/[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 172 by ph10, Tue Jun 5 10:40:13 2007 UTC revision 836 by ph10, Wed Dec 28 17:16:11 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59    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%8))
72    
73    /* Maximum length value to check against when making sure that the integer that
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    
81  /*************************************************  /*************************************************
82  *      Code parameters and static tables         *  *      Code parameters and static tables         *
83  *************************************************/  *************************************************/
# Line 72  so this number is very generous. Line 92  so this number is very generous.
92  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
93  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
94  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
95  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
96    filled up by repetitions of forward references, for example patterns like
97    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
98    that the workspace is expanded using malloc() in this situation. The value
99    below is therefore a minimum, and we put a maximum on it for safety. The
100    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
101    kicks in at the same number of forward references in all cases. */
102    
103    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
104    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
105    
106    /* The overrun tests check for a slightly smaller size so that they detect the
107    overrun before it actually does run off the end of the data block. */
108    
109    #define WORK_SIZE_SAFETY_MARGIN (100)
110    
111  #define COMPILE_WORK_SIZE (4096)  /* Private flags added to firstchar and reqchar. */
112    
113    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
114    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
115    
116    /* Repeated character flags. */
117    
118    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
119    
120  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
121  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
122  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
123  is invalid. */  is invalid. */
124    
125  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
126    
127    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
128    in UTF-8 mode. */
129    
130  static const short int escapes[] = {  static const short int escapes[] = {
131       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
132       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
133     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
134       0,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
135  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
136  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
137     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
138       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
139  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
140       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
141         -ESC_D,                  -ESC_E,
142         0,                       -ESC_G,
143         -ESC_H,                  0,
144         0,                       -ESC_K,
145         0,                       0,
146         -ESC_N,                  0,
147         -ESC_P,                  -ESC_Q,
148         -ESC_R,                  -ESC_S,
149         0,                       0,
150         -ESC_V,                  -ESC_W,
151         -ESC_X,                  0,
152         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
153         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
154         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
155         CHAR_GRAVE_ACCENT,       7,
156         -ESC_b,                  0,
157         -ESC_d,                  ESC_e,
158         ESC_f,                   0,
159         -ESC_h,                  0,
160         0,                       -ESC_k,
161         0,                       0,
162         ESC_n,                   0,
163         -ESC_p,                  0,
164         ESC_r,                   -ESC_s,
165         ESC_tee,                 0,
166         -ESC_v,                  -ESC_w,
167         0,                       0,
168         -ESC_z
169  };  };
170    
171  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
172    
173    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
174    
175  static const short int escapes[] = {  static const short int escapes[] = {
176  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
177  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
182  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
183  /*  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,
184  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
185  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
186  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
187  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
188  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
189  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
190  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
191  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
192  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
193  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
194  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
195  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
196  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
197  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
198  /*  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 200  static const short int escapes[] = {
200  #endif  #endif
201    
202    
203  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
204  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
205  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
206    string is built from string macros so that it works in UTF-8 mode on EBCDIC
207  static const char *const posix_names[] = {  platforms. */
208    "alpha", "lower", "upper",  
209    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
210    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
211      int   op;                  /* Op when no arg, or -1 if arg mandatory */
212      int   op_arg;              /* Op when arg present, or -1 if not allowed */
213    } verbitem;
214    
215    static const char verbnames[] =
216      "\0"                       /* Empty name is a shorthand for MARK */
217      STRING_MARK0
218      STRING_ACCEPT0
219      STRING_COMMIT0
220      STRING_F0
221      STRING_FAIL0
222      STRING_PRUNE0
223      STRING_SKIP0
224      STRING_THEN;
225    
226    static const verbitem verbs[] = {
227      { 0, -1,        OP_MARK },
228      { 4, -1,        OP_MARK },
229      { 6, OP_ACCEPT, -1 },
230      { 6, OP_COMMIT, -1 },
231      { 1, OP_FAIL,   -1 },
232      { 4, OP_FAIL,   -1 },
233      { 5, OP_PRUNE,  OP_PRUNE_ARG },
234      { 4, OP_SKIP,   OP_SKIP_ARG  },
235      { 4, OP_THEN,   OP_THEN_ARG  }
236    };
237    
238    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
239    
240    
241  static const uschar posix_name_lengths[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
242    now all in a single string, to reduce the number of relocations when a shared
243    library is dynamically loaded. The list of lengths is terminated by a zero
244    length entry. The first three must be alpha, lower, upper, as this is assumed
245    for handling case independence. */
246    
247    static const char posix_names[] =
248      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
249      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
250      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
251      STRING_word0  STRING_xdigit;
252    
253    static const pcre_uint8 posix_name_lengths[] = {
254    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
255    
256  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 164  static const int posix_class_maps[] = { Line 280  static const int posix_class_maps[] = {
280    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
281  };  };
282    
283    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
284    substitutes must be in the order of the names, defined above, and there are
285    both positive and negative cases. NULL means no substitute. */
286    
287    #ifdef SUPPORT_UCP
288    static const pcre_uchar string_PNd[]  = {
289      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
290      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    static const pcre_uchar string_pNd[]  = {
292      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
293      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    static const pcre_uchar string_PXsp[] = {
295      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
296      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
297    static const pcre_uchar string_pXsp[] = {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_PXwd[] = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pXwd[] = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    
307    static const pcre_uchar *substitutes[] = {
308      string_PNd,           /* \D */
309      string_pNd,           /* \d */
310      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
311      string_pXsp,          /* \s */
312      string_PXwd,          /* \W */
313      string_pXwd           /* \w */
314    };
315    
316    static const pcre_uchar string_pL[] =   {
317      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pLl[] =  {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_pLu[] =  {
323      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_pXan[] = {
326      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    static const pcre_uchar string_h[] =    {
329      CHAR_BACKSLASH, CHAR_h, '\0' };
330    static const pcre_uchar string_pXps[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_PL[] =   {
334      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_PLl[] =  {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_PLu[] =  {
340      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
341      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
342    static const pcre_uchar string_PXan[] = {
343      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_H[] =    {
346      CHAR_BACKSLASH, CHAR_H, '\0' };
347    static const pcre_uchar string_PXps[] = {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    
351    static const pcre_uchar *posix_substitutes[] = {
352      string_pL,            /* alpha */
353      string_pLl,           /* lower */
354      string_pLu,           /* upper */
355      string_pXan,          /* alnum */
356      NULL,                 /* ascii */
357      string_h,             /* blank */
358      NULL,                 /* cntrl */
359      string_pNd,           /* digit */
360      NULL,                 /* graph */
361      NULL,                 /* print */
362      NULL,                 /* punct */
363      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
364      string_pXwd,          /* word */
365      NULL,                 /* xdigit */
366      /* Negated cases */
367      string_PL,            /* ^alpha */
368      string_PLl,           /* ^lower */
369      string_PLu,           /* ^upper */
370      string_PXan,          /* ^alnum */
371      NULL,                 /* ^ascii */
372      string_H,             /* ^blank */
373      NULL,                 /* ^cntrl */
374      string_PNd,           /* ^digit */
375      NULL,                 /* ^graph */
376      NULL,                 /* ^print */
377      NULL,                 /* ^punct */
378      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
379      string_PXwd,          /* ^word */
380      NULL                  /* ^xdigit */
381    };
382    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
383    #endif
384    
385  #define STRING(a)  # a  #define STRING(a)  # a
386  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 388  static const int posix_class_maps[] = {
388  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
389  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
390  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
391  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
392    the number of relocations needed when a shared library is loaded dynamically,
393  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
394    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
395    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
396    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
397    "unrecognized character follows \\",  
398    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
399    substring, so that the whole string ends with \0\0, which can be detected when
400    counting through. */
401    
402    static const char error_texts[] =
403      "no error\0"
404      "\\ at end of pattern\0"
405      "\\c at end of pattern\0"
406      "unrecognized character follows \\\0"
407      "numbers out of order in {} quantifier\0"
408    /* 5 */    /* 5 */
409    "number too big in {} quantifier",    "number too big in {} quantifier\0"
410    "missing terminating ] for character class",    "missing terminating ] for character class\0"
411    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
412    "range out of order in character class",    "range out of order in character class\0"
413    "nothing to repeat",    "nothing to repeat\0"
414    /* 10 */    /* 10 */
415    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
416    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
417    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
418    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
419    "missing )",    "missing )\0"
420    /* 15 */    /* 15 */
421    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
422    "erroffset passed as NULL",    "erroffset passed as NULL\0"
423    "unknown option bit(s) set",    "unknown option bit(s) set\0"
424    "missing ) after comment",    "missing ) after comment\0"
425    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
426    /* 20 */    /* 20 */
427    "regular expression too large",    "regular expression is too large\0"
428    "failed to get memory",    "failed to get memory\0"
429    "unmatched parentheses",    "unmatched parentheses\0"
430    "internal error: code overflow",    "internal error: code overflow\0"
431    "unrecognized character after (?<",    "unrecognized character after (?<\0"
432    /* 25 */    /* 25 */
433    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
434    "malformed number or name after (?(",    "malformed number or name after (?(\0"
435    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
436    "assertion expected after (?(",    "assertion expected after (?(\0"
437    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
438    /* 30 */    /* 30 */
439    "unknown POSIX class name",    "unknown POSIX class name\0"
440    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
441    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
442    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
443    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
444    /* 35 */    /* 35 */
445    "invalid condition (?(0)",    "invalid condition (?(0)\0"
446    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
447    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
448    "number after (?C is > 255",    "number after (?C is > 255\0"
449    "closing ) for (?C expected",    "closing ) for (?C expected\0"
450    /* 40 */    /* 40 */
451    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
452    "unrecognized character after (?P",    "unrecognized character after (?P\0"
453    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
454    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
455    "invalid UTF-8 string",    "invalid UTF-8 string\0"
456    /* 45 */    /* 45 */
457    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
458    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
459    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
460    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
461    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
462    /* 50 */    /* 50 */
463    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
464    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
465    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
466    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
467    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
468    /* 55 */    /* 55 */
469    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
470    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
471    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
472    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
473  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
474      /* 60 */
475      "(*VERB) not recognized\0"
476      "number is too big\0"
477      "subpattern name expected\0"
478      "digit expected after (?+\0"
479      "] is an invalid data character in JavaScript compatibility mode\0"
480      /* 65 */
481      "different names for subpatterns of the same number are not allowed\0"
482      "(*MARK) must have an argument\0"
483      "this version of PCRE is not compiled with PCRE_UCP support\0"
484      "\\c must be followed by an ASCII character\0"
485      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
486      /* 70 */
487      "internal error: unknown opcode in find_fixedlength()\0"
488      "\\N is not supported in a class\0"
489      "too many forward references\0"
490      "disallowed UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
491      ;
492    
493  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
494  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 263  For convenience, we use the same bit def Line 506  For convenience, we use the same bit def
506    
507  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
508    
509  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
510  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
511    into a subtraction and unsigned comparison). */
512    
513    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
514    
515    #ifndef EBCDIC
516    
517    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
518    UTF-8 mode. */
519    
520    static const pcre_uint8 digitab[] =
521    {    {
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 299  static const unsigned char digitab[] = Line 552  static const unsigned char digitab[] =
552    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
553    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
554    
555  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
556  static const unsigned char digitab[] =  
557    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
558    
559    static const pcre_uint8 digitab[] =
560    {    {
561    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
562    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 335  static const unsigned char digitab[] = Line 591  static const unsigned char digitab[] =
591    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
592    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
593    
594  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
595    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
596    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
597    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 374  static const unsigned char ebcdic_charta Line 630  static const unsigned char ebcdic_charta
630  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
631    
632  static BOOL  static BOOL
633    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
634      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
635    
636    
637    
638    /*************************************************
639    *            Find an error text                  *
640    *************************************************/
641    
642    /* The error texts are now all in one long string, to save on relocations. As
643    some of the text is of unknown length, we can't use a table of offsets.
644    Instead, just count through the strings. This is not a performance issue
645    because it happens only when there has been a compilation error.
646    
647    Argument:   the error number
648    Returns:    pointer to the error string
649    */
650    
651    static const char *
652    find_error_text(int n)
653    {
654    const char *s = error_texts;
655    for (; n > 0; n--)
656      {
657      while (*s++ != 0) {};
658      if (*s == 0) return "Error text not found (please report)";
659      }
660    return s;
661    }
662    
663    
664    /*************************************************
665    *           Expand the workspace                 *
666    *************************************************/
667    
668    /* This function is called during the second compiling phase, if the number of
669    forward references fills the existing workspace, which is originally a block on
670    the stack. A larger block is obtained from malloc() unless the ultimate limit
671    has been reached or the increase will be rather small.
672    
673    Argument: pointer to the compile data block
674    Returns:  0 if all went well, else an error number
675    */
676    
677    static int
678    expand_workspace(compile_data *cd)
679    {
680    pcre_uchar *newspace;
681    int newsize = cd->workspace_size * 2;
682    
683    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
684    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
685        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
686     return ERR72;
687    
688    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
689    if (newspace == NULL) return ERR21;
690    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
691    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
692    if (cd->workspace_size > COMPILE_WORK_SIZE)
693      (PUBL(free))((void *)cd->start_workspace);
694    cd->start_workspace = newspace;
695    cd->workspace_size = newsize;
696    return 0;
697    }
698    
699    
700    
701    /*************************************************
702    *            Check for counted repeat            *
703    *************************************************/
704    
705    /* This function is called when a '{' is encountered in a place where it might
706    start a quantifier. It looks ahead to see if it really is a quantifier or not.
707    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
708    where the ddds are digits.
709    
710    Arguments:
711      p         pointer to the first char after '{'
712    
713    Returns:    TRUE or FALSE
714    */
715    
716    static BOOL
717    is_counted_repeat(const pcre_uchar *p)
718    {
719    if (!IS_DIGIT(*p)) return FALSE;
720    p++;
721    while (IS_DIGIT(*p)) p++;
722    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
723    
724    if (*p++ != CHAR_COMMA) return FALSE;
725    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
726    
727    if (!IS_DIGIT(*p)) return FALSE;
728    p++;
729    while (IS_DIGIT(*p)) p++;
730    
731    return (*p == CHAR_RIGHT_CURLY_BRACKET);
732    }
733    
734    
735    
# Line 400  Arguments: Line 754  Arguments:
754    
755  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
756                   negative => a special escape sequence                   negative => a special escape sequence
757                   on error, errorptr is set                   on error, errorcodeptr is set
758  */  */
759    
760  static int  static int
761  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
762    int options, BOOL isclass)    int options, BOOL isclass)
763  {  {
764  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
765  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
766  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
767    pcre_int32 c;
768    int i;
769    
770  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
771  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 418  ptr--;                            /* Set Line 774  ptr--;                            /* Set
774    
775  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
776    
777  /* 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
778  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.
779  Otherwise further processing may be required. */  Otherwise further processing may be required. */
780    
781  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
782  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
783  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
784    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
785    
786  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
787  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
788    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
789  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
790  #endif  #endif
791    
# Line 435  else if ((i = escapes[c - 0x48]) != 0) Line 793  else if ((i = escapes[c - 0x48]) != 0)
793    
794  else  else
795    {    {
796    const uschar *oldptr;    const pcre_uchar *oldptr;
797    BOOL braced, negated;    BOOL braced, negated;
798    
799    switch (c)    switch (c)
# Line 443  else Line 801  else
801      /* 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
802      error. */      error. */
803    
804      case 'l':      case CHAR_l:
805      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
806      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
807      break;      break;
808    
809      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
810      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
811      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
812      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
813      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
814      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
815            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
816      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
817      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
818        {          {
819        const uschar *p;          c = 0;
820        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
821          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
822        if (*p != 0 && *p != '}')            register int cc = *(++ptr);
823    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
824              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
825              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
826    #else           /* EBCDIC coding */
827              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
828              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
829    #endif
830              }
831            }
832          }
833        else
834          *errorcodeptr = ERR37;
835        break;
836    
837        case CHAR_U:
838        /* In JavaScript, \U is an uppercase U letter. */
839        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
840        break;
841    
842        /* In a character class, \g is just a literal "g". Outside a character
843        class, \g must be followed by one of a number of specific things:
844    
845        (1) A number, either plain or braced. If positive, it is an absolute
846        backreference. If negative, it is a relative backreference. This is a Perl
847        5.10 feature.
848    
849        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
850        is part of Perl's movement towards a unified syntax for back references. As
851        this is synonymous with \k{name}, we fudge it up by pretending it really
852        was \k.
853    
854        (3) For Oniguruma compatibility we also support \g followed by a name or a
855        number either in angle brackets or in single quotes. However, these are
856        (possibly recursive) subroutine calls, _not_ backreferences. Just return
857        the -ESC_g code (cf \k). */
858    
859        case CHAR_g:
860        if (isclass) break;
861        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
862          {
863          c = -ESC_g;
864          break;
865          }
866    
867        /* Handle the Perl-compatible cases */
868    
869        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
870          {
871          const pcre_uchar *p;
872          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
873            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
874          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
875          {          {
876          c = -ESC_k;          c = -ESC_k;
877          break;          break;
# Line 474  else Line 881  else
881        }        }
882      else braced = FALSE;      else braced = FALSE;
883    
884      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
885        {        {
886        negated = TRUE;        negated = TRUE;
887        ptr++;        ptr++;
888        }        }
889      else negated = FALSE;      else negated = FALSE;
890    
891        /* The integer range is limited by the machine's int representation. */
892      c = 0;      c = 0;
893      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
894        c = c * 10 + *(++ptr) - '0';        {
895          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
896            {
897            c = -1;
898            break;
899            }
900          c = c * 10 + *(++ptr) - CHAR_0;
901          }
902        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
903          {
904          while (IS_DIGIT(ptr[1]))
905            ptr++;
906          *errorcodeptr = ERR61;
907          break;
908          }
909    
910      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
911        {        {
912        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
913        return 0;        break;
914          }
915    
916        if (c == 0)
917          {
918          *errorcodeptr = ERR58;
919          break;
920        }        }
921    
922      if (negated)      if (negated)
# Line 496  else Line 924  else
924        if (c > bracount)        if (c > bracount)
925          {          {
926          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
927          return 0;          break;
928          }          }
929        c = bracount - (c - 1);        c = bracount - (c - 1);
930        }        }
# Line 516  else Line 944  else
944      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
945      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
946    
947      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:
948      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
949    
950      if (!isclass)      if (!isclass)
951        {        {
952        oldptr = ptr;        oldptr = ptr;
953        c -= '0';        /* The integer range is limited by the machine's int representation. */
954        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
955          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
956            {
957            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
958              {
959              c = -1;
960              break;
961              }
962            c = c * 10 + *(++ptr) - CHAR_0;
963            }
964          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
965            {
966            while (IS_DIGIT(ptr[1]))
967              ptr++;
968            *errorcodeptr = ERR61;
969            break;
970            }
971        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
972          {          {
973          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 537  else Line 980  else
980      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.
981      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
982    
983      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
984        {        {
985        ptr--;        ptr--;
986        c = 0;        c = 0;
# Line 550  else Line 993  else
993      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, but no more
994      than 3 octal digits. */      than 3 octal digits. */
995    
996      case '0':      case CHAR_0:
997      c -= '0';      c -= CHAR_0;
998      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
999          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1000      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
1001      break;      break;
1002    
1003      /* \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
1004      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.
1005      treated as a data character. */      If not, { is treated as a data character. */
1006    
1007      case 'x':      case CHAR_x:
1008      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1009        {        {
1010        const uschar *pt = ptr + 2;        /* In JavaScript, \x must be followed by two hexadecimal numbers.
1011        int count = 0;        Otherwise it is a lowercase x letter. */
1012          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1013            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1014            {
1015            c = 0;
1016            for (i = 0; i < 2; ++i)
1017              {
1018              register int cc = *(++ptr);
1019    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1020              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1021              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1022    #else           /* EBCDIC coding */
1023              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1024              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1025    #endif
1026              }
1027            }
1028          break;
1029          }
1030    
1031        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1032          {
1033          const pcre_uchar *pt = ptr + 2;
1034    
1035        c = 0;        c = 0;
1036        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1037          {          {
1038          register int cc = *pt++;          register int cc = *pt++;
1039          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1040    
1041  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1042          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1043          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1044  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1045          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1046          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1047  #endif  #endif
1048    
1049    #ifdef COMPILE_PCRE8
1050            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1051    #else
1052    #ifdef COMPILE_PCRE16
1053            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1054    #endif
1055    #endif
1056            }
1057    
1058          if (c < 0)
1059            {
1060            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1061            *errorcodeptr = ERR34;
1062          }          }
1063    
1064        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1065          {          {
1066          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1067          ptr = pt;          ptr = pt;
1068          break;          break;
1069          }          }
# Line 597  else Line 1075  else
1075      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1076    
1077      c = 0;      c = 0;
1078      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1079        {        {
1080        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1081        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1082  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1083        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1084        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1085  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1086        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1087        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1088  #endif  #endif
1089        }        }
1090      break;      break;
1091    
1092      /* 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.
1093      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
1094        coding is ASCII-specific, but then the whole concept of \cx is
1095      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1096    
1097      case 'c':      case CHAR_c:
1098      c = *(++ptr);      c = *(++ptr);
1099      if (c == 0)      if (c == 0)
1100        {        {
1101        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1102        return 0;        break;
1103        }        }
1104    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1105  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1106      if (c >= 'a' && c <= 'z') c -= 32;        {
1107          *errorcodeptr = ERR68;
1108          break;
1109          }
1110        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1111      c ^= 0x40;      c ^= 0x40;
1112  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1113      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1114      c ^= 0xC0;      c ^= 0xC0;
1115  #endif  #endif
1116      break;      break;
1117    
1118      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1119      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1120      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
1121      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
1122      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1123    
1124      default:      default:
1125      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 649  else Line 1132  else
1132      }      }
1133    }    }
1134    
1135    /* Perl supports \N{name} for character names, as well as plain \N for "not
1136    newline". PCRE does not support \N{name}. However, it does support
1137    quantification such as \N{2,3}. */
1138    
1139    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1140         !is_counted_repeat(ptr+2))
1141      *errorcodeptr = ERR37;
1142    
1143    /* If PCRE_UCP is set, we change the values for \d etc. */
1144    
1145    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1146      c -= (ESC_DU - ESC_D);
1147    
1148    /* Set the pointer to the final character before returning. */
1149    
1150  *ptrptr = ptr;  *ptrptr = ptr;
1151  return c;  return c;
1152  }  }
# Line 675  Returns:         type value from ucp_typ Line 1173  Returns:         type value from ucp_typ
1173  */  */
1174    
1175  static int  static int
1176  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1177  {  {
1178  int c, i, bot, top;  int c, i, bot, top;
1179  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1180  char name[32];  pcre_uchar name[32];
1181    
1182  c = *(++ptr);  c = *(++ptr);
1183  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 689  if (c == 0) goto ERROR_RETURN; Line 1187  if (c == 0) goto ERROR_RETURN;
1187  /* \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
1188  negation. */  negation. */
1189    
1190  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1191    {    {
1192    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1193      {      {
1194      *negptr = TRUE;      *negptr = TRUE;
1195      ptr++;      ptr++;
1196      }      }
1197    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1198      {      {
1199      c = *(++ptr);      c = *(++ptr);
1200      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1201      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1202      name[i] = c;      name[i] = c;
1203      }      }
1204    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1205    name[i] = 0;    name[i] = 0;
1206    }    }
1207    
# Line 720  else Line 1218  else
1218  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1219    
1220  bot = 0;  bot = 0;
1221  top = _pcre_utt_size;  top = PRIV(utt_size);
1222    
1223  while (bot < top)  while (bot < top)
1224    {    {
1225    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1226    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1227    if (c == 0)    if (c == 0)
1228      {      {
1229      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1230      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1231      }      }
1232    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1233    }    }
# Line 749  return -1; Line 1247  return -1;
1247    
1248    
1249  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1250  *         Read repeat counts                     *  *         Read repeat counts                     *
1251  *************************************************/  *************************************************/
1252    
# Line 800  Returns:         pointer to '}' on succe Line 1265  Returns:         pointer to '}' on succe
1265                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1266  */  */
1267    
1268  static const uschar *  static const pcre_uchar *
1269  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)
1270  {  {
1271  int min = 0;  int min = 0;
1272  int max = -1;  int max = -1;
# Line 809  int max = -1; Line 1274  int max = -1;
1274  /* 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
1275  an integer overflow. */  an integer overflow. */
1276    
1277  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1278  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1279    {    {
1280    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 819  if (min < 0 || min > 65535) Line 1284  if (min < 0 || min > 65535)
1284  /* 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.
1285  Also, max must not be less than min. */  Also, max must not be less than min. */
1286    
1287  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1288    {    {
1289    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1290      {      {
1291      max = 0;      max = 0;
1292      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1293      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1294        {        {
1295        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 849  return p; Line 1314  return p;
1314    
1315    
1316  /*************************************************  /*************************************************
1317  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1318  *************************************************/  *************************************************/
1319    
1320  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1321    top-level call starts at the beginning of the pattern. All other calls must
1322    start at a parenthesis. It scans along a pattern's text looking for capturing
1323  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
1324  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
1325  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
1326  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1327  be terminated by '>' because that is checked in the first pass.  
1328    This function was originally called only from the second pass, in which we know
1329    that if (?< or (?' or (?P< is encountered, the name will be correctly
1330    terminated because that is checked in the first pass. There is now one call to
1331    this function in the first pass, to check for a recursive back reference by
1332    name (so that we can make the whole group atomic). In this case, we need check
1333    only up to the current position in the pattern, and that is still OK because
1334    and previous occurrences will have been checked. To make this work, the test
1335    for "end of pattern" is a check against cd->end_pattern in the main loop,
1336    instead of looking for a binary zero. This means that the special first-pass
1337    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1338    processing items within the loop are OK, because afterwards the main loop will
1339    terminate.)
1340    
1341  Arguments:  Arguments:
1342    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1343    count        current count of capturing parens so far encountered    cd           compile background data
1344    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1345    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1346    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1347      utf          TRUE if we are in UTF-8 / UTF-16 mode
1348      count        pointer to the current capturing subpattern number (updated)
1349    
1350  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1351  */  */
1352    
1353  static int  static int
1354  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,
1355    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1356  {  {
1357  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1358    int start_count = *count;
1359    int hwm_count = start_count;
1360    BOOL dup_parens = FALSE;
1361    
1362    /* If the first character is a parenthesis, check on the type of group we are
1363    dealing with. The very first call may not start with a parenthesis. */
1364    
1365  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1366    {    {
1367    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1368    
1369      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1370    
1371      /* Handle a normal, unnamed capturing parenthesis. */
1372    
1373      else if (ptr[1] != CHAR_QUESTION_MARK)
1374        {
1375        *count += 1;
1376        if (name == NULL && *count == lorn) return *count;
1377        ptr++;
1378        }
1379    
1380      /* All cases now have (? at the start. Remember when we are in a group
1381      where the parenthesis numbers are duplicated. */
1382    
1383      else if (ptr[2] == CHAR_VERTICAL_LINE)
1384        {
1385        ptr += 3;
1386        dup_parens = TRUE;
1387        }
1388    
1389      /* Handle comments; all characters are allowed until a ket is reached. */
1390    
1391      else if (ptr[2] == CHAR_NUMBER_SIGN)
1392        {
1393        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1394        goto FAIL_EXIT;
1395        }
1396    
1397      /* Handle a condition. If it is an assertion, just carry on so that it
1398      is processed as normal. If not, skip to the closing parenthesis of the
1399      condition (there can't be any nested parens). */
1400    
1401      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1402        {
1403        ptr += 2;
1404        if (ptr[1] != CHAR_QUESTION_MARK)
1405          {
1406          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1407          if (*ptr != 0) ptr++;
1408          }
1409        }
1410    
1411      /* Start with (? but not a condition. */
1412    
1413      else
1414        {
1415        ptr += 2;
1416        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1417    
1418        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1419    
1420        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1421            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1422          {
1423          int term;
1424          const pcre_uchar *thisname;
1425          *count += 1;
1426          if (name == NULL && *count == lorn) return *count;
1427          term = *ptr++;
1428          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1429          thisname = ptr;
1430          while (*ptr != term) ptr++;
1431          if (name != NULL && lorn == ptr - thisname &&
1432              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1433            return *count;
1434          term++;
1435          }
1436        }
1437      }
1438    
1439    /* Past any initial parenthesis handling, scan for parentheses or vertical
1440    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1441    first-pass call when this value is temporarily adjusted to stop at the current
1442    position. So DO NOT change this to a test for binary zero. */
1443    
1444    for (; ptr < cd->end_pattern; ptr++)
1445      {
1446    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1447    
1448    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1449      {      {
1450      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1451      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1452        {        {
1453        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1454        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1455        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1456        }        }
1457      continue;      continue;
1458      }      }
1459    
1460    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1461      are handled for real. If the first character is '^', skip it. Also, if the
1462      first few characters (either before or after ^) are \Q\E or \E we skip them
1463      too. This makes for compatibility with Perl. Note the use of STR macros to
1464      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1465    
1466      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1467        {
1468        BOOL negate_class = FALSE;
1469        for (;;)
1470          {
1471          if (ptr[1] == CHAR_BACKSLASH)
1472            {
1473            if (ptr[2] == CHAR_E)
1474              ptr+= 2;
1475            else if (STRNCMP_UC_C8(ptr + 2,
1476                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1477              ptr += 4;
1478            else
1479              break;
1480            }
1481          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1482            {
1483            negate_class = TRUE;
1484            ptr++;
1485            }
1486          else break;
1487          }
1488    
1489    if (*ptr == '[')      /* If the next character is ']', it is a data character that must be
1490      {      skipped, except in JavaScript compatibility mode. */
1491      while (*(++ptr) != ']')  
1492        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1493            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1494          ptr++;
1495    
1496        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1497        {        {
1498        if (*ptr == '\\')        if (*ptr == 0) return -1;
1499          if (*ptr == CHAR_BACKSLASH)
1500          {          {
1501          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1502          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1503            {            {
1504            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1505            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1506            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1507            }            }
1508          continue;          continue;
1509          }          }
# Line 916  for (; *ptr != 0; ptr++) Line 1513  for (; *ptr != 0; ptr++)
1513    
1514    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1515    
1516    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1517      {      {
1518      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1519      if (*ptr == 0) return -1;      while (*ptr != 0)
1520          {
1521          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1522          ptr++;
1523    #ifdef SUPPORT_UTF
1524          if (utf) FORWARDCHAR(ptr);
1525    #endif
1526          }
1527        if (*ptr == 0) goto FAIL_EXIT;
1528      continue;      continue;
1529      }      }
1530    
1531    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1532    
1533    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1534      {      {
1535      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1536      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1537      continue;      if (*ptr == 0) goto FAIL_EXIT;
1538      }      }
1539    
1540    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1541    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1542        if (dup_parens && *count < hwm_count) *count = hwm_count;
1543        goto FAIL_EXIT;
1544        }
1545    
1546    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1547        {
1548        if (*count > hwm_count) hwm_count = *count;
1549        *count = start_count;
1550        }
1551      }
1552    
1553    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1554         *ptr != '\'')  *ptrptr = ptr;
1555      continue;  return -1;
1556    }
1557    
1558    
1559    
1560    
1561    /*************************************************
1562    *       Find forward referenced subpattern       *
1563    *************************************************/
1564    
1565    count++;  /* This function scans along a pattern's text looking for capturing
1566    subpatterns, and counting them. If it finds a named pattern that matches the
1567    name it is given, it returns its number. Alternatively, if the name is NULL, it
1568    returns when it reaches a given numbered subpattern. This is used for forward
1569    references to subpatterns. We used to be able to start this scan from the
1570    current compiling point, using the current count value from cd->bracount, and
1571    do it all in a single loop, but the addition of the possibility of duplicate
1572    subpattern numbers means that we have to scan from the very start, in order to
1573    take account of such duplicates, and to use a recursive function to keep track
1574    of the different types of group.
1575    
1576    Arguments:
1577      cd           compile background data
1578      name         name to seek, or NULL if seeking a numbered subpattern
1579      lorn         name length, or subpattern number if name is NULL
1580      xmode        TRUE if we are in /x mode
1581      utf          TRUE if we are in UTF-8 / UTF-16 mode
1582    
1583    Returns:       the number of the found subpattern, or -1 if not found
1584    */
1585    
1586    if (name == NULL && count == lorn) return count;  static int
1587    term = *ptr++;  find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1588    if (term == '<') term = '>';    BOOL utf)
1589    thisname = ptr;  {
1590    while (*ptr != term) ptr++;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1591    if (name != NULL && lorn == ptr - thisname &&  int count = 0;
1592        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  int rc;
1593      return count;  
1594    /* If the pattern does not start with an opening parenthesis, the first call
1595    to find_parens_sub() will scan right to the end (if necessary). However, if it
1596    does start with a parenthesis, find_parens_sub() will return when it hits the
1597    matching closing parens. That is why we have to have a loop. */
1598    
1599    for (;;)
1600      {
1601      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1602      if (rc > 0 || *ptr++ == 0) break;
1603    }    }
1604    
1605  return -1;  return rc;
1606  }  }
1607    
1608    
1609    
1610    
1611  /*************************************************  /*************************************************
1612  *      Find first significant op code            *  *      Find first significant op code            *
1613  *************************************************/  *************************************************/
1614    
1615  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1616  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
1617  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
1618  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
1619  assertions, and also the \b assertion; for others it does not.  does not.
1620    
1621  Arguments:  Arguments:
1622    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  
1623    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1624    
1625  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1626  */  */
1627    
1628  static const uschar*  static const pcre_uchar*
1629  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1630  {  {
1631  for (;;)  for (;;)
1632    {    {
1633    switch ((int)*code)    switch ((int)*code)
1634      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1635      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1636      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1637      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1638      if (!skipassert) return code;      if (!skipassert) return code;
1639      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1640      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1641      break;      break;
1642    
1643      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1008  for (;;) Line 1647  for (;;)
1647    
1648      case OP_CALLOUT:      case OP_CALLOUT:
1649      case OP_CREF:      case OP_CREF:
1650        case OP_NCREF:
1651      case OP_RREF:      case OP_RREF:
1652        case OP_NRREF:
1653      case OP_DEF:      case OP_DEF:
1654      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1655      break;      break;
1656    
1657      default:      default:
# Line 1024  for (;;) Line 1665  for (;;)
1665    
1666    
1667  /*************************************************  /*************************************************
1668  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1669  *************************************************/  *************************************************/
1670    
1671  /* 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,
1672  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.
1673  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
1674    temporarily terminated with OP_END when this function is called.
1675    
1676    This function is called when a backward assertion is encountered, so that if it
1677    fails, the error message can point to the correct place in the pattern.
1678    However, we cannot do this when the assertion contains subroutine calls,
1679    because they can be forward references. We solve this by remembering this case
1680    and doing the check at the end; a flag specifies which mode we are running in.
1681    
1682  Arguments:  Arguments:
1683    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1684    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1685      atend    TRUE if called when the pattern is complete
1686  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1687               or -2 if \C was encountered  
1688    Returns:   the fixed length,
1689                 or -1 if there is no fixed length,
1690                 or -2 if \C was encountered (in UTF-8 mode only)
1691                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1692                 or -4 if an unknown opcode was encountered (internal error)
1693  */  */
1694    
1695  static int  static int
1696  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1697  {  {
1698  int length = -1;  int length = -1;
1699    
1700  register int branchlength = 0;  register int branchlength = 0;
1701  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1702    
1703  /* 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
1704  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1053  branch, check the length against that of Line 1706  branch, check the length against that of
1706  for (;;)  for (;;)
1707    {    {
1708    int d;    int d;
1709      pcre_uchar *ce, *cs;
1710    register int op = *cc;    register int op = *cc;
1711    
1712    switch (op)    switch (op)
1713      {      {
1714        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1715        OP_BRA (normal non-capturing bracket) because the other variants of these
1716        opcodes are all concerned with unlimited repeated groups, which of course
1717        are not of fixed length. */
1718    
1719      case OP_CBRA:      case OP_CBRA:
1720      case OP_BRA:      case OP_BRA:
1721      case OP_ONCE:      case OP_ONCE:
1722        case OP_ONCE_NC:
1723      case OP_COND:      case OP_COND:
1724      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1725      if (d < 0) return d;      if (d < 0) return d;
1726      branchlength += d;      branchlength += d;
1727      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1728      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1729      break;      break;
1730    
1731      /* 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.
1732      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
1733      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
1734        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1735        because they all imply an unlimited repeat. */
1736    
1737      case OP_ALT:      case OP_ALT:
1738      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1739      case OP_END:      case OP_END:
1740        case OP_ACCEPT:
1741        case OP_ASSERT_ACCEPT:
1742      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1743        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1744      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1084  for (;;) Line 1746  for (;;)
1746      branchlength = 0;      branchlength = 0;
1747      break;      break;
1748    
1749        /* A true recursion implies not fixed length, but a subroutine call may
1750        be OK. If the subroutine is a forward reference, we can't deal with
1751        it until the end of the pattern, so return -3. */
1752    
1753        case OP_RECURSE:
1754        if (!atend) return -3;
1755        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1756        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1757        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1758        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1759        if (d < 0) return d;
1760        branchlength += d;
1761        cc += 1 + LINK_SIZE;
1762        break;
1763    
1764      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1765    
1766      case OP_ASSERT:      case OP_ASSERT:
# Line 1091  for (;;) Line 1768  for (;;)
1768      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1769      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1770      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1771      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1772        break;
1773    
1774      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1775    
1776      case OP_REVERSE:      case OP_MARK:
1777        case OP_PRUNE_ARG:
1778        case OP_SKIP_ARG:
1779        case OP_THEN_ARG:
1780        cc += cc[1] + PRIV(OP_lengths)[*cc];
1781        break;
1782    
1783        case OP_CALLOUT:
1784        case OP_CIRC:
1785        case OP_CIRCM:
1786        case OP_CLOSE:
1787        case OP_COMMIT:
1788      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1789      case OP_DEF:      case OP_DEF:
1790      case OP_OPT:      case OP_DOLL:
1791      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1792      case OP_EOD:      case OP_EOD:
1793      case OP_EODN:      case OP_EODN:
1794      case OP_CIRC:      case OP_FAIL:
1795      case OP_DOLL:      case OP_NCREF:
1796        case OP_NRREF:
1797      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1798        case OP_PRUNE:
1799        case OP_REVERSE:
1800        case OP_RREF:
1801        case OP_SET_SOM:
1802        case OP_SKIP:
1803        case OP_SOD:
1804        case OP_SOM:
1805        case OP_THEN:
1806      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1807      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1808      break;      break;
1809    
1810      /* Handle literal characters */      /* Handle literal characters */
1811    
1812      case OP_CHAR:      case OP_CHAR:
1813      case OP_CHARNC:      case OP_CHARI:
1814      case OP_NOT:      case OP_NOT:
1815        case OP_NOTI:
1816      branchlength++;      branchlength++;
1817      cc += 2;      cc += 2;
1818  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1819      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1820  #endif  #endif
1821      break;      break;
1822    
# Line 1131  for (;;) Line 1824  for (;;)
1824      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1825    
1826      case OP_EXACT:      case OP_EXACT:
1827        case OP_EXACTI:
1828        case OP_NOTEXACT:
1829        case OP_NOTEXACTI:
1830      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1831      cc += 4;      cc += 2 + IMM2_SIZE;
1832  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1833      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1834  #endif  #endif
1835      break;      break;
1836    
1837      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1838      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1839      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1840        cc += 1 + IMM2_SIZE + 1;
1841      break;      break;
1842    
1843      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1153  for (;;) Line 1847  for (;;)
1847      cc += 2;      cc += 2;
1848      /* Fall through */      /* Fall through */
1849    
1850        case OP_HSPACE:
1851        case OP_VSPACE:
1852        case OP_NOT_HSPACE:
1853        case OP_NOT_VSPACE:
1854      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1855      case OP_DIGIT:      case OP_DIGIT:
1856      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1160  for (;;) Line 1858  for (;;)
1858      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1859      case OP_WORDCHAR:      case OP_WORDCHAR:
1860      case OP_ANY:      case OP_ANY:
1861        case OP_ALLANY:
1862      branchlength++;      branchlength++;
1863      cc++;      cc++;
1864      break;      break;
1865    
1866      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1867        otherwise \C is coded as OP_ALLANY. */
1868    
1869      case OP_ANYBYTE:      case OP_ANYBYTE:
1870      return -2;      return -2;
1871    
1872      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1873    
1874  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1875      case OP_XCLASS:      case OP_XCLASS:
1876      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1877      /* Fall through */      /* Fall through */
1878  #endif  #endif
1879    
1880      case OP_CLASS:      case OP_CLASS:
1881      case OP_NCLASS:      case OP_NCLASS:
1882      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1883    
1884      switch (*cc)      switch (*cc)
1885        {        {
1886          case OP_CRPLUS:
1887          case OP_CRMINPLUS:
1888        case OP_CRSTAR:        case OP_CRSTAR:
1889        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1890        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1191  for (;;) Line 1893  for (;;)
1893    
1894        case OP_CRRANGE:        case OP_CRRANGE:
1895        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1896        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1897        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1898        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1899        break;        break;
1900    
1901        default:        default:
# Line 1203  for (;;) Line 1905  for (;;)
1905    
1906      /* Anything else is variable length */      /* Anything else is variable length */
1907    
1908      default:      case OP_ANYNL:
1909        case OP_BRAMINZERO:
1910        case OP_BRAPOS:
1911        case OP_BRAPOSZERO:
1912        case OP_BRAZERO:
1913        case OP_CBRAPOS:
1914        case OP_EXTUNI:
1915        case OP_KETRMAX:
1916        case OP_KETRMIN:
1917        case OP_KETRPOS:
1918        case OP_MINPLUS:
1919        case OP_MINPLUSI:
1920        case OP_MINQUERY:
1921        case OP_MINQUERYI:
1922        case OP_MINSTAR:
1923        case OP_MINSTARI:
1924        case OP_MINUPTO:
1925        case OP_MINUPTOI:
1926        case OP_NOTMINPLUS:
1927        case OP_NOTMINPLUSI:
1928        case OP_NOTMINQUERY:
1929        case OP_NOTMINQUERYI:
1930        case OP_NOTMINSTAR:
1931        case OP_NOTMINSTARI:
1932        case OP_NOTMINUPTO:
1933        case OP_NOTMINUPTOI:
1934        case OP_NOTPLUS:
1935        case OP_NOTPLUSI:
1936        case OP_NOTPOSPLUS:
1937        case OP_NOTPOSPLUSI:
1938        case OP_NOTPOSQUERY:
1939        case OP_NOTPOSQUERYI:
1940        case OP_NOTPOSSTAR:
1941        case OP_NOTPOSSTARI:
1942        case OP_NOTPOSUPTO:
1943        case OP_NOTPOSUPTOI:
1944        case OP_NOTQUERY:
1945        case OP_NOTQUERYI:
1946        case OP_NOTSTAR:
1947        case OP_NOTSTARI:
1948        case OP_NOTUPTO:
1949        case OP_NOTUPTOI:
1950        case OP_PLUS:
1951        case OP_PLUSI:
1952        case OP_POSPLUS:
1953        case OP_POSPLUSI:
1954        case OP_POSQUERY:
1955        case OP_POSQUERYI:
1956        case OP_POSSTAR:
1957        case OP_POSSTARI:
1958        case OP_POSUPTO:
1959        case OP_POSUPTOI:
1960        case OP_QUERY:
1961        case OP_QUERYI:
1962        case OP_REF:
1963        case OP_REFI:
1964        case OP_SBRA:
1965        case OP_SBRAPOS:
1966        case OP_SCBRA:
1967        case OP_SCBRAPOS:
1968        case OP_SCOND:
1969        case OP_SKIPZERO:
1970        case OP_STAR:
1971        case OP_STARI:
1972        case OP_TYPEMINPLUS:
1973        case OP_TYPEMINQUERY:
1974        case OP_TYPEMINSTAR:
1975        case OP_TYPEMINUPTO:
1976        case OP_TYPEPLUS:
1977        case OP_TYPEPOSPLUS:
1978        case OP_TYPEPOSQUERY:
1979        case OP_TYPEPOSSTAR:
1980        case OP_TYPEPOSUPTO:
1981        case OP_TYPEQUERY:
1982        case OP_TYPESTAR:
1983        case OP_TYPEUPTO:
1984        case OP_UPTO:
1985        case OP_UPTOI:
1986      return -1;      return -1;
1987    
1988        /* Catch unrecognized opcodes so that when new ones are added they
1989        are not forgotten, as has happened in the past. */
1990    
1991        default:
1992        return -4;
1993      }      }
1994    }    }
1995  /* Control never gets here */  /* Control never gets here */
# Line 1214  for (;;) Line 1999  for (;;)
1999    
2000    
2001  /*************************************************  /*************************************************
2002  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2003  *************************************************/  *************************************************/
2004    
2005  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2006  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2007    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2008    so that it can be called from pcre_study() when finding the minimum matching
2009    length.
2010    
2011  Arguments:  Arguments:
2012    code        points to start of expression    code        points to start of expression
2013    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2014    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2015    
2016  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
2017  */  */
2018    
2019  static const uschar *  const pcre_uchar *
2020  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2021  {  {
2022  for (;;)  for (;;)
2023    {    {
2024    register int c = *code;    register int c = *code;
2025    
2026    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2027    
2028    /* 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
2029    map. This includes negated single high-valued characters. The length in    map. This includes negated single high-valued characters. The length in
2030    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
2031    
2032    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2033    
2034      /* Handle recursion */
2035    
2036      else if (c == OP_REVERSE)
2037        {
2038        if (number < 0) return (pcre_uchar *)code;
2039        code += PRIV(OP_lengths)[c];
2040        }
2041    
2042    /* Handle capturing bracket */    /* Handle capturing bracket */
2043    
2044    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2045               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2046      {      {
2047      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2048      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2049      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2050      }      }
2051    
2052      /* Otherwise, we can get the item's length from the table, except that for
2053      repeated character types, we have to test for \p and \P, which have an extra
2054      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2055      must add in its length. */
2056    
2057      else
2058        {
2059        switch(c)
2060          {
2061          case OP_TYPESTAR:
2062          case OP_TYPEMINSTAR:
2063          case OP_TYPEPLUS:
2064          case OP_TYPEMINPLUS:
2065          case OP_TYPEQUERY:
2066          case OP_TYPEMINQUERY:
2067          case OP_TYPEPOSSTAR:
2068          case OP_TYPEPOSPLUS:
2069          case OP_TYPEPOSQUERY:
2070          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2071          break;
2072    
2073          case OP_TYPEUPTO:
2074          case OP_TYPEMINUPTO:
2075          case OP_TYPEEXACT:
2076          case OP_TYPEPOSUPTO:
2077          if (code[1 + IMM2_SIZE] == OP_PROP
2078            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2079          break;
2080    
2081          case OP_MARK:
2082          case OP_PRUNE_ARG:
2083          case OP_SKIP_ARG:
2084          code += code[1];
2085          break;
2086    
2087          case OP_THEN_ARG:
2088          code += code[1];
2089          break;
2090          }
2091    
2092        /* Add in the fixed length from the table */
2093    
2094        code += PRIV(OP_lengths)[c];
2095    
2096    /* 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
2097    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
2098    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2099    
2100    else  #ifdef SUPPORT_UTF
2101      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2102        {        {
2103        case OP_CHAR:        case OP_CHAR:
2104        case OP_CHARNC:        case OP_CHARI:
2105        case OP_EXACT:        case OP_EXACT:
2106          case OP_EXACTI:
2107        case OP_UPTO:        case OP_UPTO:
2108          case OP_UPTOI:
2109        case OP_MINUPTO:        case OP_MINUPTO:
2110          case OP_MINUPTOI:
2111        case OP_POSUPTO:        case OP_POSUPTO:
2112          case OP_POSUPTOI:
2113        case OP_STAR:        case OP_STAR:
2114          case OP_STARI:
2115        case OP_MINSTAR:        case OP_MINSTAR:
2116          case OP_MINSTARI:
2117        case OP_POSSTAR:        case OP_POSSTAR:
2118          case OP_POSSTARI:
2119        case OP_PLUS:        case OP_PLUS:
2120          case OP_PLUSI:
2121        case OP_MINPLUS:        case OP_MINPLUS:
2122          case OP_MINPLUSI:
2123        case OP_POSPLUS:        case OP_POSPLUS:
2124          case OP_POSPLUSI:
2125        case OP_QUERY:        case OP_QUERY:
2126          case OP_QUERYI:
2127        case OP_MINQUERY:        case OP_MINQUERY:
2128          case OP_MINQUERYI:
2129        case OP_POSQUERY:        case OP_POSQUERY:
2130        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2131          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2132        break;        break;
2133        }        }
2134    #else
2135        (void)(utf);  /* Keep compiler happy by referencing function argument */
2136  #endif  #endif
2137      }      }
2138    }    }
# Line 1295  instance of OP_RECURSE. Line 2149  instance of OP_RECURSE.
2149    
2150  Arguments:  Arguments:
2151    code        points to start of expression    code        points to start of expression
2152    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2153    
2154  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
2155  */  */
2156    
2157  static const uschar *  static const pcre_uchar *
2158  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2159  {  {
2160  for (;;)  for (;;)
2161    {    {
# Line 1315  for (;;) Line 2169  for (;;)
2169    
2170    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2171    
2172    /* 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
2173    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
2174    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
2175    bytes. */    must add in its length. */
2176    
2177    else    else
2178      {      {
2179      code += _pcre_OP_lengths[c];      switch(c)
2180  #ifdef SUPPORT_UTF8        {
2181      if (utf8) switch(c)        case OP_TYPESTAR:
2182          case OP_TYPEMINSTAR:
2183          case OP_TYPEPLUS:
2184          case OP_TYPEMINPLUS:
2185          case OP_TYPEQUERY:
2186          case OP_TYPEMINQUERY:
2187          case OP_TYPEPOSSTAR:
2188          case OP_TYPEPOSPLUS:
2189          case OP_TYPEPOSQUERY:
2190          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2191          break;
2192    
2193          case OP_TYPEPOSUPTO:
2194          case OP_TYPEUPTO:
2195          case OP_TYPEMINUPTO:
2196          case OP_TYPEEXACT:
2197          if (code[1 + IMM2_SIZE] == OP_PROP
2198            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2199          break;
2200    
2201          case OP_MARK:
2202          case OP_PRUNE_ARG:
2203          case OP_SKIP_ARG:
2204          code += code[1];
2205          break;
2206    
2207          case OP_THEN_ARG:
2208          code += code[1];
2209          break;
2210          }
2211    
2212        /* Add in the fixed length from the table */
2213    
2214        code += PRIV(OP_lengths)[c];
2215    
2216        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2217        by a multi-byte character. The length in the table is a minimum, so we have
2218        to arrange to skip the extra bytes. */
2219    
2220    #ifdef SUPPORT_UTF
2221        if (utf) switch(c)
2222        {        {
2223        case OP_CHAR:        case OP_CHAR:
2224        case OP_CHARNC:        case OP_CHARI:
2225        case OP_EXACT:        case OP_EXACT:
2226          case OP_EXACTI:
2227        case OP_UPTO:        case OP_UPTO:
2228          case OP_UPTOI:
2229        case OP_MINUPTO:        case OP_MINUPTO:
2230          case OP_MINUPTOI:
2231        case OP_POSUPTO:        case OP_POSUPTO:
2232          case OP_POSUPTOI:
2233        case OP_STAR:        case OP_STAR:
2234          case OP_STARI:
2235        case OP_MINSTAR:        case OP_MINSTAR:
2236          case OP_MINSTARI:
2237        case OP_POSSTAR:        case OP_POSSTAR:
2238          case OP_POSSTARI:
2239        case OP_PLUS:        case OP_PLUS:
2240          case OP_PLUSI:
2241        case OP_MINPLUS:        case OP_MINPLUS:
2242          case OP_MINPLUSI:
2243        case OP_POSPLUS:        case OP_POSPLUS:
2244          case OP_POSPLUSI:
2245        case OP_QUERY:        case OP_QUERY:
2246          case OP_QUERYI:
2247        case OP_MINQUERY:        case OP_MINQUERY:
2248          case OP_MINQUERYI:
2249        case OP_POSQUERY:        case OP_POSQUERY:
2250        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2251          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2252        break;        break;
2253        }        }
2254    #else
2255        (void)(utf);  /* Keep compiler happy by referencing function argument */
2256  #endif  #endif
2257      }      }
2258    }    }
# Line 1359  for (;;) Line 2268  for (;;)
2268  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()
2269  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
2270  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2271  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
2272  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
2273    bracket whose current branch will already have been scanned.
2274    
2275  Arguments:  Arguments:
2276    code        points to start of search    code        points to start of search
2277    endcode     points to where to stop    endcode     points to where to stop
2278    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2279      cd          contains pointers to tables etc.
2280    
2281  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2282  */  */
2283    
2284  static BOOL  static BOOL
2285  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2286      BOOL utf, compile_data *cd)
2287  {  {
2288  register int c;  register int c;
2289  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2290       code < endcode;       code < endcode;
2291       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2292    {    {
2293    const uschar *ccode;    const pcre_uchar *ccode;
2294    
2295    c = *code;    c = *code;
2296    
2297    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2298      first_significant_code() with a TRUE final argument. */
2299    
2300    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2301      {      {
     code += _pcre_OP_lengths[c];  
2302      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2303      c = *code;      c = *code;
2304      continue;      continue;
2305      }      }
2306    
2307    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2308      implies a backward reference subroutine call, we can scan it. If it's a
2309      forward reference subroutine call, we can't. To detect forward reference
2310      we have to scan up the list that is kept in the workspace. This function is
2311      called only when doing the real compile, not during the pre-compile that
2312      measures the size of the compiled pattern. */
2313    
2314    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_RECURSE)
2315      {      {
2316        const pcre_uchar *scode;
2317      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2318    
2319      /* Scan a closed bracket */      /* Test for forward reference */
2320    
2321        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2322          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2323    
2324        /* Not a forward reference, test for completed backward reference */
2325    
2326      empty_branch = FALSE;      empty_branch = FALSE;
2327        scode = cd->start_code + GET(code, 1);
2328        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2329    
2330        /* Completed backwards reference */
2331    
2332      do      do
2333        {        {
2334        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2335            {
2336          empty_branch = TRUE;          empty_branch = TRUE;
2337            break;
2338            }
2339          scode += GET(scode, 1);
2340          }
2341        while (*scode == OP_ALT);
2342    
2343        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2344        continue;
2345        }
2346    
2347      /* Groups with zero repeats can of course be empty; skip them. */
2348    
2349      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2350          c == OP_BRAPOSZERO)
2351        {
2352        code += PRIV(OP_lengths)[c];
2353        do code += GET(code, 1); while (*code == OP_ALT);
2354        c = *code;
2355        continue;
2356        }
2357    
2358      /* A nested group that is already marked as "could be empty" can just be
2359      skipped. */
2360    
2361      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2362          c == OP_SCBRA || c == OP_SCBRAPOS)
2363        {
2364        do code += GET(code, 1); while (*code == OP_ALT);
2365        c = *code;
2366        continue;
2367        }
2368    
2369      /* For other groups, scan the branches. */
2370    
2371      if (c == OP_BRA  || c == OP_BRAPOS ||
2372          c == OP_CBRA || c == OP_CBRAPOS ||
2373          c == OP_ONCE || c == OP_ONCE_NC ||
2374          c == OP_COND)
2375        {
2376        BOOL empty_branch;
2377        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2378    
2379        /* If a conditional group has only one branch, there is a second, implied,
2380        empty branch, so just skip over the conditional, because it could be empty.
2381        Otherwise, scan the individual branches of the group. */
2382    
2383        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2384        code += GET(code, 1);        code += GET(code, 1);
2385        else
2386          {
2387          empty_branch = FALSE;
2388          do
2389            {
2390            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2391              empty_branch = TRUE;
2392            code += GET(code, 1);
2393            }
2394          while (*code == OP_ALT);
2395          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2396        }        }
2397      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2398      c = *code;      c = *code;
2399      continue;      continue;
2400      }      }
# Line 1418  for (code = first_significant_code(code Line 2403  for (code = first_significant_code(code
2403    
2404    switch (c)    switch (c)
2405      {      {
2406      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2407        cannot be represented just by a bit map. This includes negated single
2408        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2409        actual length is stored in the compiled code, so we must update "code"
2410        here. */
2411    
2412  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2413      case OP_XCLASS:      case OP_XCLASS:
2414      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2415      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2416  #endif  #endif
2417    
2418      case OP_CLASS:      case OP_CLASS:
2419      case OP_NCLASS:      case OP_NCLASS:
2420      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2421    
2422  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2423      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2424  #endif  #endif
2425    
# Line 1466  for (code = first_significant_code(code Line 2455  for (code = first_significant_code(code
2455      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2456      case OP_WORDCHAR:      case OP_WORDCHAR:
2457      case OP_ANY:      case OP_ANY:
2458        case OP_ALLANY:
2459      case OP_ANYBYTE:      case OP_ANYBYTE:
2460      case OP_CHAR:      case OP_CHAR:
2461      case OP_CHARNC:      case OP_CHARI:
2462      case OP_NOT:      case OP_NOT:
2463        case OP_NOTI:
2464      case OP_PLUS:      case OP_PLUS:
2465      case OP_MINPLUS:      case OP_MINPLUS:
2466      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1484  for (code = first_significant_code(code Line 2475  for (code = first_significant_code(code
2475      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2476      return FALSE;      return FALSE;
2477    
2478        /* These are going to continue, as they may be empty, but we have to
2479        fudge the length for the \p and \P cases. */
2480    
2481        case OP_TYPESTAR:
2482        case OP_TYPEMINSTAR:
2483        case OP_TYPEPOSSTAR:
2484        case OP_TYPEQUERY:
2485        case OP_TYPEMINQUERY:
2486        case OP_TYPEPOSQUERY:
2487        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2488        break;
2489    
2490        /* Same for these */
2491    
2492        case OP_TYPEUPTO:
2493        case OP_TYPEMINUPTO:
2494        case OP_TYPEPOSUPTO:
2495        if (code[1 + IMM2_SIZE] == OP_PROP
2496          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2497        break;
2498    
2499      /* End of branch */      /* End of branch */
2500    
2501      case OP_KET:      case OP_KET:
2502      case OP_KETRMAX:      case OP_KETRMAX:
2503      case OP_KETRMIN:      case OP_KETRMIN:
2504        case OP_KETRPOS:
2505      case OP_ALT:      case OP_ALT:
2506      return TRUE;      return TRUE;
2507    
2508      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2509      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2510    
2511  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2512      case OP_STAR:      case OP_STAR:
2513        case OP_STARI:
2514      case OP_MINSTAR:      case OP_MINSTAR:
2515        case OP_MINSTARI:
2516      case OP_POSSTAR:      case OP_POSSTAR:
2517        case OP_POSSTARI:
2518      case OP_QUERY:      case OP_QUERY:
2519        case OP_QUERYI:
2520      case OP_MINQUERY:      case OP_MINQUERY:
2521        case OP_MINQUERYI:
2522      case OP_POSQUERY:      case OP_POSQUERY:
2523        case OP_POSQUERYI:
2524        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2525        break;
2526    
2527      case OP_UPTO:      case OP_UPTO:
2528        case OP_UPTOI:
2529      case OP_MINUPTO:      case OP_MINUPTO:
2530        case OP_MINUPTOI:
2531      case OP_POSUPTO:      case OP_POSUPTO:
2532      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2533        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2534      break;      break;
2535  #endif  #endif
2536    
2537        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2538        string. */
2539    
2540        case OP_MARK:
2541        case OP_PRUNE_ARG:
2542        case OP_SKIP_ARG:
2543        code += code[1];
2544        break;
2545    
2546        case OP_THEN_ARG:
2547        code += code[1];
2548        break;
2549    
2550        /* None of the remaining opcodes are required to match a character. */
2551    
2552        default:
2553        break;
2554      }      }
2555    }    }
2556    
# Line 1524  return TRUE; Line 2567  return TRUE;
2567  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
2568  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,
2569  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.
2570    This function is called only during the real compile, not during the
2571    pre-compile.
2572    
2573  Arguments:  Arguments:
2574    code        points to start of the recursion    code        points to start of the recursion
2575    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2576    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2577    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2578      cd          pointers to tables etc
2579    
2580  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2581  */  */
2582    
2583  static BOOL  static BOOL
2584  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2585    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2586  {  {
2587  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2588    {    {
2589    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2590        return FALSE;
2591    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2592    }    }
2593  return TRUE;  return TRUE;
# Line 1553  return TRUE; Line 2600  return TRUE;
2600  *************************************************/  *************************************************/
2601    
2602  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2603  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
2604  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2605  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2606    
2607    Originally, this function only recognized a sequence of letters between the
2608    terminators, but it seems that Perl recognizes any sequence of characters,
2609    though of course unknown POSIX names are subsequently rejected. Perl gives an
2610    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2611    didn't consider this to be a POSIX class. Likewise for [:1234:].
2612    
2613    The problem in trying to be exactly like Perl is in the handling of escapes. We
2614    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2615    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2616    below handles the special case of \], but does not try to do any other escape
2617    processing. This makes it different from Perl for cases such as [:l\ower:]
2618    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2619    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2620    I think.
2621    
2622    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2623    It seems that the appearance of a nested POSIX class supersedes an apparent
2624    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2625    a digit.
2626    
2627    In Perl, unescaped square brackets may also appear as part of class names. For
2628    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2629    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2630    seem right at all. PCRE does not allow closing square brackets in POSIX class
2631    names.
2632    
2633  Argument:  Arguments:
2634    ptr      pointer to the initial [    ptr      pointer to the initial [
2635    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2636    
2637  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2638  */  */
2639    
2640  static BOOL  static BOOL
2641  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2642  {  {
2643  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2644  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2645  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2646    {    {
2647    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2648    return TRUE;      ptr++;
2649      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2650      else
2651        {
2652        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2653          {
2654          *endptr = ptr;
2655          return TRUE;
2656          }
2657        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2658             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2659              ptr[1] == CHAR_EQUALS_SIGN) &&
2660            check_posix_syntax(ptr, endptr))
2661          return FALSE;
2662        }
2663    }    }
2664  return FALSE;  return FALSE;
2665  }  }
# Line 1598  Returns:     a value representing the na Line 2682  Returns:     a value representing the na
2682  */  */
2683    
2684  static int  static int
2685  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2686  {  {
2687    const char *pn = posix_names;
2688  register int yield = 0;  register int yield = 0;
2689  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2690    {    {
2691    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2692      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2693      pn += posix_name_lengths[yield] + 1;
2694    yield++;    yield++;
2695    }    }
2696  return -1;  return -1;
# Line 1619  return -1; Line 2705  return -1;
2705  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2706  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2707  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
2708  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
2709  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
2710  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
2711  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
2712  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2713    OP_END.
2714    
2715  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2716  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 1634  value in the reference (which is a group Line 2721  value in the reference (which is a group
2721  Arguments:  Arguments:
2722    group      points to the start of the group    group      points to the start of the group
2723    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2724    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2725    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2726    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
2727    
# Line 1642  Returns:     nothing Line 2729  Returns:     nothing
2729  */  */
2730    
2731  static void  static void
2732  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2733    uschar *save_hwm)    pcre_uchar *save_hwm)
2734  {  {
2735  uschar *ptr = group;  pcre_uchar *ptr = group;
2736  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2737    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2738    {    {
2739    int offset;    int offset;
2740    uschar *hc;    pcre_uchar *hc;
2741    
2742    /* 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
2743    reference. */    reference. */
# Line 1694  Arguments: Line 2782  Arguments:
2782  Returns:         new code pointer  Returns:         new code pointer
2783  */  */
2784    
2785  static uschar *  static pcre_uchar *
2786  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2787  {  {
2788  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2789  *code++ = 255;  *code++ = 255;
2790  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2791  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2792  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2793  }  }
2794    
2795    
# Line 1723  Returns:             nothing Line 2811  Returns:             nothing
2811  */  */
2812    
2813  static void  static void
2814  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2815  {  {
2816  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2817  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2818  }  }
2819    
# Line 1757  get_othercase_range(unsigned int *cptr, Line 2845  get_othercase_range(unsigned int *cptr,
2845  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2846    
2847  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2848    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2849    
2850  if (c > d) return FALSE;  if (c > d) return FALSE;
2851    
# Line 1766  next = othercase + 1; Line 2854  next = othercase + 1;
2854    
2855  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2856    {    {
2857    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2858    next++;    next++;
2859    }    }
2860    
# Line 1775  for (++c; c <= d; c++) Line 2863  for (++c; c <= d; c++)
2863    
2864  return TRUE;  return TRUE;
2865  }  }
2866    
2867    
2868    
2869    /*************************************************
2870    *        Check a character and a property        *
2871    *************************************************/
2872    
2873    /* This function is called by check_auto_possessive() when a property item
2874    is adjacent to a fixed character.
2875    
2876    Arguments:
2877      c            the character
2878      ptype        the property type
2879      pdata        the data for the type
2880      negated      TRUE if it's a negated property (\P or \p{^)
2881    
2882    Returns:       TRUE if auto-possessifying is OK
2883    */
2884    
2885    static BOOL
2886    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2887    {
2888    const ucd_record *prop = GET_UCD(c);
2889    switch(ptype)
2890      {
2891      case PT_LAMP:
2892      return (prop->chartype == ucp_Lu ||
2893              prop->chartype == ucp_Ll ||
2894              prop->chartype == ucp_Lt) == negated;
2895    
2896      case PT_GC:
2897      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2898    
2899      case PT_PC:
2900      return (pdata == prop->chartype) == negated;
2901    
2902      case PT_SC:
2903      return (pdata == prop->script) == negated;
2904    
2905      /* These are specials */
2906    
2907      case PT_ALNUM:
2908      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2909              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2910    
2911      case PT_SPACE:    /* Perl space */
2912      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2913              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2914              == negated;
2915    
2916      case PT_PXSPACE:  /* POSIX space */
2917      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2918              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2919              c == CHAR_FF || c == CHAR_CR)
2920              == negated;
2921    
2922      case PT_WORD:
2923      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2924              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2925              c == CHAR_UNDERSCORE) == negated;
2926      }
2927    return FALSE;
2928    }
2929  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2930    
2931    
# Line 1788  whether the next thing could possibly ma Line 2939  whether the next thing could possibly ma
2939  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2940    
2941  Arguments:  Arguments:
2942    op_code       the repeated op code    previous      pointer to the repeated opcode
2943    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2944    ptr           next character in pattern    ptr           next character in pattern
2945    options       options bits    options       options bits
2946    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1800  Returns:        TRUE if possessifying is Line 2949  Returns:        TRUE if possessifying is
2949  */  */
2950    
2951  static BOOL  static BOOL
2952  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2953    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2954  {  {
2955  int next;  pcre_int32 c, next;
2956    int op_code = *previous++;
2957    
2958  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2959    
# Line 1811  if ((options & PCRE_EXTENDED) != 0) Line 2961  if ((options & PCRE_EXTENDED) != 0)
2961    {    {
2962    for (;;)    for (;;)
2963      {      {
2964      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2965      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2966        {        {
2967        while (*(++ptr) != 0)        ptr++;
2968          while (*ptr != 0)
2969            {
2970          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2971            ptr++;
2972    #ifdef SUPPORT_UTF
2973            if (utf) FORWARDCHAR(ptr);
2974    #endif
2975            }
2976        }        }
2977      else break;      else break;
2978      }      }
# Line 1824  if ((options & PCRE_EXTENDED) != 0) Line 2981  if ((options & PCRE_EXTENDED) != 0)
2981  /* 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
2982  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2983    
2984  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2985    {    {
2986    int temperrorcode = 0;    int temperrorcode = 0;
2987    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2988    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2989    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2990    }    }
2991    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2992    {    {
2993  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2994    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2995  #endif  #endif
2996    next = *ptr++;    next = *ptr++;
2997    }    }
   
2998  else return FALSE;  else return FALSE;
2999    
3000  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1848  if ((options & PCRE_EXTENDED) != 0) Line 3003  if ((options & PCRE_EXTENDED) != 0)
3003    {    {
3004    for (;;)    for (;;)
3005      {      {
3006      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3007      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3008        {        {
3009        while (*(++ptr) != 0)        ptr++;
3010          while (*ptr != 0)
3011            {
3012          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3013            ptr++;
3014    #ifdef SUPPORT_UTF
3015            if (utf) FORWARDCHAR(ptr);
3016    #endif
3017            }
3018        }        }
3019      else break;      else break;
3020      }      }
# Line 1860  if ((options & PCRE_EXTENDED) != 0) Line 3022  if ((options & PCRE_EXTENDED) != 0)
3022    
3023  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3024    
3025  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3026    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3027        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
3028    
3029  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
3030    the next item is a character. */
3031    
3032  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3033    {    {
3034    case OP_CHAR:    case OP_CHAR:
3035  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3036    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3037    #else
3038      c = *previous;
3039  #endif  #endif
3040    return item != next;    return c != next;
3041    
3042    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
3043    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
3044    high-valued characters. */    high-valued characters. */
3045    
3046    case OP_CHARNC:    case OP_CHARI:
3047  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3048    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3049  #endif  #else
3050    if (item == next) return FALSE;    c = *previous;
3051  #ifdef SUPPORT_UTF8  #endif
3052    if (utf8)    if (c == next) return FALSE;
3053    #ifdef SUPPORT_UTF
3054      if (utf)
3055      {      {
3056      unsigned int othercase;      unsigned int othercase;
3057      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3058  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3059      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3060  #else  #else
3061      othercase = NOTACHAR;      othercase = NOTACHAR;
3062  #endif  #endif
3063      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3064      }      }
3065    else    else
3066  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3067    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3068    
3069    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3070      opcodes are not used for multi-byte characters, because they are coded using
3071      an XCLASS instead. */
3072    
3073    case OP_NOT:    case OP_NOT:
3074    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
3075    if (item == next) return TRUE;  
3076    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
3077  #ifdef SUPPORT_UTF8    if ((c = *previous) == next) return TRUE;
3078    if (utf8)  #ifdef SUPPORT_UTF
3079      if (utf)
3080      {      {
3081      unsigned int othercase;      unsigned int othercase;
3082      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3083  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3084      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3085  #else  #else
3086      othercase = NOTACHAR;      othercase = NOTACHAR;
3087  #endif  #endif
3088      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3089      }      }
3090    else    else
3091  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3092    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3093    
3094      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3095      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3096    
3097    case OP_DIGIT:    case OP_DIGIT:
3098    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1944  if (next >= 0) switch(op_code) Line 3112  if (next >= 0) switch(op_code)
3112    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3113    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
3114    
3115      case OP_HSPACE:
3116      case OP_NOT_HSPACE:
3117      switch(next)
3118        {
3119        case 0x09:
3120        case 0x20:
3121        case 0xa0:
3122        case 0x1680:
3123        case 0x180e:
3124        case 0x2000:
3125        case 0x2001:
3126        case 0x2002:
3127        case 0x2003:
3128        case 0x2004:
3129        case 0x2005:
3130        case 0x2006:
3131        case 0x2007:
3132        case 0x2008:
3133        case 0x2009:
3134        case 0x200A:
3135        case 0x202f:
3136        case 0x205f:
3137        case 0x3000:
3138        return op_code == OP_NOT_HSPACE;
3139        default:
3140        return op_code != OP_NOT_HSPACE;
3141        }
3142    
3143      case OP_ANYNL:
3144      case OP_VSPACE:
3145      case OP_NOT_VSPACE:
3146      switch(next)
3147        {
3148        case 0x0a:
3149        case 0x0b:
3150        case 0x0c:
3151        case 0x0d:
3152        case 0x85:
3153        case 0x2028:
3154        case 0x2029:
3155        return op_code == OP_NOT_VSPACE;
3156        default:
3157        return op_code != OP_NOT_VSPACE;
3158        }
3159    
3160    #ifdef SUPPORT_UCP
3161      case OP_PROP:
3162      return check_char_prop(next, previous[0], previous[1], FALSE);
3163    
3164      case OP_NOTPROP:
3165      return check_char_prop(next, previous[0], previous[1], TRUE);
3166    #endif
3167    
3168    default:    default:
3169    return FALSE;    return FALSE;
3170    }    }
3171    
3172    
3173  /* 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
3174    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3175    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3176    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3177    replaced by OP_PROP codes when PCRE_UCP is set. */
3178    
3179  switch(op_code)  switch(op_code)
3180    {    {
3181    case OP_CHAR:    case OP_CHAR:
3182    case OP_CHARNC:    case OP_CHARI:
3183  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3184    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3185    #else
3186      c = *previous;
3187  #endif  #endif
3188    switch(-next)    switch(-next)
3189      {      {
3190      case ESC_d:      case ESC_d:
3191      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3192    
3193      case ESC_D:      case ESC_D:
3194      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3195    
3196      case ESC_s:      case ESC_s:
3197      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3198    
3199      case ESC_S:      case ESC_S:
3200      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3201    
3202      case ESC_w:      case ESC_w:
3203      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3204    
3205      case ESC_W:      case ESC_W:
3206      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3207    
3208        case ESC_h:
3209        case ESC_H:
3210        switch(c)
3211          {
3212          case 0x09:
3213          case 0x20:
3214          case 0xa0:
3215          case 0x1680:
3216          case 0x180e:
3217          case 0x2000:
3218          case 0x2001:
3219          case 0x2002:
3220          case 0x2003:
3221          case 0x2004:
3222          case 0x2005:
3223          case 0x2006:
3224          case 0x2007:
3225          case 0x2008:
3226          case 0x2009:
3227          case 0x200A:
3228          case 0x202f:
3229          case 0x205f:
3230          case 0x3000:
3231          return -next != ESC_h;
3232          default:
3233          return -next == ESC_h;
3234          }
3235    
3236        case ESC_v:
3237        case ESC_V:
3238        switch(c)
3239          {
3240          case 0x0a:
3241          case 0x0b:
3242          case 0x0c:
3243          case 0x0d:
3244          case 0x85:
3245          case 0x2028:
3246          case 0x2029:
3247          return -next != ESC_v;
3248          default:
3249          return -next == ESC_v;
3250          }
3251    
3252        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3253        their substitutions and process them. The result will always be either
3254        -ESC_p or -ESC_P. Then fall through to process those values. */
3255    
3256    #ifdef SUPPORT_UCP
3257        case ESC_du:
3258        case ESC_DU:
3259        case ESC_wu:
3260        case ESC_WU:
3261        case ESC_su:
3262        case ESC_SU:
3263          {
3264          int temperrorcode = 0;
3265          ptr = substitutes[-next - ESC_DU];
3266          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3267          if (temperrorcode != 0) return FALSE;
3268          ptr++;    /* For compatibility */
3269          }
3270        /* Fall through */
3271    
3272        case ESC_p:
3273        case ESC_P:
3274          {
3275          int ptype, pdata, errorcodeptr;
3276          BOOL negated;
3277    
3278          ptr--;      /* Make ptr point at the p or P */
3279          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3280          if (ptype < 0) return FALSE;
3281          ptr++;      /* Point past the final curly ket */
3282    
3283          /* If the property item is optional, we have to give up. (When generated
3284          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3285          to the original \d etc. At this point, ptr will point to a zero byte. */
3286    
3287          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3288            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3289              return FALSE;
3290    
3291          /* Do the property check. */
3292    
3293          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3294          }
3295    #endif
3296    
3297      default:      default:
3298      return FALSE;      return FALSE;
3299      }      }
3300    
3301      /* In principle, support for Unicode properties should be integrated here as
3302      well. It means re-organizing the above code so as to get hold of the property
3303      values before switching on the op-code. However, I wonder how many patterns
3304      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3305      these op-codes are never generated.) */
3306    
3307    case OP_DIGIT:    case OP_DIGIT:
3308    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3309             next == -ESC_h || next == -ESC_v || next == -ESC_R;
3310    
3311    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3312    return next == -ESC_d;    return next == -ESC_d;
3313    
3314    case OP_WHITESPACE:    case OP_WHITESPACE:
3315    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3316    
3317    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3318    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3319    
3320      case OP_HSPACE:
3321      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3322             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3323    
3324      case OP_NOT_HSPACE:
3325      return next == -ESC_h;
3326    
3327      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3328      case OP_ANYNL:
3329      case OP_VSPACE:
3330      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3331    
3332      case OP_NOT_VSPACE:
3333      return next == -ESC_v || next == -ESC_R;
3334    
3335    case OP_WORDCHAR:    case OP_WORDCHAR:
3336    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3337             next == -ESC_v || next == -ESC_R;
3338    
3339    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3340    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2024  Arguments: Line 3363  Arguments:
3363    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3364    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3365    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3366    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3367    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3368    bcptr          points to current branch chain    bcptr          points to current branch chain
3369      cond_depth     conditional nesting depth
3370    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3371    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3372                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2036  Returns:         TRUE on success Line 3376  Returns:         TRUE on success
3376  */  */
3377    
3378  static BOOL  static BOOL
3379  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3380    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3381      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3382    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3383  {  {
3384  int repeat_type, op_type;  int repeat_type, op_type;
3385  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3386  int bravalue = 0;  int bravalue = 0;
3387  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3388  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3389  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3390  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3391  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3392  int after_manual_callout = 0;  int after_manual_callout = 0;
3393  int length_prevgroup = 0;  int length_prevgroup = 0;
3394  register int c;  register int c;
3395  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3396  uschar *last_code = code;  pcre_uchar *last_code = code;
3397  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3398  uschar *tempcode;  pcre_uchar *tempcode;
3399  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3400  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3401  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3402  const uschar *tempptr;  const pcre_uchar *tempptr;
3403  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3404  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3405  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3406  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3407    pcre_uint8 classbits[32];
3408  #ifdef SUPPORT_UTF8  
3409  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3410  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3411  uschar *class_utf8data;  dynamically as we process the pattern. */
3412  uschar utf8_char[6];  
3413    #ifdef SUPPORT_UTF
3414    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3415    BOOL utf = (options & PCRE_UTF8) != 0;
3416    pcre_uchar utf_chars[6];
3417  #else  #else
3418  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3419  uschar *utf8_char = NULL;  #endif
3420    
3421    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3422    
3423    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3424    BOOL xclass;
3425    pcre_uchar *class_uchardata;
3426    pcre_uchar *class_uchardata_base;
3427  #endif  #endif
3428    
3429  #ifdef DEBUG  #ifdef PCRE_DEBUG
3430  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3431  #endif  #endif
3432    
# Line 2085  greedy_non_default = greedy_default ^ 1; Line 3437  greedy_non_default = greedy_default ^ 1;
3437    
3438  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3439  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
3440  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
3441  find one.  find one.
3442    
3443  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
3444  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
3445  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3446  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3447    
3448  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3449    
3450  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3451  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
3452  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
3453  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3454    value. This is used only for ASCII characters. */
3455    
3456  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3457    
3458  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3459    
3460  for (;; ptr++)  for (;; ptr++)
3461    {    {
3462    BOOL negate_class;    BOOL negate_class;
3463      BOOL should_flip_negation;
3464    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3465    BOOL is_quantifier;    BOOL is_quantifier;
3466    BOOL is_recurse;    BOOL is_recurse;
3467    int class_charcount;    BOOL reset_bracount;
3468    int class_lastchar;    int class_has_8bitchar;
3469      int class_single_char;
3470    int newoptions;    int newoptions;
3471    int recno;    int recno;
3472    int refsign;    int refsign;
3473    int skipbytes;    int skipbytes;
3474    int subreqbyte;    int subreqchar;
3475    int subfirstbyte;    int subfirstchar;
3476    int terminator;    int terminator;
3477    int mclength;    int mclength;
3478    uschar mcbuffer[8];    int tempbracount;
3479      pcre_uchar mcbuffer[8];
3480    
3481    /* Get next byte in the pattern */    /* Get next character in the pattern */
3482    
3483    c = *ptr;    c = *ptr;
3484    
3485      /* If we are at the end of a nested substitution, revert to the outer level
3486      string. Nesting only happens one level deep. */
3487    
3488      if (c == 0 && nestptr != NULL)
3489        {
3490        ptr = nestptr;
3491        nestptr = NULL;
3492        c = *ptr;
3493        }
3494    
3495    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3496    previous cycle of this loop. */    previous cycle of this loop. */
3497    
3498    if (lengthptr != NULL)    if (lengthptr != NULL)
3499      {      {
3500  #ifdef DEBUG  #ifdef PCRE_DEBUG
3501      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3502  #endif  #endif
3503      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3504            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3505        {        {
3506        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3507        goto FAILED;        goto FAILED;
# Line 2147  for (;; ptr++) Line 3514  for (;; ptr++)
3514      */      */
3515    
3516      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
     *lengthptr += code - last_code;  
     DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));  
3517    
3518        /* Paranoid check for integer overflow */
3519    
3520        if (OFLOW_MAX - *lengthptr < code - last_code)
3521          {
3522          *errorcodeptr = ERR20;
3523          goto FAILED;
3524          }
3525    
3526        *lengthptr += (int)(code - last_code);
3527        DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3528          (int)(code - last_code), c, c));
3529    
3530      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3531      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
3532      if "previous" is NULL, reset the current code pointer to the start. */      if "previous" is NULL, reset the current code pointer to the start. */
# Line 2158  for (;; ptr++) Line 3535  for (;; ptr++)
3535        {        {
3536        if (previous > orig_code)        if (previous > orig_code)
3537          {          {
3538          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3539          code -= previous - orig_code;          code -= previous - orig_code;
3540          previous = orig_code;          previous = orig_code;
3541          }          }
# Line 2174  for (;; ptr++) Line 3551  for (;; ptr++)
3551    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3552    reference list. */    reference list. */
3553    
3554    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3555               WORK_SIZE_SAFETY_MARGIN)
3556      {      {
3557      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3558      goto FAILED;      goto FAILED;
# Line 2184  for (;; ptr++) Line 3562  for (;; ptr++)
3562    
3563    if (inescq && c != 0)    if (inescq && c != 0)
3564      {      {
3565      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3566        {        {
3567        inescq = FALSE;        inescq = FALSE;
3568        ptr++;        ptr++;
# Line 2210  for (;; ptr++) Line 3588  for (;; ptr++)
3588    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3589    a quantifier. */    a quantifier. */
3590    
3591    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3592      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3593        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3594    
3595    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3596         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2221  for (;; ptr++) Line 3600  for (;; ptr++)
3600      previous_callout = NULL;      previous_callout = NULL;
3601      }      }
3602    
3603    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3604    
3605    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3606      {      {
3607      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3608      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3609        {        {
3610        while (*(++ptr) != 0)        ptr++;
3611          while (*ptr != 0)
3612          {          {
3613          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3614            ptr++;
3615    #ifdef SUPPORT_UTF
3616            if (utf) FORWARDCHAR(ptr);
3617    #endif
3618          }          }
3619        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3620    
# Line 2251  for (;; ptr++) Line 3635  for (;; ptr++)
3635      {      {
3636      /* ===================================================================*/      /* ===================================================================*/
3637      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3638      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3639      case ')':      case CHAR_RIGHT_PARENTHESIS:
3640      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3641      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3642      *codeptr = code;      *codeptr = code;
3643      *ptrptr = ptr;      *ptrptr = ptr;
3644      if (lengthptr != NULL)      if (lengthptr != NULL)
3645        {        {
3646        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3647            {
3648            *errorcodeptr = ERR20;
3649            goto FAILED;
3650            }
3651          *lengthptr += (int)(code - last_code);   /* To include callout length */
3652        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3653        }        }
3654      return TRUE;      return TRUE;
# Line 2269  for (;; ptr++) Line 3658  for (;; ptr++)
3658      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3659      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3660    
3661      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3662        previous = NULL;
3663      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3664        {        {
3665        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3666          *code++ = OP_CIRCM;
3667        }        }
3668      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3669      break;      break;
3670    
3671      case '$':      case CHAR_DOLLAR_SIGN:
3672      previous = NULL;      previous = NULL;
3673      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3674      break;      break;
3675    
3676      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3677      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3678    
3679      case '.':      case CHAR_DOT:
3680      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3681      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3682      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3683      previous = code;      previous = code;
3684      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3685      break;      break;
3686    
3687    
# Line 2306  for (;; ptr++) Line 3696  for (;; ptr++)
3696      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3697      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3698      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3699    
3700      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3701        default (Perl) mode, it is treated as a data character. */
3702    
3703        case CHAR_RIGHT_SQUARE_BRACKET:
3704        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3705          {
3706          *errorcodeptr = ERR64;
3707          goto FAILED;
3708          }
3709        goto NORMAL_CHAR;
3710    
3711        case CHAR_LEFT_SQUARE_BRACKET:
3712      previous = code;      previous = code;
3713    
3714      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3715      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3716    
3717      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3718          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3719            check_posix_syntax(ptr, &tempptr))
3720        {        {
3721        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3722        goto FAILED;        goto FAILED;
3723        }        }
3724    
3725      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3726        if the first few characters (either before or after ^) are \Q\E or \E we
3727        skip them too. This makes for compatibility with Perl. */
3728    
3729      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3730        for (;;)
3731        {        {
       negate_class = TRUE;  
3732        c = *(++ptr);        c = *(++ptr);
3733        }        if (c == CHAR_BACKSLASH)
3734      else          {
3735        {          if (ptr[1] == CHAR_E)
3736        negate_class = FALSE;            ptr++;
3737            else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
3738              ptr += 3;
3739            else
3740              break;
3741            }
3742          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3743            negate_class = TRUE;
3744          else break;
3745          }
3746    
3747        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3748        an initial ']' is taken as a data character -- the code below handles
3749        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3750        [^] must match any character, so generate OP_ALLANY. */
3751    
3752        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3753            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3754          {
3755          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3756          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3757          zerofirstchar = firstchar;
3758          break;
3759        }        }
3760    
3761      /* Keep a count of chars with values < 256 so that we can optimize the case      /* If a class contains a negative special such as \S, we need to flip the
3762      of just a single character (as long as it's < 256). However, For higher      negation flag at the end, so that support for characters > 255 works
3763      valued UTF-8 characters, we don't yet do any optimization. */      correctly (they are all included in the class). */
3764    
3765        should_flip_negation = FALSE;
3766    
3767        /* For optimization purposes, we track some properties of the class.
3768        class_has_8bitchar will be non-zero, if the class contains at least one
3769        < 256 character. class_single_char will be 1 if the class contains only
3770        a single character. */
3771    
3772      class_charcount = 0;      class_has_8bitchar = 0;
3773      class_lastchar = -1;      class_single_char = 0;
3774    
3775      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
3776      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains only 1 character (less
3777      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3778      */      */
3779    
3780      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3781    
3782  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3783      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3784      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3785        class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3786  #endif  #endif
3787    
3788      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2358  for (;; ptr++) Line 3791  for (;; ptr++)
3791    
3792      if (c != 0) do      if (c != 0) do
3793        {        {
3794        const uschar *oldptr;        const pcre_uchar *oldptr;
3795    
3796  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3797        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3798          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3799          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3800          }          }
3801  #endif  #endif
3802    
3803    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3804          /* In the pre-compile phase, accumulate the length of any extra
3805          data and reset the pointer. This is so that very large classes that
3806          contain a zillion > 255 characters no longer overwrite the work space
3807          (which is on the stack). */
3808    
3809          if (lengthptr != NULL)
3810            {
3811            *lengthptr += class_uchardata - class_uchardata_base;
3812            class_uchardata = class_uchardata_base;
3813            }
3814    #endif
3815    
3816        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3817    
3818        if (inescq)        if (inescq)
3819          {          {
3820          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3821            {            {
3822            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3823            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2386  for (;; ptr++) Line 3832  for (;; ptr++)
3832        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3833        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3834    
3835        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3836            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3837            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3838          {          {
3839          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3840          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3841          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3842          uschar pbits[32];          pcre_uint8 pbits[32];
3843    
3844          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3845            {            {
3846            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3847            goto FAILED;            goto FAILED;
3848            }            }
3849    
3850          ptr += 2;          ptr += 2;
3851          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3852            {            {
3853            local_negate = TRUE;            local_negate = TRUE;
3854              should_flip_negation = TRUE;  /* Note negative special */
3855            ptr++;            ptr++;
3856            }            }
3857    
3858          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3859          if (posix_class < 0)          if (posix_class < 0)
3860            {            {
3861            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2422  for (;; ptr++) Line 3869  for (;; ptr++)
3869          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3870            posix_class = 0;            posix_class = 0;
3871    
3872          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3873          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3874          subtract bits that may be in the main map already. At the end we or the  
3875          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3876            if ((options & PCRE_UCP) != 0)
3877              {
3878              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3879              if (posix_substitutes[pc] != NULL)
3880                {
3881                nestptr = tempptr + 1;
3882                ptr = posix_substitutes[pc] - 1;
3883                continue;
3884                }
3885              }
3886    #endif
3887            /* In the non-UCP case, we build the bit map for the POSIX class in a
3888            chunk of local store because we may be adding and subtracting from it,
3889            and we don't want to subtract bits that may be in the main map already.
3890            At the end we or the result into the bit map that is being built. */
3891    
3892          posix_class *= 3;          posix_class *= 3;
3893    
3894          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3895    
3896          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3897            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3898    
3899          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3900    
# Line 2463  for (;; ptr++) Line 3925  for (;; ptr++)
3925            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3926    
3927          ptr = tempptr + 1;          ptr = tempptr + 1;
3928          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3929            class_has_8bitchar = 1;
3930            /* Every class contains at least two characters. */
3931            class_single_char = 2;
3932          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3933          }          }
3934    
3935        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3936        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3937        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3938        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so
3939        to or into the one we are building. We assume they have more than one        speculatively set both class_has_8bitchar and class_single_char bigger
3940        character in them, so set class_charcount bigger than one. */        than one. Unrecognized escapes fall through and are either treated
3941          as literal characters (by default), or are faulted if
3942          PCRE_EXTRA is set. */
3943    
3944        if (c == '\\')        if (c == CHAR_BACKSLASH)
3945          {          {
3946          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3947          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3948    
3949          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3950          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_N)            /* \N is not supported in a class */
3951          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */            {
3952              *errorcodeptr = ERR71;
3953              goto FAILED;
3954              }
3955          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3956            {            {
3957            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3958              {              {
3959              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3960              }              }
3961            else inescq = TRUE;            else inescq = TRUE;
3962            continue;            continue;
3963            }            }
3964            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3965    
3966          if (c < 0)          if (c < 0)
3967            {            {
3968            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3969            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
3970              class_has_8bitchar++;
3971            /* Save time by not doing this in the pre-compile phase. */            /* Every class contains at least two characters. */
3972              class_single_char += 2;
3973    
3974            if (lengthptr == NULL) switch (-c)            switch (-c)
3975              {              {
3976    #ifdef SUPPORT_UCP
3977                case ESC_du:     /* These are the values given for \d etc */
3978                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3979                case ESC_wu:     /* escape sequence with an appropriate \p */
3980                case ESC_WU:     /* or \P to test Unicode properties instead */
3981                case ESC_su:     /* of the default ASCII testing. */
3982                case ESC_SU:
3983                nestptr = ptr;
3984                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3985                class_has_8bitchar--;                /* Undo! */
3986                continue;
3987    #endif
3988              case ESC_d:              case ESC_d:
3989              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3990              continue;              continue;
3991    
3992              case ESC_D:              case ESC_D:
3993                should_flip_negation = TRUE;
3994              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3995              continue;              continue;
3996    
# Line 2514  for (;; ptr++) Line 3999  for (;; ptr++)
3999              continue;              continue;
4000    
4001              case ESC_W:              case ESC_W:
4002                should_flip_negation = TRUE;
4003              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4004              continue;              continue;
4005    
4006              case ESC_s:              /* Perl 5.004 onwards omits VT from \s, but we must preserve it
4007              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              if it was previously set by something earlier in the character
4008              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              class. */
4009    
4010                case ESC_s:
4011                classbits[0] |= cbits[cbit_space];
4012                classbits[1] |= cbits[cbit_space+1] & ~0x08;
4013                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4014                continue;
4015    
4016                case ESC_S:
4017                should_flip_negation = TRUE;
4018                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
4019                classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
4020                continue;
4021    
4022                case ESC_h:
4023                SETBIT(classbits, 0x09); /* VT */
4024                SETBIT(classbits, 0x20); /* SPACE */
4025                SETBIT(classbits, 0xa0); /* NSBP */
4026    #ifndef COMPILE_PCRE8
4027                xclass = TRUE;
4028                *class_uchardata++ = XCL_SINGLE;
4029                *class_uchardata++ = 0x1680;
4030                *class_uchardata++ = XCL_SINGLE;
4031                *class_uchardata++ = 0x180e;
4032                *class_uchardata++ = XCL_RANGE;
4033                *class_uchardata++ = 0x2000;
4034                *class_uchardata++ = 0x200a;
4035                *class_uchardata++ = XCL_SINGLE;
4036                *class_uchardata++ = 0x202f;
4037                *class_uchardata++ = XCL_SINGLE;
4038                *class_uchardata++ = 0x205f;
4039                *class_uchardata++ = XCL_SINGLE;
4040                *class_uchardata++ = 0x3000;
4041    #elif defined SUPPORT_UTF
4042                if (utf)
4043                  {
4044                  xclass = TRUE;
4045                  *class_uchardata++ = XCL_SINGLE;
4046                  class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
4047                  *class_uchardata++ = XCL_SINGLE;
4048                  class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
4049                  *class_uchardata++ = XCL_RANGE;
4050                  class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
4051                  class_uchardata += PRIV(ord2utf)(0x200a, class_uchardata);
4052                  *class_uchardata++ = XCL_SINGLE;
4053                  class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
4054                  *class_uchardata++ = XCL_SINGLE;
4055                  class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
4056                  *class_uchardata++ = XCL_SINGLE;
4057                  class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
4058                  }
4059    #endif
4060                continue;
4061    
4062                case ESC_H:
4063                for (c = 0; c < 32; c++)
4064                  {
4065                  int x = 0xff;
4066                  switch (c)
4067                    {
4068                    case 0x09/8: