/[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 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 1055 by chpe, Tue Oct 16 15:53:30 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    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* 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. */
70    
71    #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  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
94  *************************************************/  *************************************************/
# Line 72  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 COMPILE_WORK_SIZE (4096)  #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    #define REQ_MASK       (REQ_CASELESS | REQ_VARY)
127    
128    /* Repeated character flags. */
129    
130    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
131    
132  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
133  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
134  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
135  is invalid. */  is invalid. */
136    
137  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
138    
139    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
140    in UTF-8 mode. */
141    
142  static const short int escapes[] = {  static const short int escapes[] = {
143       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
144       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
145     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
146       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
147  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -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       0,      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,      0, -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 106  static const short int escapes[] = { Line 193  static const short int escapes[] = {
193  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
194  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
195  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
196  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
197  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
198  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
199  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
200  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
201  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
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 */     0,     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,     0, -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,
209  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
210  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  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 164  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 171  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 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  };    "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 262  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 298  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 334  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 370  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, int, int *,  is_counted_repeat(const pcre_uchar *p)
729      int *, branch_chain *, compile_data *, int *);  {
730    if (!IS_DIGIT(*p)) return FALSE;
731    p++;
732    while (IS_DIGIT(*p)) p++;
733    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
734    
735    if (*p++ != CHAR_COMMA) return FALSE;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (!IS_DIGIT(*p)) return FALSE;
739    p++;
740    while (IS_DIGIT(*p)) p++;
741    
742    return (*p == CHAR_RIGHT_CURLY_BRACKET);
743    }
744    
745    
746    
# Line 399  Arguments: Line 765  Arguments:
765    
766  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
767                   negative => a special escape sequence                   negative => a special escape sequence
768                   on error, errorptr is set                   on error, errorcodeptr is set
769  */  */
770    
771  static int  static int
772  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
773    int options, BOOL isclass)    int options, BOOL isclass)
774  {  {
775  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
776  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
777  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
778    pcre_int32 c;
779    int i;
780    
781  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
782  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 417  ptr--;                            /* Set Line 785  ptr--;                            /* Set
785    
786  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
787    
788  /* 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
789  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.
790  Otherwise further processing may be required. */  Otherwise further processing may be required. */
791    
792  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
793  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
794  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
795    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
796    
797  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
798  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
799    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
800  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
801  #endif  #endif
802    
# Line 434  else if ((i = escapes[c - 0x48]) != 0) Line 804  else if ((i = escapes[c - 0x48]) != 0)
804    
805  else  else
806    {    {
807    const uschar *oldptr;    const pcre_uchar *oldptr;
808    BOOL braced, negated;    BOOL braced, negated;
809    
810    switch (c)    switch (c)
# Line 442  else Line 812  else
812      /* 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
813      error. */      error. */
814    
815      case 'l':      case CHAR_l:
816      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
817      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
818      break;      break;
819    
820      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
821      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
822      This is a Perl 5.10 feature. */        {
823          /* In JavaScript, \u must be followed by four hexadecimal numbers.
824          Otherwise it is a lowercase u letter. */
825          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
826            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
827            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
828            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
829            {
830            c = 0;
831            for (i = 0; i < 4; ++i)
832              {
833              register int cc = *(++ptr);
834    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
836              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837    #else           /* EBCDIC coding */
838              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
839              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840    #endif
841              }
842    
843    #if defined COMPILE_PCRE8
844            if (c > (utf ? 0x10ffff : 0xff))
845    #elif defined COMPILE_PCRE16
846            if (c > (utf ? 0x10ffff : 0xffff))
847    #elif defined COMPILE_PCRE32
848            if (utf && c > 0x10ffff)
849    #endif
850              {
851              *errorcodeptr = ERR76;
852              }
853            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
854            }
855          }
856        else
857          *errorcodeptr = ERR37;
858        break;
859    
860        case CHAR_U:
861        /* In JavaScript, \U is an uppercase U letter. */
862        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
863        break;
864    
865        /* In a character class, \g is just a literal "g". Outside a character
866        class, \g must be followed by one of a number of specific things:
867    
868        (1) A number, either plain or braced. If positive, it is an absolute
869        backreference. If negative, it is a relative backreference. This is a Perl
870        5.10 feature.
871    
872        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
873        is part of Perl's movement towards a unified syntax for back references. As
874        this is synonymous with \k{name}, we fudge it up by pretending it really
875        was \k.
876    
877        (3) For Oniguruma compatibility we also support \g followed by a name or a
878        number either in angle brackets or in single quotes. However, these are
879        (possibly recursive) subroutine calls, _not_ backreferences. Just return
880        the -ESC_g code (cf \k). */
881    
882        case CHAR_g:
883        if (isclass) break;
884        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
885          {
886          c = -ESC_g;
887          break;
888          }
889    
890        /* Handle the Perl-compatible cases */
891    
892      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
893        {        {
894          const pcre_uchar *p;
895          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
896            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
897          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
898            {
899            c = -ESC_k;
900            break;
901            }
902        braced = TRUE;        braced = TRUE;
903        ptr++;        ptr++;
904        }        }
905      else braced = FALSE;      else braced = FALSE;
906    
907      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
908        {        {
909        negated = TRUE;        negated = TRUE;
910        ptr++;        ptr++;
911        }        }
912      else negated = FALSE;      else negated = FALSE;
913    
914        /* The integer range is limited by the machine's int representation. */
915      c = 0;      c = 0;
916      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
917        c = c * 10 + *(++ptr) - '0';        {
918          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
919            {
920            c = -1;
921            break;
922            }
923          c = c * 10 + *(++ptr) - CHAR_0;
924          }
925        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
926          {
927          while (IS_DIGIT(ptr[1]))
928            ptr++;
929          *errorcodeptr = ERR61;
930          break;
931          }
932    
933      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
934        {        {
935        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
936        return 0;        break;
937          }
938    
939        if (c == 0)
940          {
941          *errorcodeptr = ERR58;
942          break;
943        }        }
944    
945      if (negated)      if (negated)
# Line 484  else Line 947  else
947        if (c > bracount)        if (c > bracount)
948          {          {
949          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
950          return 0;          break;
951          }          }
952        c = bracount - (c - 1);        c = bracount - (c - 1);
953        }        }
# Line 504  else Line 967  else
967      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
968      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
969    
970      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:
971      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
972    
973      if (!isclass)      if (!isclass)
974        {        {
975        oldptr = ptr;        oldptr = ptr;
976        c -= '0';        /* The integer range is limited by the machine's int representation. */
977        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
978          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
979            {
980            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
981              {
982              c = -1;
983              break;
984              }
985            c = c * 10 + *(++ptr) - CHAR_0;
986            }
987          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
988            {
989            while (IS_DIGIT(ptr[1]))
990              ptr++;
991            *errorcodeptr = ERR61;
992            break;
993            }
994        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
995          {          {
996          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 1003  else
1003      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.
1004      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1005    
1006      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1007        {        {
1008        ptr--;        ptr--;
1009        c = 0;        c = 0;
# Line 535  else Line 1013  else
1013      /* \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
1014      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
1015      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
1016      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,
1017      than 3 octal digits. */      but no more than 3 octal digits. */
1018    
1019      case '0':      case CHAR_0:
1020      c -= '0';      c -= CHAR_0;
1021      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1022          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1023      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1024        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1025    #endif
1026      break;      break;
1027    
1028      /* \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
1029      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.
1030      treated as a data character. */      If not, { is treated as a data character. */
1031    
1032        case CHAR_x:
1033        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1034          {
1035          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1036          Otherwise it is a lowercase x letter. */
1037          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1038            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1039            {
1040            c = 0;
1041            for (i = 0; i < 2; ++i)
1042              {
1043              register int cc = *(++ptr);
1044    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1046              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1047    #else           /* EBCDIC coding */
1048              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1049              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050    #endif
1051              }
1052            }
1053          break;
1054          }
1055    
1056      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1057        {        {
1058        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1059    
1060        c = 0;        c = 0;
1061        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1062          {          {
1063          register int cc = *pt++;          register int cc = *pt++;
1064          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1065    
1066  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1067          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1068          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1069  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1070          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1071          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1072    #endif
1073    
1074    #if defined COMPILE_PCRE8
1075            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1076    #elif defined COMPILE_PCRE16
1077            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1078    #elif defined COMPILE_PCRE32
1079            if (utf && c > 0x10ffff) { c = -1; break; }
1080  #endif  #endif
1081          }          }
1082    
1083        if (*pt == '}')        if (c < 0)
1084            {
1085            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1086            *errorcodeptr = ERR34;
1087            }
1088    
1089          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1090          {          {
1091          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1092          ptr = pt;          ptr = pt;
1093          break;          break;
1094          }          }
# Line 585  else Line 1100  else
1100      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1101    
1102      c = 0;      c = 0;
1103      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1104        {        {
1105        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1106        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1107  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1108        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1109        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1110  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1111        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1112        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1113  #endif  #endif
1114        }        }
1115      break;      break;
1116    
1117      /* 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.
1118      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
1119        coding is ASCII-specific, but then the whole concept of \cx is
1120      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1121    
1122      case 'c':      case CHAR_c:
1123      c = *(++ptr);      c = *(++ptr);
1124      if (c == 0)      if (c == 0)
1125        {        {
1126        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1127        return 0;        break;
1128        }        }
1129    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1130  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1131      if (c >= 'a' && c <= 'z') c -= 32;        {
1132          *errorcodeptr = ERR68;
1133          break;
1134          }
1135        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1136      c ^= 0x40;      c ^= 0x40;
1137  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1138      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1139      c ^= 0xC0;      c ^= 0xC0;
1140  #endif  #endif
1141      break;      break;
1142    
1143      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1144      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1145      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
1146      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
1147      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1148    
1149      default:      default:
1150      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1157  else
1157      }      }
1158    }    }
1159    
1160    /* Perl supports \N{name} for character names, as well as plain \N for "not
1161    newline". PCRE does not support \N{name}. However, it does support
1162    quantification such as \N{2,3}. */
1163    
1164    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1165         !is_counted_repeat(ptr+2))
1166      *errorcodeptr = ERR37;
1167    
1168    /* If PCRE_UCP is set, we change the values for \d etc. */
1169    
1170    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1171      c -= (ESC_DU - ESC_D);
1172    
1173    /* Set the pointer to the final character before returning. */
1174    
1175  *ptrptr = ptr;  *ptrptr = ptr;
1176  return c;  return c;
1177  }  }
# Line 663  Returns:         type value from ucp_typ Line 1198  Returns:         type value from ucp_typ
1198  */  */
1199    
1200  static int  static int
1201  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1202  {  {
1203  int c, i, bot, top;  int c, i, bot, top;
1204  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1205  char name[32];  pcre_uchar name[32];
1206    
1207  c = *(++ptr);  c = *(++ptr);
1208  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 677  if (c == 0) goto ERROR_RETURN; Line 1212  if (c == 0) goto ERROR_RETURN;
1212  /* \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
1213  negation. */  negation. */
1214    
1215  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1216    {    {
1217    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1218      {      {
1219      *negptr = TRUE;      *negptr = TRUE;
1220      ptr++;      ptr++;
1221      }      }
1222    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1223      {      {
1224      c = *(++ptr);      c = *(++ptr);
1225      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1226      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1227      name[i] = c;      name[i] = c;
1228      }      }
1229    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1230    name[i] = 0;    name[i] = 0;
1231    }    }
1232    
# Line 708  else Line 1243  else
1243  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1244    
1245  bot = 0;  bot = 0;
1246  top = _pcre_utt_size;  top = PRIV(utt_size);
1247    
1248  while (bot < top)  while (bot < top)
1249    {    {
1250    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1251    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1252    if (c == 0)    if (c == 0)
1253      {      {
1254      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1255      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1256      }      }
1257    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1258    }    }
# Line 737  return -1; Line 1272  return -1;
1272    
1273    
1274  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1275  *         Read repeat counts                     *  *         Read repeat counts                     *
1276  *************************************************/  *************************************************/
1277    
# Line 788  Returns:         pointer to '}' on succe Line 1290  Returns:         pointer to '}' on succe
1290                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1291  */  */
1292    
1293  static const uschar *  static const pcre_uchar *
1294  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)
1295  {  {
1296  int min = 0;  int min = 0;
1297  int max = -1;  int max = -1;
# Line 797  int max = -1; Line 1299  int max = -1;
1299  /* 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
1300  an integer overflow. */  an integer overflow. */
1301    
1302  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1303  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1304    {    {
1305    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1309  if (min < 0 || min > 65535)
1309  /* 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.
1310  Also, max must not be less than min. */  Also, max must not be less than min. */
1311    
1312  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1313    {    {
1314    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1315      {      {
1316      max = 0;      max = 0;
1317      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1318      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1319        {        {
1320        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1339  return p;
1339    
1340    
1341  /*************************************************  /*************************************************
1342  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1343  *************************************************/  *************************************************/
1344    
1345  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1346    top-level call starts at the beginning of the pattern. All other calls must
1347    start at a parenthesis. It scans along a pattern's text looking for capturing
1348  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
1349  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
1350  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
1351  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1352  be terminated by '>' because that is checked in the first pass.  
1353    This function was originally called only from the second pass, in which we know
1354    that if (?< or (?' or (?P< is encountered, the name will be correctly
1355    terminated because that is checked in the first pass. There is now one call to
1356    this function in the first pass, to check for a recursive back reference by
1357    name (so that we can make the whole group atomic). In this case, we need check
1358    only up to the current position in the pattern, and that is still OK because
1359    and previous occurrences will have been checked. To make this work, the test
1360    for "end of pattern" is a check against cd->end_pattern in the main loop,
1361    instead of looking for a binary zero. This means that the special first-pass
1362    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1363    processing items within the loop are OK, because afterwards the main loop will
1364    terminate.)
1365    
1366  Arguments:  Arguments:
1367    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1368    count        current count of capturing parens so far encountered    cd           compile background data
1369    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1370    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1371    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1372      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1373      count        pointer to the current capturing subpattern number (updated)
1374    
1375  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1376  */  */
1377    
1378  static int  static int
1379  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,
1380    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1381  {  {
1382  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1383    int start_count = *count;
1384    int hwm_count = start_count;
1385    BOOL dup_parens = FALSE;
1386    
1387    /* If the first character is a parenthesis, check on the type of group we are
1388    dealing with. The very first call may not start with a parenthesis. */
1389    
1390  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1391    {    {
1392    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1393    
1394      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1395    
1396      /* Handle a normal, unnamed capturing parenthesis. */
1397    
1398      else if (ptr[1] != CHAR_QUESTION_MARK)
1399        {
1400        *count += 1;
1401        if (name == NULL && *count == lorn) return *count;
1402        ptr++;
1403        }
1404    
1405      /* All cases now have (? at the start. Remember when we are in a group
1406      where the parenthesis numbers are duplicated. */
1407    
1408      else if (ptr[2] == CHAR_VERTICAL_LINE)
1409        {
1410        ptr += 3;
1411        dup_parens = TRUE;
1412        }
1413    
1414      /* Handle comments; all characters are allowed until a ket is reached. */
1415    
1416      else if (ptr[2] == CHAR_NUMBER_SIGN)
1417        {
1418        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1419        goto FAIL_EXIT;
1420        }
1421    
1422      /* Handle a condition. If it is an assertion, just carry on so that it
1423      is processed as normal. If not, skip to the closing parenthesis of the
1424      condition (there can't be any nested parens). */
1425    
1426      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1427        {
1428        ptr += 2;
1429        if (ptr[1] != CHAR_QUESTION_MARK)
1430          {
1431          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1432          if (*ptr != 0) ptr++;
1433          }
1434        }
1435    
1436      /* Start with (? but not a condition. */
1437    
1438      else
1439        {
1440        ptr += 2;
1441        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1442    
1443        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1444    
1445        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1446            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1447          {
1448          int term;
1449          const pcre_uchar *thisname;
1450          *count += 1;
1451          if (name == NULL && *count == lorn) return *count;
1452          term = *ptr++;
1453          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1454          thisname = ptr;
1455          while (*ptr != term) ptr++;
1456          if (name != NULL && lorn == ptr - thisname &&
1457              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1458            return *count;
1459          term++;
1460          }
1461        }
1462      }
1463    
1464    /* Past any initial parenthesis handling, scan for parentheses or vertical
1465    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1466    first-pass call when this value is temporarily adjusted to stop at the current
1467    position. So DO NOT change this to a test for binary zero. */
1468    
1469    for (; ptr < cd->end_pattern; ptr++)
1470      {
1471    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1472    
1473    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1474      {      {
1475      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1476      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1477        {        {
1478        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1479        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1480        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1481        }        }
1482      continue;      continue;
1483      }      }
1484    
1485    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1486      are handled for real. If the first character is '^', skip it. Also, if the
1487      first few characters (either before or after ^) are \Q\E or \E we skip them
1488      too. This makes for compatibility with Perl. Note the use of STR macros to
1489      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1490    
1491    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1492      {      {
1493      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1494        for (;;)
1495          {
1496          if (ptr[1] == CHAR_BACKSLASH)
1497            {
1498            if (ptr[2] == CHAR_E)
1499              ptr+= 2;
1500            else if (STRNCMP_UC_C8(ptr + 2,
1501                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1502              ptr += 4;
1503            else
1504              break;
1505            }
1506          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1507            {
1508            negate_class = TRUE;
1509            ptr++;
1510            }
1511          else break;
1512          }
1513    
1514        /* If the next character is ']', it is a data character that must be
1515        skipped, except in JavaScript compatibility mode. */
1516    
1517        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1518            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1519          ptr++;
1520    
1521        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1522        {        {
1523        if (*ptr == '\\')        if (*ptr == 0) return -1;
1524          if (*ptr == CHAR_BACKSLASH)
1525          {          {
1526          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1527          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1528            {            {
1529            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1530            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1531            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1532            }            }
1533          continue;          continue;
1534          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1538  for (; *ptr != 0; ptr++)
1538    
1539    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1540    
1541    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1542      {      {
1543      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1544      if (*ptr == 0) return -1;      while (*ptr != 0)
1545          {
1546          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1547          ptr++;
1548    #ifdef SUPPORT_UTF
1549          if (utf) FORWARDCHAR(ptr);
1550    #endif
1551          }
1552        if (*ptr == 0) goto FAIL_EXIT;
1553      continue;      continue;
1554      }      }
1555    
1556    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1557    
1558    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1559      {      {
1560      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1561      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1562      continue;      if (*ptr == 0) goto FAIL_EXIT;
1563      }      }
1564    
1565    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1566    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1567        if (dup_parens && *count < hwm_count) *count = hwm_count;
1568    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1569        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1570    
1571    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1572    term = *ptr++;      {
1573    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1574    thisname = ptr;      *count = start_count;
1575    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1576    }    }
1577    
1578    FAIL_EXIT:
1579    *ptrptr = ptr;
1580  return -1;  return -1;
1581  }  }
1582    
1583    
1584    
1585    
1586  /*************************************************  /*************************************************
1587  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1588  *************************************************/  *************************************************/
1589    
1590  /* This is called by several functions that scan a compiled expression looking  /* This function scans along a pattern's text looking for capturing
1591    subpatterns, and counting them. If it finds a named pattern that matches the
1592    name it is given, it returns its number. Alternatively, if the name is NULL, it
1593    returns when it reaches a given numbered subpattern. This is used for forward
1594    references to subpatterns. We used to be able to start this scan from the
1595    current compiling point, using the current count value from cd->bracount, and
1596    do it all in a single loop, but the addition of the possibility of duplicate
1597    subpattern numbers means that we have to scan from the very start, in order to
1598    take account of such duplicates, and to use a recursive function to keep track
1599    of the different types of group.
1600    
1601    Arguments:
1602      cd           compile background data
1603      name         name to seek, or NULL if seeking a numbered subpattern
1604      lorn         name length, or subpattern number if name is NULL
1605      xmode        TRUE if we are in /x mode
1606      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1607    
1608    Returns:       the number of the found subpattern, or -1 if not found
1609    */
1610    
1611    static int
1612    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1613      BOOL utf)
1614    {
1615    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1616    int count = 0;
1617    int rc;
1618    
1619    /* If the pattern does not start with an opening parenthesis, the first call
1620    to find_parens_sub() will scan right to the end (if necessary). However, if it
1621    does start with a parenthesis, find_parens_sub() will return when it hits the
1622    matching closing parens. That is why we have to have a loop. */
1623    
1624    for (;;)
1625      {
1626      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1627      if (rc > 0 || *ptr++ == 0) break;
1628      }
1629    
1630    return rc;
1631    }
1632    
1633    
1634    
1635    
1636    /*************************************************
1637    *      Find first significant op code            *
1638    *************************************************/
1639    
1640    /* This is called by several functions that scan a compiled expression looking
1641  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
1642  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
1643  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
1644  assertions, and also the \b assertion; for others it does not.  does not.
1645    
1646  Arguments:  Arguments:
1647    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  
1648    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1649    
1650  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1651  */  */
1652    
1653  static const uschar*  static const pcre_uchar*
1654  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1655  {  {
1656  for (;;)  for (;;)
1657    {    {
1658    switch ((int)*code)    switch ((int)*code)
1659      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1660      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1661      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1662      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1663      if (!skipassert) return code;      if (!skipassert) return code;
1664      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1665      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1666      break;      break;
1667    
1668      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1672  for (;;)
1672    
1673      case OP_CALLOUT:      case OP_CALLOUT:
1674      case OP_CREF:      case OP_CREF:
1675        case OP_NCREF:
1676      case OP_RREF:      case OP_RREF:
1677        case OP_NRREF:
1678      case OP_DEF:      case OP_DEF:
1679      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1680      break;      break;
1681    
1682      default:      default:
# Line 1012  for (;;) Line 1690  for (;;)
1690    
1691    
1692  /*************************************************  /*************************************************
1693  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1694  *************************************************/  *************************************************/
1695    
1696  /* 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,
1697  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.
1698  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
1699    temporarily terminated with OP_END when this function is called.
1700    
1701    This function is called when a backward assertion is encountered, so that if it
1702    fails, the error message can point to the correct place in the pattern.
1703    However, we cannot do this when the assertion contains subroutine calls,
1704    because they can be forward references. We solve this by remembering this case
1705    and doing the check at the end; a flag specifies which mode we are running in.
1706    
1707  Arguments:  Arguments:
1708    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1709    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1710      atend    TRUE if called when the pattern is complete
1711  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1712               or -2 if \C was encountered  
1713    Returns:   the fixed length,
1714                 or -1 if there is no fixed length,
1715                 or -2 if \C was encountered (in UTF-8 mode only)
1716                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1717                 or -4 if an unknown opcode was encountered (internal error)
1718  */  */
1719    
1720  static int  static int
1721  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1722  {  {
1723  int length = -1;  int length = -1;
1724    
1725  register int branchlength = 0;  register int branchlength = 0;
1726  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1727    
1728  /* 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
1729  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1041  branch, check the length against that of Line 1731  branch, check the length against that of
1731  for (;;)  for (;;)
1732    {    {
1733    int d;    int d;
1734      pcre_uchar *ce, *cs;
1735    register int op = *cc;    register int op = *cc;
1736    
1737    switch (op)    switch (op)
1738      {      {
1739        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1740        OP_BRA (normal non-capturing bracket) because the other variants of these
1741        opcodes are all concerned with unlimited repeated groups, which of course
1742        are not of fixed length. */
1743    
1744      case OP_CBRA:      case OP_CBRA:
1745      case OP_BRA:      case OP_BRA:
1746      case OP_ONCE:      case OP_ONCE:
1747        case OP_ONCE_NC:
1748      case OP_COND:      case OP_COND:
1749      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1750      if (d < 0) return d;      if (d < 0) return d;
1751      branchlength += d;      branchlength += d;
1752      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1753      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1754      break;      break;
1755    
1756      /* 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.
1757      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
1758      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
1759        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1760        because they all imply an unlimited repeat. */
1761    
1762      case OP_ALT:      case OP_ALT:
1763      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1764      case OP_END:      case OP_END:
1765        case OP_ACCEPT:
1766        case OP_ASSERT_ACCEPT:
1767      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1768        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1769      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1771  for (;;)
1771      branchlength = 0;      branchlength = 0;
1772      break;      break;
1773    
1774        /* A true recursion implies not fixed length, but a subroutine call may
1775        be OK. If the subroutine is a forward reference, we can't deal with
1776        it until the end of the pattern, so return -3. */
1777    
1778        case OP_RECURSE:
1779        if (!atend) return -3;
1780        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1781        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1782        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1783        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1784        if (d < 0) return d;
1785        branchlength += d;
1786        cc += 1 + LINK_SIZE;
1787        break;
1788    
1789      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1790    
1791      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1793  for (;;)
1793      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1794      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1795      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1796      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1797        break;
1798    
1799      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1800    
1801      case OP_REVERSE:      case OP_MARK:
1802        case OP_PRUNE_ARG:
1803        case OP_SKIP_ARG:
1804        case OP_THEN_ARG:
1805        cc += cc[1] + PRIV(OP_lengths)[*cc];
1806        break;
1807    
1808        case OP_CALLOUT:
1809        case OP_CIRC:
1810        case OP_CIRCM:
1811        case OP_CLOSE:
1812        case OP_COMMIT:
1813      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1814      case OP_DEF:      case OP_DEF:
1815      case OP_OPT:      case OP_DOLL:
1816      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1817      case OP_EOD:      case OP_EOD:
1818      case OP_EODN:      case OP_EODN:
1819      case OP_CIRC:      case OP_FAIL:
1820      case OP_DOLL:      case OP_NCREF:
1821        case OP_NRREF:
1822      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1823        case OP_PRUNE:
1824        case OP_REVERSE:
1825        case OP_RREF:
1826        case OP_SET_SOM:
1827        case OP_SKIP:
1828        case OP_SOD:
1829        case OP_SOM:
1830        case OP_THEN:
1831      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1832      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1833      break;      break;
1834    
1835      /* Handle literal characters */      /* Handle literal characters */
1836    
1837      case OP_CHAR:      case OP_CHAR:
1838      case OP_CHARNC:      case OP_CHARI:
1839      case OP_NOT:      case OP_NOT:
1840        case OP_NOTI:
1841      branchlength++;      branchlength++;
1842      cc += 2;      cc += 2;
1843  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1844      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1845  #endif  #endif
1846      break;      break;
1847    
# Line 1119  for (;;) Line 1849  for (;;)
1849      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1850    
1851      case OP_EXACT:      case OP_EXACT:
1852        case OP_EXACTI:
1853        case OP_NOTEXACT:
1854        case OP_NOTEXACTI:
1855      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1856      cc += 4;      cc += 2 + IMM2_SIZE;
1857  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1858      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1859  #endif  #endif
1860      break;      break;
1861    
1862      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1863      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1864      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1865          cc += 2;
1866        cc += 1 + IMM2_SIZE + 1;
1867      break;      break;
1868    
1869      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1873  for (;;)
1873      cc += 2;      cc += 2;
1874      /* Fall through */      /* Fall through */
1875    
1876        case OP_HSPACE:
1877        case OP_VSPACE:
1878        case OP_NOT_HSPACE:
1879        case OP_NOT_VSPACE:
1880      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1881      case OP_DIGIT:      case OP_DIGIT:
1882      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1884  for (;;)
1884      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1885      case OP_WORDCHAR:      case OP_WORDCHAR:
1886      case OP_ANY:      case OP_ANY:
1887        case OP_ALLANY:
1888      branchlength++;      branchlength++;
1889      cc++;      cc++;
1890      break;      break;
1891    
1892      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1893        otherwise \C is coded as OP_ALLANY. */
1894    
1895      case OP_ANYBYTE:      case OP_ANYBYTE:
1896      return -2;      return -2;
1897    
1898      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1899    
1900  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1901      case OP_XCLASS:      case OP_XCLASS:
1902      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1903      /* Fall through */      /* Fall through */
1904  #endif  #endif
1905    
1906      case OP_CLASS:      case OP_CLASS:
1907      case OP_NCLASS:      case OP_NCLASS:
1908      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1909    
1910      switch (*cc)      switch (*cc)
1911        {        {
1912          case OP_CRPLUS:
1913          case OP_CRMINPLUS:
1914        case OP_CRSTAR:        case OP_CRSTAR:
1915        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1916        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1179  for (;;) Line 1919  for (;;)
1919    
1920        case OP_CRRANGE:        case OP_CRRANGE:
1921        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1922        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1923        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1924        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1925        break;        break;
1926    
1927        default:        default:
# Line 1191  for (;;) Line 1931  for (;;)
1931    
1932      /* Anything else is variable length */      /* Anything else is variable length */
1933    
1934      default:      case OP_ANYNL:
1935        case OP_BRAMINZERO:
1936        case OP_BRAPOS:
1937        case OP_BRAPOSZERO:
1938        case OP_BRAZERO:
1939        case OP_CBRAPOS:
1940        case OP_EXTUNI:
1941        case OP_KETRMAX:
1942        case OP_KETRMIN:
1943        case OP_KETRPOS:
1944        case OP_MINPLUS:
1945        case OP_MINPLUSI:
1946        case OP_MINQUERY:
1947        case OP_MINQUERYI:
1948        case OP_MINSTAR:
1949        case OP_MINSTARI:
1950        case OP_MINUPTO:
1951        case OP_MINUPTOI:
1952        case OP_NOTMINPLUS:
1953        case OP_NOTMINPLUSI:
1954        case OP_NOTMINQUERY:
1955        case OP_NOTMINQUERYI:
1956        case OP_NOTMINSTAR:
1957        case OP_NOTMINSTARI:
1958        case OP_NOTMINUPTO:
1959        case OP_NOTMINUPTOI:
1960        case OP_NOTPLUS:
1961        case OP_NOTPLUSI:
1962        case OP_NOTPOSPLUS:
1963        case OP_NOTPOSPLUSI:
1964        case OP_NOTPOSQUERY:
1965        case OP_NOTPOSQUERYI:
1966        case OP_NOTPOSSTAR:
1967        case OP_NOTPOSSTARI:
1968        case OP_NOTPOSUPTO:
1969        case OP_NOTPOSUPTOI:
1970        case OP_NOTQUERY:
1971        case OP_NOTQUERYI:
1972        case OP_NOTSTAR:
1973        case OP_NOTSTARI:
1974        case OP_NOTUPTO:
1975        case OP_NOTUPTOI:
1976        case OP_PLUS:
1977        case OP_PLUSI:
1978        case OP_POSPLUS:
1979        case OP_POSPLUSI:
1980        case OP_POSQUERY:
1981        case OP_POSQUERYI:
1982        case OP_POSSTAR:
1983        case OP_POSSTARI:
1984        case OP_POSUPTO:
1985        case OP_POSUPTOI:
1986        case OP_QUERY:
1987        case OP_QUERYI:
1988        case OP_REF:
1989        case OP_REFI:
1990        case OP_SBRA:
1991        case OP_SBRAPOS:
1992        case OP_SCBRA:
1993        case OP_SCBRAPOS:
1994        case OP_SCOND:
1995        case OP_SKIPZERO:
1996        case OP_STAR:
1997        case OP_STARI:
1998        case OP_TYPEMINPLUS:
1999        case OP_TYPEMINQUERY:
2000        case OP_TYPEMINSTAR:
2001        case OP_TYPEMINUPTO:
2002        case OP_TYPEPLUS:
2003        case OP_TYPEPOSPLUS:
2004        case OP_TYPEPOSQUERY:
2005        case OP_TYPEPOSSTAR:
2006        case OP_TYPEPOSUPTO:
2007        case OP_TYPEQUERY:
2008        case OP_TYPESTAR:
2009        case OP_TYPEUPTO:
2010        case OP_UPTO:
2011        case OP_UPTOI:
2012      return -1;      return -1;
2013    
2014        /* Catch unrecognized opcodes so that when new ones are added they
2015        are not forgotten, as has happened in the past. */
2016    
2017        default:
2018        return -4;
2019      }      }
2020    }    }
2021  /* Control never gets here */  /* Control never gets here */
# Line 1202  for (;;) Line 2025  for (;;)
2025    
2026    
2027  /*************************************************  /*************************************************
2028  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2029  *************************************************/  *************************************************/
2030    
2031  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2032  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2033    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2034    so that it can be called from pcre_study() when finding the minimum matching
2035    length.
2036    
2037  Arguments:  Arguments:
2038    code        points to start of expression    code        points to start of expression
2039    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2040    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2041    
2042  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
2043  */  */
2044    
2045  static const uschar *  const pcre_uchar *
2046  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2047  {  {
2048  for (;;)  for (;;)
2049    {    {
2050    register int c = *code;    register int c = *code;
2051    
2052    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2053    
2054    /* 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 1230  for (;;) Line 2057  for (;;)
2057    
2058    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2059    
2060      /* Handle recursion */
2061    
2062      else if (c == OP_REVERSE)
2063        {
2064        if (number < 0) return (pcre_uchar *)code;
2065        code += PRIV(OP_lengths)[c];
2066        }
2067    
2068    /* Handle capturing bracket */    /* Handle capturing bracket */
2069    
2070    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2071               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2072      {      {
2073      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2074      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2075      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2076      }      }
2077    
2078      /* Otherwise, we can get the item's length from the table, except that for
2079      repeated character types, we have to test for \p and \P, which have an extra
2080      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2081      must add in its length. */
2082    
2083      else
2084        {
2085        switch(c)
2086          {
2087          case OP_TYPESTAR:
2088          case OP_TYPEMINSTAR:
2089          case OP_TYPEPLUS:
2090          case OP_TYPEMINPLUS:
2091          case OP_TYPEQUERY:
2092          case OP_TYPEMINQUERY:
2093          case OP_TYPEPOSSTAR:
2094          case OP_TYPEPOSPLUS:
2095          case OP_TYPEPOSQUERY:
2096          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2097          break;
2098    
2099          case OP_TYPEUPTO:
2100          case OP_TYPEMINUPTO:
2101          case OP_TYPEEXACT:
2102          case OP_TYPEPOSUPTO:
2103          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2104            code += 2;
2105          break;
2106    
2107          case OP_MARK:
2108          case OP_PRUNE_ARG:
2109          case OP_SKIP_ARG:
2110          code += code[1];
2111          break;
2112    
2113          case OP_THEN_ARG:
2114          code += code[1];
2115          break;
2116          }
2117    
2118        /* Add in the fixed length from the table */
2119    
2120        code += PRIV(OP_lengths)[c];
2121    
2122    /* 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
2123    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
2124    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2125    
2126    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2127      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2128        {        {
2129        case OP_CHAR:        case OP_CHAR:
2130        case OP_CHARNC:        case OP_CHARI:
2131        case OP_EXACT:        case OP_EXACT:
2132          case OP_EXACTI:
2133        case OP_UPTO:        case OP_UPTO:
2134          case OP_UPTOI:
2135        case OP_MINUPTO:        case OP_MINUPTO:
2136          case OP_MINUPTOI:
2137        case OP_POSUPTO:        case OP_POSUPTO:
2138          case OP_POSUPTOI:
2139        case OP_STAR:        case OP_STAR:
2140          case OP_STARI:
2141        case OP_MINSTAR:        case OP_MINSTAR:
2142          case OP_MINSTARI:
2143        case OP_POSSTAR:        case OP_POSSTAR:
2144          case OP_POSSTARI:
2145        case OP_PLUS:        case OP_PLUS:
2146          case OP_PLUSI:
2147        case OP_MINPLUS:        case OP_MINPLUS:
2148          case OP_MINPLUSI:
2149        case OP_POSPLUS:        case OP_POSPLUS:
2150          case OP_POSPLUSI:
2151        case OP_QUERY:        case OP_QUERY:
2152          case OP_QUERYI:
2153        case OP_MINQUERY:        case OP_MINQUERY:
2154          case OP_MINQUERYI:
2155        case OP_POSQUERY:        case OP_POSQUERY:
2156        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2157          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2158        break;        break;
2159        }        }
2160    #else
2161        (void)(utf);  /* Keep compiler happy by referencing function argument */
2162  #endif  #endif
2163      }      }
2164    }    }
# Line 1283  instance of OP_RECURSE. Line 2175  instance of OP_RECURSE.
2175    
2176  Arguments:  Arguments:
2177    code        points to start of expression    code        points to start of expression
2178    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2179    
2180  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
2181  */  */
2182    
2183  static const uschar *  static const pcre_uchar *
2184  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2185  {  {
2186  for (;;)  for (;;)
2187    {    {
# Line 1303  for (;;) Line 2195  for (;;)
2195    
2196    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2197    
2198    /* 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
2199    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
2200    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
2201    bytes. */    must add in its length. */
2202    
2203    else    else
2204      {      {
2205      code += _pcre_OP_lengths[c];      switch(c)
2206  #ifdef SUPPORT_UTF8        {
2207      if (utf8) switch(c)        case OP_TYPESTAR:
2208          case OP_TYPEMINSTAR:
2209          case OP_TYPEPLUS:
2210          case OP_TYPEMINPLUS:
2211          case OP_TYPEQUERY:
2212          case OP_TYPEMINQUERY:
2213          case OP_TYPEPOSSTAR:
2214          case OP_TYPEPOSPLUS:
2215          case OP_TYPEPOSQUERY:
2216          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2217          break;
2218    
2219          case OP_TYPEPOSUPTO:
2220          case OP_TYPEUPTO:
2221          case OP_TYPEMINUPTO:
2222          case OP_TYPEEXACT:
2223          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2224            code += 2;
2225          break;
2226    
2227          case OP_MARK:
2228          case OP_PRUNE_ARG:
2229          case OP_SKIP_ARG:
2230          code += code[1];
2231          break;
2232    
2233          case OP_THEN_ARG:
2234          code += code[1];
2235          break;
2236          }
2237    
2238        /* Add in the fixed length from the table */
2239    
2240        code += PRIV(OP_lengths)[c];
2241    
2242        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2243        by a multi-byte character. The length in the table is a minimum, so we have
2244        to arrange to skip the extra bytes. */
2245    
2246    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2247        if (utf) switch(c)
2248        {        {
2249        case OP_CHAR:        case OP_CHAR:
2250        case OP_CHARNC:        case OP_CHARI:
2251          case OP_NOT:
2252          case OP_NOTI:
2253        case OP_EXACT:        case OP_EXACT:
2254          case OP_EXACTI:
2255          case OP_NOTEXACT:
2256          case OP_NOTEXACTI:
2257        case OP_UPTO:        case OP_UPTO:
2258          case OP_UPTOI:
2259          case OP_NOTUPTO:
2260          case OP_NOTUPTOI:
2261        case OP_MINUPTO:        case OP_MINUPTO:
2262          case OP_MINUPTOI:
2263          case OP_NOTMINUPTO:
2264          case OP_NOTMINUPTOI:
2265        case OP_POSUPTO:        case OP_POSUPTO:
2266          case OP_POSUPTOI:
2267          case OP_NOTPOSUPTO:
2268          case OP_NOTPOSUPTOI:
2269        case OP_STAR:        case OP_STAR:
2270          case OP_STARI:
2271          case OP_NOTSTAR:
2272          case OP_NOTSTARI:
2273        case OP_MINSTAR:        case OP_MINSTAR:
2274          case OP_MINSTARI:
2275          case OP_NOTMINSTAR:
2276          case OP_NOTMINSTARI:
2277        case OP_POSSTAR:        case OP_POSSTAR:
2278          case OP_POSSTARI:
2279          case OP_NOTPOSSTAR:
2280          case OP_NOTPOSSTARI:
2281        case OP_PLUS:        case OP_PLUS:
2282          case OP_PLUSI:
2283          case OP_NOTPLUS:
2284          case OP_NOTPLUSI:
2285        case OP_MINPLUS:        case OP_MINPLUS:
2286          case OP_MINPLUSI:
2287          case OP_NOTMINPLUS:
2288          case OP_NOTMINPLUSI:
2289        case OP_POSPLUS:        case OP_POSPLUS:
2290          case OP_POSPLUSI:
2291          case OP_NOTPOSPLUS:
2292          case OP_NOTPOSPLUSI:
2293        case OP_QUERY:        case OP_QUERY:
2294          case OP_QUERYI:
2295          case OP_NOTQUERY:
2296          case OP_NOTQUERYI:
2297        case OP_MINQUERY:        case OP_MINQUERY:
2298          case OP_MINQUERYI:
2299          case OP_NOTMINQUERY:
2300          case OP_NOTMINQUERYI:
2301        case OP_POSQUERY:        case OP_POSQUERY:
2302        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2303          case OP_NOTPOSQUERY:
2304          case OP_NOTPOSQUERYI:
2305          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2306        break;        break;
2307        }        }
2308    #else
2309        (void)(utf);  /* Keep compiler happy by referencing function argument */
2310  #endif  #endif
2311      }      }
2312    }    }
# Line 1347  for (;;) Line 2322  for (;;)
2322  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()
2323  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
2324  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2325  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
2326  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
2327    bracket whose current branch will already have been scanned.
2328    
2329  Arguments:  Arguments:
2330    code        points to start of search    code        points to start of search
2331    endcode     points to where to stop    endcode     points to where to stop
2332    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2333      cd          contains pointers to tables etc.
2334    
2335  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2336  */  */
2337    
2338  static BOOL  static BOOL
2339  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2340      BOOL utf, compile_data *cd)
2341  {  {
2342  register int c;  register int c;
2343  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346    {    {
2347    const uschar *ccode;    const pcre_uchar *ccode;
2348    
2349    c = *code;    c = *code;
2350    
2351    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2352      first_significant_code() with a TRUE final argument. */
2353    
2354      if (c == OP_ASSERT)
2355      {      {
2356        do code += GET(code, 1); while (*code == OP_ALT);
2357        c = *code;
2358        continue;
2359        }
2360    
2361      /* For a recursion/subroutine call, if its end has been reached, which
2362      implies a backward reference subroutine call, we can scan it. If it's a
2363      forward reference subroutine call, we can't. To detect forward reference
2364      we have to scan up the list that is kept in the workspace. This function is
2365      called only when doing the real compile, not during the pre-compile that
2366      measures the size of the compiled pattern. */
2367    
2368      if (c == OP_RECURSE)
2369        {
2370        const pcre_uchar *scode;
2371      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2372    
2373      /* Scan a closed bracket */      /* Test for forward reference */
2374    
2375        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2376          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2377    
2378        /* Not a forward reference, test for completed backward reference */
2379    
2380      empty_branch = FALSE;      empty_branch = FALSE;
2381        scode = cd->start_code + GET(code, 1);
2382        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383    
2384        /* Completed backwards reference */
2385    
2386      do      do
2387        {        {
2388        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2389            {
2390          empty_branch = TRUE;          empty_branch = TRUE;
2391        code += GET(code, 1);          break;
2392            }
2393          scode += GET(scode, 1);
2394        }        }
2395      while (*code == OP_ALT);      while (*scode == OP_ALT);
2396      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2397        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2398        continue;
2399        }
2400    
2401      /* Groups with zero repeats can of course be empty; skip them. */
2402    
2403      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2404          c == OP_BRAPOSZERO)
2405        {
2406        code += PRIV(OP_lengths)[c];
2407        do code += GET(code, 1); while (*code == OP_ALT);
2408        c = *code;
2409        continue;
2410        }
2411    
2412      /* A nested group that is already marked as "could be empty" can just be
2413      skipped. */
2414    
2415      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2416          c == OP_SCBRA || c == OP_SCBRAPOS)
2417        {
2418        do code += GET(code, 1); while (*code == OP_ALT);
2419        c = *code;
2420        continue;
2421        }
2422    
2423      /* For other groups, scan the branches. */
2424    
2425      if (c == OP_BRA  || c == OP_BRAPOS ||
2426          c == OP_CBRA || c == OP_CBRAPOS ||
2427          c == OP_ONCE || c == OP_ONCE_NC ||
2428          c == OP_COND)
2429        {
2430        BOOL empty_branch;
2431        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2432    
2433        /* If a conditional group has only one branch, there is a second, implied,
2434        empty branch, so just skip over the conditional, because it could be empty.
2435        Otherwise, scan the individual branches of the group. */
2436    
2437      /* Move past the KET and fudge things so that the increment in the "for"      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2438      above has no effect. */        code += GET(code, 1);
2439        else
2440          {
2441          empty_branch = FALSE;
2442          do
2443            {
2444            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2445              empty_branch = TRUE;
2446            code += GET(code, 1);
2447            }
2448          while (*code == OP_ALT);
2449          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2450          }
2451    
2452      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2453      continue;      continue;
2454      }      }
2455    
# Line 1399  for (code = first_significant_code(code Line 2457  for (code = first_significant_code(code
2457    
2458    switch (c)    switch (c)
2459      {      {
2460      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2461        cannot be represented just by a bit map. This includes negated single
2462        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2463        actual length is stored in the compiled code, so we must update "code"
2464        here. */
2465    
2466  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2467      case OP_XCLASS:      case OP_XCLASS:
2468      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2469      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2470  #endif  #endif
2471    
2472      case OP_CLASS:      case OP_CLASS:
2473      case OP_NCLASS:      case OP_NCLASS:
2474      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2475    
2476  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2477      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2478  #endif  #endif
2479    
# Line 1447  for (code = first_significant_code(code Line 2509  for (code = first_significant_code(code
2509      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2510      case OP_WORDCHAR:      case OP_WORDCHAR:
2511      case OP_ANY:      case OP_ANY:
2512        case OP_ALLANY:
2513      case OP_ANYBYTE:      case OP_ANYBYTE:
2514      case OP_CHAR:      case OP_CHAR:
2515      case OP_CHARNC:      case OP_CHARI:
2516      case OP_NOT:      case OP_NOT:
2517        case OP_NOTI:
2518      case OP_PLUS:      case OP_PLUS:
2519      case OP_MINPLUS:      case OP_MINPLUS:
2520      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1465  for (code = first_significant_code(code Line 2529  for (code = first_significant_code(code
2529      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2530      return FALSE;      return FALSE;
2531    
2532        /* These are going to continue, as they may be empty, but we have to
2533        fudge the length for the \p and \P cases. */
2534    
2535        case OP_TYPESTAR:
2536        case OP_TYPEMINSTAR:
2537        case OP_TYPEPOSSTAR:
2538        case OP_TYPEQUERY:
2539        case OP_TYPEMINQUERY:
2540        case OP_TYPEPOSQUERY:
2541        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2542        break;
2543    
2544        /* Same for these */
2545    
2546        case OP_TYPEUPTO:
2547        case OP_TYPEMINUPTO:
2548        case OP_TYPEPOSUPTO:
2549        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2550          code += 2;
2551        break;
2552    
2553      /* End of branch */      /* End of branch */
2554    
2555      case OP_KET:      case OP_KET:
2556      case OP_KETRMAX:      case OP_KETRMAX:
2557      case OP_KETRMIN:      case OP_KETRMIN:
2558        case OP_KETRPOS:
2559      case OP_ALT:      case OP_ALT:
2560      return TRUE;      return TRUE;
2561    
2562      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2563      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2564    
2565  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2566      case OP_STAR:      case OP_STAR:
2567        case OP_STARI:
2568      case OP_MINSTAR:      case OP_MINSTAR:
2569        case OP_MINSTARI:
2570      case OP_POSSTAR:      case OP_POSSTAR:
2571        case OP_POSSTARI:
2572      case OP_QUERY:      case OP_QUERY:
2573        case OP_QUERYI:
2574      case OP_MINQUERY:      case OP_MINQUERY:
2575        case OP_MINQUERYI:
2576      case OP_POSQUERY:      case OP_POSQUERY:
2577        case OP_POSQUERYI:
2578        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2579        break;
2580    
2581      case OP_UPTO:      case OP_UPTO:
2582        case OP_UPTOI:
2583      case OP_MINUPTO:      case OP_MINUPTO:
2584        case OP_MINUPTOI:
2585      case OP_POSUPTO:      case OP_POSUPTO:
2586      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2587        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2588      break;      break;
2589  #endif  #endif
2590    
2591        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2592        string. */
2593    
2594        case OP_MARK:
2595        case OP_PRUNE_ARG:
2596        case OP_SKIP_ARG:
2597        code += code[1];
2598        break;
2599    
2600        case OP_THEN_ARG:
2601        code += code[1];
2602        break;
2603    
2604        /* None of the remaining opcodes are required to match a character. */
2605    
2606        default:
2607        break;
2608      }      }
2609    }    }
2610    
# Line 1505  return TRUE; Line 2621  return TRUE;
2621  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
2622  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,
2623  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.
2624    This function is called only during the real compile, not during the
2625    pre-compile.
2626    
2627  Arguments:  Arguments:
2628    code        points to start of the recursion    code        points to start of the recursion
2629    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2630    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2631    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2632      cd          pointers to tables etc
2633    
2634  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2635  */  */
2636    
2637  static BOOL  static BOOL
2638  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2639    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2640  {  {
2641  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2642    {    {
2643    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2644        return FALSE;
2645    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2646    }    }
2647  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2654  return TRUE;
2654  *************************************************/  *************************************************/
2655    
2656  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2657  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
2658  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2659  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2660    
2661    Originally, this function only recognized a sequence of letters between the
2662    terminators, but it seems that Perl recognizes any sequence of characters,
2663    though of course unknown POSIX names are subsequently rejected. Perl gives an
2664    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2665    didn't consider this to be a POSIX class. Likewise for [:1234:].
2666    
2667    The problem in trying to be exactly like Perl is in the handling of escapes. We
2668    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2669    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2670    below handles the special case of \], but does not try to do any other escape
2671    processing. This makes it different from Perl for cases such as [:l\ower:]
2672    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2673    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2674    I think.
2675    
2676    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2677    It seems that the appearance of a nested POSIX class supersedes an apparent
2678    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2679    a digit.
2680    
2681    In Perl, unescaped square brackets may also appear as part of class names. For
2682    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2683    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2684    seem right at all. PCRE does not allow closing square brackets in POSIX class
2685    names.
2686    
2687  Argument:  Arguments:
2688    ptr      pointer to the initial [    ptr      pointer to the initial [
2689    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2690    
2691  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2692  */  */
2693    
2694  static BOOL  static BOOL
2695  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2696  {  {
2697  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2698  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2699  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2700    {    {
2701    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2702    return TRUE;      ptr++;
2703      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2704      else
2705        {
2706        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2707          {
2708          *endptr = ptr;
2709          return TRUE;
2710          }
2711        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2712             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2713              ptr[1] == CHAR_EQUALS_SIGN) &&
2714            check_posix_syntax(ptr, endptr))
2715          return FALSE;
2716        }
2717    }    }
2718  return FALSE;  return FALSE;
2719  }  }
# Line 1579  Returns:     a value representing the na Line 2736  Returns:     a value representing the na
2736  */  */
2737    
2738  static int  static int
2739  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2740  {  {
2741    const char *pn = posix_names;
2742  register int yield = 0;  register int yield = 0;
2743  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2744    {    {
2745    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2746      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2747      pn += posix_name_lengths[yield] + 1;
2748    yield++;    yield++;
2749    }    }
2750  return -1;  return -1;
# Line 1600  return -1; Line 2759  return -1;
2759  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2760  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2761  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
2762  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
2763  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
2764  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
2765  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
2766  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2767    OP_END.
2768    
2769  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2770  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 1615  value in the reference (which is a group Line 2775  value in the reference (which is a group
2775  Arguments:  Arguments:
2776    group      points to the start of the group    group      points to the start of the group
2777    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2778    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2779    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2780    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
2781    
# Line 1623  Returns:     nothing Line 2783  Returns:     nothing
2783  */  */
2784    
2785  static void  static void
2786  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2787    uschar *save_hwm)    pcre_uchar *save_hwm)
2788  {  {
2789  uschar *ptr = group;  pcre_uchar *ptr = group;
2790  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2791    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2792    {    {
2793    int offset;    int offset;
2794    uschar *hc;    pcre_uchar *hc;
2795    
2796    /* 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
2797    reference. */    reference. */
# Line 1675  Arguments: Line 2836  Arguments:
2836  Returns:         new code pointer  Returns:         new code pointer
2837  */  */
2838    
2839  static uschar *  static pcre_uchar *
2840  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2841  {  {
2842  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2843  *code++ = 255;  *code++ = 255;
2844  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2845  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2846  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2847  }  }
2848    
2849    
# Line 1704  Returns:             nothing Line 2865  Returns:             nothing
2865  */  */
2866    
2867  static void  static void
2868  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2869  {  {
2870  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2871  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2872  }  }
2873    
# Line 1718  PUT(previous_callout, 2 + LINK_SIZE, len Line 2879  PUT(previous_callout, 2 + LINK_SIZE, len
2879  *************************************************/  *************************************************/
2880    
2881  /* 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
2882  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
2883  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
2884  start address.  start address. A character with multiple other cases is returned on its own
2885    with a special return value.
2886    
2887  Arguments:  Arguments:
2888    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1728  Arguments: Line 2890  Arguments:
2890    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2891    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2892    
2893  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2894                   0 when a range is returned
2895                  >0 the CASESET offset for char with multiple other cases
2896                    in this case, ocptr contains the original
2897  */  */
2898    
2899  static BOOL  static int
2900  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2901    unsigned int *odptr)    unsigned int *odptr)
2902  {  {
2903  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2904    int co;
2905    
2906    /* Find the first character that has an other case. If it has multiple other
2907    cases, return its case offset value. */
2908    
2909  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2910    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2911      if ((co = UCD_CASESET(c)) != 0)
2912        {
2913        *ocptr = c++;   /* Character that has the set */
2914        *cptr = c;      /* Rest of input range */
2915        return co;
2916        }
2917      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2918      }
2919    
2920  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2921    
2922  *ocptr = othercase;  *ocptr = othercase;
2923  next = othercase + 1;  next = othercase + 1;
2924    
2925  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2926    {    {
2927    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2928    next++;    next++;
2929    }    }
2930    
2931  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2932  *cptr = c;  *cptr = c;             /* Rest of input range */
2933    return 0;
2934    }
2935    
2936    
2937  return TRUE;  
2938    /*************************************************
2939    *        Check a character and a property        *
2940    *************************************************/
2941    
2942    /* This function is called by check_auto_possessive() when a property item
2943    is adjacent to a fixed character.
2944    
2945    Arguments:
2946      c            the character
2947      ptype        the property type
2948      pdata        the data for the type
2949      negated      TRUE if it's a negated property (\P or \p{^)
2950    
2951    Returns:       TRUE if auto-possessifying is OK
2952    */
2953    
2954    static BOOL
2955    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2956    {
2957    #ifdef SUPPORT_UCP
2958    const pcre_uint32 *p;
2959    #endif
2960    
2961    const ucd_record *prop = GET_UCD(c);
2962    
2963    switch(ptype)
2964      {
2965      case PT_LAMP:
2966      return (prop->chartype == ucp_Lu ||
2967              prop->chartype == ucp_Ll ||
2968              prop->chartype == ucp_Lt) == negated;
2969    
2970      case PT_GC:
2971      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2972    
2973      case PT_PC:
2974      return (pdata == prop->chartype) == negated;
2975    
2976      case PT_SC:
2977      return (pdata == prop->script) == negated;
2978    
2979      /* These are specials */
2980    
2981      case PT_ALNUM:
2982      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2983              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2984    
2985      case PT_SPACE:    /* Perl space */
2986      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2987              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2988              == negated;
2989    
2990      case PT_PXSPACE:  /* POSIX space */
2991      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2992              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2993              c == CHAR_FF || c == CHAR_CR)
2994              == negated;
2995    
2996      case PT_WORD:
2997      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2998              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2999              c == CHAR_UNDERSCORE) == negated;
3000    
3001    #ifdef SUPPORT_UCP
3002      case PT_CLIST:
3003      p = PRIV(ucd_caseless_sets) + prop->caseset;
3004      for (;;)
3005        {
3006        if ((unsigned int)c < *p) return !negated;
3007        if ((unsigned int)c == *p++) return negated;
3008        }
3009      break;  /* Control never reaches here */
3010    #endif
3011      }
3012    
3013    return FALSE;
3014  }  }
3015  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3016    
# Line 1769  whether the next thing could possibly ma Line 3025  whether the next thing could possibly ma
3025  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3026    
3027  Arguments:  Arguments:
3028    op_code       the repeated op code    previous      pointer to the repeated opcode
3029    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  
3030    ptr           next character in pattern    ptr           next character in pattern
3031    options       options bits    options       options bits
3032    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns:        TRUE if possessifying is Line 3035  Returns:        TRUE if possessifying is
3035  */  */
3036    
3037  static BOOL  static BOOL
3038  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3039    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3040  {  {
3041  int next;  pcre_int32 c = NOTACHAR;
3042    pcre_int32 next;
3043    int op_code = *previous++;
3044    
3045  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3046    
# Line 1792  if ((options & PCRE_EXTENDED) != 0) Line 3048  if ((options & PCRE_EXTENDED) != 0)
3048    {    {
3049    for (;;)    for (;;)
3050      {      {
3051      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3052      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3053        {        {
3054        while (*(++ptr) != 0)        ptr++;
3055          while (*ptr != 0)
3056            {
3057          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3058            ptr++;
3059    #ifdef SUPPORT_UTF
3060            if (utf) FORWARDCHAR(ptr);
3061    #endif
3062            }
3063        }        }
3064      else break;      else break;
3065      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 3068  if ((options & PCRE_EXTENDED) != 0)
3068  /* 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
3069  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3070    
3071  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3072    {    {
3073    int temperrorcode = 0;    int temperrorcode = 0;
3074    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
3075    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3076    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3077    }    }
3078    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3079    {    {
3080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3081    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
3082  #endif  #endif
3083    next = *ptr++;    next = *ptr++;
3084    }    }
   
3085  else return FALSE;  else return FALSE;
3086    
3087  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 3090  if ((options & PCRE_EXTENDED) != 0)
3090    {    {
3091    for (;;)    for (;;)
3092      {      {
3093      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3094      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3095        {        {
3096        while (*(++ptr) != 0)        ptr++;
3097          while (*ptr != 0)
3098            {
3099          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3100            ptr++;
3101    #ifdef SUPPORT_UTF
3102            if (utf) FORWARDCHAR(ptr);
3103    #endif
3104            }
3105        }        }
3106      else break;      else break;
3107      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 3109  if ((options & PCRE_EXTENDED) != 0)
3109    
3110  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3111    
3112  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3113    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3114        return FALSE;
3115    
3116  /* 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. */  
3117    
3118    if (op_code == OP_CHAR || op_code == OP_CHARI ||
3119        op_code == OP_NOT || op_code == OP_NOTI)
3120      {
3121    #ifdef SUPPORT_UTF
3122      GETCHARTEST(c, previous);
3123    #else
3124      c = *previous;
3125    #endif
3126      }
3127    
3128  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3129    the next item is a character. */
3130    
3131  if (next >= 0) switch(op_code)  if (next >= 0)
3132    {    {
3133    case OP_CHAR:    /* For a caseless UTF match, the next character may have more than one other
3134  #ifdef SUPPORT_UTF8    case, which maps to the special PT_CLIST property. Check this first. */
3135    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
3136    #ifdef SUPPORT_UCP
3137      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3138        {
3139        int ocs = UCD_CASESET(next);
3140        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3141        }
3142  #endif  #endif
   return item != next;  
   
   /* For CHARNC (caseless character) we must check the other case. If we have  
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
3143    
3144    case OP_CHARNC:    switch(op_code)
 #ifdef SUPPORT_UTF8  
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   if (item == next) return FALSE;  
 #ifdef SUPPORT_UTF8  
   if (utf8)  
3145      {      {
3146      unsigned int othercase;      case OP_CHAR:
3147      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3148    
3149        /* For CHARI (caseless character) we must check the other case. If we have
3150        Unicode property support, we can use it to test the other case of
3151        high-valued characters. We know that next can have only one other case,
3152        because multi-other-case characters are dealt with above. */
3153    
3154        case OP_CHARI:
3155        if (c == next) return FALSE;
3156    #ifdef SUPPORT_UTF
3157        if (utf)
3158          {
3159          unsigned int othercase;
3160          if (next < 128) othercase = cd->fcc[next]; else
3161  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3162      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3163  #else  #else
3164      othercase = NOTACHAR;        othercase = NOTACHAR;
3165  #endif  #endif
3166      return (unsigned int)item != othercase;        return (unsigned int)c != othercase;
3167      }        }
3168    else      else
3169  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3170    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3171    
3172    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3173        return c == next;
3174    case OP_NOT:  
3175    if (next < 0) return FALSE;  /* Not a character */      case OP_NOTI:
3176    if (item == next) return TRUE;      if (c == next) return TRUE;
3177    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3178  #ifdef SUPPORT_UTF8      if (utf)
3179    if (utf8)        {
3180      {        unsigned int othercase;
3181      unsigned int othercase;        if (next < 128) othercase = cd->fcc[next]; else
     if (next < 128) othercase = cd->fcc[next]; else  
3182  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3183      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3184  #else  #else
3185      othercase = NOTACHAR;        othercase = NOTACHAR;
3186  #endif  #endif
3187      return (unsigned int)item == othercase;        return (unsigned int)c == othercase;
3188      }        }
3189    else      else
3190  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3191    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3192    
3193    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3194    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3195    
3196    case OP_NOT_DIGIT:      case OP_DIGIT:
3197    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3198    
3199    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3200    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3201    
3202    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3203    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3204    
3205    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3206    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3207    
3208    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3209    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3210    
3211    default:      case OP_NOT_WORDCHAR:
3212    return FALSE;      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3213    }  
3214        case OP_HSPACE:
3215        case OP_NOT_HSPACE:
3216        switch(next)
3217          {
3218          HSPACE_CASES:
3219          return op_code == OP_NOT_HSPACE;
3220    
3221          default:
3222          return op_code != OP_NOT_HSPACE;
3223          }
3224    
3225        case OP_ANYNL:
3226        case OP_VSPACE:
3227        case OP_NOT_VSPACE:
3228        switch(next)
3229          {
3230          VSPACE_CASES:
3231          return op_code == OP_NOT_VSPACE;
3232    
3233          default:
3234          return op_code != OP_NOT_VSPACE;
3235          }
3236    
3237    #ifdef SUPPORT_UCP
3238        case OP_PROP:
3239        return check_char_prop(next, previous[0], previous[1], FALSE);
3240    
3241        case OP_NOTPROP:
3242        return check_char_prop(next, previous[0], previous[1], TRUE);
3243    #endif
3244    
3245        default:
3246        return FALSE;
3247        }
3248      }
3249    
3250  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3251    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3252    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3253    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3254    replaced by OP_PROP codes when PCRE_UCP is set. */
3255    
3256  switch(op_code)  switch(op_code)
3257    {    {
3258    case OP_CHAR:    case OP_CHAR:
3259    case OP_CHARNC:    case OP_CHARI:
 #ifdef SUPPORT_UTF8  
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
3260    switch(-next)    switch(-next)
3261      {      {
3262      case ESC_d:      case ESC_d:
3263      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3264    
3265      case ESC_D:      case ESC_D:
3266      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3267    
3268      case ESC_s:      case ESC_s:
3269      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3270    
3271      case ESC_S:      case ESC_S:
3272      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3273    
3274      case ESC_w:      case ESC_w:
3275      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3276    
3277      case ESC_W:      case ESC_W:
3278      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3279    
3280        case ESC_h:
3281        case ESC_H:
3282        switch(c)
3283          {
3284          HSPACE_CASES:
3285          return -next != ESC_h;
3286    
3287          default:
3288          return -next == ESC_h;
3289          }
3290    
3291        case ESC_v:
3292        case ESC_V:
3293        switch(c)
3294          {
3295          VSPACE_CASES:
3296          return -next != ESC_v;
3297    
3298          default:
3299          return -next == ESC_v;
3300          }
3301    
3302        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3303        their substitutions and process them. The result will always be either
3304        -ESC_p or -ESC_P. Then fall through to process those values. */
3305    
3306    #ifdef SUPPORT_UCP
3307        case ESC_du:
3308        case ESC_DU:
3309        case ESC_wu:
3310        case ESC_WU:
3311        case ESC_su:
3312        case ESC_SU:
3313          {
3314          int temperrorcode = 0;
3315          ptr = substitutes[-next - ESC_DU];
3316          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3317          if (temperrorcode != 0) return FALSE;
3318          ptr++;    /* For compatibility */
3319          }
3320        /* Fall through */
3321    
3322        case ESC_p:
3323        case ESC_P:
3324          {
3325          int ptype, pdata, errorcodeptr;
3326          BOOL negated;
3327    
3328          ptr--;      /* Make ptr point at the p or P */
3329          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3330          if (ptype < 0) return FALSE;
3331          ptr++;      /* Point past the final curly ket */
3332    
3333          /* If the property item is optional, we have to give up. (When generated
3334          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3335          to the original \d etc. At this point, ptr will point to a zero byte. */
3336    
3337          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3338            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3339              return FALSE;
3340    
3341          /* Do the property check. */
3342    
3343          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3344          }
3345    #endif
3346    
3347      default:      default:
3348      return FALSE;      return FALSE;
3349      }      }
3350    
3351      /* In principle, support for Unicode properties should be integrated here as
3352      well. It means re-organizing the above code so as to get hold of the property
3353      values before switching on the op-code. However, I wonder how many patterns
3354      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3355      these op-codes are never generated.) */
3356    
3357    case OP_DIGIT:    case OP_DIGIT:
3358    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3359             next == -ESC_h || next == -ESC_v || next == -ESC_R;
3360    
3361    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3362    return next == -ESC_d;    return next == -ESC_d;
# Line 1973  switch(op_code) Line 3365  switch(op_code)
3365    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
3366    
3367    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3368    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;
3369    
3370      case OP_HSPACE:
3371      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3372             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3373    
3374      case OP_NOT_HSPACE:
3375      return next == -ESC_h;
3376    
3377      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3378      case OP_ANYNL:
3379      case OP_VSPACE:
3380      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3381    
3382      case OP_NOT_VSPACE:
3383      return next == -ESC_v || next == -ESC_R;
3384    
3385    case OP_WORDCHAR:    case OP_WORDCHAR:
3386    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3387             next == -ESC_v || next == -ESC_R;
3388    
3389    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3390    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 1991  switch(op_code) Line 3399  switch(op_code)
3399    
3400    
3401  /*************************************************  /*************************************************
3402    *        Add a character or range to a class     *
3403    *************************************************/
3404    
3405    /* This function packages up the logic of adding a character or range of
3406    characters to a class. The character values in the arguments will be within the
3407    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3408    mutually recursive with the function immediately below.
3409    
3410    Arguments:
3411      classbits     the bit map for characters < 256
3412      uchardptr     points to the pointer for extra data
3413      options       the options word
3414      cd            contains pointers to tables etc.
3415      start         start of range character
3416      end           end of range character
3417    
3418    Returns:        the number of < 256 characters added
3419                    the pointer to extra data is updated
3420    */
3421    
3422    static int
3423    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3424      compile_data *cd, unsigned int start, unsigned int end)
3425    {
3426    unsigned int c;
3427    int n8 = 0;
3428    
3429    /* If caseless matching is required, scan the range and process alternate
3430    cases. In Unicode, there are 8-bit characters that have alternate cases that
3431    are greater than 255 and vice-versa. Sometimes we can just extend the original
3432    range. */
3433    
3434    if ((options & PCRE_CASELESS) != 0)
3435      {
3436    #ifdef SUPPORT_UCP
3437      if ((options & PCRE_UTF8) != 0)
3438        {
3439        int rc;
3440        unsigned int oc, od;
3441    
3442        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3443        c = start;
3444    
3445        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3446          {
3447          /* Handle a single character that has more than one other case. */
3448    
3449          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3450            PRIV(ucd_caseless_sets) + rc, oc);
3451    
3452          /* Do nothing if the other case range is within the original range. */
3453    
3454          else if (oc >= start && od <= end) continue;
3455    
3456          /* Extend the original range if there is overlap, noting that if oc < c, we
3457          can't have od > end because a subrange is always shorter than the basic
3458          range. Otherwise, use a recursive call to add the additional range. */
3459    
3460          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3461          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3462          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3463          }
3464        }
3465      else
3466    #endif  /* SUPPORT_UCP */
3467    
3468      /* Not UTF-mode, or no UCP */
3469    
3470      for (c = start; c <= end && c < 256; c++)
3471        {
3472        SETBIT(classbits, cd->fcc[c]);
3473        n8++;
3474        }
3475      }
3476    
3477    /* Now handle the original range. Adjust the final value according to the bit
3478    length - this means that the same lists of (e.g.) horizontal spaces can be used
3479    in all cases. */
3480    
3481    #if defined COMPILE_PCRE8
3482    #ifdef SUPPORT_UTF
3483      if ((options & PCRE_UTF8) == 0)
3484    #endif
3485      if (end > 0xff) end = 0xff;
3486    
3487    #elif defined COMPILE_PCRE16
3488    #ifdef SUPPORT_UTF
3489      if ((options & PCRE_UTF16) == 0)
3490    #endif
3491      if (end > 0xffff) end = 0xffff;
3492    
3493    #elif defined COMPILE_PCRE32
3494    #ifdef SUPPORT_UTF
3495      if ((options & PCRE_UTF32) == 0)
3496        if (end > 0xffffu) end = 0xffffu; // FIXMEchpe rebase fix this
3497    #endif
3498    #endif /* COMPILE_PCRE[8|16|32] */
3499    
3500    /* If all characters are less than 256, use the bit map. Otherwise use extra
3501    data. */
3502    
3503    if (end < 0x100)
3504      {
3505      for (c = start; c <= end; c++)
3506        {
3507        n8++;
3508        SETBIT(classbits, c);
3509        }
3510      }
3511    
3512    else
3513      {
3514      pcre_uchar *uchardata = *uchardptr;
3515    
3516    #ifdef SUPPORT_UTF
3517      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3518        {
3519        if (start < end)
3520          {
3521          *uchardata++ = XCL_RANGE;
3522          uchardata += PRIV(ord2utf)(start, uchardata);
3523          uchardata += PRIV(ord2utf)(end, uchardata);
3524          }
3525        else if (start == end)
3526          {
3527          *uchardata++ = XCL_SINGLE;
3528          uchardata += PRIV(ord2utf)(start, uchardata);
3529          }
3530        }
3531      else
3532    #endif  /* SUPPORT_UTF */
3533    
3534      /* Without UTF support, character values are constrained by the bit length,
3535      and can only be > 256 for 16-bit and 32-bit libraries. */
3536    
3537    #ifdef COMPILE_PCRE8
3538        {}
3539    #else
3540      if (start < end)
3541        {
3542        *uchardata++ = XCL_RANGE;
3543        *uchardata++ = start;
3544        *uchardata++ = end;
3545        }
3546      else if (start == end)
3547        {
3548        *uchardata++ = XCL_SINGLE;
3549        *uchardata++ = start;
3550        }
3551    #endif
3552    
3553      *uchardptr = uchardata;   /* Updata extra data pointer */
3554      }
3555    
3556    return n8;    /* Number of 8-bit characters */
3557    }
3558    
3559    
3560    
3561    
3562    /*************************************************
3563    *        Add a list of characters to a class     *
3564    *************************************************/
3565    
3566    /* This function is used for adding a list of case-equivalent characters to a
3567    class, and also for adding a list of horizontal or vertical whitespace. If the
3568    list is in order (which it should be), ranges of characters are detected and
3569    handled appropriately. This function is mutually recursive with the function
3570    above.
3571    
3572    Arguments:
3573      classbits     the bit map for characters < 256
3574      uchardptr     points to the pointer for extra data
3575      options       the options word
3576      cd            contains pointers to tables etc.
3577      p             points to row of 32-bit values, terminated by NOTACHAR
3578      except        character to omit; this is used when adding lists of
3579                      case-equivalent characters to avoid including the one we
3580                      already know about
3581    
3582    Returns:        the number of < 256 characters added
3583                    the pointer to extra data is updated
3584    */
3585    
3586    static int
3587    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3588      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3589    {
3590    int n8 = 0;
3591    while (p[0] < NOTACHAR)
3592      {
3593      int n = 0;
3594      if (p[0] != except)
3595        {
3596        while(p[n+1] == p[0] + n + 1) n++;
3597        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3598        }
3599      p += n + 1;
3600      }
3601    return n8;
3602    }
3603    
3604    
3605    
3606    /*************************************************
3607    *    Add characters not in a list to a class     *
3608    *************************************************/
3609    
3610    /* This function is used for adding the complement of a list of horizontal or
3611    vertical whitespace to a class. The list must be in order.
3612    
3613    Arguments:
3614      classbits     the bit map for characters < 256
3615      uchardptr     points to the pointer for extra data
3616      options       the options word
3617      cd            contains pointers to tables etc.
3618      p             points to row of 32-bit values, terminated by NOTACHAR
3619    
3620    Returns:        the number of < 256 characters added
3621                    the pointer to extra data is updated
3622    */
3623    
3624    static int
3625    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3626      int options, compile_data *cd, const pcre_uint32 *p)
3627    {
3628    int n8 = 0;
3629    if (p[0] > 0)
3630      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3631    while (p[0] < NOTACHAR)
3632      {
3633      while (p[1] == p[0] + 1) p++;
3634      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3635        (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);
3636      p++;
3637      }
3638    return n8;
3639    }
3640    
3641    
3642    
3643    /*************************************************
3644  *           Compile one branch                   *  *           Compile one branch                   *
3645  *************************************************/  *************************************************/
3646    
# Line 2005  Arguments: Line 3655  Arguments:
3655    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3656    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3657    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3658    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3659    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3660    bcptr          points to current branch chain    bcptr          points to current branch chain
3661      cond_depth     conditional nesting depth
3662    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3663    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3664                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2017  Returns:         TRUE on success Line 3668  Returns:         TRUE on success
3668  */  */
3669    
3670  static BOOL  static BOOL
3671  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3672    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3673      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3674    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3675  {  {
3676  int repeat_type, op_type;  int repeat_type, op_type;
3677  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3678  int bravalue = 0;  int bravalue = 0;
3679  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3680  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3681  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3682  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3683  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3684  int after_manual_callout = 0;  int after_manual_callout = 0;
3685  int length_prevgroup = 0;  int length_prevgroup = 0;
3686  register int c;  register int c;
3687  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3688  uschar *last_code = code;  pcre_uchar *last_code = code;
3689  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3690  uschar *tempcode;  pcre_uchar *tempcode;
3691  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3692  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3693  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3694  const uschar *tempptr;  const pcre_uchar *tempptr;
3695  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3696  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3697  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3698  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3699    pcre_uint8 classbits[32];
3700  #ifdef SUPPORT_UTF8  
3701  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3702  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3703  uschar *class_utf8data;  dynamically as we process the pattern. */
3704  uschar utf8_char[6];  
3705    #ifdef SUPPORT_UTF
3706    /* PCRE_UTF(16|32) have the same value as PCRE_UTF8. */
3707    BOOL utf = (options & PCRE_UTF8) != 0;
3708    pcre_uchar utf_chars[6];
3709  #else  #else
3710  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3711  uschar *utf8_char = NULL;  #endif
3712    
3713    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3714    class_uchardata always so that it can be passed to add_to_class() always,
3715    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3716    alternative calls for the different cases. */
3717    
3718    pcre_uchar *class_uchardata;
3719    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3720    BOOL xclass;
3721    pcre_uchar *class_uchardata_base;
3722  #endif  #endif
3723    
3724  #ifdef DEBUG  #ifdef PCRE_DEBUG
3725  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3726  #endif  #endif
3727    
# Line 2066  greedy_non_default = greedy_default ^ 1; Line 3732  greedy_non_default = greedy_default ^ 1;
3732    
3733  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3734  matching encountered yet". It gets changed to REQ_NONE if we hit something that  matching encountered yet". It gets changed to REQ_NONE if we hit something that
3735  matches a non-fixed char first char; reqbyte just remains unset if we never  matches a non-fixed char first char; reqchar just remains unset if we never
3736  find one.  find one.
3737    
3738  When we hit a repeat whose minimum is zero, we may have to adjust these values  When we hit a repeat whose minimum is zero, we may have to adjust these values
3739  to take the zero repeat into account. This is implemented by setting them to  to take the zero repeat into account. This is implemented by setting them to
3740  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3741  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3742    
3743  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3744    
3745  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3746  according to the current setting of the caseless flag. REQ_CASELESS is a bit  or zero, according to the current setting of the caseless flag. The
3747  value > 255. It is added into the firstbyte or reqbyte variables to record the  REQ_CASELESS leaves the lower 28 bit empty. It is added into the
3748  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3749    value. This is used only for ASCII characters. */
3750    
3751  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3752    
3753  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3754    
3755  for (;; ptr++)  for (;; ptr++)
3756    {    {
3757    BOOL negate_class;    BOOL negate_class;
3758      BOOL should_flip_negation;
3759    BOOL possessive_quantifier;    BOOL possessive_quantifier;