/[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 341 by ph10, Sat Apr 19 16:41:04 2008 UTC revision 1069 by chpe, Tue Oct 16 15:54:29 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 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,
88        int, int, int *, int *, branch_chain *, compile_data *, int *);
89    
90    
91    
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 103  so this number is very generous.
103  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
104  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
105  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
106  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
107    filled up by repetitions of forward references, for example patterns like
108    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
109    that the workspace is expanded using malloc() in this situation. The value
110    below is therefore a minimum, and we put a maximum on it for safety. The
111    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
112    kicks in at the same number of forward references in all cases. */
113    
114    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
115    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
116    
117  #define COMPILE_WORK_SIZE (4096)  /* The overrun tests check for a slightly smaller size so that they detect the
118    overrun before it actually does run off the end of the data block. */
119    
120    #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    #define REQ_MASK       (REQ_CASELESS | REQ_VARY)
127    
128    /* Repeated character flags. */
129    
130    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
131    
132  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
133  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
134  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
135  is invalid. */  is invalid. */
136    
137  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
138    
139    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
140    in UTF-8 mode. */
141    
142  static const short int escapes[] = {  static const short int escapes[] = {
143       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
144       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
145     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
146  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
147  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
148  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
149     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
150  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
151  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
152       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
153         -ESC_D,                  -ESC_E,
154         0,                       -ESC_G,
155         -ESC_H,                  0,
156         0,                       -ESC_K,
157         0,                       0,
158         -ESC_N,                  0,
159         -ESC_P,                  -ESC_Q,
160         -ESC_R,                  -ESC_S,
161         0,                       0,
162         -ESC_V,                  -ESC_W,
163         -ESC_X,                  0,
164         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
165         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
166         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
167         CHAR_GRAVE_ACCENT,       7,
168         -ESC_b,                  0,
169         -ESC_d,                  ESC_e,
170         ESC_f,                   0,
171         -ESC_h,                  0,
172         0,                       -ESC_k,
173         0,                       0,
174         ESC_n,                   0,
175         -ESC_p,                  0,
176         ESC_r,                   -ESC_s,
177         ESC_tee,                 0,
178         -ESC_v,                  -ESC_w,
179         0,                       0,
180         -ESC_z
181  };  };
182    
183  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
184    
185    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
186    
187  static const short int escapes[] = {  static const short int escapes[] = {
188  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
189  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 202  static const short int escapes[] = {
202  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
203  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
204  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
205  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
206  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
207  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
208  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 214  static const short int escapes[] = {
214    
215  /* 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
216  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
217  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
218    string is built from string macros so that it works in UTF-8 mode on EBCDIC
219    platforms. */
220    
221  typedef struct verbitem {  typedef struct verbitem {
222    int   len;    int   len;                 /* Length of verb name */
223    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
224      int   op_arg;              /* Op when arg present, or -1 if not allowed */
225  } verbitem;  } verbitem;
226    
227  static const char verbnames[] =  static const char verbnames[] =
228    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
229    "COMMIT\0"    STRING_MARK0
230    "F\0"    STRING_ACCEPT0
231    "FAIL\0"    STRING_COMMIT0
232    "PRUNE\0"    STRING_F0
233    "SKIP\0"    STRING_FAIL0
234    "THEN";    STRING_PRUNE0
235      STRING_SKIP0
236      STRING_THEN;
237    
238  static const verbitem verbs[] = {  static const verbitem verbs[] = {
239    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
240    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
241    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
242    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
243    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
244    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
245    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
246      { 4, OP_SKIP,   OP_SKIP_ARG  },
247      { 4, OP_THEN,   OP_THEN_ARG  }
248  };  };
249    
250  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 178  length entry. The first three must be al Line 257  length entry. The first three must be al
257  for handling case independence. */  for handling case independence. */
258    
259  static const char posix_names[] =  static const char posix_names[] =
260    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
261    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
262    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
263      STRING_word0  STRING_xdigit;
264    
265  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
266    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
267    
268  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 212  static const int posix_class_maps[] = { Line 292  static const int posix_class_maps[] = {
292    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
293  };  };
294    
295    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
296    substitutes must be in the order of the names, defined above, and there are
297    both positive and negative cases. NULL means no substitute. */
298    
299    #ifdef SUPPORT_UCP
300    static const pcre_uchar string_PNd[]  = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pNd[]  = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_PXsp[] = {
307      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_pXsp[] = {
310      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_PXwd[] = {
313      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_pXwd[] = {
316      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    
319    static const pcre_uchar *substitutes[] = {
320      string_PNd,           /* \D */
321      string_pNd,           /* \d */
322      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
323      string_pXsp,          /* \s */
324      string_PXwd,          /* \W */
325      string_pXwd           /* \w */
326    };
327    
328    static const pcre_uchar string_pL[] =   {
329      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pLl[] =  {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLu[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXan[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_h[] =    {
341      CHAR_BACKSLASH, CHAR_h, '\0' };
342    static const pcre_uchar string_pXps[] = {
343      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_PL[] =   {
346      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PLl[] =  {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLu[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PXan[] = {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_H[] =    {
358      CHAR_BACKSLASH, CHAR_H, '\0' };
359    static const pcre_uchar string_PXps[] = {
360      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
361      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
362    
363    static const pcre_uchar *posix_substitutes[] = {
364      string_pL,            /* alpha */
365      string_pLl,           /* lower */
366      string_pLu,           /* upper */
367      string_pXan,          /* alnum */
368      NULL,                 /* ascii */
369      string_h,             /* blank */
370      NULL,                 /* cntrl */
371      string_pNd,           /* digit */
372      NULL,                 /* graph */
373      NULL,                 /* print */
374      NULL,                 /* punct */
375      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
376      string_pXwd,          /* word */
377      NULL,                 /* xdigit */
378      /* Negated cases */
379      string_PL,            /* ^alpha */
380      string_PLl,           /* ^lower */
381      string_PLu,           /* ^upper */
382      string_PXan,          /* ^alnum */
383      NULL,                 /* ^ascii */
384      string_H,             /* ^blank */
385      NULL,                 /* ^cntrl */
386      string_PNd,           /* ^digit */
387      NULL,                 /* ^graph */
388      NULL,                 /* ^print */
389      NULL,                 /* ^punct */
390      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
391      string_PXwd,          /* ^word */
392      NULL                  /* ^xdigit */
393    };
394    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
395    #endif
396    
397  #define STRING(a)  # a  #define STRING(a)  # a
398  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 405  the number of relocations needed when a
405  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
406  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
407  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
408  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
409    
410    Each substring ends with \0 to insert a null character. This includes the final
411    substring, so that the whole string ends with \0\0, which can be detected when
412    counting through. */
413    
414  static const char error_texts[] =  static const char error_texts[] =
415    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 450  static const char error_texts[] =
450    /* 30 */    /* 30 */
451    "unknown POSIX class name\0"    "unknown POSIX class name\0"
452    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
453    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
454    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
455    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
456    /* 35 */    /* 35 */
457    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
458    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
459    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
460    "number after (?C is > 255\0"    "number after (?C is > 255\0"
461    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
462    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 473  static const char error_texts[] =
473    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
474    /* 50 */    /* 50 */
475    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
476    "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"
477    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
478    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
479    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
480    /* 55 */    /* 55 */
481    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
482    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
483    "\\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"
484    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
485    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
486    /* 60 */    /* 60 */
487    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
488    "number is too big\0"    "number is too big\0"
489    "subpattern name expected\0"    "subpattern name expected\0"
490    "digit expected after (?+\0"    "digit expected after (?+\0"
491    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
492      /* 65 */
493      "different names for subpatterns of the same number are not allowed\0"
494      "(*MARK) must have an argument\0"
495      "this version of PCRE is not compiled with Unicode property support\0"
496      "\\c must be followed by an ASCII character\0"
497      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
498      /* 70 */
499      "internal error: unknown opcode in find_fixedlength()\0"
500      "\\N is not supported in a class\0"
501      "too many forward references\0"
502      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
503      "invalid UTF-16 string\0"
504      /* 75 */
505      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
506      "character value in \\u.... sequence is too large\0"
507      "invalid UTF-32 string\0"
508      ;
509    
510  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
511  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 322  For convenience, we use the same bit def Line 523  For convenience, we use the same bit def
523    
524  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
525    
526  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
527  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
528    into a subtraction and unsigned comparison). */
529    
530    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
531    
532    #ifndef EBCDIC
533    
534    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
535    UTF-8 mode. */
536    
537    static const pcre_uint8 digitab[] =
538    {    {
539    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
540    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 358  static const unsigned char digitab[] = Line 569  static const unsigned char digitab[] =
569    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
570    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
571    
572  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
573  static const unsigned char digitab[] =  
574    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
575    
576    static const pcre_uint8 digitab[] =
577    {    {
578    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
579    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 394  static const unsigned char digitab[] = Line 608  static const unsigned char digitab[] =
608    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
609    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
610    
611  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
612    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
613    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
614    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 430  static const unsigned char ebcdic_charta Line 644  static const unsigned char ebcdic_charta
644  #endif  #endif
645    
646    
 /* 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 *);  
   
647    
648    
649  /*************************************************  /*************************************************
# Line 455  static const char * Line 663  static const char *
663  find_error_text(int n)  find_error_text(int n)
664  {  {
665  const char *s = error_texts;  const char *s = error_texts;
666  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
667      {
668      while (*s++ != 0) {};
669      if (*s == 0) return "Error text not found (please report)";
670      }
671  return s;  return s;
672  }  }
673    
674    
675  /*************************************************  /*************************************************
676    *           Expand the workspace                 *
677    *************************************************/
678    
679    /* This function is called during the second compiling phase, if the number of
680    forward references fills the existing workspace, which is originally a block on
681    the stack. A larger block is obtained from malloc() unless the ultimate limit
682    has been reached or the increase will be rather small.
683    
684    Argument: pointer to the compile data block
685    Returns:  0 if all went well, else an error number
686    */
687    
688    static int
689    expand_workspace(compile_data *cd)
690    {
691    pcre_uchar *newspace;
692    int newsize = cd->workspace_size * 2;
693    
694    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
695    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
696        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
697     return ERR72;
698    
699    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
700    if (newspace == NULL) return ERR21;
701    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
702    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
703    if (cd->workspace_size > COMPILE_WORK_SIZE)
704      (PUBL(free))((void *)cd->start_workspace);
705    cd->start_workspace = newspace;
706    cd->workspace_size = newsize;
707    return 0;
708    }
709    
710    
711    
712    /*************************************************
713    *            Check for counted repeat            *
714    *************************************************/
715    
716    /* This function is called when a '{' is encountered in a place where it might
717    start a quantifier. It looks ahead to see if it really is a quantifier or not.
718    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
719    where the ddds are digits.
720    
721    Arguments:
722      p         pointer to the first char after '{'
723    
724    Returns:    TRUE or FALSE
725    */
726    
727    static BOOL
728    is_counted_repeat(const pcre_uchar *p)
729    {
730    if (!IS_DIGIT(*p)) return FALSE;
731    p++;
732    while (IS_DIGIT(*p)) p++;
733    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
734    
735    if (*p++ != CHAR_COMMA) return FALSE;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (!IS_DIGIT(*p)) return FALSE;
739    p++;
740    while (IS_DIGIT(*p)) p++;
741    
742    return (*p == CHAR_RIGHT_CURLY_BRACKET);
743    }
744    
745    
746    
747    /*************************************************
748  *            Handle escapes                      *  *            Handle escapes                      *
749  *************************************************/  *************************************************/
750    
751  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
752  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character
753  encodes one of the more complicated things such as \d. A backreference to group  which will be placed in chptr. A backreference to group n is returned as
754  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  negative n. When UTF-8 is enabled, a positive value greater than 255 may
755  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  be returned in chptr.
756  ptr is pointing at the \. On exit, it is on the final character of the escape  On entry,ptr is pointing at the \. On exit, it is on the final character of the
757  sequence.  escape sequence.
758    
759  Arguments:  Arguments:
760    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
761      chptr          points to the data character
762    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
763    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
764    options        the options bits    options        the options bits
765    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
766    
767  Returns:         zero or positive => a data character  Returns:         zero => a data character
768                   negative => a special escape sequence                   positive => a special escape sequence
769                     negative => a back reference
770                   on error, errorcodeptr is set                   on error, errorcodeptr is set
771  */  */
772    
773  static int  static int
774  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
775    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
776  {  {
777  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
778  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
779  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
780    pcre_uint32 c;
781    int escape = 0;
782    int i;
783    
784  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
785  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 503  if (c == 0) *errorcodeptr = ERR1; Line 792  if (c == 0) *errorcodeptr = ERR1;
792  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.
793  Otherwise further processing may be required. */  Otherwise further processing may be required. */
794    
795  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
796  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
797  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
798    else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
799    
800  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
801  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
802  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
803    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
804  #endif  #endif
805    
806  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
807    
808  else  else
809    {    {
810    const uschar *oldptr;    const pcre_uchar *oldptr;
811    BOOL braced, negated;    BOOL braced, negated, overflow;
812      int s;
813    
814    switch (c)    switch (c)
815      {      {
816      /* 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
817      error. */      error. */
818    
819      case 'l':      case CHAR_l:
820      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
821      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
822      break;      break;
823    
824      /* \g must be followed by one of a number of specific things:      case CHAR_u:
825        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
826          {
827          /* In JavaScript, \u must be followed by four hexadecimal numbers.
828          Otherwise it is a lowercase u letter. */
829          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
830            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
831            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
832            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
833            {
834            c = 0;
835            for (i = 0; i < 4; ++i)
836              {
837              register pcre_uint32 cc = *(++ptr);
838    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
839              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
840              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
841    #else           /* EBCDIC coding */
842              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
843              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
844    #endif
845              }
846    
847    #if defined COMPILE_PCRE8
848            if (c > (utf ? 0x10ffff : 0xff))
849    #elif defined COMPILE_PCRE16
850            if (c > (utf ? 0x10ffff : 0xffff))
851    #elif defined COMPILE_PCRE32
852            if (utf && c > 0x10ffff)
853    #endif
854              {
855              *errorcodeptr = ERR76;
856              }
857            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
858            }
859          }
860        else
861          *errorcodeptr = ERR37;
862        break;
863    
864        case CHAR_U:
865        /* In JavaScript, \U is an uppercase U letter. */
866        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
867        break;
868    
869        /* In a character class, \g is just a literal "g". Outside a character
870        class, \g must be followed by one of a number of specific things:
871    
872      (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
873      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
874      5.10 feature.      5.10 feature.
875    
876      (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
877      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
878      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
879      was \k.      was \k.
   
     (3) For Oniguruma compatibility we also support \g followed by a name or a  
     number either in angle brackets or in single quotes. However, these are  
     (possibly recursive) subroutine calls, _not_ backreferences. Just return  
     the -ESC_g code (cf \k). */  
880    
881      case 'g':      (3) For Oniguruma compatibility we also support \g followed by a name or a
882      if (ptr[1] == '<' || ptr[1] == '\'')      number either in angle brackets or in single quotes. However, these are
883        (possibly recursive) subroutine calls, _not_ backreferences. Just return
884        the ESC_g code (cf \k). */
885    
886        case CHAR_g:
887        if (isclass) break;
888        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
889        {        {
890        c = -ESC_g;        escape = ESC_g;
891        break;        break;
892        }        }
893    
894      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
895    
896      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
897        {        {
898        const uschar *p;        const pcre_uchar *p;
899        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
900          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
901        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
902          {          {
903          c = -ESC_k;          escape = ESC_k;
904          break;          break;
905          }          }
906        braced = TRUE;        braced = TRUE;
# Line 572  else Line 908  else
908        }        }
909      else braced = FALSE;      else braced = FALSE;
910    
911      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
912        {        {
913        negated = TRUE;        negated = TRUE;
914        ptr++;        ptr++;
915        }        }
916      else negated = FALSE;      else negated = FALSE;
917    
918      c = 0;      /* The integer range is limited by the machine's int representation. */
919      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
920        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
921        while (IS_DIGIT(ptr[1]))
922      if (c < 0)   /* Integer overflow */        {
923          if (s > INT_MAX / 10 - 1) /* Integer overflow */
924            {
925            overflow = TRUE;
926            break;
927            }
928          s = s * 10 + (int)(*(++ptr) - CHAR_0);
929          }
930        if (overflow) /* Integer overflow */
931        {        {
932          while (IS_DIGIT(ptr[1]))
933            ptr++;
934        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
935        break;        break;
936        }        }
937    
938      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
939        {        {
940        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
941        break;        break;
942        }        }
943    
944      if (c == 0)      if (s == 0)
945        {        {
946        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
947        break;        break;
948        }        }
949    
950      if (negated)      if (negated)
951        {        {
952        if (c > bracount)        if (s > bracount)
953          {          {
954          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
955          break;          break;
956          }          }
957        c = bracount - (c - 1);        s = bracount - (s - 1);
958        }        }
959    
960      c = -(ESC_REF + c);      escape = -s;
961      break;      break;
962    
963      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 626  else Line 972  else
972      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
973      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
974    
975      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:
976      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
977    
978      if (!isclass)      if (!isclass)
979        {        {
980        oldptr = ptr;        oldptr = ptr;
981        c -= '0';        /* The integer range is limited by the machine's int representation. */
982        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
983          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
984        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
985          {          {
986            if (s > INT_MAX / 10 - 1) /* Integer overflow */
987              {
988              overflow = TRUE;
989              break;
990              }
991            s = s * 10 + (int)(*(++ptr) - CHAR_0);
992            }
993          if (overflow) /* Integer overflow */
994            {
995            while (IS_DIGIT(ptr[1]))
996              ptr++;
997          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
998          break;          break;
999          }          }
1000        if (c < 10 || c <= bracount)        if (s < 10 || s <= bracount)
1001          {          {
1002          c = -(ESC_REF + c);          escape = -s;
1003          break;          break;
1004          }          }
1005        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 652  else Line 1009  else
1009      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.
1010      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1011    
1012      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1013        {        {
1014        ptr--;        ptr--;
1015        c = 0;        c = 0;
# Line 662  else Line 1019  else
1019      /* \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
1020      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
1021      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
1022      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,
1023      than 3 octal digits. */      but no more than 3 octal digits. */
1024    
1025      case '0':      case CHAR_0:
1026      c -= '0';      c -= CHAR_0;
1027      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1028          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1029      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1030        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1031    #endif
1032      break;      break;
1033    
1034      /* \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
1035      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.
1036      treated as a data character. */      If not, { is treated as a data character. */
1037    
1038        case CHAR_x:
1039        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1040          {
1041          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1042          Otherwise it is a lowercase x letter. */
1043          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1044            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1045            {
1046            c = 0;
1047            for (i = 0; i < 2; ++i)
1048              {
1049              register pcre_uint32 cc = *(++ptr);
1050    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1051              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1052              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1053    #else           /* EBCDIC coding */
1054              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1055              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1056    #endif
1057              }
1058            }
1059          break;
1060          }
1061    
1062      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1063        {        {
1064        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1065    
1066        c = 0;        c = 0;
1067        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1068          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1069          {          {
1070          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1071          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1072          count++;  
1073    #ifdef COMPILE_PCRE32
1074  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1075          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1076          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1077    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1078            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1079            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1080  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1081          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1082          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1083    #endif
1084    
1085    #if defined COMPILE_PCRE8
1086            if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }
1087    #elif defined COMPILE_PCRE16
1088            if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }
1089    #elif defined COMPILE_PCRE32
1090            if (utf && c > 0x10ffff) { overflow = TRUE; break; }
1091  #endif  #endif
1092          }          }
1093    
1094        if (*pt == '}')        if (overflow)
1095          {          {
1096          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1097            *errorcodeptr = ERR34;
1098            }
1099    
1100          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1101            {
1102            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1103          ptr = pt;          ptr = pt;
1104          break;          break;
1105          }          }
# Line 712  else Line 1111  else
1111      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1112    
1113      c = 0;      c = 0;
1114      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1115        {        {
1116        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1117        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1118  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1119        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1120        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1121  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1122        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1123        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1124  #endif  #endif
1125        }        }
1126      break;      break;
1127    
1128      /* 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.
1129      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
1130        coding is ASCII-specific, but then the whole concept of \cx is
1131      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1132    
1133      case 'c':      case CHAR_c:
1134      c = *(++ptr);      c = *(++ptr);
1135      if (c == 0)      if (c == 0)
1136        {        {
1137        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1138        break;        break;
1139        }        }
1140    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1141  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1142      if (c >= 'a' && c <= 'z') c -= 32;        {
1143          *errorcodeptr = ERR68;
1144          break;
1145          }
1146        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1147      c ^= 0x40;      c ^= 0x40;
1148  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1149      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1150      c ^= 0xC0;      c ^= 0xC0;
1151  #endif  #endif
1152      break;      break;
# Line 764  else Line 1168  else
1168      }      }
1169    }    }
1170    
1171  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1172  return c;  newline". PCRE does not support \N{name}. However, it does support
1173  }  quantification such as \N{2,3}. */
1174    
1175    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1176         !is_counted_repeat(ptr+2))
1177      *errorcodeptr = ERR37;
1178    
1179    /* If PCRE_UCP is set, we change the values for \d etc. */
1180    
1181    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1182      escape += (ESC_DU - ESC_D);
1183    
1184    /* Set the pointer to the final character before returning. */
1185    
1186    *ptrptr = ptr;
1187    *chptr = c;
1188    return escape;
1189    }
1190    
1191  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1192  /*************************************************  /*************************************************
# Line 790  Returns:         type value from ucp_typ Line 1208  Returns:         type value from ucp_typ
1208  */  */
1209    
1210  static int  static int
1211  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1212  {  {
1213  int c, i, bot, top;  pcre_uchar c;
1214  const uschar *ptr = *ptrptr;  int i, bot, top;
1215  char name[32];  const pcre_uchar *ptr = *ptrptr;
1216    pcre_uchar name[32];
1217    
1218  c = *(++ptr);  c = *(++ptr);
1219  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 804  if (c == 0) goto ERROR_RETURN; Line 1223  if (c == 0) goto ERROR_RETURN;
1223  /* \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
1224  negation. */  negation. */
1225    
1226  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1227    {    {
1228    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1229      {      {
1230      *negptr = TRUE;      *negptr = TRUE;
1231      ptr++;      ptr++;
1232      }      }
1233    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1234      {      {
1235      c = *(++ptr);      c = *(++ptr);
1236      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1237      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1238      name[i] = c;      name[i] = c;
1239      }      }
1240    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1241    name[i] = 0;    name[i] = 0;
1242    }    }
1243    
# Line 835  else Line 1254  else
1254  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1255    
1256  bot = 0;  bot = 0;
1257  top = _pcre_utt_size;  top = PRIV(utt_size);
1258    
1259  while (bot < top)  while (bot < top)
1260    {    {
1261      int r;
1262    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1263    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1264    if (c == 0)    if (r == 0)
1265      {      {
1266      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1267      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1268      }      }
1269    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1270    }    }
1271    
1272  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
# Line 864  return -1; Line 1284  return -1;
1284    
1285    
1286  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1287  *         Read repeat counts                     *  *         Read repeat counts                     *
1288  *************************************************/  *************************************************/
1289    
# Line 915  Returns:         pointer to '}' on succe Line 1302  Returns:         pointer to '}' on succe
1302                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1303  */  */
1304    
1305  static const uschar *  static const pcre_uchar *
1306  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)
1307  {  {
1308  int min = 0;  int min = 0;
1309  int max = -1;  int max = -1;
# Line 924  int max = -1; Line 1311  int max = -1;
1311  /* 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
1312  an integer overflow. */  an integer overflow. */
1313    
1314  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1315  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1316    {    {
1317    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 934  if (min < 0 || min > 65535) Line 1321  if (min < 0 || min > 65535)
1321  /* 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.
1322  Also, max must not be less than min. */  Also, max must not be less than min. */
1323    
1324  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1325    {    {
1326    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1327      {      {
1328      max = 0;      max = 0;
1329      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1330      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1331        {        {
1332        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 964  return p; Line 1351  return p;
1351    
1352    
1353  /*************************************************  /*************************************************
1354  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1355  *************************************************/  *************************************************/
1356    
1357  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1358    top-level call starts at the beginning of the pattern. All other calls must
1359    start at a parenthesis. It scans along a pattern's text looking for capturing
1360  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
1361  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
1362  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
1363  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1364  be terminated by '>' because that is checked in the first pass.  
1365    This function was originally called only from the second pass, in which we know
1366    that if (?< or (?' or (?P< is encountered, the name will be correctly
1367    terminated because that is checked in the first pass. There is now one call to
1368    this function in the first pass, to check for a recursive back reference by
1369    name (so that we can make the whole group atomic). In this case, we need check
1370    only up to the current position in the pattern, and that is still OK because
1371    and previous occurrences will have been checked. To make this work, the test
1372    for "end of pattern" is a check against cd->end_pattern in the main loop,
1373    instead of looking for a binary zero. This means that the special first-pass
1374    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1375    processing items within the loop are OK, because afterwards the main loop will
1376    terminate.)
1377    
1378  Arguments:  Arguments:
1379    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1380    cd           compile background data    cd           compile background data
1381    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1382    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1383    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1384      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1385      count        pointer to the current capturing subpattern number (updated)
1386    
1387  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1388  */  */
1389    
1390  static int  static int
1391  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1392    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1393  {  {
1394  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1395  int count = cd->bracount;  int start_count = *count;
1396    int hwm_count = start_count;
1397    BOOL dup_parens = FALSE;
1398    
1399  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1400    dealing with. The very first call may not start with a parenthesis. */
1401    
1402    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1403    {    {
1404    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1405    
1406      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1407    
1408      /* Handle a normal, unnamed capturing parenthesis. */
1409    
1410      else if (ptr[1] != CHAR_QUESTION_MARK)
1411        {
1412        *count += 1;
1413        if (name == NULL && *count == lorn) return *count;
1414        ptr++;
1415        }
1416    
1417      /* All cases now have (? at the start. Remember when we are in a group
1418      where the parenthesis numbers are duplicated. */
1419    
1420      else if (ptr[2] == CHAR_VERTICAL_LINE)
1421        {
1422        ptr += 3;
1423        dup_parens = TRUE;
1424        }
1425    
1426      /* Handle comments; all characters are allowed until a ket is reached. */
1427    
1428      else if (ptr[2] == CHAR_NUMBER_SIGN)
1429        {
1430        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1431        goto FAIL_EXIT;
1432        }
1433    
1434      /* Handle a condition. If it is an assertion, just carry on so that it
1435      is processed as normal. If not, skip to the closing parenthesis of the
1436      condition (there can't be any nested parens). */
1437    
1438      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1439        {
1440        ptr += 2;
1441        if (ptr[1] != CHAR_QUESTION_MARK)
1442          {
1443          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1444          if (*ptr != 0) ptr++;
1445          }
1446        }
1447    
1448      /* Start with (? but not a condition. */
1449    
1450      else
1451        {
1452        ptr += 2;
1453        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1454    
1455        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1456    
1457        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1458            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1459          {
1460          pcre_uchar term;
1461          const pcre_uchar *thisname;
1462          *count += 1;
1463          if (name == NULL && *count == lorn) return *count;
1464          term = *ptr++;
1465          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1466          thisname = ptr;
1467          while (*ptr != term) ptr++;
1468          if (name != NULL && lorn == (int)(ptr - thisname) &&
1469              STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
1470            return *count;
1471          term++;
1472          }
1473        }
1474      }
1475    
1476    /* Past any initial parenthesis handling, scan for parentheses or vertical
1477    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1478    first-pass call when this value is temporarily adjusted to stop at the current
1479    position. So DO NOT change this to a test for binary zero. */
1480    
1481    for (; ptr < cd->end_pattern; ptr++)
1482      {
1483    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1484    
1485    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1486      {      {
1487      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1488      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1489        {        {
1490        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1491        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1492        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1493        }        }
1494      continue;      continue;
1495      }      }
# Line 1012  for (; *ptr != 0; ptr++) Line 1497  for (; *ptr != 0; ptr++)
1497    /* Skip over character classes; this logic must be similar to the way they    /* Skip over character classes; this logic must be similar to the way they
1498    are handled for real. If the first character is '^', skip it. Also, if the    are handled for real. If the first character is '^', skip it. Also, if the
1499    first few characters (either before or after ^) are \Q\E or \E we skip them    first few characters (either before or after ^) are \Q\E or \E we skip them
1500    too. This makes for compatibility with Perl. */    too. This makes for compatibility with Perl. Note the use of STR macros to
1501      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1502    
1503    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1504      {      {
1505      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1506      for (;;)      for (;;)
1507        {        {
1508        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == '\\')  
1509          {          {
1510          if (ptr[1] == 'E') ptr++;          if (ptr[2] == CHAR_E)
1511            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr+= 2;
1512              else break;          else if (STRNCMP_UC_C8(ptr + 2,
1513                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1514              ptr += 4;
1515            else
1516              break;
1517          }          }
1518        else if (!negate_class && c == '^')        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1519            {
1520          negate_class = TRUE;          negate_class = TRUE;
1521            ptr++;
1522            }
1523        else break;        else break;
1524        }        }
1525    
1526      /* If the next character is ']', it is a data character that must be      /* If the next character is ']', it is a data character that must be
1527      skipped, except in JavaScript compatibility mode. */      skipped, except in JavaScript compatibility mode. */
1528    
1529      if (ptr[1] == ']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1530        ptr++;          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1531          ptr++;
1532      while (*(++ptr) != ']')  
1533        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1534        {        {
1535        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1536        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1537          {          {
1538          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1539          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1540            {            {
1541            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1542            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1543            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1544            }            }
1545          continue;          continue;
1546          }          }
# Line 1057  for (; *ptr != 0; ptr++) Line 1550  for (; *ptr != 0; ptr++)
1550    
1551    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1552    
1553    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1554      {      {
1555      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1556      if (*ptr == 0) return -1;      while (*ptr != 0)
1557          {
1558          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1559          ptr++;
1560    #ifdef SUPPORT_UTF
1561          if (utf) FORWARDCHAR(ptr);
1562    #endif
1563          }
1564        if (*ptr == 0) goto FAIL_EXIT;
1565      continue;      continue;
1566      }      }
1567    
1568    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1569    
1570    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1571      {      {
1572      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1573      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1574      continue;      if (*ptr == 0) goto FAIL_EXIT;
1575      }      }
1576    
1577    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1578    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1579        if (dup_parens && *count < hwm_count) *count = hwm_count;
1580    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1581        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1582    
1583    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1584    term = *ptr++;      {
1585    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1586    thisname = ptr;      *count = start_count;
1587    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1588    }    }
1589    
1590    FAIL_EXIT:
1591    *ptrptr = ptr;
1592  return -1;  return -1;
1593  }  }
1594    
1595    
1596    
1597    
1598  /*************************************************  /*************************************************
1599  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1600  *************************************************/  *************************************************/
1601    
1602  /* This is called by several functions that scan a compiled expression looking  /* This function scans along a pattern's text looking for capturing
1603  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
1604  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
1605  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
1606  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
1607    current compiling point, using the current count value from cd->bracount, and
1608    do it all in a single loop, but the addition of the possibility of duplicate
1609    subpattern numbers means that we have to scan from the very start, in order to
1610    take account of such duplicates, and to use a recursive function to keep track
1611    of the different types of group.
1612    
1613  Arguments:  Arguments:
1614    code         pointer to the start of the group    cd           compile background data
1615    options      pointer to external options    name         name to seek, or NULL if seeking a numbered subpattern
1616    optbit       the option bit whose changing is significant, or    lorn         name length, or subpattern number if name is NULL
1617                   zero if none are    xmode        TRUE if we are in /x mode
1618    skipassert   TRUE if certain assertions are to be skipped    utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1619    
1620  Returns:       pointer to the first significant opcode  Returns:       the number of the found subpattern, or -1 if not found
1621  */  */
1622    
1623  static const uschar*  static int
1624  first_significant_code(const uschar *code, int *options, int optbit,  find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1625    BOOL skipassert)    BOOL utf)
1626  {  {
1627    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1628    int count = 0;
1629    int rc;
1630    
1631    /* If the pattern does not start with an opening parenthesis, the first call
1632    to find_parens_sub() will scan right to the end (if necessary). However, if it
1633    does start with a parenthesis, find_parens_sub() will return when it hits the
1634    matching closing parens. That is why we have to have a loop. */
1635    
1636  for (;;)  for (;;)
1637    {    {
1638    switch ((int)*code)    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1639      {    if (rc > 0 || *ptr++ == 0) break;
1640      case OP_OPT:    }
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
1641    
1642    return rc;
1643    }
1644    
1645    
1646    
1647    
1648    /*************************************************
1649    *      Find first significant op code            *
1650    *************************************************/
1651    
1652    /* This is called by several functions that scan a compiled expression looking
1653    for a fixed first character, or an anchoring op code etc. It skips over things
1654    that do not influence this. For some calls, it makes sense to skip negative
1655    forward and all backward assertions, and also the \b assertion; for others it
1656    does not.
1657    
1658    Arguments:
1659      code         pointer to the start of the group
1660      skipassert   TRUE if certain assertions are to be skipped
1661    
1662    Returns:       pointer to the first significant opcode
1663    */
1664    
1665    static const pcre_uchar*
1666    first_significant_code(const pcre_uchar *code, BOOL skipassert)
1667    {
1668    for (;;)
1669      {
1670      switch ((int)*code)
1671        {
1672      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1673      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1674      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1675      if (!skipassert) return code;      if (!skipassert) return code;
1676      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1677      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1678      break;      break;
1679    
1680      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1149  for (;;) Line 1684  for (;;)
1684    
1685      case OP_CALLOUT:      case OP_CALLOUT:
1686      case OP_CREF:      case OP_CREF:
1687        case OP_NCREF:
1688      case OP_RREF:      case OP_RREF:
1689        case OP_NRREF:
1690      case OP_DEF:      case OP_DEF:
1691      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1692      break;      break;
1693    
1694      default:      default:
# Line 1165  for (;;) Line 1702  for (;;)
1702    
1703    
1704  /*************************************************  /*************************************************
1705  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1706  *************************************************/  *************************************************/
1707    
1708  /* 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,
1709  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.
1710  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
1711    temporarily terminated with OP_END when this function is called.
1712    
1713    This function is called when a backward assertion is encountered, so that if it
1714    fails, the error message can point to the correct place in the pattern.
1715    However, we cannot do this when the assertion contains subroutine calls,
1716    because they can be forward references. We solve this by remembering this case
1717    and doing the check at the end; a flag specifies which mode we are running in.
1718    
1719  Arguments:  Arguments:
1720    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1721    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1722      atend    TRUE if called when the pattern is complete
1723  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1724               or -2 if \C was encountered  
1725    Returns:   the fixed length,
1726                 or -1 if there is no fixed length,
1727                 or -2 if \C was encountered (in UTF-8 mode only)
1728                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1729                 or -4 if an unknown opcode was encountered (internal error)
1730  */  */
1731    
1732  static int  static int
1733  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1734  {  {
1735  int length = -1;  int length = -1;
1736    
1737  register int branchlength = 0;  register int branchlength = 0;
1738  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1739    
1740  /* 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
1741  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1194  branch, check the length against that of Line 1743  branch, check the length against that of
1743  for (;;)  for (;;)
1744    {    {
1745    int d;    int d;
1746    register int op = *cc;    pcre_uchar *ce, *cs;
1747      register pcre_uchar op = *cc;
1748    
1749    switch (op)    switch (op)
1750      {      {
1751        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1752        OP_BRA (normal non-capturing bracket) because the other variants of these
1753        opcodes are all concerned with unlimited repeated groups, which of course
1754        are not of fixed length. */
1755    
1756      case OP_CBRA:      case OP_CBRA:
1757      case OP_BRA:      case OP_BRA:
1758      case OP_ONCE:      case OP_ONCE:
1759        case OP_ONCE_NC:
1760      case OP_COND:      case OP_COND:
1761      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1762      if (d < 0) return d;      if (d < 0) return d;
1763      branchlength += d;      branchlength += d;
1764      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1765      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1766      break;      break;
1767    
1768      /* 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.
1769      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
1770      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
1771        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1772        because they all imply an unlimited repeat. */
1773    
1774      case OP_ALT:      case OP_ALT:
1775      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1776      case OP_END:      case OP_END:
1777        case OP_ACCEPT:
1778        case OP_ASSERT_ACCEPT:
1779      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1780        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1781      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1224  for (;;) Line 1783  for (;;)
1783      branchlength = 0;      branchlength = 0;
1784      break;      break;
1785    
1786        /* A true recursion implies not fixed length, but a subroutine call may
1787        be OK. If the subroutine is a forward reference, we can't deal with
1788        it until the end of the pattern, so return -3. */
1789    
1790        case OP_RECURSE:
1791        if (!atend) return -3;
1792        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1793        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1794        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1795        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1796        if (d < 0) return d;
1797        branchlength += d;
1798        cc += 1 + LINK_SIZE;
1799        break;
1800    
1801      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1802    
1803      case OP_ASSERT:      case OP_ASSERT:
# Line 1231  for (;;) Line 1805  for (;;)
1805      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1806      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1807      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1808      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1809        break;
1810    
1811      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1812    
1813      case OP_REVERSE:      case OP_MARK:
1814        case OP_PRUNE_ARG:
1815        case OP_SKIP_ARG:
1816        case OP_THEN_ARG:
1817        cc += cc[1] + PRIV(OP_lengths)[*cc];
1818        break;
1819    
1820        case OP_CALLOUT:
1821        case OP_CIRC:
1822        case OP_CIRCM:
1823        case OP_CLOSE:
1824        case OP_COMMIT:
1825      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1826      case OP_DEF:      case OP_DEF:
1827      case OP_OPT:      case OP_DOLL:
1828      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1829      case OP_EOD:      case OP_EOD:
1830      case OP_EODN:      case OP_EODN:
1831      case OP_CIRC:      case OP_FAIL:
1832      case OP_DOLL:      case OP_NCREF:
1833        case OP_NRREF:
1834      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1835        case OP_PRUNE:
1836        case OP_REVERSE:
1837        case OP_RREF:
1838        case OP_SET_SOM:
1839        case OP_SKIP:
1840        case OP_SOD:
1841        case OP_SOM:
1842        case OP_THEN:
1843      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1844      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1845      break;      break;
1846    
1847      /* Handle literal characters */      /* Handle literal characters */
1848    
1849      case OP_CHAR:      case OP_CHAR:
1850      case OP_CHARNC:      case OP_CHARI:
1851      case OP_NOT:      case OP_NOT:
1852        case OP_NOTI:
1853      branchlength++;      branchlength++;
1854      cc += 2;      cc += 2;
1855  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1856      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1857  #endif  #endif
1858      break;      break;
1859    
# Line 1271  for (;;) Line 1861  for (;;)
1861      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1862    
1863      case OP_EXACT:      case OP_EXACT:
1864      branchlength += GET2(cc,1);      case OP_EXACTI:
1865      cc += 4;      case OP_NOTEXACT:
1866  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1867      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1868        {      cc += 2 + IMM2_SIZE;
1869        while((*cc & 0x80) == 0x80) cc++;  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1870        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1871  #endif  #endif
1872      break;      break;
1873    
1874      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1875      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1876      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1877      cc += 4;        cc += 2;
1878        cc += 1 + IMM2_SIZE + 1;
1879      break;      break;
1880    
1881      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1294  for (;;) Line 1885  for (;;)
1885      cc += 2;      cc += 2;
1886      /* Fall through */      /* Fall through */
1887    
1888        case OP_HSPACE:
1889        case OP_VSPACE:
1890        case OP_NOT_HSPACE:
1891        case OP_NOT_VSPACE:
1892      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1893      case OP_DIGIT:      case OP_DIGIT:
1894      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1301  for (;;) Line 1896  for (;;)
1896      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1897      case OP_WORDCHAR:      case OP_WORDCHAR:
1898      case OP_ANY:      case OP_ANY:
1899        case OP_ALLANY:
1900      branchlength++;      branchlength++;
1901      cc++;      cc++;
1902      break;      break;
1903    
1904      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1905        otherwise \C is coded as OP_ALLANY. */
1906    
1907      case OP_ANYBYTE:      case OP_ANYBYTE:
1908      return -2;      return -2;
1909    
1910      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1911    
1912  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1913      case OP_XCLASS:      case OP_XCLASS:
1914      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1915      /* Fall through */      /* Fall through */
1916  #endif  #endif
1917    
1918      case OP_CLASS:      case OP_CLASS:
1919      case OP_NCLASS:      case OP_NCLASS:
1920      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1921    
1922      switch (*cc)      switch (*cc)
1923        {        {
1924          case OP_CRPLUS:
1925          case OP_CRMINPLUS:
1926        case OP_CRSTAR:        case OP_CRSTAR:
1927        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1928        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1332  for (;;) Line 1931  for (;;)
1931    
1932        case OP_CRRANGE:        case OP_CRRANGE:
1933        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1934        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1935        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1936        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1937        break;        break;
1938    
1939        default:        default:
# Line 1344  for (;;) Line 1943  for (;;)
1943    
1944      /* Anything else is variable length */      /* Anything else is variable length */
1945    
1946      default:      case OP_ANYNL:
1947        case OP_BRAMINZERO:
1948        case OP_BRAPOS:
1949        case OP_BRAPOSZERO:
1950        case OP_BRAZERO:
1951        case OP_CBRAPOS:
1952        case OP_EXTUNI:
1953        case OP_KETRMAX:
1954        case OP_KETRMIN:
1955        case OP_KETRPOS:
1956        case OP_MINPLUS:
1957        case OP_MINPLUSI:
1958        case OP_MINQUERY:
1959        case OP_MINQUERYI:
1960        case OP_MINSTAR:
1961        case OP_MINSTARI:
1962        case OP_MINUPTO:
1963        case OP_MINUPTOI:
1964        case OP_NOTMINPLUS:
1965        case OP_NOTMINPLUSI:
1966        case OP_NOTMINQUERY:
1967        case OP_NOTMINQUERYI:
1968        case OP_NOTMINSTAR:
1969        case OP_NOTMINSTARI:
1970        case OP_NOTMINUPTO:
1971        case OP_NOTMINUPTOI:
1972        case OP_NOTPLUS:
1973        case OP_NOTPLUSI:
1974        case OP_NOTPOSPLUS:
1975        case OP_NOTPOSPLUSI:
1976        case OP_NOTPOSQUERY:
1977        case OP_NOTPOSQUERYI:
1978        case OP_NOTPOSSTAR:
1979        case OP_NOTPOSSTARI:
1980        case OP_NOTPOSUPTO:
1981        case OP_NOTPOSUPTOI:
1982        case OP_NOTQUERY:
1983        case OP_NOTQUERYI:
1984        case OP_NOTSTAR:
1985        case OP_NOTSTARI:
1986        case OP_NOTUPTO:
1987        case OP_NOTUPTOI:
1988        case OP_PLUS:
1989        case OP_PLUSI:
1990        case OP_POSPLUS:
1991        case OP_POSPLUSI:
1992        case OP_POSQUERY:
1993        case OP_POSQUERYI:
1994        case OP_POSSTAR:
1995        case OP_POSSTARI:
1996        case OP_POSUPTO:
1997        case OP_POSUPTOI:
1998        case OP_QUERY:
1999        case OP_QUERYI:
2000        case OP_REF:
2001        case OP_REFI:
2002        case OP_SBRA:
2003        case OP_SBRAPOS:
2004        case OP_SCBRA:
2005        case OP_SCBRAPOS:
2006        case OP_SCOND:
2007        case OP_SKIPZERO:
2008        case OP_STAR:
2009        case OP_STARI:
2010        case OP_TYPEMINPLUS:
2011        case OP_TYPEMINQUERY:
2012        case OP_TYPEMINSTAR:
2013        case OP_TYPEMINUPTO:
2014        case OP_TYPEPLUS:
2015        case OP_TYPEPOSPLUS:
2016        case OP_TYPEPOSQUERY:
2017        case OP_TYPEPOSSTAR:
2018        case OP_TYPEPOSUPTO:
2019        case OP_TYPEQUERY:
2020        case OP_TYPESTAR:
2021        case OP_TYPEUPTO:
2022        case OP_UPTO:
2023        case OP_UPTOI:
2024      return -1;      return -1;
2025    
2026        /* Catch unrecognized opcodes so that when new ones are added they
2027        are not forgotten, as has happened in the past. */
2028    
2029        default:
2030        return -4;
2031      }      }
2032    }    }
2033  /* Control never gets here */  /* Control never gets here */
# Line 1355  for (;;) Line 2037  for (;;)
2037    
2038    
2039  /*************************************************  /*************************************************
2040  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2041  *************************************************/  *************************************************/
2042    
2043  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2044  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2045    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2046    so that it can be called from pcre_study() when finding the minimum matching
2047    length.
2048    
2049  Arguments:  Arguments:
2050    code        points to start of expression    code        points to start of expression
2051    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2052    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2053    
2054  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
2055  */  */
2056    
2057  static const uschar *  const pcre_uchar *
2058  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2059  {  {
2060  for (;;)  for (;;)
2061    {    {
2062    register int c = *code;    register pcre_uchar c = *code;
2063    
2064    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2065    
2066    /* 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 1383  for (;;) Line 2069  for (;;)
2069    
2070    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2071    
2072      /* Handle recursion */
2073    
2074      else if (c == OP_REVERSE)
2075        {
2076        if (number < 0) return (pcre_uchar *)code;
2077        code += PRIV(OP_lengths)[c];
2078        }
2079    
2080    /* Handle capturing bracket */    /* Handle capturing bracket */
2081    
2082    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2083               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2084      {      {
2085      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2086      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2087      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2088      }      }
2089    
2090    /* 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
2091    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
2092    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2093      must add in its length. */
2094    
2095    else    else
2096      {      {
# Line 1416  for (;;) Line 2112  for (;;)
2112        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2113        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2114        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2115        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2116            code += 2;
2117          break;
2118    
2119          case OP_MARK:
2120          case OP_PRUNE_ARG:
2121          case OP_SKIP_ARG:
2122          code += code[1];
2123          break;
2124    
2125          case OP_THEN_ARG:
2126          code += code[1];
2127        break;        break;
2128        }        }
2129    
2130      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2131    
2132      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2133    
2134    /* 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
2135    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
2136    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2137    
2138  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2139      if (utf8) switch(c)      if (utf) switch(c)
2140        {        {
2141        case OP_CHAR:        case OP_CHAR:
2142        case OP_CHARNC:        case OP_CHARI:
2143        case OP_EXACT:        case OP_EXACT:
2144          case OP_EXACTI:
2145        case OP_UPTO:        case OP_UPTO:
2146          case OP_UPTOI:
2147        case OP_MINUPTO:        case OP_MINUPTO:
2148          case OP_MINUPTOI:
2149        case OP_POSUPTO:        case OP_POSUPTO:
2150          case OP_POSUPTOI:
2151        case OP_STAR:        case OP_STAR:
2152          case OP_STARI:
2153        case OP_MINSTAR:        case OP_MINSTAR:
2154          case OP_MINSTARI:
2155        case OP_POSSTAR:        case OP_POSSTAR:
2156          case OP_POSSTARI:
2157        case OP_PLUS:        case OP_PLUS:
2158          case OP_PLUSI:
2159        case OP_MINPLUS:        case OP_MINPLUS:
2160          case OP_MINPLUSI:
2161        case OP_POSPLUS:        case OP_POSPLUS:
2162          case OP_POSPLUSI:
2163        case OP_QUERY:        case OP_QUERY:
2164          case OP_QUERYI:
2165        case OP_MINQUERY:        case OP_MINQUERY:
2166          case OP_MINQUERYI:
2167        case OP_POSQUERY:        case OP_POSQUERY:
2168        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2169          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2170        break;        break;
2171        }        }
2172    #else
2173        (void)(utf);  /* Keep compiler happy by referencing function argument */
2174  #endif  #endif
2175      }      }
2176    }    }
# Line 1465  instance of OP_RECURSE. Line 2187  instance of OP_RECURSE.
2187    
2188  Arguments:  Arguments:
2189    code        points to start of expression    code        points to start of expression
2190    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2191    
2192  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
2193  */  */
2194    
2195  static const uschar *  static const pcre_uchar *
2196  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2197  {  {
2198  for (;;)  for (;;)
2199    {    {
# Line 1487  for (;;) Line 2209  for (;;)
2209    
2210    /* 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
2211    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
2212    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2213      must add in its length. */
2214    
2215    else    else
2216      {      {
# Line 1509  for (;;) Line 2232  for (;;)
2232        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2233        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2234        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2235        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2236            code += 2;
2237          break;
2238    
2239          case OP_MARK:
2240          case OP_PRUNE_ARG:
2241          case OP_SKIP_ARG:
2242          code += code[1];
2243          break;
2244    
2245          case OP_THEN_ARG:
2246          code += code[1];
2247        break;        break;
2248        }        }
2249    
2250      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2251    
2252      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2253    
2254      /* 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
2255      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
2256      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2257    
2258  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2259      if (utf8) switch(c)      if (utf) switch(c)
2260        {        {
2261        case OP_CHAR:        case OP_CHAR:
2262        case OP_CHARNC:        case OP_CHARI:
2263          case OP_NOT:
2264          case OP_NOTI:
2265        case OP_EXACT:        case OP_EXACT:
2266          case OP_EXACTI:
2267          case OP_NOTEXACT:
2268          case OP_NOTEXACTI:
2269        case OP_UPTO:        case OP_UPTO:
2270          case OP_UPTOI:
2271          case OP_NOTUPTO:
2272          case OP_NOTUPTOI:
2273        case OP_MINUPTO:        case OP_MINUPTO:
2274          case OP_MINUPTOI:
2275          case OP_NOTMINUPTO:
2276          case OP_NOTMINUPTOI:
2277        case OP_POSUPTO:        case OP_POSUPTO:
2278          case OP_POSUPTOI:
2279          case OP_NOTPOSUPTO:
2280          case OP_NOTPOSUPTOI:
2281        case OP_STAR:        case OP_STAR:
2282          case OP_STARI:
2283          case OP_NOTSTAR:
2284          case OP_NOTSTARI:
2285        case OP_MINSTAR:        case OP_MINSTAR:
2286          case OP_MINSTARI:
2287          case OP_NOTMINSTAR:
2288          case OP_NOTMINSTARI:
2289        case OP_POSSTAR:        case OP_POSSTAR:
2290          case OP_POSSTARI:
2291          case OP_NOTPOSSTAR:
2292          case OP_NOTPOSSTARI:
2293        case OP_PLUS:        case OP_PLUS:
2294          case OP_PLUSI:
2295          case OP_NOTPLUS:
2296          case OP_NOTPLUSI:
2297        case OP_MINPLUS:        case OP_MINPLUS:
2298          case OP_MINPLUSI:
2299          case OP_NOTMINPLUS:
2300          case OP_NOTMINPLUSI:
2301        case OP_POSPLUS:        case OP_POSPLUS:
2302          case OP_POSPLUSI:
2303          case OP_NOTPOSPLUS:
2304          case OP_NOTPOSPLUSI:
2305        case OP_QUERY:        case OP_QUERY:
2306          case OP_QUERYI:
2307          case OP_NOTQUERY:
2308          case OP_NOTQUERYI:
2309        case OP_MINQUERY:        case OP_MINQUERY:
2310          case OP_MINQUERYI:
2311          case OP_NOTMINQUERY:
2312          case OP_NOTMINQUERYI:
2313        case OP_POSQUERY:        case OP_POSQUERY:
2314        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2315          case OP_NOTPOSQUERY:
2316          case OP_NOTPOSQUERYI:
2317          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2318        break;        break;
2319        }        }
2320    #else
2321        (void)(utf);  /* Keep compiler happy by referencing function argument */
2322  #endif  #endif
2323      }      }
2324    }    }
# Line 1564  bracket whose current branch will alread Line 2341  bracket whose current branch will alread
2341  Arguments:  Arguments:
2342    code        points to start of search    code        points to start of search
2343    endcode     points to where to stop    endcode     points to where to stop
2344    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2345      cd          contains pointers to tables etc.
2346    
2347  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2348  */  */
2349    
2350  static BOOL  static BOOL
2351  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2352      BOOL utf, compile_data *cd)
2353  {  {
2354  register int c;  register int c;
2355  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2356       code < endcode;       code < endcode;
2357       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2358    {    {
2359    const uschar *ccode;    const pcre_uchar *ccode;
2360    
2361    c = *code;    c = *code;
2362    
# Line 1591  for (code = first_significant_code(code Line 2370  for (code = first_significant_code(code
2370      continue;      continue;
2371      }      }
2372    
2373      /* For a recursion/subroutine call, if its end has been reached, which
2374      implies a backward reference subroutine call, we can scan it. If it's a
2375      forward reference subroutine call, we can't. To detect forward reference
2376      we have to scan up the list that is kept in the workspace. This function is
2377      called only when doing the real compile, not during the pre-compile that
2378      measures the size of the compiled pattern. */
2379    
2380      if (c == OP_RECURSE)
2381        {
2382        const pcre_uchar *scode;
2383        BOOL empty_branch;
2384    
2385        /* Test for forward reference */
2386    
2387        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2388          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2389    
2390        /* Not a forward reference, test for completed backward reference */
2391    
2392        empty_branch = FALSE;
2393        scode = cd->start_code + GET(code, 1);
2394        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2395    
2396        /* Completed backwards reference */
2397    
2398        do
2399          {
2400          if (could_be_empty_branch(scode, endcode, utf, cd))
2401            {
2402            empty_branch = TRUE;
2403            break;
2404            }
2405          scode += GET(scode, 1);
2406          }
2407        while (*scode == OP_ALT);
2408    
2409        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2410        continue;
2411        }
2412    
2413    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2414    
2415    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2416          c == OP_BRAPOSZERO)
2417        {
2418        code += PRIV(OP_lengths)[c];
2419        do code += GET(code, 1); while (*code == OP_ALT);
2420        c = *code;
2421        continue;
2422        }
2423    
2424      /* A nested group that is already marked as "could be empty" can just be
2425      skipped. */
2426    
2427      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2428          c == OP_SCBRA || c == OP_SCBRAPOS)
2429      {      {
     code += _pcre_OP_lengths[c];  
2430      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2431      c = *code;      c = *code;
2432      continue;      continue;
# Line 1603  for (code = first_significant_code(code Line 2434  for (code = first_significant_code(code
2434    
2435    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2436    
2437    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2438          c == OP_CBRA || c == OP_CBRAPOS ||
2439          c == OP_ONCE || c == OP_ONCE_NC ||
2440          c == OP_COND)
2441      {      {
2442      BOOL empty_branch;      BOOL empty_branch;
2443      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2444    
2445      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2446        empty branch, so just skip over the conditional, because it could be empty.
2447        Otherwise, scan the individual branches of the group. */
2448    
2449      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;  
2450        code += GET(code, 1);        code += GET(code, 1);
2451        else
2452          {
2453          empty_branch = FALSE;
2454          do
2455            {
2456            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2457              empty_branch = TRUE;
2458            code += GET(code, 1);
2459            }
2460          while (*code == OP_ALT);
2461          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2462        }        }
2463      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2464      c = *code;      c = *code;
2465      continue;      continue;
2466      }      }
# Line 1629  for (code = first_significant_code(code Line 2471  for (code = first_significant_code(code
2471      {      {
2472      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2473      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2474      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2475      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"
2476      here. */      here. */
2477    
2478  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2479      case OP_XCLASS:      case OP_XCLASS:
2480      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2481      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1641  for (code = first_significant_code(code Line 2483  for (code = first_significant_code(code
2483    
2484      case OP_CLASS:      case OP_CLASS:
2485      case OP_NCLASS:      case OP_NCLASS:
2486      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2487    
2488  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2489      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2490  #endif  #endif
2491    
# Line 1679  for (code = first_significant_code(code Line 2521  for (code = first_significant_code(code
2521      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2522      case OP_WORDCHAR:      case OP_WORDCHAR:
2523      case OP_ANY:      case OP_ANY:
2524        case OP_ALLANY:
2525      case OP_ANYBYTE:      case OP_ANYBYTE:
2526      case OP_CHAR:      case OP_CHAR:
2527      case OP_CHARNC:      case OP_CHARI:
2528      case OP_NOT:      case OP_NOT:
2529        case OP_NOTI:
2530      case OP_PLUS:      case OP_PLUS:
2531      case OP_MINPLUS:      case OP_MINPLUS:
2532      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1714  for (code = first_significant_code(code Line 2558  for (code = first_significant_code(code
2558      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2559      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2560      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2561      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2562          code += 2;
2563      break;      break;
2564    
2565      /* End of branch */      /* End of branch */
# Line 1722  for (code = first_significant_code(code Line 2567  for (code = first_significant_code(code
2567      case OP_KET:      case OP_KET:
2568      case OP_KETRMAX:      case OP_KETRMAX:
2569      case OP_KETRMIN:      case OP_KETRMIN:
2570        case OP_KETRPOS:
2571      case OP_ALT:      case OP_ALT:
2572      return TRUE;      return TRUE;
2573    
2574      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2575      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2576    
2577  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2578      case OP_STAR:      case OP_STAR:
2579        case OP_STARI:
2580      case OP_MINSTAR:      case OP_MINSTAR:
2581        case OP_MINSTARI:
2582      case OP_POSSTAR:      case OP_POSSTAR:
2583        case OP_POSSTARI:
2584      case OP_QUERY:      case OP_QUERY:
2585        case OP_QUERYI:
2586      case OP_MINQUERY:      case OP_MINQUERY:
2587        case OP_MINQUERYI:
2588      case OP_POSQUERY:      case OP_POSQUERY:
2589        case OP_POSQUERYI:
2590        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2591        break;
2592    
2593      case OP_UPTO:      case OP_UPTO:
2594        case OP_UPTOI:
2595      case OP_MINUPTO:      case OP_MINUPTO:
2596        case OP_MINUPTOI:
2597      case OP_POSUPTO:      case OP_POSUPTO:
2598      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2599        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2600      break;      break;
2601  #endif  #endif
2602    
2603        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2604        string. */
2605    
2606        case OP_MARK:
2607        case OP_PRUNE_ARG:
2608        case OP_SKIP_ARG:
2609        code += code[1];
2610        break;
2611    
2612        case OP_THEN_ARG:
2613        code += code[1];
2614        break;
2615    
2616        /* None of the remaining opcodes are required to match a character. */
2617    
2618        default:
2619        break;
2620      }      }
2621    }    }
2622    
# Line 1757  return TRUE; Line 2633  return TRUE;
2633  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
2634  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,
2635  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.
2636    This function is called only during the real compile, not during the
2637    pre-compile.
2638    
2639  Arguments:  Arguments:
2640    code        points to start of the recursion    code        points to start of the recursion
2641    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2642    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2643    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2644      cd          pointers to tables etc
2645    
2646  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2647  */  */
2648    
2649  static BOOL  static BOOL
2650  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2651    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2652  {  {
2653  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2654    {    {
2655    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2656        return FALSE;
2657    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2658    }    }
2659  return TRUE;  return TRUE;
# Line 1805  where Perl recognizes it as the POSIX cl Line 2685  where Perl recognizes it as the POSIX cl
2685  "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,
2686  I think.  I think.
2687    
2688    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2689    It seems that the appearance of a nested POSIX class supersedes an apparent
2690    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2691    a digit.
2692    
2693    In Perl, unescaped square brackets may also appear as part of class names. For
2694    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2695    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2696    seem right at all. PCRE does not allow closing square brackets in POSIX class
2697    names.
2698    
2699  Arguments:  Arguments:
2700    ptr      pointer to the initial [    ptr      pointer to the initial [
2701    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1813  Returns:   TRUE or FALSE Line 2704  Returns:   TRUE or FALSE
2704  */  */
2705    
2706  static BOOL  static BOOL
2707  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2708  {  {
2709  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2710  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2711  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2712    {    {
2713    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2714        ptr++;
2715      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2716      else
2717      {      {
2718      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2719        {        {
2720        *endptr = ptr;        *endptr = ptr;
2721        return TRUE;        return TRUE;
2722        }        }
2723        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2724             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2725              ptr[1] == CHAR_EQUALS_SIGN) &&
2726            check_posix_syntax(ptr, endptr))
2727          return FALSE;
2728      }      }
2729    }    }
2730  return FALSE;  return FALSE;
# Line 1850  Returns:     a value representing the na Line 2748  Returns:     a value representing the na
2748  */  */
2749    
2750  static int  static int
2751  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2752  {  {
2753  const char *pn = posix_names;  const char *pn = posix_names;
2754  register int yield = 0;  register int yield = 0;
2755  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2756    {    {
2757    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2758      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2759    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2760    yield++;    yield++;
2761    }    }
# Line 1889  value in the reference (which is a group Line 2787  value in the reference (which is a group
2787  Arguments:  Arguments:
2788    group      points to the start of the group    group      points to the start of the group
2789    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2790    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2791    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2792    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
2793    
# Line 1897  Returns:     nothing Line 2795  Returns:     nothing
2795  */  */
2796    
2797  static void  static void
2798  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2799    uschar *save_hwm)    pcre_uchar *save_hwm)
2800  {  {
2801  uschar *ptr = group;  pcre_uchar *ptr = group;
2802    
2803  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2804    {    {
2805    int offset;    int offset;
2806    uschar *hc;    pcre_uchar *hc;
2807    
2808    /* 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
2809    reference. */    reference. */
2810    
2811    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2812      {      {
2813      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1950  Arguments: Line 2848  Arguments:
2848  Returns:         new code pointer  Returns:         new code pointer
2849  */  */
2850    
2851  static uschar *  static pcre_uchar *
2852  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2853  {  {
2854  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2855  *code++ = 255;  *code++ = 255;
2856  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2857  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2858  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2859  }  }
2860    
2861    
# Line 1979  Returns:             nothing Line 2877  Returns:             nothing
2877  */  */
2878    
2879  static void  static void
2880  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2881  {  {
2882  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2883  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2884  }  }
2885    
# Line 1993  PUT(previous_callout, 2 + LINK_SIZE, len Line 2891  PUT(previous_callout, 2 + LINK_SIZE, len
2891  *************************************************/  *************************************************/
2892    
2893  /* 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
2894  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
2895  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
2896  start address.  start address. A character with multiple other cases is returned on its own
2897    with a special return value.
2898    
2899  Arguments:  Arguments:
2900    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2003  Arguments: Line 2902  Arguments:
2902    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2903    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2904    
2905  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2906                   0 when a range is returned
2907                  >0 the CASESET offset for char with multiple other cases
2908                    in this case, ocptr contains the original
2909  */  */
2910    
2911  static BOOL  static int
2912  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
2913    unsigned int *odptr)    pcre_uint32 *odptr)
2914  {  {
2915  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
2916    int co;
2917    
2918    /* Find the first character that has an other case. If it has multiple other
2919    cases, return its case offset value. */
2920    
2921  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2922    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2923      if ((co = UCD_CASESET(c)) != 0)
2924        {
2925        *ocptr = c++;   /* Character that has the set */
2926        *cptr = c;      /* Rest of input range */
2927        return co;
2928        }
2929      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2930      }
2931    
2932  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2933    
2934  *ocptr = othercase;  *ocptr = othercase;
2935  next = othercase + 1;  next = othercase + 1;
2936    
2937  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2938    {    {
2939    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2940    next++;    next++;
2941    }    }
2942    
2943  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2944  *cptr = c;  *cptr = c;             /* Rest of input range */
2945    return 0;
2946    }
2947    
2948    
2949  return TRUE;  
2950    /*************************************************
2951    *        Check a character and a property        *
2952    *************************************************/
2953    
2954    /* This function is called by check_auto_possessive() when a property item
2955    is adjacent to a fixed character.
2956    
2957    Arguments:
2958      c            the character
2959      ptype        the property type
2960      pdata        the data for the type
2961      negated      TRUE if it's a negated property (\P or \p{^)
2962    
2963    Returns:       TRUE if auto-possessifying is OK
2964    */
2965    
2966    static BOOL
2967    check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)
2968    {
2969    #ifdef SUPPORT_UCP
2970    const pcre_uint32 *p;
2971    #endif
2972    
2973    const ucd_record *prop = GET_UCD(c);
2974    
2975    switch(ptype)
2976      {
2977      case PT_LAMP:
2978      return (prop->chartype == ucp_Lu ||
2979              prop->chartype == ucp_Ll ||
2980              prop->chartype == ucp_Lt) == negated;
2981    
2982      case PT_GC:
2983      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2984    
2985      case PT_PC:
2986      return (pdata == prop->chartype) == negated;
2987    
2988      case PT_SC:
2989      return (pdata == prop->script) == negated;
2990    
2991      /* These are specials */
2992    
2993      case PT_ALNUM:
2994      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2995              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2996    
2997      case PT_SPACE:    /* Perl space */
2998      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2999              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
3000              == negated;
3001    
3002      case PT_PXSPACE:  /* POSIX space */
3003      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3004              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3005              c == CHAR_FF || c == CHAR_CR)
3006              == negated;
3007    
3008      case PT_WORD:
3009      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3010              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3011              c == CHAR_UNDERSCORE) == negated;
3012    
3013    #ifdef SUPPORT_UCP
3014      case PT_CLIST:
3015      p = PRIV(ucd_caseless_sets) + prop->caseset;
3016      for (;;)
3017        {
3018        if ((unsigned int)c < *p) return !negated;
3019        if ((unsigned int)c == *p++) return negated;
3020        }
3021      break;  /* Control never reaches here */
3022    #endif
3023      }
3024    
3025    return FALSE;
3026  }  }
3027  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3028    
# Line 2044  whether the next thing could possibly ma Line 3037  whether the next thing could possibly ma
3037  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3038    
3039  Arguments:  Arguments:
3040    op_code       the repeated op code    previous      pointer to the repeated opcode
3041    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  
3042    ptr           next character in pattern    ptr           next character in pattern
3043    options       options bits    options       options bits
3044    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2056  Returns:        TRUE if possessifying is Line 3047  Returns:        TRUE if possessifying is
3047  */  */
3048    
3049  static BOOL  static BOOL
3050  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3051    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3052  {  {
3053  int next;  pcre_uint32 c = NOTACHAR;
3054    pcre_uint32 next;
3055    int escape;
3056    int op_code = *previous++;
3057    
3058  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3059    
# Line 2067  if ((options & PCRE_EXTENDED) != 0) Line 3061  if ((options & PCRE_EXTENDED) != 0)
3061    {    {
3062    for (;;)    for (;;)
3063      {      {
3064      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3065      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3066        {        {
3067        while (*(++ptr) != 0)        ptr++;
3068          while (*ptr != 0)
3069            {
3070          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3071            ptr++;
3072    #ifdef SUPPORT_UTF
3073            if (utf) FORWARDCHAR(ptr);
3074    #endif
3075            }
3076        }        }
3077      else break;      else break;
3078      }      }
# Line 2080  if ((options & PCRE_EXTENDED) != 0) Line 3081  if ((options & PCRE_EXTENDED) != 0)
3081  /* 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
3082  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3083    
3084  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3085    {    {
3086    int temperrorcode = 0;    int temperrorcode = 0;
3087    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3088    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3089    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3090    }    }
3091    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3092    {    {
3093  #ifdef SUPPORT_UTF8    escape = 0;
3094    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3095      if (utf) { GETCHARINC(next, ptr); } else
3096  #endif  #endif
3097    next = *ptr++;    next = *ptr++;
3098    }    }
   
3099  else return FALSE;  else return FALSE;
3100    
3101  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2104  if ((options & PCRE_EXTENDED) != 0) Line 3104  if ((options & PCRE_EXTENDED) != 0)
3104    {    {
3105    for (;;)    for (;;)
3106      {      {
3107      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3108      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3109        {        {
3110        while (*(++ptr) != 0)        ptr++;
3111          while (*ptr != 0)
3112            {
3113          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3114            ptr++;
3115    #ifdef SUPPORT_UTF
3116            if (utf) FORWARDCHAR(ptr);
3117    #endif
3118            }
3119        }        }
3120      else break;      else break;
3121      }      }
# Line 2116  if ((options & PCRE_EXTENDED) != 0) Line 3123  if ((options & PCRE_EXTENDED) != 0)
3123    
3124  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3125    
3126  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3127    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3128        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. */  
   
3129    
3130  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3131    
3132  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3133        op_code == OP_NOT || op_code == OP_NOTI)
3134      //if (escape == 0) switch(op_code)
3135    {    {
3136    case OP_CHAR:  #ifdef SUPPORT_UTF
3137  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3138    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3139      c = *previous;
3140  #endif  #endif
3141    return item != next;    }
3142    
3143    /* For CHARNC (caseless character) we must check the other case. If we have  /* Now compare the next item with the previous opcode. First, handle cases when
3144    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3145    
3146    case OP_CHARNC:  if (escape == 0)
3147  #ifdef SUPPORT_UTF8    {
3148    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
3149      case, which maps to the special PT_CLIST property. Check this first. */
3150    
3151    #ifdef SUPPORT_UCP
3152      if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3153        {
3154        int ocs = UCD_CASESET(next);
3155        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3156        }
3157  #endif  #endif
3158    if (item == next) return FALSE;  
3159  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3160      {      {
3161      unsigned int othercase;      case OP_CHAR:
3162      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3163    
3164        /* For CHARI (caseless character) we must check the other case. If we have
3165        Unicode property support, we can use it to test the other case of
3166        high-valued characters. We know that next can have only one other case,
3167        because multi-other-case characters are dealt with above. */
3168    
3169        case OP_CHARI:
3170        if (c == next) return FALSE;
3171    #ifdef SUPPORT_UTF
3172        if (utf)
3173          {
3174          pcre_uint32 othercase;
3175          if (next < 128) othercase = cd->fcc[next]; else
3176  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3177      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE(next);
3178  #else  #else
3179      othercase = NOTACHAR;        othercase = NOTACHAR;
3180  #endif  #endif
3181      return (unsigned int)item != othercase;        return c != othercase;
3182      }        }
3183    else      else
3184  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3185    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3186    
3187    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3188        return c == next;
3189    case OP_NOT:  
3190    if (item == next) return TRUE;      case OP_NOTI:
3191    if ((options & PCRE_CASELESS) == 0) return FALSE;      if (c == next) return TRUE;
3192  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3193    if (utf8)      if (utf)
3194      {        {
3195      unsigned int othercase;        pcre_uint32 othercase;
3196      if (next < 128) othercase = cd->fcc[next]; else        if (next < 128) othercase = cd->fcc[next]; else
3197  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3198      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE(next);
3199  #else  #else
3200      othercase = NOTACHAR;        othercase = NOTACHAR;
3201  #endif  #endif
3202      return (unsigned int)item == othercase;        return c == othercase;
3203      }        }
3204    else      else
3205  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3206    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3207    
3208    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3209    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3210    
3211    case OP_NOT_DIGIT:      case OP_DIGIT:
3212    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3213    
3214    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3215    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3216    
3217    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3218    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3219    
3220    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3221    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3222    
3223    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3224    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3225    
3226    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3227    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3228    switch(next)  
3229      {      case OP_HSPACE:
3230      case 0x09:      case OP_NOT_HSPACE:
3231      case 0x20:      switch(next)
3232      case 0xa0:        {
3233      case 0x1680:        HSPACE_CASES:
3234      case 0x180e:        return op_code == OP_NOT_HSPACE;
3235      case 0x2000:  
3236      case 0x2001:        default:
3237      case 0x2002:        return op_code != OP_NOT_HSPACE;
3238      case 0x2003:        }
3239      case 0x2004:  
3240      case 0x2005:      case OP_ANYNL:
3241      case 0x2006:      case OP_VSPACE:
3242      case 0x2007:      case OP_NOT_VSPACE:
3243      case 0x2008:      switch(next)
3244      case 0x2009:        {
3245      case 0x200A:        VSPACE_CASES:
3246      case 0x202f:        return op_code == OP_NOT_VSPACE;
3247      case 0x205f:  
3248      case 0x3000:        default:
3249      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3250      default:        }
3251      return op_code == OP_HSPACE;  
3252      }  #ifdef SUPPORT_UCP
3253        case OP_PROP:
3254        return check_char_prop(next, previous[0], previous[1], FALSE);
3255    
3256        case OP_NOTPROP:
3257        return check_char_prop(next, previous[0], previous[1], TRUE);
3258    #endif
3259    
   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;  
3260      default:      default:
3261      return op_code == OP_VSPACE;      return FALSE;
3262      }      }
   
   default:  
   return FALSE;  
3263    }    }
3264    
3265    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3266  /* 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
3267    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3268    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3269    replaced by OP_PROP codes when PCRE_UCP is set. */
3270    
3271  switch(op_code)  switch(op_code)
3272    {    {
3273    case OP_CHAR:    case OP_CHAR:
3274    case OP_CHARNC:    case OP_CHARI:
3275  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3276      {      {
3277      case ESC_d:      case ESC_d:
3278      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3279    
3280      case ESC_D:      case ESC_D:
3281      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3282    
3283      case ESC_s:      case ESC_s:
3284      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3285    
3286      case ESC_S:      case ESC_S:
3287      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3288    
3289      case ESC_w:      case ESC_w:
3290      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3291    
3292      case ESC_W:      case ESC_W:
3293      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3294    
3295      case ESC_h:      case ESC_h:
3296      case ESC_H:      case ESC_H:
3297      switch(item)      switch(c)
3298        {        {
3299        case 0x09:        HSPACE_CASES:
3300        case 0x20:        return escape != ESC_h;
3301        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;  
3302        default:        default:
3303        return -next == ESC_h;        return escape == ESC_h;
3304        }        }
3305    
3306      case ESC_v:      case ESC_v:
3307      case ESC_V:      case ESC_V:
3308      switch(item)      switch(c)
3309        {        {
3310        case 0x0a:        VSPACE_CASES:
3311        case 0x0b:        return escape != ESC_v;
3312        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3313        default:        default:
3314        return -next == ESC_v;        return escape == ESC_v;
3315        }        }
3316    
3317        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3318        their substitutions and process them. The result will always be either
3319        ESC_p or ESC_P. Then fall through to process those values. */
3320    
3321    #ifdef SUPPORT_UCP
3322        case ESC_du:
3323        case ESC_DU:
3324        case ESC_wu:
3325        case ESC_WU:
3326        case ESC_su:
3327        case ESC_SU:
3328          {
3329          int temperrorcode = 0;
3330          ptr = substitutes[escape - ESC_DU];
3331          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3332          if (temperrorcode != 0) return FALSE;
3333          ptr++;    /* For compatibility */
3334          }
3335        /* Fall through */
3336    
3337        case ESC_p:
3338        case ESC_P:
3339          {
3340          int ptype, pdata, errorcodeptr;
3341          BOOL negated;
3342    
3343          ptr--;      /* Make ptr point at the p or P */
3344          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3345          if (ptype < 0) return FALSE;
3346          ptr++;      /* Point past the final curly ket */
3347    
3348          /* If the property item is optional, we have to give up. (When generated
3349          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3350          to the original \d etc. At this point, ptr will point to a zero byte. */
3351    
3352          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3353            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3354              return FALSE;
3355    
3356          /* Do the property check. */
3357    
3358          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3359          }
3360    #endif
3361    
3362      default:      default:
3363      return FALSE;      return FALSE;
3364      }      }
3365    
3366      /* In principle, support for Unicode properties should be integrated here as
3367      well. It means re-organizing the above code so as to get hold of the property
3368      values before switching on the op-code. However, I wonder how many patterns
3369      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3370      these op-codes are never generated.) */
3371    
3372    case OP_DIGIT:    case OP_DIGIT:
3373    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3374           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3375    
3376    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3377    return next == -ESC_d;    return escape == ESC_d;
3378    
3379    case OP_WHITESPACE:    case OP_WHITESPACE:
3380    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3381    
3382    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3383    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3384    
3385    case OP_HSPACE:    case OP_HSPACE:
3386    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3387             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3388    
3389    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3390    return next == -ESC_h;    return escape == ESC_h;
3391    
3392    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3393      case OP_ANYNL:
3394    case OP_VSPACE:    case OP_VSPACE:
3395    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3396    
3397    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3398    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3399    
3400    case OP_WORDCHAR:    case OP_WORDCHAR:
3401    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3402             escape == ESC_v || escape == ESC_R;
3403    
3404    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3405    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3406    
3407    default:    default:
3408    return FALSE;    return FALSE;
# Line 2367  switch(op_code) Line 3414  switch(op_code)
3414    
3415    
3416  /*************************************************  /*************************************************
3417    *        Add a character or range to a class     *
3418    *************************************************/
3419    
3420    /* This function packages up the logic of adding a character or range of
3421    characters to a class. The character values in the arguments will be within the
3422    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3423    mutually recursive with the function immediately below.
3424    
3425    Arguments:
3426      classbits     the bit map for characters < 256
3427      uchardptr     points to the pointer for extra data
3428      options       the options word
3429      cd            contains pointers to tables etc.
3430      start         start of range character
3431      end           end of range character
3432    
3433    Returns:        the number of < 256 characters added
3434                    the pointer to extra data is updated
3435    */
3436    
3437    static int
3438    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3439      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3440    {
3441    pcre_uint32 c;
3442    int n8 = 0;
3443    
3444    /* If caseless matching is required, scan the range and process alternate
3445    cases. In Unicode, there are 8-bit characters that have alternate cases that
3446    are greater than 255 and vice-versa. Sometimes we can just extend the original
3447    range. */
3448    
3449    if ((options & PCRE_CASELESS) != 0)
3450      {
3451    #ifdef SUPPORT_UCP
3452      if ((options & PCRE_UTF8) != 0)
3453        {
3454        int rc;
3455        pcre_uint32 oc, od;
3456    
3457        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3458        c = start;
3459    
3460        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3461          {
3462          /* Handle a single character that has more than one other case. */
3463    
3464          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3465            PRIV(ucd_caseless_sets) + rc, oc);
3466    
3467          /* Do nothing if the other case range is within the original range. */
3468    
3469          else if (oc >= start && od <= end) continue;
3470    
3471          /* Extend the original range if there is overlap, noting that if oc < c, we
3472          can't have od > end because a subrange is always shorter than the basic
3473          range. Otherwise, use a recursive call to add the additional range. */
3474    
3475          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3476          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3477          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3478          }
3479        }
3480      else
3481    #endif  /* SUPPORT_UCP */
3482    
3483      /* Not UTF-mode, or no UCP */
3484    
3485      for (c = start; c <= end && c < 256; c++)
3486        {
3487        SETBIT(classbits, cd->fcc[c]);
3488        n8++;
3489        }
3490      }
3491    
3492    /* Now handle the original range. Adjust the final value according to the bit
3493    length - this means that the same lists of (e.g.) horizontal spaces can be used
3494    in all cases. */
3495    
3496    #if defined COMPILE_PCRE8
3497    #ifdef SUPPORT_UTF
3498      if ((options & PCRE_UTF8) == 0)
3499    #endif
3500      if (end > 0xff) end = 0xff;
3501    
3502    #elif defined COMPILE_PCRE16
3503    #ifdef SUPPORT_UTF
3504      if ((options & PCRE_UTF16) == 0)
3505    #endif
3506      if (end > 0xffff) end = 0xffff;
3507    
3508    #endif /* COMPILE_PCRE[8|16] */
3509    
3510    /* If all characters are less than 256, use the bit map. Otherwise use extra
3511    data. */
3512    
3513    if (end < 0x100)
3514      {
3515      for (c = start; c <= end; c++)
3516        {
3517        n8++;
3518        SETBIT(classbits, c);
3519        }
3520      }
3521    
3522    else
3523      {
3524      pcre_uchar *uchardata = *uchardptr;
3525    
3526    #ifdef SUPPORT_UTF
3527      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3528        {
3529        if (start < end)
3530          {
3531          *uchardata++ = XCL_RANGE;
3532          uchardata += PRIV(ord2utf)(start, uchardata);
3533          uchardata += PRIV(ord2utf)(end, uchardata);
3534          }
3535        else if (start == end)
3536          {
3537          *uchardata++ = XCL_SINGLE;
3538          uchardata += PRIV(ord2utf)(start, uchardata);
3539          }
3540        }
3541      else
3542    #endif  /* SUPPORT_UTF */
3543    
3544      /* Without UTF support, character values are constrained by the bit length,
3545      and can only be > 256 for 16-bit and 32-bit libraries. */
3546    
3547    #ifdef COMPILE_PCRE8
3548        {}
3549    #else
3550      if (start < end)
3551        {
3552        *uchardata++ = XCL_RANGE;
3553        *uchardata++ = start;
3554        *uchardata++ = end;
3555        }
3556      else if (start == end)
3557        {
3558        *uchardata++ = XCL_SINGLE;
3559        *uchardata++ = start;
3560        }
3561    #endif
3562    
3563      *uchardptr = uchardata;   /* Updata extra data pointer */
3564      }
3565    
3566    return n8;    /* Number of 8-bit characters */
3567    }
3568    
3569    
3570    
3571    
3572    /*************************************************
3573    *        Add a list of characters to a class     *
3574    *************************************************/
3575    
3576    /* This function is used for adding a list of case-equivalent characters to a
3577    class, and also for adding a list of horizontal or vertical whitespace. If the
3578    list is in order (which it should be), ranges of characters are detected and
3579    handled appropriately. This function is mutually recursive with the function
3580    above.
3581    
3582    Arguments:
3583      classbits     the bit map for characters < 256
3584      uchardptr     points to the pointer for extra data
3585      options       the options word
3586      cd            contains pointers to tables etc.
3587      p             points to row of 32-bit values, terminated by NOTACHAR
3588      except        character to omit; this is used when adding lists of
3589                      case-equivalent characters to avoid including the one we
3590                      already know about
3591    
3592    Returns:        the number of < 256 characters added
3593                    the pointer to extra data is updated
3594    */
3595    
3596    static int
3597    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3598      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3599    {
3600    int n8 = 0;
3601    while (p[0] < NOTACHAR)
3602      {
3603      int n = 0;
3604      if (p[0] != except)
3605        {
3606        while(p[n+1] == p[0] + n + 1) n++;
3607        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3608        }
3609      p += n + 1;
3610      }
3611    return n8;
3612    }
3613    
3614    
3615    
3616    /*************************************************
3617    *    Add characters not in a list to a class     *
3618    *************************************************/
3619    
3620    /* This function is used for adding the complement of a list of horizontal or
3621    vertical whitespace to a class. The list must be in order.
3622    
3623    Arguments:
3624      classbits     the bit map for characters < 256
3625      uchardptr     points to the pointer for extra data
3626      options       the options word
3627      cd            contains pointers to tables etc.
3628      p             points to row of 32-bit values, terminated by NOTACHAR
3629    
3630    Returns:        the number of < 256 characters added
3631                    the pointer to extra data is updated
3632    */
3633    
3634    static int
3635    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3636      int options, compile_data *cd, const pcre_uint32 *p)
3637    {
3638    BOOL utf = (options & PCRE_UTF8) != 0;
3639    int n8 = 0;
3640    if (p[0] > 0)
3641      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3642    while (p[0] < NOTACHAR)
3643      {
3644      while (p[1] == p[0] + 1) p++;
3645      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3646        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3647      p++;
3648      }
3649    return n8;
3650    }
3651    
3652    
3653    
3654    /*************************************************
3655  *           Compile one branch                   *  *           Compile one branch                   *
3656  *************************************************/  *************************************************/
3657    
# Line 2381  Arguments: Line 3666  Arguments:
3666    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3667    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3668    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3669    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3670    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3671    bcptr          points to current branch chain    bcptr          points to current branch chain
3672      cond_depth     conditional nesting depth
3673    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3674    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3675                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2393  Returns:         TRUE on success Line 3679  Returns:         TRUE on success
3679  */  */
3680    
3681  static BOOL  static BOOL
3682  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3683    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3684      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3685    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3686  {  {
3687  int repeat_type, op_type;  int repeat_type, op_type;
3688  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3689  int bravalue = 0;  int bravalue = 0;
3690  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3691  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3692  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3693  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3694  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3695  int after_manual_callout = 0;  int after_manual_callout = 0;
3696  int length_prevgroup = 0;  int length_prevgroup = 0;
3697  register int c;  register pcre_uint32 c;
3698  register uschar *code = *codeptr;  int escape;
3699  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3700  uschar *orig_code = code;  pcre_uchar *last_code = code;
3701  uschar *tempcode;  pcre_uchar *orig_code = code;
3702    pcre_uchar *tempcode;
3703  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3704  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3705  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3706  const uschar *tempptr;  const pcre_uchar *tempptr;
3707  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3708  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3709  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3710  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3711    pcre_uint8 classbits[32];
3712  #ifdef SUPPORT_UTF8  
3713  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3714  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3715  uschar *class_utf8data;  dynamically as we process the pattern. */
3716  uschar *class_utf8data_base;  
3717  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3718    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3719    BOOL utf = (options & PCRE_UTF8) != 0;
3720    pcre_uchar utf_chars[6];
3721  #else  #else
3722  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3723  uschar *utf8_char = NULL;  #endif
3724    
3725    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3726    class_uchardata always so that it can be passed to add_to_class() always,
3727    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3728    alternative calls for the different cases. */
3729    
3730    pcre_uchar *class_uchardata;
3731    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3732    BOOL xclass;
3733    pcre_uchar *class_uchardata_base;
3734  #endif  #endif
3735    
3736  #ifdef DEBUG  #ifdef PCRE_DEBUG
3737  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3738  #endif  #endif
3739    
# Line 2443  greedy_non_default = greedy_default ^ 1; Line 3744  greedy_non_default = greedy_default ^ 1;
3744    
3745  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3746  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
3747  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
3748  find one.  find one.
3749    
3750  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
3751  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
3752  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3753  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3754    
3755  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3756    
3757  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3758  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
3759  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
3760  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3761    value. This is used only for ASCII characters. */
3762    
3763  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3764    
3765  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3766    
# Line 2470  for (;; ptr++) Line 3772  for (;; ptr++)
3772    BOOL is_quantifier;    BOOL is_quantifier;
3773    BOOL is_recurse;    BOOL is_recurse;
3774    BOOL reset_bracount;    BOOL reset_bracount;
3775    int class_charcount;    int class_has_8bitchar;
3776    int class_lastchar;    int class_one_char;
3777    int newoptions;    int newoptions;
3778    int recno;    int recno;
3779    int refsign;    int refsign;
3780    int skipbytes;    int skipbytes;
3781    int subreqbyte;    int subreqchar;
3782    int subfirstbyte;    int subfirstchar;
3783    int terminator;    int terminator;
3784    int mclength;    int mclength;
3785    uschar mcbuffer[8];    int tempbracount;
3786      int ec; // FIXMEchpe pcre_uint32
3787      pcre_uchar mcbuffer[8];
3788    
3789    /* Get next byte in the pattern */    /* Get next character in the pattern */
3790    
3791    c = *ptr;    c = *ptr;
3792    
3793      /* If we are at the end of a nested substitution, revert to the outer level
3794      string. Nesting only happens one level deep. */
3795    
3796      if (c == 0 && nestptr != NULL)
3797        {
3798        ptr = nestptr;
3799        nestptr = NULL;
3800        c = *ptr;
3801        }
3802    
3803    /* 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
3804    previous cycle of this loop. */    previous cycle of this loop. */
3805    
3806    if (lengthptr != NULL)    if (lengthptr != NULL)
3807      {      {
3808  #ifdef DEBUG  #ifdef PCRE_DEBUG
3809      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3810  #endif  #endif
3811      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3812            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3813        {        {
3814        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3815        goto FAILED;        goto FAILED;
# Line 2516  for (;; ptr++) Line 3831  for (;; ptr++)
3831        goto FAILED;        goto FAILED;
3832        }        }
3833    
3834      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3835      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3836          (int)(code - last_code), c, c));
3837    
3838      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3839      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2527  for (;; ptr++) Line 3843  for (;; ptr++)
3843        {        {
3844        if (previous > orig_code)        if (previous > orig_code)
3845          {          {
3846          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3847          code -= previous - orig_code;          code -= previous - orig_code;
3848          previous = orig_code;          previous = orig_code;
3849          }          }
# Line 2543  for (;; ptr++) Line 3859  for (;; ptr++)
3859    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3860    reference list. */    reference list. */
3861    
3862    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3863               WORK_SIZE_SAFETY_MARGIN)
3864      {      {
3865      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3866      goto FAILED;      goto FAILED;
# Line 2553  for (;; ptr++) Line 3870  for (;; ptr++)
3870    
3871    if (inescq && c != 0)    if (inescq && c != 0)
3872      {      {
3873      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3874        {        {
3875        inescq = FALSE;        inescq = FALSE;
3876        ptr++;        ptr++;
# Line 2579  for (;; ptr++) Line 3896  for (;; ptr++)
3896    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3897    a quantifier. */    a quantifier. */
3898    
3899    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3900      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3901        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3902    
3903    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3904         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2590  for (;; ptr++) Line 3908  for (;; ptr++)
3908      previous_callout = NULL;      previous_callout = NULL;
3909      }      }
3910    
3911    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3912    
3913    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3914      {      {
3915      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3916      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3917        {        {
3918        while (*(++ptr) != 0)        ptr++;
3919          while (*ptr != 0)
3920          {          {
3921          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3922            ptr++;
3923    #ifdef SUPPORT_UTF
3924            if (utf) FORWARDCHAR(ptr);
3925    #endif
3926          }          }
3927        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3928    
# Line 2620  for (;; ptr++) Line 3943  for (;; ptr++)
3943      {      {
3944      /* ===================================================================*/      /* ===================================================================*/
3945      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3946      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3947      case ')':      case CHAR_RIGHT_PARENTHESIS:
3948      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3949      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3950      *codeptr = code;      *codeptr = code;
3951      *ptrptr = ptr;      *ptrptr = ptr;
3952      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2633  for (;; ptr++) Line 3956  for (;; ptr++)
3956          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3957          goto FAILED;          goto FAILED;
3958          }          }
3959        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3960        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));