/[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 334 by ph10, Fri Apr 11 15:48:14 2008 UTC revision 1266 by ph10, Sun Mar 3 11:14:26 2013 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) 1997-2008 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 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  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
106  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
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118  #define COMPILE_WORK_SIZE (4096)  /* The overrun tests check for a slightly smaller size so that they detect the
119    overrun before it actually does run off the end of the data block. */
120    
121    #define WORK_SIZE_SAFETY_MARGIN (100)
122    
123    /* Private flags added to firstchar and reqchar. */
124    
125    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
126    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
127    /* Negative values for the firstchar and reqchar flags */
128    #define REQ_UNSET       (-2)
129    #define REQ_NONE        (-1)
130    
131    /* Repeated character flags. */
132    
133    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
134    
135  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
136  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
137  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
138  is invalid. */  is invalid. */
139    
140  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
141    
142    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
143    in UTF-8 mode. */
144    
145  static const short int escapes[] = {  static const short int escapes[] = {
146       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
147       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
148     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
149  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
150  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
151  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
152     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
153  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
154  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
155       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
156         -ESC_D,                  -ESC_E,
157         0,                       -ESC_G,
158         -ESC_H,                  0,
159         0,                       -ESC_K,
160         0,                       0,
161         -ESC_N,                  0,
162         -ESC_P,                  -ESC_Q,
163         -ESC_R,                  -ESC_S,
164         0,                       0,
165         -ESC_V,                  -ESC_W,
166         -ESC_X,                  0,
167         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
168         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
169         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
170         CHAR_GRAVE_ACCENT,       7,
171         -ESC_b,                  0,
172         -ESC_d,                  ESC_e,
173         ESC_f,                   0,
174         -ESC_h,                  0,
175         0,                       -ESC_k,
176         0,                       0,
177         ESC_n,                   0,
178         -ESC_p,                  0,
179         ESC_r,                   -ESC_s,
180         ESC_tee,                 0,
181         -ESC_v,                  -ESC_w,
182         0,                       0,
183         -ESC_z
184  };  };
185    
186  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
187    
188    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
189    
190  static const short int escapes[] = {  static const short int escapes[] = {
191  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
192  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 205  static const short int escapes[] = {
205  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
206  /*  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,
207  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
208  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
209  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
210  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
211  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 217  static const short int escapes[] = {
217    
218  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
219  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
220  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
221    string is built from string macros so that it works in UTF-8 mode on EBCDIC
222    platforms. */
223    
224  typedef struct verbitem {  typedef struct verbitem {
225    int   len;    int   len;                 /* Length of verb name */
226    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
227      int   op_arg;              /* Op when arg present, or -1 if not allowed */
228  } verbitem;  } verbitem;
229    
230  static const char verbnames[] =  static const char verbnames[] =
231    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
232    "COMMIT\0"    STRING_MARK0
233    "F\0"    STRING_ACCEPT0
234    "FAIL\0"    STRING_COMMIT0
235    "PRUNE\0"    STRING_F0
236    "SKIP\0"    STRING_FAIL0
237    "THEN";    STRING_PRUNE0
238      STRING_SKIP0
239      STRING_THEN;
240    
241  static const verbitem verbs[] = {  static const verbitem verbs[] = {
242    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
243    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
244    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
245    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
246    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
247    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
248    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
249      { 4, OP_SKIP,   OP_SKIP_ARG  },
250      { 4, OP_THEN,   OP_THEN_ARG  }
251  };  };
252    
253  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 178  length entry. The first three must be al Line 260  length entry. The first three must be al
260  for handling case independence. */  for handling case independence. */
261    
262  static const char posix_names[] =  static const char posix_names[] =
263    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
264    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
265    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
266      STRING_word0  STRING_xdigit;
267    
268  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
269    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 };
270    
271  /* 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 212  static const int posix_class_maps[] = { Line 295  static const int posix_class_maps[] = {
295    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
296  };  };
297    
298    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
299    substitutes must be in the order of the names, defined above, and there are
300    both positive and negative cases. NULL means no substitute. */
301    
302    #ifdef SUPPORT_UCP
303    static const pcre_uchar string_PNd[]  = {
304      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_pNd[]  = {
307      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_PXsp[] = {
310      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_pXsp[] = {
313      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_PXwd[] = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pXwd[] = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    
322    static const pcre_uchar *substitutes[] = {
323      string_PNd,           /* \D */
324      string_pNd,           /* \d */
325      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
326      string_pXsp,          /* \s */
327      string_PXwd,          /* \W */
328      string_pXwd           /* \w */
329    };
330    
331    static const pcre_uchar string_pL[] =   {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLl[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pLu[] =  {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_pXan[] = {
341      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_h[] =    {
344      CHAR_BACKSLASH, CHAR_h, '\0' };
345    static const pcre_uchar string_pXps[] = {
346      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PL[] =   {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLl[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PLu[] =  {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_PXan[] = {
358      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
359      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
360    static const pcre_uchar string_H[] =    {
361      CHAR_BACKSLASH, CHAR_H, '\0' };
362    static const pcre_uchar string_PXps[] = {
363      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    
366    static const pcre_uchar *posix_substitutes[] = {
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      /* Negated cases */
382      string_PL,            /* ^alpha */
383      string_PLl,           /* ^lower */
384      string_PLu,           /* ^upper */
385      string_PXan,          /* ^alnum */
386      NULL,                 /* ^ascii */
387      string_H,             /* ^blank */
388      NULL,                 /* ^cntrl */
389      string_PNd,           /* ^digit */
390      NULL,                 /* ^graph */
391      NULL,                 /* ^print */
392      NULL,                 /* ^punct */
393      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
394      string_PXwd,          /* ^word */
395      NULL                  /* ^xdigit */
396    };
397    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
398    #endif
399    
400  #define STRING(a)  # a  #define STRING(a)  # a
401  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 408  the number of relocations needed when a
408  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
409  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
410  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
411  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
412    
413    Each substring ends with \0 to insert a null character. This includes the final
414    substring, so that the whole string ends with \0\0, which can be detected when
415    counting through. */
416    
417  static const char error_texts[] =  static const char error_texts[] =
418    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 453  static const char error_texts[] =
453    /* 30 */    /* 30 */
454    "unknown POSIX class name\0"    "unknown POSIX class name\0"
455    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
456    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
457    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
458    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
459    /* 35 */    /* 35 */
460    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
461    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
462    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
463    "number after (?C is > 255\0"    "number after (?C is > 255\0"
464    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
465    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 476  static const char error_texts[] =
476    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
477    /* 50 */    /* 50 */
478    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
479    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
480    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
481    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
482    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
483    /* 55 */    /* 55 */
484    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
485    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
486    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
487    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
488    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
489    /* 60 */    /* 60 */
490    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
491    "number is too big\0"    "number is too big\0"
492    "subpattern name expected\0"    "subpattern name expected\0"
493    "digit expected after (?+";    "digit expected after (?+\0"
494      "] is an invalid data character in JavaScript compatibility mode\0"
495      /* 65 */
496      "different names for subpatterns of the same number are not allowed\0"
497      "(*MARK) must have an argument\0"
498      "this version of PCRE is not compiled with Unicode property support\0"
499      "\\c must be followed by an ASCII character\0"
500      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
501      /* 70 */
502      "internal error: unknown opcode in find_fixedlength()\0"
503      "\\N is not supported in a class\0"
504      "too many forward references\0"
505      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
506      "invalid UTF-16 string\0"
507      /* 75 */
508      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
509      "character value in \\u.... sequence is too large\0"
510      "invalid UTF-32 string\0"
511      ;
512    
513  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
514  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 321  For convenience, we use the same bit def Line 526  For convenience, we use the same bit def
526    
527  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
528    
529  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
530  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
531    into a subtraction and unsigned comparison). */
532    
533    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
534    
535    #ifndef EBCDIC
536    
537    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
538    UTF-8 mode. */
539    
540    static const pcre_uint8 digitab[] =
541    {    {
542    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
543    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 357  static const unsigned char digitab[] = Line 572  static const unsigned char digitab[] =
572    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
573    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
574    
575  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
576  static const unsigned char digitab[] =  
577    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
578    
579    static const pcre_uint8 digitab[] =
580    {    {
581    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
582    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 393  static const unsigned char digitab[] = Line 611  static const unsigned char digitab[] =
611    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
612    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
613    
614  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
615    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
616    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
617    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 429  static const unsigned char ebcdic_charta Line 647  static const unsigned char ebcdic_charta
647  #endif  #endif
648    
649    
 /* Definition to allow mutual recursion */  
   
 static BOOL  
   compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  
     int *, int *, branch_chain *, compile_data *, int *);  
   
   
650    
651  /*************************************************  /*************************************************
652  *            Find an error text                  *  *            Find an error text                  *
# Line 454  static const char * Line 665  static const char *
665  find_error_text(int n)  find_error_text(int n)
666  {  {
667  const char *s = error_texts;  const char *s = error_texts;
668  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
669      {
670      while (*s++ != CHAR_NULL) {};
671      if (*s == CHAR_NULL) return "Error text not found (please report)";
672      }
673  return s;  return s;
674  }  }
675    
676    
677  /*************************************************  /*************************************************
678    *           Expand the workspace                 *
679    *************************************************/
680    
681    /* This function is called during the second compiling phase, if the number of
682    forward references fills the existing workspace, which is originally a block on
683    the stack. A larger block is obtained from malloc() unless the ultimate limit
684    has been reached or the increase will be rather small.
685    
686    Argument: pointer to the compile data block
687    Returns:  0 if all went well, else an error number
688    */
689    
690    static int
691    expand_workspace(compile_data *cd)
692    {
693    pcre_uchar *newspace;
694    int newsize = cd->workspace_size * 2;
695    
696    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
697    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
698        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
699     return ERR72;
700    
701    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
702    if (newspace == NULL) return ERR21;
703    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
704    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
705    if (cd->workspace_size > COMPILE_WORK_SIZE)
706      (PUBL(free))((void *)cd->start_workspace);
707    cd->start_workspace = newspace;
708    cd->workspace_size = newsize;
709    return 0;
710    }
711    
712    
713    
714    /*************************************************
715    *            Check for counted repeat            *
716    *************************************************/
717    
718    /* This function is called when a '{' is encountered in a place where it might
719    start a quantifier. It looks ahead to see if it really is a quantifier or not.
720    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
721    where the ddds are digits.
722    
723    Arguments:
724      p         pointer to the first char after '{'
725    
726    Returns:    TRUE or FALSE
727    */
728    
729    static BOOL
730    is_counted_repeat(const pcre_uchar *p)
731    {
732    if (!IS_DIGIT(*p)) return FALSE;
733    p++;
734    while (IS_DIGIT(*p)) p++;
735    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
736    
737    if (*p++ != CHAR_COMMA) return FALSE;
738    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
739    
740    if (!IS_DIGIT(*p)) return FALSE;
741    p++;
742    while (IS_DIGIT(*p)) p++;
743    
744    return (*p == CHAR_RIGHT_CURLY_BRACKET);
745    }
746    
747    
748    
749    /*************************************************
750  *            Handle escapes                      *  *            Handle escapes                      *
751  *************************************************/  *************************************************/
752    
753  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
754  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character
755  encodes one of the more complicated things such as \d. A backreference to group  which will be placed in chptr. A backreference to group n is returned as
756  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  negative n. When UTF-8 is enabled, a positive value greater than 255 may
757  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  be returned in chptr.
758  ptr is pointing at the \. On exit, it is on the final character of the escape  On entry,ptr is pointing at the \. On exit, it is on the final character of the
759  sequence.  escape sequence.
760    
761  Arguments:  Arguments:
762    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
763      chptr          points to the data character
764    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
765    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
766    options        the options bits    options        the options bits
767    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
768    
769  Returns:         zero or positive => a data character  Returns:         zero => a data character
770                   negative => a special escape sequence                   positive => a special escape sequence
771                     negative => a back reference
772                   on error, errorcodeptr is set                   on error, errorcodeptr is set
773  */  */
774    
775  static int  static int
776  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
777    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
778  {  {
779  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
780  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
781  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
782    pcre_uint32 c;
783    int escape = 0;
784    int i;
785    
786  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
787  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
788    
789  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
790    
791  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
792    
793  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
794  in 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.
795  Otherwise further processing may be required. */  Otherwise further processing may be required. */
796    
797  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
798  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
799  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
800    else if ((i = escapes[c - CHAR_0]) != 0)
801      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
802    
803  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
804  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
805  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
806    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
807  #endif  #endif
808    
809  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
810    
811  else  else
812    {    {
813    const uschar *oldptr;    const pcre_uchar *oldptr;
814    BOOL braced, negated;    BOOL braced, negated, overflow;
815      int s;
816    
817    switch (c)    switch (c)
818      {      {
819      /* 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
820      error. */      error. */
821    
822      case 'l':      case CHAR_l:
823      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
824      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
825      break;      break;
826    
827      /* \g must be followed by one of a number of specific things:      case CHAR_u:
828        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
829          {
830          /* In JavaScript, \u must be followed by four hexadecimal numbers.
831          Otherwise it is a lowercase u letter. */
832          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
833            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
834            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
835            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
836            {
837            c = 0;
838            for (i = 0; i < 4; ++i)
839              {
840              register pcre_uint32 cc = *(++ptr);
841    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
842              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
843              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
844    #else           /* EBCDIC coding */
845              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
846              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
847    #endif
848              }
849    
850    #if defined COMPILE_PCRE8
851            if (c > (utf ? 0x10ffffU : 0xffU))
852    #elif defined COMPILE_PCRE16
853            if (c > (utf ? 0x10ffffU : 0xffffU))
854    #elif defined COMPILE_PCRE32
855            if (utf && c > 0x10ffffU)
856    #endif
857              {
858              *errorcodeptr = ERR76;
859              }
860            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
861            }
862          }
863        else
864          *errorcodeptr = ERR37;
865        break;
866    
867        case CHAR_U:
868        /* In JavaScript, \U is an uppercase U letter. */
869        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
870        break;
871    
872        /* In a character class, \g is just a literal "g". Outside a character
873        class, \g must be followed by one of a number of specific things:
874    
875      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
876      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
877      5.10 feature.      5.10 feature.
878    
879      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
880      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
881      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
882      was \k.      was \k.
883    
884      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
885      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
886      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
887      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
888    
889      case 'g':      case CHAR_g:
890      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
891        {      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
892        c = -ESC_g;        {
893        break;        escape = ESC_g;
894        }        break;
895          }
896    
897      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
898    
899      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
900        {        {
901        const uschar *p;        const pcre_uchar *p;
902        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
903          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
904        if (*p != 0 && *p != '}')        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
905          {          {
906          c = -ESC_k;          escape = ESC_k;
907          break;          break;
908          }          }
909        braced = TRUE;        braced = TRUE;
# Line 571  else Line 911  else
911        }        }
912      else braced = FALSE;      else braced = FALSE;
913    
914      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
915        {        {
916        negated = TRUE;        negated = TRUE;
917        ptr++;        ptr++;
918        }        }
919      else negated = FALSE;      else negated = FALSE;
920    
921      c = 0;      /* The integer range is limited by the machine's int representation. */
922      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
923        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
924        while (IS_DIGIT(ptr[1]))
925      if (c < 0)   /* Integer overflow */        {
926          if (s > INT_MAX / 10 - 1) /* Integer overflow */
927            {
928            overflow = TRUE;
929            break;
930            }
931          s = s * 10 + (int)(*(++ptr) - CHAR_0);
932          }
933        if (overflow) /* Integer overflow */
934        {        {
935          while (IS_DIGIT(ptr[1]))
936            ptr++;
937        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
938        break;        break;
939        }        }
940    
941      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
942        {        {
943        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
944        break;        break;
945        }        }
946    
947      if (c == 0)      if (s == 0)
948        {        {
949        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
950        break;        break;
951        }        }
952    
953      if (negated)      if (negated)
954        {        {
955        if (c > bracount)        if (s > bracount)
956          {          {
957          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
958          break;          break;
959          }          }
960        c = bracount - (c - 1);        s = bracount - (s - 1);
961        }        }
962    
963      c = -(ESC_REF + c);      escape = -s;
964      break;      break;
965    
966      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 625  else Line 975  else
975      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
976      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
977    
978      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:
979      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
980    
981      if (!isclass)      if (!isclass)
982        {        {
983        oldptr = ptr;        oldptr = ptr;
984        c -= '0';        /* The integer range is limited by the machine's int representation. */
985        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
986          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
987        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
988            {
989            if (s > INT_MAX / 10 - 1) /* Integer overflow */
990              {
991              overflow = TRUE;
992              break;
993              }
994            s = s * 10 + (int)(*(++ptr) - CHAR_0);
995            }
996          if (overflow) /* Integer overflow */
997          {          {
998            while (IS_DIGIT(ptr[1]))
999              ptr++;
1000          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1001          break;          break;
1002          }          }
1003        if (c < 10 || c <= bracount)        if (s < 10 || s <= bracount)
1004          {          {
1005          c = -(ESC_REF + c);          escape = -s;
1006          break;          break;
1007          }          }
1008        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 651  else Line 1012  else
1012      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.
1013      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1014    
1015      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1016        {        {
1017        ptr--;        ptr--;
1018        c = 0;        c = 0;
# Line 661  else Line 1022  else
1022      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1023      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1024      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1025      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1026      than 3 octal digits. */      but no more than 3 octal digits. */
1027    
1028      case '0':      case CHAR_0:
1029      c -= '0';      c -= CHAR_0;
1030      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1031          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1032      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1033        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1034    #endif
1035      break;      break;
1036    
1037      /* \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
1038      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.
1039      treated as a data character. */      If not, { is treated as a data character. */
1040    
1041        case CHAR_x:
1042        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1043          {
1044          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1045          Otherwise it is a lowercase x letter. */
1046          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1047            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1048            {
1049            c = 0;
1050            for (i = 0; i < 2; ++i)
1051              {
1052              register pcre_uint32 cc = *(++ptr);
1053    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1054              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1055              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1056    #else           /* EBCDIC coding */
1057              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1058              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1059    #endif
1060              }
1061            }
1062          break;
1063          }
1064    
1065      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1066        {        {
1067        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1068    
1069        c = 0;        c = 0;
1070        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1071          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1072          {          {
1073          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1074          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1075          count++;  
1076    #ifdef COMPILE_PCRE32
1077  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1078          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1079          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1080    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1081            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1082            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1083  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1084          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1085          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1086    #endif
1087    
1088    #if defined COMPILE_PCRE8
1089            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1090    #elif defined COMPILE_PCRE16
1091            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1092    #elif defined COMPILE_PCRE32
1093            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1094  #endif  #endif
1095          }          }
1096    
1097        if (*pt == '}')        if (overflow)
1098            {
1099            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1100            *errorcodeptr = ERR34;
1101            }
1102    
1103          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1104          {          {
1105          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1106          ptr = pt;          ptr = pt;
1107          break;          break;
1108          }          }
# Line 711  else Line 1114  else
1114      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1115    
1116      c = 0;      c = 0;
1117      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1118        {        {
1119        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1120        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1121  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1122        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1123        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1124  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1125        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1126        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1127  #endif  #endif
1128        }        }
1129      break;      break;
1130    
1131      /* 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.
1132      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
1133        coding is ASCII-specific, but then the whole concept of \cx is
1134      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1135    
1136      case 'c':      case CHAR_c:
1137      c = *(++ptr);      c = *(++ptr);
1138      if (c == 0)      if (c == CHAR_NULL)
1139        {        {
1140        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1141        break;        break;
1142        }        }
1143    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1144  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1145      if (c >= 'a' && c <= 'z') c -= 32;        {
1146          *errorcodeptr = ERR68;
1147          break;
1148          }
1149        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1150      c ^= 0x40;      c ^= 0x40;
1151  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1152      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1153      c ^= 0xC0;      c ^= 0xC0;
1154  #endif  #endif
1155      break;      break;
# Line 763  else Line 1171  else
1171      }      }
1172    }    }
1173    
1174  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1175  return c;  newline". PCRE does not support \N{name}. However, it does support
1176  }  quantification such as \N{2,3}. */
1177    
1178    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1179         !is_counted_repeat(ptr+2))
1180      *errorcodeptr = ERR37;
1181    
1182    /* If PCRE_UCP is set, we change the values for \d etc. */
1183    
1184    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1185      escape += (ESC_DU - ESC_D);
1186    
1187    /* Set the pointer to the final character before returning. */
1188    
1189    *ptrptr = ptr;
1190    *chptr = c;
1191    return escape;
1192    }
1193    
1194  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1195  /*************************************************  /*************************************************
# Line 782  escape sequence. Line 1204  escape sequence.
1204  Argument:  Argument:
1205    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1206    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1207    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1208      pdataptr       points to an unsigned int that is set to the detailed property value
1209    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1210    
1211  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1212  */  */
1213    
1214  static int  static BOOL
1215  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1216      unsigned int *pdataptr, int *errorcodeptr)
1217  {  {
1218  int c, i, bot, top;  pcre_uchar c;
1219  const uschar *ptr = *ptrptr;  int i, bot, top;
1220  char name[32];  const pcre_uchar *ptr = *ptrptr;
1221    pcre_uchar name[32];
1222    
1223  c = *(++ptr);  c = *(++ptr);
1224  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1225    
1226  *negptr = FALSE;  *negptr = FALSE;
1227    
1228  /* \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
1229  negation. */  negation. */
1230    
1231  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1232    {    {
1233    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1234      {      {
1235      *negptr = TRUE;      *negptr = TRUE;
1236      ptr++;      ptr++;
1237      }      }
1238    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1239      {      {
1240      c = *(++ptr);      c = *(++ptr);
1241      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1242      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1243      name[i] = c;      name[i] = c;
1244      }      }
1245    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1246    name[i] = 0;    name[i] = 0;
1247    }    }
1248    
# Line 834  else Line 1259  else
1259  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1260    
1261  bot = 0;  bot = 0;
1262  top = _pcre_utt_size;  top = PRIV(utt_size);
1263    
1264  while (bot < top)  while (bot < top)
1265    {    {
1266      int r;
1267    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1268    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1269    if (c == 0)    if (r == 0)
1270      {      {
1271      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1272      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1273        return TRUE;
1274      }      }
1275    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1276    }    }
1277    
1278  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1279  *ptrptr = ptr;  *ptrptr = ptr;
1280  return -1;  return FALSE;
1281    
1282  ERROR_RETURN:  ERROR_RETURN:
1283  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1284  *ptrptr = ptr;  *ptrptr = ptr;
1285  return -1;  return FALSE;
1286  }  }
1287  #endif  #endif
1288    
# Line 863  return -1; Line 1290  return -1;
1290    
1291    
1292  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1293  *         Read repeat counts                     *  *         Read repeat counts                     *
1294  *************************************************/  *************************************************/
1295    
# Line 914  Returns:         pointer to '}' on succe Line 1308  Returns:         pointer to '}' on succe
1308                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1309  */  */
1310    
1311  static const uschar *  static const pcre_uchar *
1312  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)
1313  {  {
1314  int min = 0;  int min = 0;
1315  int max = -1;  int max = -1;
# Line 923  int max = -1; Line 1317  int max = -1;
1317  /* 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
1318  an integer overflow. */  an integer overflow. */
1319    
1320  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1321  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1322    {    {
1323    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1327  if (min < 0 || min > 65535)
1327  /* 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.
1328  Also, max must not be less than min. */  Also, max must not be less than min. */
1329    
1330  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1331    {    {
1332    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1333      {      {
1334      max = 0;      max = 0;
1335      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1336      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1337        {        {
1338        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1357  return p;
1357    
1358    
1359  /*************************************************  /*************************************************
1360  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1361  *************************************************/  *************************************************/
1362    
1363  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1364    top-level call starts at the beginning of the pattern. All other calls must
1365    start at a parenthesis. It scans along a pattern's text looking for capturing
1366  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
1367  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
1368  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
1369  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1370  be terminated by '>' because that is checked in the first pass.  
1371    This function was originally called only from the second pass, in which we know
1372    that if (?< or (?' or (?P< is encountered, the name will be correctly
1373    terminated because that is checked in the first pass. There is now one call to
1374    this function in the first pass, to check for a recursive back reference by
1375    name (so that we can make the whole group atomic). In this case, we need check
1376    only up to the current position in the pattern, and that is still OK because
1377    and previous occurrences will have been checked. To make this work, the test
1378    for "end of pattern" is a check against cd->end_pattern in the main loop,
1379    instead of looking for a binary zero. This means that the special first-pass
1380    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1381    processing items within the loop are OK, because afterwards the main loop will
1382    terminate.)
1383    
1384  Arguments:  Arguments:
1385    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1386    count        current count of capturing parens so far encountered    cd           compile background data
1387    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1388    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1389    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1390      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1391      count        pointer to the current capturing subpattern number (updated)
1392    
1393  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1394  */  */
1395    
1396  static int  static int
1397  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,
1398    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1399  {  {
1400  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1401    int start_count = *count;
1402    int hwm_count = start_count;
1403    BOOL dup_parens = FALSE;
1404    
1405  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1406    dealing with. The very first call may not start with a parenthesis. */
1407    
1408    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1409    {    {
1410    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1411    
1412      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1413    
1414      /* Handle a normal, unnamed capturing parenthesis. */
1415    
1416      else if (ptr[1] != CHAR_QUESTION_MARK)
1417        {
1418        *count += 1;
1419        if (name == NULL && *count == lorn) return *count;
1420        ptr++;
1421        }
1422    
1423      /* All cases now have (? at the start. Remember when we are in a group
1424      where the parenthesis numbers are duplicated. */
1425    
1426      else if (ptr[2] == CHAR_VERTICAL_LINE)
1427        {
1428        ptr += 3;
1429        dup_parens = TRUE;
1430        }
1431    
1432      /* Handle comments; all characters are allowed until a ket is reached. */
1433    
1434      else if (ptr[2] == CHAR_NUMBER_SIGN)
1435        {
1436        for (ptr += 3; *ptr != CHAR_NULL; ptr++)
1437          if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1438        goto FAIL_EXIT;
1439        }
1440    
1441      /* Handle a condition. If it is an assertion, just carry on so that it
1442      is processed as normal. If not, skip to the closing parenthesis of the
1443      condition (there can't be any nested parens). */
1444    
1445      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1446        {
1447        ptr += 2;
1448        if (ptr[1] != CHAR_QUESTION_MARK)
1449          {
1450          while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1451          if (*ptr != CHAR_NULL) ptr++;
1452          }
1453        }
1454    
1455      /* Start with (? but not a condition. */
1456    
1457      else
1458        {
1459        ptr += 2;
1460        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1461    
1462        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1463    
1464        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1465            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1466          {
1467          pcre_uchar term;
1468          const pcre_uchar *thisname;
1469          *count += 1;
1470          if (name == NULL && *count == lorn) return *count;
1471          term = *ptr++;
1472          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1473          thisname = ptr;
1474          while (*ptr != term) ptr++;
1475          if (name != NULL && lorn == (int)(ptr - thisname) &&
1476              STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
1477            return *count;
1478          term++;
1479          }
1480        }
1481      }
1482    
1483    /* Past any initial parenthesis handling, scan for parentheses or vertical
1484    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1485    first-pass call when this value is temporarily adjusted to stop at the current
1486    position. So DO NOT change this to a test for binary zero. */
1487    
1488    for (; ptr < cd->end_pattern; ptr++)
1489      {
1490    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1491    
1492    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1493      {      {
1494      if (*(++ptr) == 0) return -1;      if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1495      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1496        {        {
1497        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1498        if (*ptr == 0) return -1;        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1499        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1500        }        }
1501      continue;      continue;
1502      }      }
1503    
1504    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1505      are handled for real. If the first character is '^', skip it. Also, if the
1506      first few characters (either before or after ^) are \Q\E or \E we skip them
1507      too. This makes for compatibility with Perl. Note the use of STR macros to
1508      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1509    
1510    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1511      {      {
1512      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1513        for (;;)
1514          {
1515          if (ptr[1] == CHAR_BACKSLASH)
1516            {
1517            if (ptr[2] == CHAR_E)
1518              ptr+= 2;
1519            else if (STRNCMP_UC_C8(ptr + 2,
1520                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1521              ptr += 4;
1522            else
1523              break;
1524            }
1525          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1526            {
1527            negate_class = TRUE;
1528            ptr++;
1529            }
1530          else break;
1531          }
1532    
1533        /* If the next character is ']', it is a data character that must be
1534        skipped, except in JavaScript compatibility mode. */
1535    
1536        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1537            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1538          ptr++;
1539    
1540        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1541        {        {
1542        if (*ptr == 0) return -1;        if (*ptr == CHAR_NULL) return -1;
1543        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1544          {          {
1545          if (*(++ptr) == 0) return -1;          if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1546          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1547            {            {
1548            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1549            if (*ptr == 0) return -1;            if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1550            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1551            }            }
1552          continue;          continue;
1553          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1557  for (; *ptr != 0; ptr++)
1557    
1558    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1559    
1560    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1561      {      {
1562      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1563      if (*ptr == 0) return -1;      while (*ptr != CHAR_NULL)
1564          {
1565          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1566          ptr++;
1567    #ifdef SUPPORT_UTF
1568          if (utf) FORWARDCHAR(ptr);
1569    #endif
1570          }
1571        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1572      continue;      continue;
1573      }      }
1574    
1575    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1576    
1577    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1578      {      {
1579      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1580      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1581      continue;      if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1582      }      }
1583    
1584    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1585    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1586        if (dup_parens && *count < hwm_count) *count = hwm_count;
1587    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1588        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1589    
1590    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1591    term = *ptr++;      {
1592    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1593    thisname = ptr;      *count = start_count;
1594    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1595    }    }
1596    
1597    FAIL_EXIT:
1598    *ptrptr = ptr;
1599  return -1;  return -1;
1600  }  }
1601    
1602    
1603    
1604    
1605  /*************************************************  /*************************************************
1606  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1607  *************************************************/  *************************************************/
1608    
1609  /* This is called by several functions that scan a compiled expression looking  /* This function scans along a pattern's text looking for capturing
1610  for a fixed first character, or an anchoring op code etc. It skips over things  subpatterns, and counting them. If it finds a named pattern that matches the
1611  that do not influence this. For some calls, a change of option is important.  name it is given, it returns its number. Alternatively, if the name is NULL, it
1612  For some calls, it makes sense to skip negative forward and all backward  returns when it reaches a given numbered subpattern. This is used for forward
1613  assertions, and also the \b assertion; for others it does not.  references to subpatterns. We used to be able to start this scan from the
1614    current compiling point, using the current count value from cd->bracount, and
1615    do it all in a single loop, but the addition of the possibility of duplicate
1616    subpattern numbers means that we have to scan from the very start, in order to
1617    take account of such duplicates, and to use a recursive function to keep track
1618    of the different types of group.
1619    
1620  Arguments:  Arguments:
1621    code         pointer to the start of the group    cd           compile background data
1622    options      pointer to external options    name         name to seek, or NULL if seeking a numbered subpattern
1623    optbit       the option bit whose changing is significant, or    lorn         name length, or subpattern number if name is NULL
1624                   zero if none are    xmode        TRUE if we are in /x mode
1625    skipassert   TRUE if certain assertions are to be skipped    utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1626    
1627    Returns:       the number of the found subpattern, or -1 if not found
1628    */
1629    
1630    static int
1631    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1632      BOOL utf)
1633    {
1634    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1635    int count = 0;
1636    int rc;
1637    
1638    /* If the pattern does not start with an opening parenthesis, the first call
1639    to find_parens_sub() will scan right to the end (if necessary). However, if it
1640    does start with a parenthesis, find_parens_sub() will return when it hits the
1641    matching closing parens. That is why we have to have a loop. */
1642    
1643    for (;;)
1644      {
1645      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1646      if (rc > 0 || *ptr++ == CHAR_NULL) break;
1647      }
1648    
1649    return rc;
1650    }
1651    
1652    
1653    
1654    
1655    /*************************************************
1656    *      Find first significant op code            *
1657    *************************************************/
1658    
1659    /* This is called by several functions that scan a compiled expression looking
1660    for a fixed first character, or an anchoring op code etc. It skips over things
1661    that do not influence this. For some calls, it makes sense to skip negative
1662    forward and all backward assertions, and also the \b assertion; for others it
1663    does not.
1664    
1665    Arguments:
1666      code         pointer to the start of the group
1667      skipassert   TRUE if certain assertions are to be skipped
1668    
1669  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1670  */  */
1671    
1672  static const uschar*  static const pcre_uchar*
1673  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1674  {  {
1675  for (;;)  for (;;)
1676    {    {
1677    switch ((int)*code)    switch ((int)*code)
1678      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1679      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1680      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1681      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1682      if (!skipassert) return code;      if (!skipassert) return code;
1683      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1684      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1685      break;      break;
1686    
1687      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1123  for (;;) Line 1691  for (;;)
1691    
1692      case OP_CALLOUT:      case OP_CALLOUT:
1693      case OP_CREF:      case OP_CREF:
1694        case OP_NCREF:
1695      case OP_RREF:      case OP_RREF:
1696        case OP_NRREF:
1697      case OP_DEF:      case OP_DEF:
1698      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1699      break;      break;
1700    
1701      default:      default:
# Line 1139  for (;;) Line 1709  for (;;)
1709    
1710    
1711  /*************************************************  /*************************************************
1712  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1713  *************************************************/  *************************************************/
1714    
1715  /* 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,
1716  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.
1717  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
1718    temporarily terminated with OP_END when this function is called.
1719    
1720    This function is called when a backward assertion is encountered, so that if it
1721    fails, the error message can point to the correct place in the pattern.
1722    However, we cannot do this when the assertion contains subroutine calls,
1723    because they can be forward references. We solve this by remembering this case
1724    and doing the check at the end; a flag specifies which mode we are running in.
1725    
1726  Arguments:  Arguments:
1727    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1728    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1729      atend    TRUE if called when the pattern is complete
1730  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1731               or -2 if \C was encountered  
1732    Returns:   the fixed length,
1733                 or -1 if there is no fixed length,
1734                 or -2 if \C was encountered (in UTF-8 mode only)
1735                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1736                 or -4 if an unknown opcode was encountered (internal error)
1737  */  */
1738    
1739  static int  static int
1740  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1741  {  {
1742  int length = -1;  int length = -1;
1743    
1744  register int branchlength = 0;  register int branchlength = 0;
1745  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1746    
1747  /* 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
1748  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1168  branch, check the length against that of Line 1750  branch, check the length against that of
1750  for (;;)  for (;;)
1751    {    {
1752    int d;    int d;
1753    register int op = *cc;    pcre_uchar *ce, *cs;
1754      register pcre_uchar op = *cc;
1755    
1756    switch (op)    switch (op)
1757      {      {
1758        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1759        OP_BRA (normal non-capturing bracket) because the other variants of these
1760        opcodes are all concerned with unlimited repeated groups, which of course
1761        are not of fixed length. */
1762    
1763      case OP_CBRA:      case OP_CBRA:
1764      case OP_BRA:      case OP_BRA:
1765      case OP_ONCE:      case OP_ONCE:
1766        case OP_ONCE_NC:
1767      case OP_COND:      case OP_COND:
1768      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1769      if (d < 0) return d;      if (d < 0) return d;
1770      branchlength += d;      branchlength += d;
1771      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1772      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1773      break;      break;
1774    
1775      /* 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.
1776      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
1777      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
1778        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1779        because they all imply an unlimited repeat. */
1780    
1781      case OP_ALT:      case OP_ALT:
1782      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1783      case OP_END:      case OP_END:
1784        case OP_ACCEPT:
1785        case OP_ASSERT_ACCEPT:
1786      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1787        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1788      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1790  for (;;)
1790      branchlength = 0;      branchlength = 0;
1791      break;      break;
1792    
1793        /* A true recursion implies not fixed length, but a subroutine call may
1794        be OK. If the subroutine is a forward reference, we can't deal with
1795        it until the end of the pattern, so return -3. */
1796    
1797        case OP_RECURSE:
1798        if (!atend) return -3;
1799        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1800        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1801        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1802        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1803        if (d < 0) return d;
1804        branchlength += d;
1805        cc += 1 + LINK_SIZE;
1806        break;
1807    
1808      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1809    
1810      case OP_ASSERT:      case OP_ASSERT:
# Line 1205  for (;;) Line 1812  for (;;)
1812      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1813      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1814      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1815      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1816        break;
1817    
1818      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1819    
1820      case OP_REVERSE:      case OP_MARK:
1821        case OP_PRUNE_ARG:
1822        case OP_SKIP_ARG:
1823        case OP_THEN_ARG:
1824        cc += cc[1] + PRIV(OP_lengths)[*cc];
1825        break;
1826    
1827        case OP_CALLOUT:
1828        case OP_CIRC:
1829        case OP_CIRCM:
1830        case OP_CLOSE:
1831        case OP_COMMIT:
1832      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1833      case OP_DEF:      case OP_DEF:
1834      case OP_OPT:      case OP_DOLL:
1835      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1836      case OP_EOD:      case OP_EOD:
1837      case OP_EODN:      case OP_EODN:
1838      case OP_CIRC:      case OP_FAIL:
1839      case OP_DOLL:      case OP_NCREF:
1840        case OP_NRREF:
1841      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1842        case OP_PRUNE:
1843        case OP_REVERSE:
1844        case OP_RREF:
1845        case OP_SET_SOM:
1846        case OP_SKIP:
1847        case OP_SOD:
1848        case OP_SOM:
1849        case OP_THEN:
1850      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1851      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1852      break;      break;
1853    
1854      /* Handle literal characters */      /* Handle literal characters */
1855    
1856      case OP_CHAR:      case OP_CHAR:
1857      case OP_CHARNC:      case OP_CHARI:
1858      case OP_NOT:      case OP_NOT:
1859        case OP_NOTI:
1860      branchlength++;      branchlength++;
1861      cc += 2;      cc += 2;
1862  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1863      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1864  #endif  #endif
1865      break;      break;
1866    
# Line 1245  for (;;) Line 1868  for (;;)
1868      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1869    
1870      case OP_EXACT:      case OP_EXACT:
1871      branchlength += GET2(cc,1);      case OP_EXACTI:
1872      cc += 4;      case OP_NOTEXACT:
1873  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1874      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1875        {      cc += 2 + IMM2_SIZE;
1876        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1877        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1878  #endif  #endif
1879      break;      break;
1880    
1881      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1882      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1883      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1884      cc += 4;        cc += 2;
1885        cc += 1 + IMM2_SIZE + 1;
1886      break;      break;
1887    
1888      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1268  for (;;) Line 1892  for (;;)
1892      cc += 2;      cc += 2;
1893      /* Fall through */      /* Fall through */
1894    
1895        case OP_HSPACE:
1896        case OP_VSPACE:
1897        case OP_NOT_HSPACE:
1898        case OP_NOT_VSPACE:
1899      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1900      case OP_DIGIT:      case OP_DIGIT:
1901      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1903  for (;;)
1903      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1904      case OP_WORDCHAR:      case OP_WORDCHAR:
1905      case OP_ANY:      case OP_ANY:
1906        case OP_ALLANY:
1907      branchlength++;      branchlength++;
1908      cc++;      cc++;
1909      break;      break;
1910    
1911      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1912        otherwise \C is coded as OP_ALLANY. */
1913    
1914      case OP_ANYBYTE:      case OP_ANYBYTE:
1915      return -2;      return -2;
1916    
1917      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1918    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1919      case OP_CLASS:      case OP_CLASS:
1920      case OP_NCLASS:      case OP_NCLASS:
1921      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1922        case OP_XCLASS:
1923        /* The original code caused an unsigned overflow in 64 bit systems,
1924        so now we use a conditional statement. */
1925        if (op == OP_XCLASS)
1926          cc += GET(cc, 1);
1927        else
1928          cc += PRIV(OP_lengths)[OP_CLASS];
1929    #else
1930        cc += PRIV(OP_lengths)[OP_CLASS];
1931    #endif
1932    
1933      switch (*cc)      switch (*cc)
1934        {        {
1935          case OP_CRPLUS:
1936          case OP_CRMINPLUS:
1937        case OP_CRSTAR:        case OP_CRSTAR:
1938        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1939        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1306  for (;;) Line 1942  for (;;)
1942    
1943        case OP_CRRANGE:        case OP_CRRANGE:
1944        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1945        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1946        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1947        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1948        break;        break;
1949    
1950        default:        default:
# Line 1318  for (;;) Line 1954  for (;;)
1954    
1955      /* Anything else is variable length */      /* Anything else is variable length */
1956    
1957      default:      case OP_ANYNL:
1958        case OP_BRAMINZERO:
1959        case OP_BRAPOS:
1960        case OP_BRAPOSZERO:
1961        case OP_BRAZERO:
1962        case OP_CBRAPOS:
1963        case OP_EXTUNI:
1964        case OP_KETRMAX:
1965        case OP_KETRMIN:
1966        case OP_KETRPOS:
1967        case OP_MINPLUS:
1968        case OP_MINPLUSI:
1969        case OP_MINQUERY:
1970        case OP_MINQUERYI:
1971        case OP_MINSTAR:
1972        case OP_MINSTARI:
1973        case OP_MINUPTO:
1974        case OP_MINUPTOI:
1975        case OP_NOTMINPLUS:
1976        case OP_NOTMINPLUSI:
1977        case OP_NOTMINQUERY:
1978        case OP_NOTMINQUERYI:
1979        case OP_NOTMINSTAR:
1980        case OP_NOTMINSTARI:
1981        case OP_NOTMINUPTO:
1982        case OP_NOTMINUPTOI:
1983        case OP_NOTPLUS:
1984        case OP_NOTPLUSI:
1985        case OP_NOTPOSPLUS:
1986        case OP_NOTPOSPLUSI:
1987        case OP_NOTPOSQUERY:
1988        case OP_NOTPOSQUERYI:
1989        case OP_NOTPOSSTAR:
1990        case OP_NOTPOSSTARI:
1991        case OP_NOTPOSUPTO:
1992        case OP_NOTPOSUPTOI:
1993        case OP_NOTQUERY:
1994        case OP_NOTQUERYI:
1995        case OP_NOTSTAR:
1996        case OP_NOTSTARI:
1997        case OP_NOTUPTO:
1998        case OP_NOTUPTOI:
1999        case OP_PLUS:
2000        case OP_PLUSI:
2001        case OP_POSPLUS:
2002        case OP_POSPLUSI:
2003        case OP_POSQUERY:
2004        case OP_POSQUERYI:
2005        case OP_POSSTAR:
2006        case OP_POSSTARI:
2007        case OP_POSUPTO:
2008        case OP_POSUPTOI:
2009        case OP_QUERY:
2010        case OP_QUERYI:
2011        case OP_REF:
2012        case OP_REFI:
2013        case OP_SBRA:
2014        case OP_SBRAPOS:
2015        case OP_SCBRA:
2016        case OP_SCBRAPOS:
2017        case OP_SCOND:
2018        case OP_SKIPZERO:
2019        case OP_STAR:
2020        case OP_STARI:
2021        case OP_TYPEMINPLUS:
2022        case OP_TYPEMINQUERY:
2023        case OP_TYPEMINSTAR:
2024        case OP_TYPEMINUPTO:
2025        case OP_TYPEPLUS:
2026        case OP_TYPEPOSPLUS:
2027        case OP_TYPEPOSQUERY:
2028        case OP_TYPEPOSSTAR:
2029        case OP_TYPEPOSUPTO:
2030        case OP_TYPEQUERY:
2031        case OP_TYPESTAR:
2032        case OP_TYPEUPTO:
2033        case OP_UPTO:
2034        case OP_UPTOI:
2035      return -1;      return -1;
2036    
2037        /* Catch unrecognized opcodes so that when new ones are added they
2038        are not forgotten, as has happened in the past. */
2039    
2040        default:
2041        return -4;
2042      }      }
2043    }    }
2044  /* Control never gets here */  /* Control never gets here */
# Line 1329  for (;;) Line 2048  for (;;)
2048    
2049    
2050  /*************************************************  /*************************************************
2051  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2052  *************************************************/  *************************************************/
2053    
2054  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2055  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2056    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2057    so that it can be called from pcre_study() when finding the minimum matching
2058    length.
2059    
2060  Arguments:  Arguments:
2061    code        points to start of expression    code        points to start of expression
2062    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2063    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2064    
2065  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
2066  */  */
2067    
2068  static const uschar *  const pcre_uchar *
2069  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2070  {  {
2071  for (;;)  for (;;)
2072    {    {
2073    register int c = *code;    register pcre_uchar c = *code;
2074    
2075    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2076    
2077    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1357  for (;;) Line 2080  for (;;)
2080    
2081    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2082    
2083      /* Handle recursion */
2084    
2085      else if (c == OP_REVERSE)
2086        {
2087        if (number < 0) return (pcre_uchar *)code;
2088        code += PRIV(OP_lengths)[c];
2089        }
2090    
2091    /* Handle capturing bracket */    /* Handle capturing bracket */
2092    
2093    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2094               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2095      {      {
2096      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2097      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2098      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2099      }      }
2100    
2101    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2102    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2103    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2104      must add in its length. */
2105    
2106    else    else
2107      {      {
# Line 1390  for (;;) Line 2123  for (;;)
2123        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2124        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2125        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2126        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2127            code += 2;
2128          break;
2129    
2130          case OP_MARK:
2131          case OP_PRUNE_ARG:
2132          case OP_SKIP_ARG:
2133          code += code[1];
2134          break;
2135    
2136          case OP_THEN_ARG:
2137          code += code[1];
2138        break;        break;
2139        }        }
2140    
2141      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2142    
2143      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2144    
2145    /* 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
2146    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
2147    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2148    
2149  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2150      if (utf8) switch(c)      if (utf) switch(c)
2151        {        {
2152        case OP_CHAR:        case OP_CHAR:
2153        case OP_CHARNC:        case OP_CHARI:
2154        case OP_EXACT:        case OP_EXACT:
2155          case OP_EXACTI:
2156        case OP_UPTO:        case OP_UPTO:
2157          case OP_UPTOI:
2158        case OP_MINUPTO:        case OP_MINUPTO:
2159          case OP_MINUPTOI:
2160        case OP_POSUPTO:        case OP_POSUPTO:
2161          case OP_POSUPTOI:
2162        case OP_STAR:        case OP_STAR:
2163          case OP_STARI:
2164        case OP_MINSTAR:        case OP_MINSTAR:
2165          case OP_MINSTARI:
2166        case OP_POSSTAR:        case OP_POSSTAR:
2167          case OP_POSSTARI:
2168        case OP_PLUS:        case OP_PLUS:
2169          case OP_PLUSI:
2170        case OP_MINPLUS:        case OP_MINPLUS:
2171          case OP_MINPLUSI:
2172        case OP_POSPLUS:        case OP_POSPLUS:
2173          case OP_POSPLUSI:
2174        case OP_QUERY:        case OP_QUERY:
2175          case OP_QUERYI:
2176        case OP_MINQUERY:        case OP_MINQUERY:
2177          case OP_MINQUERYI:
2178        case OP_POSQUERY:        case OP_POSQUERY:
2179        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2180          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2181        break;        break;
2182        }        }
2183    #else
2184        (void)(utf);  /* Keep compiler happy by referencing function argument */
2185  #endif  #endif
2186      }      }
2187    }    }
# Line 1439  instance of OP_RECURSE. Line 2198  instance of OP_RECURSE.
2198    
2199  Arguments:  Arguments:
2200    code        points to start of expression    code        points to start of expression
2201    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2202    
2203  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
2204  */  */
2205    
2206  static const uschar *  static const pcre_uchar *
2207  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2208  {  {
2209  for (;;)  for (;;)
2210    {    {
2211    register int c = *code;    register pcre_uchar c = *code;
2212    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2213    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2214    
# Line 1461  for (;;) Line 2220  for (;;)
2220    
2221    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2222    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2223    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2224      must add in its length. */
2225    
2226    else    else
2227      {      {
# Line 1483  for (;;) Line 2243  for (;;)
2243        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2244        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2245        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2246        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2247            code += 2;
2248          break;
2249    
2250          case OP_MARK:
2251          case OP_PRUNE_ARG:
2252          case OP_SKIP_ARG:
2253          code += code[1];
2254          break;
2255    
2256          case OP_THEN_ARG:
2257          code += code[1];
2258        break;        break;
2259        }        }
2260    
2261      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2262    
2263      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2264    
2265      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2266      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2267      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2268    
2269  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2270      if (utf8) switch(c)      if (utf) switch(c)
2271        {        {
2272        case OP_CHAR:        case OP_CHAR:
2273        case OP_CHARNC:        case OP_CHARI:
2274          case OP_NOT:
2275          case OP_NOTI:
2276        case OP_EXACT:        case OP_EXACT:
2277          case OP_EXACTI:
2278          case OP_NOTEXACT:
2279          case OP_NOTEXACTI:
2280        case OP_UPTO:        case OP_UPTO:
2281          case OP_UPTOI:
2282          case OP_NOTUPTO:
2283          case OP_NOTUPTOI:
2284        case OP_MINUPTO:        case OP_MINUPTO:
2285          case OP_MINUPTOI:
2286          case OP_NOTMINUPTO:
2287          case OP_NOTMINUPTOI:
2288        case OP_POSUPTO:        case OP_POSUPTO:
2289          case OP_POSUPTOI:
2290          case OP_NOTPOSUPTO:
2291          case OP_NOTPOSUPTOI:
2292        case OP_STAR:        case OP_STAR:
2293          case OP_STARI:
2294          case OP_NOTSTAR:
2295          case OP_NOTSTARI:
2296        case OP_MINSTAR:        case OP_MINSTAR:
2297          case OP_MINSTARI:
2298          case OP_NOTMINSTAR:
2299          case OP_NOTMINSTARI:
2300        case OP_POSSTAR:        case OP_POSSTAR:
2301          case OP_POSSTARI:
2302          case OP_NOTPOSSTAR:
2303          case OP_NOTPOSSTARI:
2304        case OP_PLUS:        case OP_PLUS:
2305          case OP_PLUSI:
2306          case OP_NOTPLUS:
2307          case OP_NOTPLUSI:
2308        case OP_MINPLUS:        case OP_MINPLUS:
2309          case OP_MINPLUSI:
2310          case OP_NOTMINPLUS:
2311          case OP_NOTMINPLUSI:
2312        case OP_POSPLUS:        case OP_POSPLUS:
2313          case OP_POSPLUSI:
2314          case OP_NOTPOSPLUS:
2315          case OP_NOTPOSPLUSI:
2316        case OP_QUERY:        case OP_QUERY:
2317          case OP_QUERYI:
2318          case OP_NOTQUERY:
2319          case OP_NOTQUERYI:
2320        case OP_MINQUERY:        case OP_MINQUERY:
2321          case OP_MINQUERYI:
2322          case OP_NOTMINQUERY:
2323          case OP_NOTMINQUERYI:
2324        case OP_POSQUERY:        case OP_POSQUERY:
2325        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2326          case OP_NOTPOSQUERY:
2327          case OP_NOTPOSQUERYI:
2328          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2329        break;        break;
2330        }        }
2331    #else
2332        (void)(utf);  /* Keep compiler happy by referencing function argument */
2333  #endif  #endif
2334      }      }
2335    }    }
# Line 1538  bracket whose current branch will alread Line 2352  bracket whose current branch will alread
2352  Arguments:  Arguments:
2353    code        points to start of search    code        points to start of search
2354    endcode     points to where to stop    endcode     points to where to stop
2355    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2356      cd          contains pointers to tables etc.
2357    
2358  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2359  */  */
2360    
2361  static BOOL  static BOOL
2362  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2363      BOOL utf, compile_data *cd)
2364  {  {
2365  register int c;  register pcre_uchar c;
2366  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2367       code < endcode;       code < endcode;
2368       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2369    {    {
2370    const uschar *ccode;    const pcre_uchar *ccode;
2371    
2372    c = *code;    c = *code;
2373    
# Line 1565  for (code = first_significant_code(code Line 2381  for (code = first_significant_code(code
2381      continue;      continue;
2382      }      }
2383    
2384      /* For a recursion/subroutine call, if its end has been reached, which
2385      implies a backward reference subroutine call, we can scan it. If it's a
2386      forward reference subroutine call, we can't. To detect forward reference
2387      we have to scan up the list that is kept in the workspace. This function is
2388      called only when doing the real compile, not during the pre-compile that
2389      measures the size of the compiled pattern. */
2390    
2391      if (c == OP_RECURSE)
2392        {
2393        const pcre_uchar *scode;
2394        BOOL empty_branch;
2395    
2396        /* Test for forward reference */
2397    
2398        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2399          if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2400    
2401        /* Not a forward reference, test for completed backward reference */
2402    
2403        empty_branch = FALSE;
2404        scode = cd->start_code + GET(code, 1);
2405        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2406    
2407        /* Completed backwards reference */
2408    
2409        do
2410          {
2411          if (could_be_empty_branch(scode, endcode, utf, cd))
2412            {
2413            empty_branch = TRUE;
2414            break;
2415            }
2416          scode += GET(scode, 1);
2417          }
2418        while (*scode == OP_ALT);
2419    
2420        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2421        continue;
2422        }
2423    
2424    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2425    
2426    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2427          c == OP_BRAPOSZERO)
2428        {
2429        code += PRIV(OP_lengths)[c];
2430        do code += GET(code, 1); while (*code == OP_ALT);
2431        c = *code;
2432        continue;
2433        }
2434    
2435      /* A nested group that is already marked as "could be empty" can just be
2436      skipped. */
2437    
2438      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2439          c == OP_SCBRA || c == OP_SCBRAPOS)
2440      {      {
     code += _pcre_OP_lengths[c];  
2441      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2442      c = *code;      c = *code;
2443      continue;      continue;
# Line 1577  for (code = first_significant_code(code Line 2445  for (code = first_significant_code(code
2445    
2446    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2447    
2448    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2449          c == OP_CBRA || c == OP_CBRAPOS ||
2450          c == OP_ONCE || c == OP_ONCE_NC ||
2451          c == OP_COND)
2452      {      {
2453      BOOL empty_branch;      BOOL empty_branch;
2454      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2455    
2456      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2457        empty branch, so just skip over the conditional, because it could be empty.
2458        Otherwise, scan the individual branches of the group. */
2459    
2460      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
2461        code += GET(code, 1);        code += GET(code, 1);
2462        else
2463          {
2464          empty_branch = FALSE;
2465          do
2466            {
2467            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2468              empty_branch = TRUE;
2469            code += GET(code, 1);
2470            }
2471          while (*code == OP_ALT);
2472          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2473        }        }
2474      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2475      c = *code;      c = *code;
2476      continue;      continue;
2477      }      }
# Line 1603  for (code = first_significant_code(code Line 2482  for (code = first_significant_code(code
2482      {      {
2483      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2484      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2485      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2486      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2487      here. */      here. */
2488    
2489  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2490      case OP_XCLASS:      case OP_XCLASS:
2491      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2492      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1615  for (code = first_significant_code(code Line 2494  for (code = first_significant_code(code
2494    
2495      case OP_CLASS:      case OP_CLASS:
2496      case OP_NCLASS:      case OP_NCLASS:
2497      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2498    
2499  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2500      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2501  #endif  #endif
2502    
# Line 1653  for (code = first_significant_code(code Line 2532  for (code = first_significant_code(code
2532      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2533      case OP_WORDCHAR:      case OP_WORDCHAR:
2534      case OP_ANY:      case OP_ANY:
2535        case OP_ALLANY:
2536      case OP_ANYBYTE:      case OP_ANYBYTE:
2537      case OP_CHAR:      case OP_CHAR:
2538      case OP_CHARNC:      case OP_CHARI:
2539      case OP_NOT:      case OP_NOT:
2540        case OP_NOTI:
2541      case OP_PLUS:      case OP_PLUS:
2542      case OP_MINPLUS:      case OP_MINPLUS:
2543      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1688  for (code = first_significant_code(code Line 2569  for (code = first_significant_code(code
2569      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2570      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2571      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2572      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2573          code += 2;
2574      break;      break;
2575    
2576      /* End of branch */      /* End of branch */
# Line 1696  for (code = first_significant_code(code Line 2578  for (code = first_significant_code(code
2578      case OP_KET:      case OP_KET:
2579      case OP_KETRMAX:      case OP_KETRMAX:
2580      case OP_KETRMIN:      case OP_KETRMIN:
2581        case OP_KETRPOS:
2582      case OP_ALT:      case OP_ALT:
2583      return TRUE;      return TRUE;
2584    
2585      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2586      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2587    
2588  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2589      case OP_STAR:      case OP_STAR:
2590        case OP_STARI:
2591      case OP_MINSTAR:      case OP_MINSTAR:
2592        case OP_MINSTARI:
2593      case OP_POSSTAR:      case OP_POSSTAR:
2594        case OP_POSSTARI:
2595      case OP_QUERY:      case OP_QUERY:
2596        case OP_QUERYI:
2597      case OP_MINQUERY:      case OP_MINQUERY:
2598        case OP_MINQUERYI:
2599      case OP_POSQUERY:      case OP_POSQUERY:
2600        case OP_POSQUERYI:
2601        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2602        break;
2603    
2604      case OP_UPTO:      case OP_UPTO:
2605        case OP_UPTOI:
2606      case OP_MINUPTO:      case OP_MINUPTO:
2607        case OP_MINUPTOI:
2608      case OP_POSUPTO:      case OP_POSUPTO:
2609      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2610        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2611      break;      break;
2612  #endif  #endif
2613    
2614        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2615        string. */
2616    
2617        case OP_MARK:
2618        case OP_PRUNE_ARG:
2619        case OP_SKIP_ARG:
2620        code += code[1];
2621        break;
2622    
2623        case OP_THEN_ARG:
2624        code += code[1];
2625        break;
2626    
2627        /* None of the remaining opcodes are required to match a character. */
2628    
2629        default:
2630        break;
2631      }      }
2632    }    }
2633    
# Line 1731  return TRUE; Line 2644  return TRUE;
2644  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
2645  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,
2646  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.
2647    This function is called only during the real compile, not during the
2648    pre-compile.
2649    
2650  Arguments:  Arguments:
2651    code        points to start of the recursion    code        points to start of the recursion
2652    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2653    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2654    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2655      cd          pointers to tables etc
2656    
2657  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2658  */  */
2659    
2660  static BOOL  static BOOL
2661  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2662    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2663  {  {
2664  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2665    {    {
2666    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2667        return FALSE;
2668    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2669    }    }
2670  return TRUE;  return TRUE;
# Line 1779  where Perl recognizes it as the POSIX cl Line 2696  where Perl recognizes it as the POSIX cl
2696  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2697  I think.  I think.
2698    
2699    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2700    It seems that the appearance of a nested POSIX class supersedes an apparent
2701    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2702    a digit.
2703    
2704    In Perl, unescaped square brackets may also appear as part of class names. For
2705    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2706    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2707    seem right at all. PCRE does not allow closing square brackets in POSIX class
2708    names.
2709    
2710  Arguments:  Arguments:
2711    ptr      pointer to the initial [    ptr      pointer to the initial [
2712    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1787  Returns:   TRUE or FALSE Line 2715  Returns:   TRUE or FALSE
2715  */  */
2716    
2717  static BOOL  static BOOL
2718  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2719  {  {
2720  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
2721  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2722  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != CHAR_NULL; ptr++)
2723    {    {
2724    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2725        ptr++;
2726      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2727      else
2728      {      {
2729      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2730        {        {
2731        *endptr = ptr;        *endptr = ptr;
2732        return TRUE;        return TRUE;
2733        }        }
2734        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2735             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2736              ptr[1] == CHAR_EQUALS_SIGN) &&
2737            check_posix_syntax(ptr, endptr))
2738          return FALSE;
2739      }      }
2740    }    }
2741  return FALSE;  return FALSE;
# Line 1824  Returns:     a value representing the na Line 2759  Returns:     a value representing the na
2759  */  */
2760    
2761  static int  static int
2762  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2763  {  {
2764  const char *pn = posix_names;  const char *pn = posix_names;
2765  register int yield = 0;  register int yield = 0;
2766  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2767    {    {
2768    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2769      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
2770    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2771    yield++;    yield++;
2772    }    }
# Line 1847  return -1; Line 2782  return -1;
2782  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2783  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2784  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
2785  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
2786  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
2787  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
2788  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
2789  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2790    OP_END.
2791    
2792  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2793  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 1862  value in the reference (which is a group Line 2798  value in the reference (which is a group
2798  Arguments:  Arguments:
2799    group      points to the start of the group    group      points to the start of the group
2800    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2801    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2802    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2803    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
2804    
# Line 1870  Returns:     nothing Line 2806  Returns:     nothing
2806  */  */
2807    
2808  static void  static void
2809  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2810    uschar *save_hwm)    pcre_uchar *save_hwm)
2811  {  {
2812  uschar *ptr = group;  pcre_uchar *ptr = group;
2813    
2814  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2815    {    {
2816    int offset;    int offset;
2817    uschar *hc;    pcre_uchar *hc;
2818    
2819    /* 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
2820    reference. */    reference. */
2821    
2822    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2823      {      {
2824      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
2825      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
2826        {        {
2827        PUT(hc, 0, offset + adjust);        PUT(hc, 0, offset + adjust);
# Line 1898  while ((ptr = (uschar *)find_recurse(ptr Line 2834  while ((ptr = (uschar *)find_recurse(ptr
2834    
2835    if (hc >= cd->hwm)    if (hc >= cd->hwm)
2836      {      {
2837      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
2838      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2839      }      }
2840    
# Line 1923  Arguments: Line 2859  Arguments:
2859  Returns:         new code pointer  Returns:         new code pointer
2860  */  */
2861    
2862  static uschar *  static pcre_uchar *
2863  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2864  {  {
2865  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2866  *code++ = 255;  *code++ = 255;
2867  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2868  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2869  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2870  }  }
2871    
2872    
# Line 1952  Returns:             nothing Line 2888  Returns:             nothing
2888  */  */
2889    
2890  static void  static void
2891  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2892  {  {
2893  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2894  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2895  }  }
2896    
# Line 1966  PUT(previous_callout, 2 + LINK_SIZE, len Line 2902  PUT(previous_callout, 2 + LINK_SIZE, len
2902  *************************************************/  *************************************************/
2903    
2904  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
2905  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
2906  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
2907  start address.  start address. A character with multiple other cases is returned on its own
2908    with a special return value.
2909    
2910  Arguments:  Arguments:
2911    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1976  Arguments: Line 2913  Arguments:
2913    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2914    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2915    
2916  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2917                   0 when a range is returned
2918                  >0 the CASESET offset for char with multiple other cases
2919                    in this case, ocptr contains the original
2920  */  */
2921    
2922  static BOOL  static int
2923  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
2924    unsigned int *odptr)    pcre_uint32 *odptr)
2925  {  {
2926  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
2927    unsigned int co;
2928    
2929    /* Find the first character that has an other case. If it has multiple other
2930    cases, return its case offset value. */
2931    
2932  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2933    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2934      if ((co = UCD_CASESET(c)) != 0)
2935        {
2936        *ocptr = c++;   /* Character that has the set */
2937        *cptr = c;      /* Rest of input range */
2938        return (int)co;
2939        }
2940      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2941      }
2942    
2943  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2944    
2945  *ocptr = othercase;  *ocptr = othercase;
2946  next = othercase + 1;  next = othercase + 1;
2947    
2948  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2949    {    {
2950    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2951    next++;    next++;
2952    }    }
2953    
2954  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2955  *cptr = c;  *cptr = c;             /* Rest of input range */
2956    return 0;
2957    }
2958    
2959  return TRUE;  
2960    
2961    /*************************************************
2962    *        Check a character and a property        *
2963    *************************************************/
2964    
2965    /* This function is called by check_auto_possessive() when a property item
2966    is adjacent to a fixed character.
2967    
2968    Arguments:
2969      c            the character
2970      ptype        the property type
2971      pdata        the data for the type
2972      negated      TRUE if it's a negated property (\P or \p{^)
2973    
2974    Returns:       TRUE if auto-possessifying is OK
2975    */
2976    
2977    static BOOL
2978    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)
2979    {
2980    #ifdef SUPPORT_UCP
2981    const pcre_uint32 *p;
2982    #endif
2983    
2984    const ucd_record *prop = GET_UCD(c);
2985    
2986    switch(ptype)
2987      {
2988      case PT_LAMP:
2989      return (prop->chartype == ucp_Lu ||
2990              prop->chartype == ucp_Ll ||
2991              prop->chartype == ucp_Lt) == negated;
2992    
2993      case PT_GC:
2994      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2995    
2996      case PT_PC:
2997      return (pdata == prop->chartype) == negated;
2998    
2999      case PT_SC:
3000      return (pdata == prop->script) == negated;
3001    
3002      /* These are specials */
3003    
3004      case PT_ALNUM:
3005      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3006              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
3007    
3008      case PT_SPACE:    /* Perl space */
3009      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3010              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
3011              == negated;
3012    
3013      case PT_PXSPACE:  /* POSIX space */
3014      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3015              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3016              c == CHAR_FF || c == CHAR_CR)
3017              == negated;
3018    
3019      case PT_WORD:
3020      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3021              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3022              c == CHAR_UNDERSCORE) == negated;
3023    
3024    #ifdef SUPPORT_UCP
3025      case PT_CLIST:
3026      p = PRIV(ucd_caseless_sets) + prop->caseset;
3027      for (;;)
3028        {
3029        if (c < *p) return !negated;
3030        if (c == *p++) return negated;
3031        }
3032      break;  /* Control never reaches here */
3033    #endif
3034      }
3035    
3036    return FALSE;
3037  }  }
3038  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3039    
# Line 2017  whether the next thing could possibly ma Line 3048  whether the next thing could possibly ma
3048  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3049    
3050  Arguments:  Arguments:
3051    op_code       the repeated op code    previous      pointer to the repeated opcode
3052    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
3053    ptr           next character in pattern    ptr           next character in pattern
3054    options       options bits    options       options bits
3055    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2029  Returns:        TRUE if possessifying is Line 3058  Returns:        TRUE if possessifying is
3058  */  */
3059    
3060  static BOOL  static BOOL
3061  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3062    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3063  {  {
3064  int next;  pcre_uint32 c = NOTACHAR;
3065    pcre_uint32 next;
3066    int escape;
3067    pcre_uchar op_code = *previous++;
3068    
3069  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3070    
# Line 2040  if ((options & PCRE_EXTENDED) != 0) Line 3072  if ((options & PCRE_EXTENDED) != 0)
3072    {    {
3073    for (;;)    for (;;)
3074      {      {
3075      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3076      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3077        {        {
3078        while (*(++ptr) != 0)        ptr++;
3079          while (*ptr != CHAR_NULL)
3080            {
3081          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3082            ptr++;
3083    #ifdef SUPPORT_UTF
3084            if (utf) FORWARDCHAR(ptr);
3085    #endif
3086            }
3087        }        }
3088      else break;      else break;
3089      }      }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 3092  if ((options & PCRE_EXTENDED) != 0)
3092  /* 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
3093  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3094    
3095  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3096    {    {
3097    int temperrorcode = 0;    int temperrorcode = 0;
3098    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,
3099        FALSE);
3100    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3101    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3102    }    }
3103    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3104    {    {
3105  #ifdef SUPPORT_UTF8    escape = 0;
3106    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3107      if (utf) { GETCHARINC(next, ptr); } else
3108  #endif  #endif
3109    next = *ptr++;    next = *ptr++;
3110    }    }
   
3111  else return FALSE;  else return FALSE;
3112    
3113  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2077  if ((options & PCRE_EXTENDED) != 0) Line 3116  if ((options & PCRE_EXTENDED) != 0)
3116    {    {
3117    for (;;)    for (;;)
3118      {      {
3119      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3120      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3121        {        {
3122        while (*(++ptr) != 0)        ptr++;
3123          while (*ptr != CHAR_NULL)
3124            {
3125          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3126            ptr++;
3127    #ifdef SUPPORT_UTF
3128            if (utf) FORWARDCHAR(ptr);
3129    #endif
3130            }
3131        }        }
3132      else break;      else break;
3133      }      }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 3135  if ((options & PCRE_EXTENDED) != 0)
3135    
3136  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3137    
3138  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3139    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3140        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. */  
   
3141    
3142  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3143    
3144  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3145        op_code == OP_NOT || op_code == OP_NOTI)
3146    {    {
3147    case OP_CHAR:  #ifdef SUPPORT_UTF
3148  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3149    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3150      c = *previous;
3151  #endif  #endif
3152    return item != next;    }
3153    
3154    /* Now compare the next item with the previous opcode. First, handle cases when
3155    the next item is a character. */
3156    
3157    /* For CHARNC (caseless character) we must check the other case. If we have  if (escape == 0)
3158    Unicode property support, we can use it to test the other case of    {
3159    high-valued characters. */    /* For a caseless UTF match, the next character may have more than one other
3160      case, which maps to the special PT_CLIST property. Check this first. */
3161    
3162    case OP_CHARNC:  #ifdef SUPPORT_UCP
3163  #ifdef SUPPORT_UTF8    if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3164    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }      {
3165        unsigned int ocs = UCD_CASESET(next);
3166        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3167        }
3168  #endif  #endif
3169    if (item == next) return FALSE;  
3170  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3171      {      {
3172      unsigned int othercase;      case OP_CHAR:
3173      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3174    
3175        /* For CHARI (caseless character) we must check the other case. If we have
3176        Unicode property support, we can use it to test the other case of
3177        high-valued characters. We know that next can have only one other case,
3178        because multi-other-case characters are dealt with above. */
3179    
3180        case OP_CHARI:
3181        if (c == next) return FALSE;
3182    #ifdef SUPPORT_UTF
3183        if (utf)
3184          {
3185          pcre_uint32 othercase;
3186          if (next < 128) othercase = cd->fcc[next]; else
3187  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3188      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE(next);
3189  #else  #else
3190      othercase = NOTACHAR;        othercase = NOTACHAR;
3191  #endif  #endif
3192      return (unsigned int)item != othercase;        return c != othercase;
3193      }        }
3194    else      else
3195  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3196    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3197    
3198    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3199        return c == next;
3200    
3201    case OP_NOT:      case OP_NOTI:
3202    if (item == next) return TRUE;      if (c == next) return TRUE;
3203    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3204  #ifdef SUPPORT_UTF8      if (utf)
3205    if (utf8)        {
3206      {        pcre_uint32 othercase;
3207      unsigned int othercase;        if (next < 128) othercase = cd->fcc[next]; else
     if (next < 128) othercase = cd->fcc[next]; else  
3208  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3209      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE(next);
3210  #else  #else
3211      othercase = NOTACHAR;        othercase = NOTACHAR;
3212  #endif  #endif
3213      return (unsigned int)item == othercase;        return c == othercase;
3214      }        }
3215    else      else
3216  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3217    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
   
   case OP_DIGIT:  
   return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;  
3218    
3219    case OP_NOT_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3220    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3221    
3222    case OP_WHITESPACE:      case OP_DIGIT:
3223    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3224    
3225    case OP_NOT_WHITESPACE:      case OP_NOT_DIGIT:
3226    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3227    
3228    case OP_WORDCHAR:      case OP_WHITESPACE:
3229    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3230    
3231    case OP_NOT_WORDCHAR:      case OP_NOT_WHITESPACE:
3232    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3233    
3234    case OP_HSPACE:      case OP_WORDCHAR:
3235    case OP_NOT_HSPACE:      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3236    switch(next)  
3237      {      case OP_NOT_WORDCHAR:
3238      case 0x09:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3239      case 0x20:  
3240      case 0xa0:      case OP_HSPACE:
3241      case 0x1680:      case OP_NOT_HSPACE:
3242      case 0x180e:      switch(next)
3243      case 0x2000:        {
3244      case 0x2001:        HSPACE_CASES:
3245      case 0x2002:        return op_code == OP_NOT_HSPACE;
3246      case 0x2003:  
3247      case 0x2004:        default:
3248      case 0x2005:        return op_code != OP_NOT_HSPACE;
3249      case 0x2006:        }
3250      case 0x2007:  
3251      case 0x2008:      case OP_ANYNL:
3252      case 0x2009:      case OP_VSPACE:
3253      case 0x200A:      case OP_NOT_VSPACE:
3254      case 0x202f:      switch(next)
3255      case 0x205f:        {
3256      case 0x3000:        VSPACE_CASES:
3257      return op_code != OP_HSPACE;        return op_code == OP_NOT_VSPACE;
3258      default:  
3259      return op_code == OP_HSPACE;        default:
3260      }        return op_code != OP_NOT_VSPACE;
3261          }
3262    
3263    #ifdef SUPPORT_UCP
3264        case OP_PROP:
3265        return check_char_prop(next, previous[0], previous[1], FALSE);
3266    
3267        case OP_NOTPROP:
3268        return check_char_prop(next, previous[0], previous[1], TRUE);
3269    #endif
3270    
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
     {  
     case 0x0a:  
     case 0x0b:  
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code != OP_VSPACE;  
3271      default:      default:
3272      return op_code == OP_VSPACE;      return FALSE;
3273      }      }
   
   default:  
   return FALSE;  
3274    }    }
3275    
3276    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3277  /* Handle the case when the next item is \d, \s, etc. */  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3278    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3279    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3280    replaced by OP_PROP codes when PCRE_UCP is set. */
3281    
3282  switch(op_code)  switch(op_code)
3283    {    {
3284    case OP_CHAR:    case OP_CHAR:
3285    case OP_CHARNC:    case OP_CHARI:
3286  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3287      {      {
3288      case ESC_d:      case ESC_d:
3289      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3290    
3291      case ESC_D:      case ESC_D:
3292      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3293    
3294      case ESC_s:      case ESC_s:
3295      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3296    
3297      case ESC_S:      case ESC_S:
3298      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3299    
3300      case ESC_w:      case ESC_w:
3301      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3302    
3303      case ESC_W:      case ESC_W:
3304      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3305    
3306      case ESC_h:      case ESC_h:
3307      case ESC_H:      case ESC_H:
3308      switch(item)      switch(c)
3309        {        {
3310        case 0x09:        HSPACE_CASES:
3311        case 0x20:        return escape != ESC_h;
3312        case 0xa0:  
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
3313        default:        default:
3314        return -next == ESC_h;        return escape == ESC_h;
3315        }        }
3316    
3317      case ESC_v:      case ESC_v:
3318      case ESC_V:      case ESC_V:
3319      switch(item)      switch(c)
3320        {        {
3321        case 0x0a:        VSPACE_CASES:
3322        case 0x0b:        return escape != ESC_v;
3323        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3324        default:        default:
3325        return -next == ESC_v;        return escape == ESC_v;
3326        }        }
3327    
3328        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3329        their substitutions and process them. The result will always be either
3330        ESC_p or ESC_P. Then fall through to process those values. */
3331    
3332    #ifdef SUPPORT_UCP
3333        case ESC_du:
3334        case ESC_DU:
3335        case ESC_wu:
3336        case ESC_WU:
3337        case ESC_su:
3338        case ESC_SU:
3339          {
3340          int temperrorcode = 0;
3341          ptr = substitutes[escape - ESC_DU];
3342          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3343          if (temperrorcode != 0) return FALSE;
3344          ptr++;    /* For compatibility */
3345          }
3346        /* Fall through */
3347    
3348        case ESC_p:
3349        case ESC_P:
3350          {
3351          unsigned int ptype = 0, pdata = 0;
3352          int errorcodeptr;
3353          BOOL negated;
3354    
3355          ptr--;      /* Make ptr point at the p or P */
3356          if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))
3357            return FALSE;
3358          ptr++;      /* Point past the final curly ket */
3359    
3360          /* If the property item is optional, we have to give up. (When generated
3361          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3362          to the original \d etc. At this point, ptr will point to a zero byte. */
3363    
3364          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3365            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3366              return FALSE;
3367    
3368          /* Do the property check. */
3369    
3370          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3371          }
3372    #endif
3373    
3374      default:      default:
3375      return FALSE;      return FALSE;
3376      }      }
3377    
3378      /* In principle, support for Unicode properties should be integrated here as
3379      well. It means re-organizing the above code so as to get hold of the property
3380      values before switching on the op-code. However, I wonder how many patterns
3381      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3382      these op-codes are never generated.) */
3383    
3384    case OP_DIGIT:    case OP_DIGIT:
3385    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3386           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3387    
3388    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3389    return next == -ESC_d;    return escape == ESC_d;
3390    
3391    case OP_WHITESPACE:    case OP_WHITESPACE:
3392    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3393    
3394    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3395    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3396    
3397    case OP_HSPACE:    case OP_HSPACE:
3398    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3399             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3400    
3401    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3402    return next == -ESC_h;    return escape == ESC_h;
3403    
3404    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3405      case OP_ANYNL:
3406    case OP_VSPACE:    case OP_VSPACE:
3407    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3408    
3409    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3410    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3411    
3412    case OP_WORDCHAR:    case OP_WORDCHAR:
3413    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3414             escape == ESC_v || escape == ESC_R;
3415    
3416    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3417    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3418    
3419    default:    default:
3420    return FALSE;    return FALSE;
# Line 2340  switch(op_code) Line 3426  switch(op_code)
3426    
3427    
3428  /*************************************************  /*************************************************
3429    *        Add a character or range to a class     *
3430    *************************************************/
3431    
3432    /* This function packages up the logic of adding a character or range of
3433    characters to a class. The character values in the arguments will be within the
3434    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3435    mutually recursive with the function immediately below.
3436    
3437    Arguments:
3438      classbits     the bit map for characters < 256
3439      uchardptr     points to the pointer for extra data
3440      options       the options word
3441      cd            contains pointers to tables etc.
3442      start         start of range character
3443      end           end of range character
3444    
3445    Returns:        the number of < 256 characters added
3446                    the pointer to extra data is updated
3447    */
3448    
3449    static int
3450    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3451      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3452    {
3453    pcre_uint32 c;
3454    int n8 = 0;
3455    
3456    /* If caseless matching is required, scan the range and process alternate
3457    cases. In Unicode, there are 8-bit characters that have alternate cases that
3458    are greater than 255 and vice-versa. Sometimes we can just extend the original
3459    range. */
3460    
3461    if ((options & PCRE_CASELESS) != 0)
3462      {
3463    #ifdef SUPPORT_UCP
3464      if ((options & PCRE_UTF8) != 0)
3465        {
3466        int rc;
3467        pcre_uint32 oc, od;
3468    
3469        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3470        c = start;
3471    
3472        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3473          {
3474          /* Handle a single character that has more than one other case. */
3475    
3476          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3477            PRIV(ucd_caseless_sets) + rc, oc);
3478    
3479          /* Do nothing if the other case range is within the original range. */
3480    
3481          else if (oc >= start && od <= end) continue;
3482    
3483          /* Extend the original range if there is overlap, noting that if oc < c, we
3484          can't have od > end because a subrange is always shorter than the basic
3485          range. Otherwise, use a recursive call to add the additional range. */
3486    
3487          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3488          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3489          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3490          }
3491        }
3492      else
3493    #endif  /* SUPPORT_UCP */
3494    
3495      /* Not UTF-mode, or no UCP */
3496    
3497      for (c = start; c <= end && c < 256; c++)
3498        {
3499        SETBIT(classbits, cd->fcc[c]);
3500        n8++;
3501        }
3502      }
3503    
3504    /* Now handle the original range. Adjust the final value according to the bit
3505    length - this means that the same lists of (e.g.) horizontal spaces can be used
3506    in all cases. */
3507    
3508    #if defined COMPILE_PCRE8
3509    #ifdef SUPPORT_UTF
3510      if ((options & PCRE_UTF8) == 0)
3511    #endif
3512      if (end > 0xff) end = 0xff;
3513    
3514    #elif defined COMPILE_PCRE16
3515    #ifdef SUPPORT_UTF
3516      if ((options & PCRE_UTF16) == 0)
3517    #endif
3518      if (end > 0xffff) end = 0xffff;
3519    
3520    #endif /* COMPILE_PCRE[8|16] */
3521    
3522    /* If all characters are less than 256, use the bit map. Otherwise use extra
3523    data. */
3524    
3525    if (end < 0x100)
3526      {
3527      for (c = start; c <= end; c++)
3528        {
3529        n8++;
3530        SETBIT(classbits, c);
3531        }
3532      }
3533    
3534    else
3535      {
3536      pcre_uchar *uchardata = *uchardptr;
3537    
3538    #ifdef SUPPORT_UTF
3539      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3540        {
3541        if (start < end)
3542          {
3543          *uchardata++ = XCL_RANGE;
3544          uchardata += PRIV(ord2utf)(start, uchardata);
3545          uchardata += PRIV(ord2utf)(end, uchardata);
3546          }
3547        else if (start == end)
3548          {
3549          *uchardata++ = XCL_SINGLE;
3550          uchardata += PRIV(ord2utf)(start, uchardata);
3551          }
3552        }
3553      else
3554    #endif  /* SUPPORT_UTF */
3555    
3556      /* Without UTF support, character values are constrained by the bit length,
3557      and can only be > 256 for 16-bit and 32-bit libraries. */
3558    
3559    #ifdef COMPILE_PCRE8
3560        {}
3561    #else
3562      if (start < end)
3563        {
3564        *uchardata++ = XCL_RANGE;
3565        *uchardata++ = start;
3566        *uchardata++ = end;
3567        }
3568      else if (start == end)
3569        {
3570        *uchardata++ = XCL_SINGLE;
3571        *uchardata++ = start;
3572        }
3573    #endif
3574    
3575      *uchardptr = uchardata;   /* Updata extra data pointer */
3576      }
3577    
3578    return n8;    /* Number of 8-bit characters */
3579    }
3580    
3581    
3582    
3583    
3584    /*************************************************
3585    *        Add a list of characters to a class     *
3586    *************************************************/
3587    
3588    /* This function is used for adding a list of case-equivalent characters to a
3589    class, and also for adding a list of horizontal or vertical whitespace. If the
3590    list is in order (which it should be), ranges of characters are detected and
3591    handled appropriately. This function is mutually recursive with the function
3592    above.
3593    
3594    Arguments:
3595      classbits     the bit map for characters < 256
3596      uchardptr     points to the pointer for extra data
3597      options       the options word
3598      cd            contains pointers to tables etc.
3599      p             points to row of 32-bit values, terminated by NOTACHAR
3600      except        character to omit; this is used when adding lists of
3601                      case-equivalent characters to avoid including the one we
3602                      already know about
3603    
3604    Returns:        the number of < 256 characters added
3605                    the pointer to extra data is updated
3606    */
3607    
3608    static int
3609    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3610      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3611    {
3612    int n8 = 0;
3613    while (p[0] < NOTACHAR)
3614      {
3615      int n = 0;
3616      if (p[0] != except)
3617        {
3618        while(p[n+1] == p[0] + n + 1) n++;
3619        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3620        }
3621      p += n + 1;
3622      }
3623    return n8;
3624    }
3625    
3626    
3627    
3628    /*************************************************
3629    *    Add characters not in a list to a class     *
3630    *************************************************/
3631    
3632    /* This function is used for adding the complement of a list of horizontal or
3633    vertical whitespace to a class. The list must be in order.
3634    
3635    Arguments:
3636      classbits     the bit map for characters < 256
3637      uchardptr     points to the pointer for extra data
3638      options       the options word
3639      cd            contains pointers to tables etc.
3640      p             points to row of 32-bit values, terminated by NOTACHAR
3641    
3642    Returns:        the number of < 256 characters added
3643                    the pointer to extra data is updated
3644    */
3645    
3646    static int
3647    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3648      int options, compile_data *cd, const pcre_uint32 *p)
3649    {
3650    BOOL utf = (options & PCRE_UTF8) != 0;
3651    int n8 = 0;
3652    if (p[0] > 0)
3653      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3654    while (p[0] < NOTACHAR)
3655      {
3656      while (p[1] == p[0] + 1) p++;
3657      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3658        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3659      p++;
3660      }
3661    return n8;
3662    }
3663    
3664    
3665    
3666    /*************************************************
3667  *           Compile one branch                   *  *           Compile one branch                   *
3668  *************************************************/  *************************************************/
3669    
# Line 2354  Arguments: Line 3678  Arguments:
3678    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3679    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3680    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3681    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr    place to put the first required character
3682    reqbyteptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
3683      reqcharptr     place to put the last required character
3684      reqcharflagsptr place to put the last required character flags, or a negative number
3685    bcptr          points to current branch chain    bcptr          points to current branch chain
3686      cond_depth     conditional nesting depth
3687    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3688    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3689                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2366  Returns:         TRUE on success Line 3693  Returns:         TRUE on success
3693  */  */
3694    
3695  static BOOL  static BOOL
3696  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3697    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
3698      pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
3699      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
3700      branch_chain *bcptr, int cond_depth,
3701    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3702  {  {
3703  int repeat_type, op_type;  int repeat_type, op_type;
3704  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3705  int bravalue = 0;  int bravalue = 0;
3706  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3707  int firstbyte, reqbyte;  pcre_uint32 firstchar, reqchar;
3708  int zeroreqbyte, zerofirstbyte;  pcre_int32 firstcharflags, reqcharflags;
3709  int req_caseopt, reqvary, tempreqvary;  pcre_uint32 zeroreqchar, zerofirstchar;
3710  int options = *optionsptr;  pcre_int32 zeroreqcharflags, zerofirstcharflags;
3711    pcre_int32 req_caseopt, reqvary, tempreqvary;
3712    int options = *optionsptr;               /* May change dynamically */
3713  int after_manual_callout = 0;  int after_manual_callout = 0;
3714  int length_prevgroup = 0;  int length_prevgroup = 0;
3715  register int c;  register pcre_uint32 c;
3716  register uschar *code = *codeptr;  int escape;
3717  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3718  uschar *orig_code = code;  pcre_uchar *last_code = code;
3719  uschar *tempcode;  pcre_uchar *orig_code = code;
3720    pcre_uchar *tempcode;
3721  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3722  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3723  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3724  const uschar *tempptr;  const pcre_uchar *tempptr;
3725  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3726  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3727  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3728  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3729    pcre_uint8 classbits[32];
3730  #ifdef SUPPORT_UTF8  
3731  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3732  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3733  uschar *class_utf8data;  dynamically as we process the pattern. */
3734  uschar *class_utf8data_base;  
3735  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3736    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3737    BOOL utf = (options & PCRE_UTF8) != 0;
3738    #ifndef COMPILE_PCRE32
3739    pcre_uchar utf_chars[6];
3740    #endif
3741  #else  #else
3742  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3743  uschar *utf8_char = NULL;  #endif
3744    
3745    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3746    class_uchardata always so that it can be passed to add_to_class() always,
3747    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3748    alternative calls for the different cases. */
3749    
3750    pcre_uchar *class_uchardata;
3751    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3752    BOOL xclass;
3753    pcre_uchar *class_uchardata_base;
3754  #endif  #endif
3755    
3756  #ifdef DEBUG  #ifdef PCRE_DEBUG
3757  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3758  #endif  #endif
3759    
# Line 2416  greedy_non_default = greedy_default ^ 1; Line 3764  greedy_non_default = greedy_default ^ 1;
3764    
3765  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3766  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
3767  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
3768  find one.  find one.
3769    
3770  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
3771  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
3772  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3773  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3774    
3775  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
3776    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
3777    
3778  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3779  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
3780  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
3781  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3782    value. This is used only for ASCII characters. */
3783    
3784  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3785    
3786  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3787    
# Line 2443  for (;; ptr++) Line 3793  for (;; ptr++)
3793    BOOL is_quantifier;    BOOL is_quantifier;
3794    BOOL is_recurse;    BOOL is_recurse;
3795    BOOL reset_bracount;    BOOL reset_bracount;
3796    int class_charcount;    int class_has_8bitchar;
3797    int class_lastchar;    int class_one_char;
3798    int newoptions;    int newoptions;
3799    int recno;    int recno;
3800    int refsign;    int refsign;
3801    int skipbytes;    int skipbytes;
3802    int subreqbyte;    pcre_uint32 subreqchar, subfirstchar;
3803    int subfirstbyte;    pcre_int32 subreqcharflags, subfirstcharflags;
3804    int terminator;    int terminator;
3805    int mclength;    unsigned int mclength;
3806    uschar mcbuffer[8];    unsigned int tempbracount;
3807      pcre_uint32 ec;
3808      pcre_uchar mcbuffer[8];
3809    
3810    /* Get next byte in the pattern */    /* Get next character in the pattern */
3811    
3812    c = *ptr;    c = *ptr;
3813    
3814      /* If we are at the end of a nested substitution, revert to the outer level
3815      string. Nesting only happens one level deep. */
3816    
3817      if (c == CHAR_NULL && nestptr != NULL)
3818        {
3819        ptr = nestptr;
3820        nestptr = NULL;
3821        c = *ptr;
3822        }
3823    
3824    /* 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
3825    previous cycle of this loop. */    previous cycle of this loop. */
3826    
3827    if (lengthptr != NULL)    if (lengthptr != NULL)
3828      {      {
3829  #ifdef DEBUG  #ifdef PCRE_DEBUG
3830      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3831  #endif  #endif
3832      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3833            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3834        {        {
3835        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3836        goto FAILED;        goto FAILED;
# Line 2489  for (;; ptr++) Line 3852  for (;; ptr++)
3852        goto FAILED;        goto FAILED;
3853        }        }
3854    
3855      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3856      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3857          (int)(code - last_code), c, c));
3858    
3859