcode;
3699  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3700  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3701  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3702  const uschar *tempptr;  const pcre_uchar *tempptr;
3703  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3704  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3705  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3706  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3707    pcre_uint8 classbits[32];
3708  #ifdef SUPPORT_UTF8  
3709  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3710  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3711  uschar *class_utf8data;  dynamically as we process the pattern. */
3712  uschar utf8_char[6];  
3713    #ifdef SUPPORT_UTF
3714    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3715    BOOL utf = (options & PCRE_UTF8) != 0;
3716    pcre_uchar utf_chars[6];
3717  #else  #else
3718  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3719  uschar *utf8_char = NULL;  #endif
3720