/[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 327 by ph10, Sat Mar 8 19:38:30 2008 UTC revision 1348 by ph10, Fri Jul 5 10:38:37 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118  #define COMPILE_WORK_SIZE (4096)  /* The overrun tests check for a slightly smaller size so that they detect the
119    overrun before it actually does run off the end of the data block. */
120    
121    #define WORK_SIZE_SAFETY_MARGIN (100)
122    
123    /* Private flags added to firstchar and reqchar. */
124    
125    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
126    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
127    /* Negative values for the firstchar and reqchar flags */
128    #define REQ_UNSET       (-2)
129    #define REQ_NONE        (-1)
130    
131    /* Repeated character flags. */
132    
133    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
134    
135  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
136  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
137  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
138  is invalid. */  is invalid. */
139    
140  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
141    
142    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
143    in UTF-8 mode. */
144    
145  static const short int escapes[] = {  static const short int escapes[] = {
146       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
147       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
148     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
149  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
150  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
151  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
152     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
153  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
154  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
155       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
156         -ESC_D,                  -ESC_E,
157         0,                       -ESC_G,
158         -ESC_H,                  0,
159         0,                       -ESC_K,
160         0,                       0,
161         -ESC_N,                  0,
162         -ESC_P,                  -ESC_Q,
163         -ESC_R,                  -ESC_S,
164         0,                       0,
165         -ESC_V,                  -ESC_W,
166         -ESC_X,                  0,
167         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
168         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
169         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
170         CHAR_GRAVE_ACCENT,       7,
171         -ESC_b,                  0,
172         -ESC_d,                  ESC_e,
173         ESC_f,                   0,
174         -ESC_h,                  0,
175         0,                       -ESC_k,
176         0,                       0,
177         ESC_n,                   0,
178         -ESC_p,                  0,
179         ESC_r,                   -ESC_s,
180         ESC_tee,                 0,
181         -ESC_v,                  -ESC_w,
182         0,                       0,
183         -ESC_z
184  };  };
185    
186  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
187    
188    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
189    
190  static const short int escapes[] = {  static const short int escapes[] = {
191  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
192  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 205  static const short int escapes[] = {
205  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
206  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
207  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
208  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
209  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
210  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
211  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 217  static const short int escapes[] = {
217    
218  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
219  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
220  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
221    string is built from string macros so that it works in UTF-8 mode on EBCDIC
222    platforms. */
223    
224  typedef struct verbitem {  typedef struct verbitem {
225    int   len;    int   len;                 /* Length of verb name */
226    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
227      int   op_arg;              /* Op when arg present, or -1 if not allowed */
228  } verbitem;  } verbitem;
229    
230  static const char verbnames[] =  static const char verbnames[] =
231    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
232    "COMMIT\0"    STRING_MARK0
233    "F\0"    STRING_ACCEPT0
234    "FAIL\0"    STRING_COMMIT0
235    "PRUNE\0"    STRING_F0
236    "SKIP\0"    STRING_FAIL0
237    "THEN";    STRING_PRUNE0
238      STRING_SKIP0
239      STRING_THEN;
240    
241  static const verbitem verbs[] = {  static const verbitem verbs[] = {
242    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
243    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
244    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
245    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
246    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
247    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
248    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
249      { 4, OP_SKIP,   OP_SKIP_ARG  },
250      { 4, OP_THEN,   OP_THEN_ARG  }
251  };  };
252    
253  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 178  length entry. The first three must be al Line 260  length entry. The first three must be al
260  for handling case independence. */  for handling case independence. */
261    
262  static const char posix_names[] =  static const char posix_names[] =
263    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
264    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
265    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
266      STRING_word0  STRING_xdigit;
267    
268  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
269    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
270    
271  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 212  static const int posix_class_maps[] = { Line 295  static const int posix_class_maps[] = {
295    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
296  };  };
297    
298    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
299    substitutes must be in the order of the names, defined above, and there are
300    both positive and negative cases. NULL means no substitute. */
301    
302    #ifdef SUPPORT_UCP
303    static const pcre_uchar string_PNd[]  = {
304      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_pNd[]  = {
307      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_PXsp[] = {
310      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_pXsp[] = {
313      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_PXwd[] = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pXwd[] = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    
322    static const pcre_uchar *substitutes[] = {
323      string_PNd,           /* \D */
324      string_pNd,           /* \d */
325      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
326      string_pXsp,          /* \s */
327      string_PXwd,          /* \W */
328      string_pXwd           /* \w */
329    };
330    
331    static const pcre_uchar string_pL[] =   {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLl[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pLu[] =  {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_pXan[] = {
341      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_h[] =    {
344      CHAR_BACKSLASH, CHAR_h, '\0' };
345    static const pcre_uchar string_pXps[] = {
346      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PL[] =   {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLl[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PLu[] =  {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_PXan[] = {
358      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
359      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
360    static const pcre_uchar string_H[] =    {
361      CHAR_BACKSLASH, CHAR_H, '\0' };
362    static const pcre_uchar string_PXps[] = {
363      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    
366    static const pcre_uchar *posix_substitutes[] = {
367      string_pL,            /* alpha */
368      string_pLl,           /* lower */
369      string_pLu,           /* upper */
370      string_pXan,          /* alnum */
371      NULL,                 /* ascii */
372      string_h,             /* blank */
373      NULL,                 /* cntrl */
374      string_pNd,           /* digit */
375      NULL,                 /* graph */
376      NULL,                 /* print */
377      NULL,                 /* punct */
378      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
379      string_pXwd,          /* word */
380      NULL,                 /* xdigit */
381      /* Negated cases */
382      string_PL,            /* ^alpha */
383      string_PLl,           /* ^lower */
384      string_PLu,           /* ^upper */
385      string_PXan,          /* ^alnum */
386      NULL,                 /* ^ascii */
387      string_H,             /* ^blank */
388      NULL,                 /* ^cntrl */
389      string_PNd,           /* ^digit */
390      NULL,                 /* ^graph */
391      NULL,                 /* ^print */
392      NULL,                 /* ^punct */
393      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
394      string_PXwd,          /* ^word */
395      NULL                  /* ^xdigit */
396    };
397    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
398    #endif
399    
400  #define STRING(a)  # a  #define STRING(a)  # a
401  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 408  the number of relocations needed when a
408  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
409  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
410  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
411  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
412    
413    Each substring ends with \0 to insert a null character. This includes the final
414    substring, so that the whole string ends with \0\0, which can be detected when
415    counting through. */
416    
417  static const char error_texts[] =  static const char error_texts[] =
418    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 453  static const char error_texts[] =
453    /* 30 */    /* 30 */
454    "unknown POSIX class name\0"    "unknown POSIX class name\0"
455    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
456    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
457    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
458    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
459    /* 35 */    /* 35 */
460    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
461    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
462    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
463    "number after (?C is > 255\0"    "number after (?C is > 255\0"
464    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
465    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 476  static const char error_texts[] =
476    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
477    /* 50 */    /* 50 */
478    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
479    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
480    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
481    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
482    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
483    /* 55 */    /* 55 */
484    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
485    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
486    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
487    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
488    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
489    /* 60 */    /* 60 */
490    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
491    "number is too big\0"    "number is too big\0"
492    "subpattern name expected\0"    "subpattern name expected\0"
493    "digit expected after (?+";    "digit expected after (?+\0"
494      "] is an invalid data character in JavaScript compatibility mode\0"
495      /* 65 */
496      "different names for subpatterns of the same number are not allowed\0"
497      "(*MARK) must have an argument\0"
498      "this version of PCRE is not compiled with Unicode property support\0"
499      "\\c must be followed by an ASCII character\0"
500      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
501      /* 70 */
502      "internal error: unknown opcode in find_fixedlength()\0"
503      "\\N is not supported in a class\0"
504      "too many forward references\0"
505      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
506      "invalid UTF-16 string\0"
507      /* 75 */
508      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
509      "character value in \\u.... sequence is too large\0"
510      "invalid UTF-32 string\0"
511      "setting UTF is disabled by the application\0"
512      ;
513    
514  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
515  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 527  For convenience, we use the same bit def
527    
528  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
529    
530  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
531  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
532    into a subtraction and unsigned comparison). */
533    
534    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
535    
536    #ifndef EBCDIC
537    
538    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
539    UTF-8 mode. */
540    
541    static const pcre_uint8 digitab[] =
542    {    {
543    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
544    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 357  static const unsigned char digitab[] = Line 573  static const unsigned char digitab[] =
573    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
574    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
575    
576  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
577  static const unsigned char digitab[] =  
578    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
579    
580    static const pcre_uint8 digitab[] =
581    {    {
582    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
583    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 393  static const unsigned char digitab[] = Line 612  static const unsigned char digitab[] =
612    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
613    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
614    
615  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
616    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
617    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
618    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 429  static const unsigned char ebcdic_charta Line 648  static const unsigned char ebcdic_charta
648  #endif  #endif
649    
650    
 /* 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 *);  
   
   
651    
652  /*************************************************  /*************************************************
653  *            Find an error text                  *  *            Find an error text                  *
# Line 454  static const char * Line 666  static const char *
666  find_error_text(int n)  find_error_text(int n)
667  {  {
668  const char *s = error_texts;  const char *s = error_texts;
669  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
670      {
671      while (*s++ != CHAR_NULL) {};
672      if (*s == CHAR_NULL) return "Error text not found (please report)";
673      }
674  return s;  return s;
675  }  }
676    
677    
678  /*************************************************  /*************************************************
679    *           Expand the workspace                 *
680    *************************************************/
681    
682    /* This function is called during the second compiling phase, if the number of
683    forward references fills the existing workspace, which is originally a block on
684    the stack. A larger block is obtained from malloc() unless the ultimate limit
685    has been reached or the increase will be rather small.
686    
687    Argument: pointer to the compile data block
688    Returns:  0 if all went well, else an error number
689    */
690    
691    static int
692    expand_workspace(compile_data *cd)
693    {
694    pcre_uchar *newspace;
695    int newsize = cd->workspace_size * 2;
696    
697    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
698    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
699        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
700     return ERR72;
701    
702    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
703    if (newspace == NULL) return ERR21;
704    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
705    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
706    if (cd->workspace_size > COMPILE_WORK_SIZE)
707      (PUBL(free))((void *)cd->start_workspace);
708    cd->start_workspace = newspace;
709    cd->workspace_size = newsize;
710    return 0;
711    }
712    
713    
714    
715    /*************************************************
716    *            Check for counted repeat            *
717    *************************************************/
718    
719    /* This function is called when a '{' is encountered in a place where it might
720    start a quantifier. It looks ahead to see if it really is a quantifier or not.
721    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
722    where the ddds are digits.
723    
724    Arguments:
725      p         pointer to the first char after '{'
726    
727    Returns:    TRUE or FALSE
728    */
729    
730    static BOOL
731    is_counted_repeat(const pcre_uchar *p)
732    {
733    if (!IS_DIGIT(*p)) return FALSE;
734    p++;
735    while (IS_DIGIT(*p)) p++;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (*p++ != CHAR_COMMA) return FALSE;
739    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
740    
741    if (!IS_DIGIT(*p)) return FALSE;
742    p++;
743    while (IS_DIGIT(*p)) p++;
744    
745    return (*p == CHAR_RIGHT_CURLY_BRACKET);
746    }
747    
748    
749    
750    /*************************************************
751  *            Handle escapes                      *  *            Handle escapes                      *
752  *************************************************/  *************************************************/
753    
754  /* 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
755  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
756  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
757  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
758  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  be returned in chptr.
759  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
760  sequence.  escape sequence.
761    
762  Arguments:  Arguments:
763    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
764      chptr          points to the data character
765    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
766    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
767    options        the options bits    options        the options bits
768    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
769    
770  Returns:         zero or positive => a data character  Returns:         zero => a data character
771                   negative => a special escape sequence                   positive => a special escape sequence
772                     negative => a back reference
773                   on error, errorcodeptr is set                   on error, errorcodeptr is set
774  */  */
775    
776  static int  static int
777  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
778    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
779  {  {
780  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
781  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
782  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
783    pcre_uint32 c;
784    int escape = 0;
785    int i;
786    
787  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
788  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
789    
790  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
791    
792  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
793    
794  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
795  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.
796  Otherwise further processing may be required. */  Otherwise further processing may be required. */
797    
798  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
799  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
800  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
801    else if ((i = escapes[c - CHAR_0]) != 0)
802      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
803    
804  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
805  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
806  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
807    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
808  #endif  #endif
809    
810  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
811    
812  else  else
813    {    {
814    const uschar *oldptr;    const pcre_uchar *oldptr;
815    BOOL braced, negated;    BOOL braced, negated, overflow;
816      int s;
817    
818    switch (c)    switch (c)
819      {      {
820      /* 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
821      error. */      error. */
822    
823      case 'l':      case CHAR_l:
824      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
825      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
826      break;      break;
827    
828      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
829      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
830      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
831      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
832      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
833      fudge it up by pretending it really was \k. */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
834            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
835      case 'g':          && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
836      if (ptr[1] == '{')          && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
837        {          {
838        const uschar *p;          c = 0;
839        for (p = ptr+2; *p != 0 && *p != '}'; p++)          for (i = 0; i < 4; ++i)
840          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;            {
841        if (*p != 0 && *p != '}')            register pcre_uint32 cc = *(++ptr);
842    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
843              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
844              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
845    #else           /* EBCDIC coding */
846              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
847              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
848    #endif
849              }
850    
851    #if defined COMPILE_PCRE8
852            if (c > (utf ? 0x10ffffU : 0xffU))
853    #elif defined COMPILE_PCRE16
854            if (c > (utf ? 0x10ffffU : 0xffffU))
855    #elif defined COMPILE_PCRE32
856            if (utf && c > 0x10ffffU)
857    #endif
858              {
859              *errorcodeptr = ERR76;
860              }
861            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
862            }
863          }
864        else
865          *errorcodeptr = ERR37;
866        break;
867    
868        case CHAR_U:
869        /* In JavaScript, \U is an uppercase U letter. */
870        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
871        break;
872    
873        /* In a character class, \g is just a literal "g". Outside a character
874        class, \g must be followed by one of a number of specific things:
875    
876        (1) A number, either plain or braced. If positive, it is an absolute
877        backreference. If negative, it is a relative backreference. This is a Perl
878        5.10 feature.
879    
880        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
881        is part of Perl's movement towards a unified syntax for back references. As
882        this is synonymous with \k{name}, we fudge it up by pretending it really
883        was \k.
884    
885        (3) For Oniguruma compatibility we also support \g followed by a name or a
886        number either in angle brackets or in single quotes. However, these are
887        (possibly recursive) subroutine calls, _not_ backreferences. Just return
888        the ESC_g code (cf \k). */
889    
890        case CHAR_g:
891        if (isclass) break;
892        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
893          {
894          escape = ESC_g;
895          break;
896          }
897    
898        /* Handle the Perl-compatible cases */
899    
900        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
901          {
902          const pcre_uchar *p;
903          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
904            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
905          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
906          {          {
907          c = -ESC_k;          escape = ESC_k;
908          break;          break;
909          }          }
910        braced = TRUE;        braced = TRUE;
# Line 554  else Line 912  else
912        }        }
913      else braced = FALSE;      else braced = FALSE;
914    
915      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
916        {        {
917        negated = TRUE;        negated = TRUE;
918        ptr++;        ptr++;
919        }        }
920      else negated = FALSE;      else negated = FALSE;
921    
922      c = 0;      /* The integer range is limited by the machine's int representation. */
923      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
924        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
925        while (IS_DIGIT(ptr[1]))
926      if (c < 0)        {
927          if (s > INT_MAX / 10 - 1) /* Integer overflow */
928            {
929            overflow = TRUE;
930            break;
931            }
932          s = s * 10 + (int)(*(++ptr) - CHAR_0);
933          }
934        if (overflow) /* Integer overflow */
935        {        {
936          while (IS_DIGIT(ptr[1]))
937            ptr++;
938        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
939        break;        break;
940        }        }
941    
942      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
943        {        {
944        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
945        break;        break;
946        }        }
947    
948        if (s == 0)
949          {
950          *errorcodeptr = ERR58;
951          break;
952          }
953    
954      if (negated)      if (negated)
955        {        {
956        if (c > bracount)        if (s > bracount)
957          {          {
958          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
959          break;          break;
960          }          }
961        c = bracount - (c - 1);        s = bracount - (s - 1);
962        }        }
963    
964      c = -(ESC_REF + c);      escape = -s;
965      break;      break;
966    
967      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 602  else Line 976  else
976      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
977      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
978    
979      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:
980      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
981    
982      if (!isclass)      if (!isclass)
983        {        {
984        oldptr = ptr;        oldptr = ptr;
985        c -= '0';        /* The integer range is limited by the machine's int representation. */
986        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
987          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
988        if (c < 0)        while (IS_DIGIT(ptr[1]))
989            {
990            if (s > INT_MAX / 10 - 1) /* Integer overflow */
991              {
992              overflow = TRUE;
993              break;
994              }
995            s = s * 10 + (int)(*(++ptr) - CHAR_0);
996            }
997          if (overflow) /* Integer overflow */
998          {          {
999            while (IS_DIGIT(ptr[1]))
1000              ptr++;
1001          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1002          break;          break;
1003          }          }
1004        if (c < 10 || c <= bracount)        if (s < 10 || s <= bracount)
1005          {          {
1006          c = -(ESC_REF + c);          escape = -s;
1007          break;          break;
1008          }          }
1009        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 628  else Line 1013  else
1013      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.
1014      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1015    
1016      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1017        {        {
1018        ptr--;        ptr--;
1019        c = 0;        c = 0;
# Line 638  else Line 1023  else
1023      /* \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
1024      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
1025      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
1026      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,
1027      than 3 octal digits. */      but no more than 3 octal digits. */
1028    
1029      case '0':      case CHAR_0:
1030      c -= '0';      c -= CHAR_0;
1031      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1032          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1033      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1034        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1035    #endif
1036      break;      break;
1037    
1038      /* \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
1039      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.
1040      treated as a data character. */      If not, { is treated as a data character. */
1041    
1042        case CHAR_x:
1043        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1044          {
1045          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1046          Otherwise it is a lowercase x letter. */
1047          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1048            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1049            {
1050            c = 0;
1051            for (i = 0; i < 2; ++i)
1052              {
1053              register pcre_uint32 cc = *(++ptr);
1054    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1055              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1056              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1057    #else           /* EBCDIC coding */
1058              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1059              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1060    #endif
1061              }
1062            }
1063          break;
1064          }
1065    
1066      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1067        {        {
1068        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1069    
1070        c = 0;        c = 0;
1071        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1072          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1073          {          {
1074          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1075          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1076          count++;  
1077    #ifdef COMPILE_PCRE32
1078  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1079          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1080          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1081    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1082            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1083            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1084  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1085          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1086          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1087    #endif
1088    
1089    #if defined COMPILE_PCRE8
1090            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1091    #elif defined COMPILE_PCRE16
1092            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1093    #elif defined COMPILE_PCRE32
1094            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1095  #endif  #endif
1096          }          }
1097    
1098        if (*pt == '}')        if (overflow)
1099          {          {
1100          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1101            *errorcodeptr = ERR34;
1102            }
1103    
1104          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1105            {
1106            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1107          ptr = pt;          ptr = pt;
1108          break;          break;
1109          }          }
# Line 688  else Line 1115  else
1115      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1116    
1117      c = 0;      c = 0;
1118      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1119        {        {
1120        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1121        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1122  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1123        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1124        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1125  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1126        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1127        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1128  #endif  #endif
1129        }        }
1130      break;      break;
1131    
1132      /* 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.
1133      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
1134        coding is ASCII-specific, but then the whole concept of \cx is
1135      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1136    
1137      case 'c':      case CHAR_c:
1138      c = *(++ptr);      c = *(++ptr);
1139      if (c == 0)      if (c == CHAR_NULL)
1140        {        {
1141        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1142        break;        break;
1143        }        }
1144    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1145  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1146      if (c >= 'a' && c <= 'z') c -= 32;        {
1147          *errorcodeptr = ERR68;
1148          break;
1149          }
1150        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1151      c ^= 0x40;      c ^= 0x40;
1152  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1153      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1154      c ^= 0xC0;      c ^= 0xC0;
1155  #endif  #endif
1156      break;      break;
# Line 740  else Line 1172  else
1172      }      }
1173    }    }
1174    
1175  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1176  return c;  newline". PCRE does not support \N{name}. However, it does support
1177  }  quantification such as \N{2,3}. */
1178    
1179    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1180         !is_counted_repeat(ptr+2))
1181      *errorcodeptr = ERR37;
1182    
1183    /* If PCRE_UCP is set, we change the values for \d etc. */
1184    
1185    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1186      escape += (ESC_DU - ESC_D);
1187    
1188    /* Set the pointer to the final character before returning. */
1189    
1190    *ptrptr = ptr;
1191    *chptr = c;
1192    return escape;
1193    }
1194    
1195  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1196  /*************************************************  /*************************************************
# Line 759  escape sequence. Line 1205  escape sequence.
1205  Argument:  Argument:
1206    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1207    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1208    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1209      pdataptr       points to an unsigned int that is set to the detailed property value
1210    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1211    
1212  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1213  */  */
1214    
1215  static int  static BOOL
1216  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1217      unsigned int *pdataptr, int *errorcodeptr)
1218  {  {
1219  int c, i, bot, top;  pcre_uchar c;
1220  const uschar *ptr = *ptrptr;  int i, bot, top;
1221  char name[32];  const pcre_uchar *ptr = *ptrptr;
1222    pcre_uchar name[32];
1223    
1224  c = *(++ptr);  c = *(++ptr);
1225  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1226    
1227  *negptr = FALSE;  *negptr = FALSE;
1228    
1229  /* \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
1230  negation. */  negation. */
1231    
1232  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1233    {    {
1234    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1235      {      {
1236      *negptr = TRUE;      *negptr = TRUE;
1237      ptr++;      ptr++;
1238      }      }
1239    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1240      {      {
1241      c = *(++ptr);      c = *(++ptr);
1242      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1243      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1244      name[i] = c;      name[i] = c;
1245      }      }
1246    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1247    name[i] = 0;    name[i] = 0;
1248    }    }
1249    
# Line 811  else Line 1260  else
1260  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1261    
1262  bot = 0;  bot = 0;
1263  top = _pcre_utt_size;  top = PRIV(utt_size);
1264    
1265  while (bot < top)  while (bot < top)
1266    {    {
1267      int r;
1268    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1269    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1270    if (c == 0)    if (r == 0)
1271      {      {
1272      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1273      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1274        return TRUE;
1275      }      }
1276    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1277    }    }
1278    
1279  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1280  *ptrptr = ptr;  *ptrptr = ptr;
1281  return -1;  return FALSE;
1282    
1283  ERROR_RETURN:  ERROR_RETURN:
1284  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1285  *ptrptr = ptr;  *ptrptr = ptr;
1286  return -1;  return FALSE;
1287  }  }
1288  #endif  #endif
1289    
# Line 840  return -1; Line 1291  return -1;
1291    
1292    
1293  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1294  *         Read repeat counts                     *  *         Read repeat counts                     *
1295  *************************************************/  *************************************************/
1296    
# Line 891  Returns:         pointer to '}' on succe Line 1309  Returns:         pointer to '}' on succe
1309                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1310  */  */
1311    
1312  static const uschar *  static const pcre_uchar *
1313  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)
1314  {  {
1315  int min = 0;  int min = 0;
1316  int max = -1;  int max = -1;
# Line 900  int max = -1; Line 1318  int max = -1;
1318  /* 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
1319  an integer overflow. */  an integer overflow. */
1320    
1321  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1322  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1323    {    {
1324    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1328  if (min < 0 || min > 65535)
1328  /* 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.
1329  Also, max must not be less than min. */  Also, max must not be less than min. */
1330    
1331  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1332    {    {
1333    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1334      {      {
1335      max = 0;      max = 0;
1336      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1337      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1338        {        {
1339        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1358  return p;
1358    
1359    
1360  /*************************************************  /*************************************************
1361  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1362  *************************************************/  *************************************************/
1363    
1364  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1365    top-level call starts at the beginning of the pattern. All other calls must
1366    start at a parenthesis. It scans along a pattern's text looking for capturing
1367  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
1368  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
1369  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
1370  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1371  be terminated by '>' because that is checked in the first pass.  
1372    This function was originally called only from the second pass, in which we know
1373    that if (?< or (?' or (?P< is encountered, the name will be correctly
1374    terminated because that is checked in the first pass. There is now one call to
1375    this function in the first pass, to check for a recursive back reference by
1376    name (so that we can make the whole group atomic). In this case, we need check
1377    only up to the current position in the pattern, and that is still OK because
1378    and previous occurrences will have been checked. To make this work, the test
1379    for "end of pattern" is a check against cd->end_pattern in the main loop,
1380    instead of looking for a binary zero. This means that the special first-pass
1381    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1382    processing items within the loop are OK, because afterwards the main loop will
1383    terminate.)
1384    
1385  Arguments:  Arguments:
1386    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1387    count        current count of capturing parens so far encountered    cd           compile background data
1388    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1389    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1390    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1391      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1392      count        pointer to the current capturing subpattern number (updated)
1393    
1394  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1395  */  */
1396    
1397  static int  static int
1398  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1399    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1400  {  {
1401  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1402    int start_count = *count;
1403    int hwm_count = start_count;
1404    BOOL dup_parens = FALSE;
1405    
1406    /* If the first character is a parenthesis, check on the type of group we are
1407    dealing with. The very first call may not start with a parenthesis. */
1408    
1409  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1410    {    {
1411    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1412    
1413      if (ptr[1] == CHAR_ASTERISK)
1414        {
1415        ptr += 2;
1416        while (ptr < cd->end_pattern && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1417        }
1418    
1419      /* Handle a normal, unnamed capturing parenthesis. */
1420    
1421      else if (ptr[1] != CHAR_QUESTION_MARK)
1422        {
1423        *count += 1;
1424        if (name == NULL && *count == lorn) return *count;
1425        ptr++;
1426        }
1427    
1428      /* All cases now have (? at the start. Remember when we are in a group
1429      where the parenthesis numbers are duplicated. */
1430    
1431      else if (ptr[2] == CHAR_VERTICAL_LINE)
1432        {
1433        ptr += 3;
1434        dup_parens = TRUE;
1435        }
1436    
1437      /* Handle comments; all characters are allowed until a ket is reached. */
1438    
1439      else if (ptr[2] == CHAR_NUMBER_SIGN)
1440        {
1441        for (ptr += 3; *ptr != CHAR_NULL; ptr++)
1442          if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1443        goto FAIL_EXIT;
1444        }
1445    
1446      /* Handle a condition. If it is an assertion, just carry on so that it
1447      is processed as normal. If not, skip to the closing parenthesis of the
1448      condition (there can't be any nested parens). */
1449    
1450      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1451        {
1452        ptr += 2;
1453        if (ptr[1] != CHAR_QUESTION_MARK)
1454          {
1455          while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1456          if (*ptr != CHAR_NULL) ptr++;
1457          }
1458        }
1459    
1460      /* Start with (? but not a condition. */
1461    
1462      else
1463        {
1464        ptr += 2;
1465        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1466    
1467        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1468    
1469        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1470            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1471          {
1472          pcre_uchar term;
1473          const pcre_uchar *thisname;
1474          *count += 1;
1475          if (name == NULL && *count == lorn) return *count;
1476          term = *ptr++;
1477          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1478          thisname = ptr;
1479          while (*ptr != term) ptr++;
1480          if (name != NULL && lorn == (int)(ptr - thisname) &&
1481              STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
1482            return *count;
1483          term++;
1484          }
1485        }
1486      }
1487    
1488    /* Past any initial parenthesis handling, scan for parentheses or vertical
1489    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1490    first-pass call when this value is temporarily adjusted to stop at the current
1491    position. So DO NOT change this to a test for binary zero. */
1492    
1493    for (; ptr < cd->end_pattern; ptr++)
1494      {
1495    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1496    
1497    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1498      {      {
1499      if (*(++ptr) == 0) return -1;      if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1500      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1501        {        {
1502        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1503        if (*ptr == 0) return -1;        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1504        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1505        }        }
1506      continue;      continue;
1507      }      }
1508    
1509    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1510      are handled for real. If the first character is '^', skip it. Also, if the
1511      first few characters (either before or after ^) are \Q\E or \E we skip them
1512      too. This makes for compatibility with Perl. Note the use of STR macros to
1513      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1514    
1515    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1516      {      {
1517      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1518        for (;;)
1519          {
1520          if (ptr[1] == CHAR_BACKSLASH)
1521            {
1522            if (ptr[2] == CHAR_E)
1523              ptr+= 2;
1524            else if (STRNCMP_UC_C8(ptr + 2,
1525                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1526              ptr += 4;
1527            else
1528              break;
1529            }
1530          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1531            {
1532            negate_class = TRUE;
1533            ptr++;
1534            }
1535          else break;
1536          }
1537    
1538        /* If the next character is ']', it is a data character that must be
1539        skipped, except in JavaScript compatibility mode. */
1540    
1541        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1542            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1543          ptr++;
1544    
1545        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1546        {        {
1547        if (*ptr == 0) return -1;        if (*ptr == CHAR_NULL) return -1;
1548        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1549          {          {
1550          if (*(++ptr) == 0) return -1;          if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1551          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1552            {            {
1553            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1554            if (*ptr == 0) return -1;            if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1555            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1556            }            }
1557          continue;          continue;
1558          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1562  for (; *ptr != 0; ptr++)
1562    
1563    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1564    
1565    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1566      {      {
1567      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1568      if (*ptr == 0) return -1;      while (*ptr != CHAR_NULL)
1569          {
1570          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1571          ptr++;
1572    #ifdef SUPPORT_UTF
1573          if (utf) FORWARDCHAR(ptr);
1574    #endif
1575          }
1576        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1577      continue;      continue;
1578      }      }
1579    
1580    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1581    
1582    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1583      {      {
1584      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1585      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1586      continue;      if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1587      }      }
1588    
1589    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1590    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1591        if (dup_parens && *count < hwm_count) *count = hwm_count;
1592    /* We have to disambiguate (?<! and (?<= from (?<name> */      goto FAIL_EXIT;
1593        }
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
1594    
1595    if (name == NULL && count == lorn) return count;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1596    term = *ptr++;      {
1597    if (term == '<') term = '>';      if (*count > hwm_count) hwm_count = *count;
1598    thisname = ptr;      *count = start_count;
1599    while (*ptr != term) ptr++;      }
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1600    }    }
1601    
1602    FAIL_EXIT:
1603    *ptrptr = ptr;
1604  return -1;  return -1;
1605  }  }
1606    
1607    
1608    
1609    
1610  /*************************************************  /*************************************************
1611  *      Find first significant op code            *  *       Find forward referenced subpattern       *
1612  *************************************************/  *************************************************/
1613    
1614  /* This is called by several functions that scan a compiled expression looking  /* This function scans along a pattern's text looking for capturing
1615  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
1616  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
1617  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
1618  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
1619    current compiling point, using the current count value from cd->bracount, and
1620    do it all in a single loop, but the addition of the possibility of duplicate
1621    subpattern numbers means that we have to scan from the very start, in order to
1622    take account of such duplicates, and to use a recursive function to keep track
1623    of the different types of group.
1624    
1625  Arguments:  Arguments:
1626    code         pointer to the start of the group    cd           compile background data
1627    options      pointer to external options    name         name to seek, or NULL if seeking a numbered subpattern
1628    optbit       the option bit whose changing is significant, or    lorn         name length, or subpattern number if name is NULL
1629                   zero if none are    xmode        TRUE if we are in /x mode
1630    skipassert   TRUE if certain assertions are to be skipped    utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1631    
1632    Returns:       the number of the found subpattern, or -1 if not found
1633    */
1634    
1635    static int
1636    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1637      BOOL utf)
1638    {
1639    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1640    int count = 0;
1641    int rc;
1642    
1643    /* If the pattern does not start with an opening parenthesis, the first call
1644    to find_parens_sub() will scan right to the end (if necessary). However, if it
1645    does start with a parenthesis, find_parens_sub() will return when it hits the
1646    matching closing parens. That is why we have to have a loop. */
1647    
1648    for (;;)
1649      {
1650      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1651      if (rc > 0 || *ptr++ == CHAR_NULL) break;
1652      }
1653    
1654    return rc;
1655    }
1656    
1657    
1658    
1659    
1660    /*************************************************
1661    *      Find first significant op code            *
1662    *************************************************/
1663    
1664    /* This is called by several functions that scan a compiled expression looking
1665    for a fixed first character, or an anchoring op code etc. It skips over things
1666    that do not influence this. For some calls, it makes sense to skip negative
1667    forward and all backward assertions, and also the \b assertion; for others it
1668    does not.
1669    
1670    Arguments:
1671      code         pointer to the start of the group
1672      skipassert   TRUE if certain assertions are to be skipped
1673    
1674  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1675  */  */
1676    
1677  static const uschar*  static const pcre_uchar*
1678  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1679  {  {
1680  for (;;)  for (;;)
1681    {    {
1682    switch ((int)*code)    switch ((int)*code)
1683      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1684      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1685      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1686      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1687      if (!skipassert) return code;      if (!skipassert) return code;
1688      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1689      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1690      break;      break;
1691    
1692      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1100  for (;;) Line 1696  for (;;)
1696    
1697      case OP_CALLOUT:      case OP_CALLOUT:
1698      case OP_CREF:      case OP_CREF:
1699        case OP_NCREF:
1700      case OP_RREF:      case OP_RREF:
1701        case OP_NRREF:
1702      case OP_DEF:      case OP_DEF:
1703      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1704      break;      break;
1705    
1706      default:      default:
# Line 1116  for (;;) Line 1714  for (;;)
1714    
1715    
1716  /*************************************************  /*************************************************
1717  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1718  *************************************************/  *************************************************/
1719    
1720  /* 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,
1721  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.
1722  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
1723    temporarily terminated with OP_END when this function is called.
1724    
1725    This function is called when a backward assertion is encountered, so that if it
1726    fails, the error message can point to the correct place in the pattern.
1727    However, we cannot do this when the assertion contains subroutine calls,
1728    because they can be forward references. We solve this by remembering this case
1729    and doing the check at the end; a flag specifies which mode we are running in.
1730    
1731  Arguments:  Arguments:
1732    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1733    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1734      atend    TRUE if called when the pattern is complete
1735  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1736               or -2 if \C was encountered  
1737    Returns:   the fixed length,
1738                 or -1 if there is no fixed length,
1739                 or -2 if \C was encountered (in UTF-8 mode only)
1740                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1741                 or -4 if an unknown opcode was encountered (internal error)
1742  */  */
1743    
1744  static int  static int
1745  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1746  {  {
1747  int length = -1;  int length = -1;
1748    
1749  register int branchlength = 0;  register int branchlength = 0;
1750  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1751    
1752  /* 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
1753  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1145  branch, check the length against that of Line 1755  branch, check the length against that of
1755  for (;;)  for (;;)
1756    {    {
1757    int d;    int d;
1758    register int op = *cc;    pcre_uchar *ce, *cs;
1759      register pcre_uchar op = *cc;
1760    
1761    switch (op)    switch (op)
1762      {      {
1763        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1764        OP_BRA (normal non-capturing bracket) because the other variants of these
1765        opcodes are all concerned with unlimited repeated groups, which of course
1766        are not of fixed length. */
1767    
1768      case OP_CBRA:      case OP_CBRA:
1769      case OP_BRA:      case OP_BRA:
1770      case OP_ONCE:      case OP_ONCE:
1771        case OP_ONCE_NC:
1772      case OP_COND:      case OP_COND:
1773      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1774      if (d < 0) return d;      if (d < 0) return d;
1775      branchlength += d;      branchlength += d;
1776      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1777      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1778      break;      break;
1779    
1780      /* 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.
1781      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
1782      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
1783        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1784        because they all imply an unlimited repeat. */
1785    
1786      case OP_ALT:      case OP_ALT:
1787      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1788      case OP_END:      case OP_END:
1789        case OP_ACCEPT:
1790        case OP_ASSERT_ACCEPT:
1791      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1792        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1793      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1175  for (;;) Line 1795  for (;;)
1795      branchlength = 0;      branchlength = 0;
1796      break;      break;
1797    
1798        /* A true recursion implies not fixed length, but a subroutine call may
1799        be OK. If the subroutine is a forward reference, we can't deal with
1800        it until the end of the pattern, so return -3. */
1801    
1802        case OP_RECURSE:
1803        if (!atend) return -3;
1804        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1805        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1806        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1807        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1808        if (d < 0) return d;
1809        branchlength += d;
1810        cc += 1 + LINK_SIZE;
1811        break;
1812    
1813      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1814    
1815      case OP_ASSERT:      case OP_ASSERT:
# Line 1182  for (;;) Line 1817  for (;;)
1817      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1818      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1819      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1820      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1821        break;
1822    
1823      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1824    
1825      case OP_REVERSE:      case OP_MARK:
1826        case OP_PRUNE_ARG:
1827        case OP_SKIP_ARG:
1828        case OP_THEN_ARG:
1829        cc += cc[1] + PRIV(OP_lengths)[*cc];
1830        break;
1831    
1832        case OP_CALLOUT:
1833        case OP_CIRC:
1834        case OP_CIRCM:
1835        case OP_CLOSE:
1836        case OP_COMMIT:
1837      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1838      case OP_DEF:      case OP_DEF:
1839      case OP_OPT:      case OP_DOLL:
1840      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1841      case OP_EOD:      case OP_EOD:
1842      case OP_EODN:      case OP_EODN:
1843      case OP_CIRC:      case OP_FAIL:
1844      case OP_DOLL:      case OP_NCREF:
1845        case OP_NRREF:
1846      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1847        case OP_PRUNE:
1848        case OP_REVERSE:
1849        case OP_RREF:
1850        case OP_SET_SOM:
1851        case OP_SKIP:
1852        case OP_SOD:
1853        case OP_SOM:
1854        case OP_THEN:
1855      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1856      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1857      break;      break;
1858    
1859      /* Handle literal characters */      /* Handle literal characters */
1860    
1861      case OP_CHAR:      case OP_CHAR:
1862      case OP_CHARNC:      case OP_CHARI:
1863      case OP_NOT:      case OP_NOT:
1864        case OP_NOTI:
1865      branchlength++;      branchlength++;
1866      cc += 2;      cc += 2;
1867  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1868      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1869  #endif  #endif
1870      break;      break;
1871    
# Line 1222  for (;;) Line 1873  for (;;)
1873      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1874    
1875      case OP_EXACT:      case OP_EXACT:
1876      branchlength += GET2(cc,1);      case OP_EXACTI:
1877      cc += 4;      case OP_NOTEXACT:
1878  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1879      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1880        {      cc += 2 + IMM2_SIZE;
1881        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1882        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1883  #endif  #endif
1884      break;      break;
1885    
1886      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1887      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1888      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1889      cc += 4;        cc += 2;
1890        cc += 1 + IMM2_SIZE + 1;
1891      break;      break;
1892    
1893      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1245  for (;;) Line 1897  for (;;)
1897      cc += 2;      cc += 2;
1898      /* Fall through */      /* Fall through */
1899    
1900        case OP_HSPACE:
1901        case OP_VSPACE:
1902        case OP_NOT_HSPACE:
1903        case OP_NOT_VSPACE:
1904      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1905      case OP_DIGIT:      case OP_DIGIT:
1906      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1252  for (;;) Line 1908  for (;;)
1908      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1909      case OP_WORDCHAR:      case OP_WORDCHAR:
1910      case OP_ANY:      case OP_ANY:
1911        case OP_ALLANY:
1912      branchlength++;      branchlength++;
1913      cc++;      cc++;
1914      break;      break;
1915    
1916      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1917        otherwise \C is coded as OP_ALLANY. */
1918    
1919      case OP_ANYBYTE:      case OP_ANYBYTE:
1920      return -2;      return -2;
1921    
1922      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1923    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1924      case OP_CLASS:      case OP_CLASS:
1925      case OP_NCLASS:      case OP_NCLASS:
1926      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1927        case OP_XCLASS:
1928        /* The original code caused an unsigned overflow in 64 bit systems,
1929        so now we use a conditional statement. */
1930        if (op == OP_XCLASS)
1931          cc += GET(cc, 1);
1932        else
1933          cc += PRIV(OP_lengths)[OP_CLASS];
1934    #else
1935        cc += PRIV(OP_lengths)[OP_CLASS];
1936    #endif
1937    
1938      switch (*cc)      switch (*cc)
1939        {        {
1940          case OP_CRPLUS:
1941          case OP_CRMINPLUS:
1942        case OP_CRSTAR:        case OP_CRSTAR:
1943        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1944        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1283  for (;;) Line 1947  for (;;)
1947    
1948        case OP_CRRANGE:        case OP_CRRANGE:
1949        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1950        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1951        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1952        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1953        break;        break;
1954    
1955        default:        default:
# Line 1295  for (;;) Line 1959  for (;;)
1959    
1960      /* Anything else is variable length */      /* Anything else is variable length */
1961    
1962      default:      case OP_ANYNL:
1963        case OP_BRAMINZERO:
1964        case OP_BRAPOS:
1965        case OP_BRAPOSZERO:
1966        case OP_BRAZERO:
1967        case OP_CBRAPOS:
1968        case OP_EXTUNI:
1969        case OP_KETRMAX:
1970        case OP_KETRMIN:
1971        case OP_KETRPOS:
1972        case OP_MINPLUS:
1973        case OP_MINPLUSI:
1974        case OP_MINQUERY:
1975        case OP_MINQUERYI:
1976        case OP_MINSTAR:
1977        case OP_MINSTARI:
1978        case OP_MINUPTO:
1979        case OP_MINUPTOI:
1980        case OP_NOTMINPLUS:
1981        case OP_NOTMINPLUSI:
1982        case OP_NOTMINQUERY:
1983        case OP_NOTMINQUERYI:
1984        case OP_NOTMINSTAR:
1985        case OP_NOTMINSTARI:
1986        case OP_NOTMINUPTO:
1987        case OP_NOTMINUPTOI:
1988        case OP_NOTPLUS:
1989        case OP_NOTPLUSI:
1990        case OP_NOTPOSPLUS:
1991        case OP_NOTPOSPLUSI:
1992        case OP_NOTPOSQUERY:
1993        case OP_NOTPOSQUERYI:
1994        case OP_NOTPOSSTAR:
1995        case OP_NOTPOSSTARI:
1996        case OP_NOTPOSUPTO:
1997        case OP_NOTPOSUPTOI:
1998        case OP_NOTQUERY:
1999        case OP_NOTQUERYI:
2000        case OP_NOTSTAR:
2001        case OP_NOTSTARI:
2002        case OP_NOTUPTO:
2003        case OP_NOTUPTOI:
2004        case OP_PLUS:
2005        case OP_PLUSI:
2006        case OP_POSPLUS:
2007        case OP_POSPLUSI:
2008        case OP_POSQUERY:
2009        case OP_POSQUERYI:
2010        case OP_POSSTAR:
2011        case OP_POSSTARI:
2012        case OP_POSUPTO:
2013        case OP_POSUPTOI:
2014        case OP_QUERY:
2015        case OP_QUERYI:
2016        case OP_REF:
2017        case OP_REFI:
2018        case OP_SBRA:
2019        case OP_SBRAPOS:
2020        case OP_SCBRA:
2021        case OP_SCBRAPOS:
2022        case OP_SCOND:
2023        case OP_SKIPZERO:
2024        case OP_STAR:
2025        case OP_STARI:
2026        case OP_TYPEMINPLUS:
2027        case OP_TYPEMINQUERY:
2028        case OP_TYPEMINSTAR:
2029        case OP_TYPEMINUPTO:
2030        case OP_TYPEPLUS:
2031        case OP_TYPEPOSPLUS:
2032        case OP_TYPEPOSQUERY:
2033        case OP_TYPEPOSSTAR:
2034        case OP_TYPEPOSUPTO:
2035        case OP_TYPEQUERY:
2036        case OP_TYPESTAR:
2037        case OP_TYPEUPTO:
2038        case OP_UPTO:
2039        case OP_UPTOI:
2040      return -1;      return -1;
2041    
2042        /* Catch unrecognized opcodes so that when new ones are added they
2043        are not forgotten, as has happened in the past. */
2044    
2045        default:
2046        return -4;
2047      }      }
2048    }    }
2049  /* Control never gets here */  /* Control never gets here */
# Line 1306  for (;;) Line 2053  for (;;)
2053    
2054    
2055  /*************************************************  /*************************************************
2056  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2057  *************************************************/  *************************************************/
2058    
2059  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2060  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2061    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2062    so that it can be called from pcre_study() when finding the minimum matching
2063    length.
2064    
2065  Arguments:  Arguments:
2066    code        points to start of expression    code        points to start of expression
2067    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2068    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2069    
2070  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
2071  */  */
2072    
2073  static const uschar *  const pcre_uchar *
2074  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2075  {  {
2076  for (;;)  for (;;)
2077    {    {
2078    register int c = *code;    register pcre_uchar c = *code;
2079    
2080    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2081    
2082    /* 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 1334  for (;;) Line 2085  for (;;)
2085    
2086    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2087    
2088      /* Handle recursion */
2089    
2090      else if (c == OP_REVERSE)
2091        {
2092        if (number < 0) return (pcre_uchar *)code;
2093        code += PRIV(OP_lengths)[c];
2094        }
2095    
2096    /* Handle capturing bracket */    /* Handle capturing bracket */
2097    
2098    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2099               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2100      {      {
2101      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2102      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2103      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2104      }      }
2105    
2106    /* 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
2107    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
2108    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2109      must add in its length. */
2110    
2111    else    else
2112      {      {
# Line 1367  for (;;) Line 2128  for (;;)
2128        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2129        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2130        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2131        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2132            code += 2;
2133          break;
2134    
2135          case OP_MARK:
2136          case OP_PRUNE_ARG:
2137          case OP_SKIP_ARG:
2138          case OP_THEN_ARG:
2139          code += code[1];
2140        break;        break;
2141        }        }
2142    
2143      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2144    
2145      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2146    
2147    /* 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
2148    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
2149    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2150    
2151  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2152      if (utf8) switch(c)      if (utf) switch(c)
2153        {        {
2154        case OP_CHAR:        case OP_CHAR:
2155        case OP_CHARNC:        case OP_CHARI:
2156        case OP_EXACT:        case OP_EXACT:
2157          case OP_EXACTI:
2158        case OP_UPTO:        case OP_UPTO:
2159          case OP_UPTOI:
2160        case OP_MINUPTO:        case OP_MINUPTO:
2161          case OP_MINUPTOI:
2162        case OP_POSUPTO:        case OP_POSUPTO:
2163          case OP_POSUPTOI:
2164        case OP_STAR:        case OP_STAR:
2165          case OP_STARI:
2166        case OP_MINSTAR:        case OP_MINSTAR:
2167          case OP_MINSTARI:
2168        case OP_POSSTAR:        case OP_POSSTAR:
2169          case OP_POSSTARI:
2170        case OP_PLUS:        case OP_PLUS:
2171          case OP_PLUSI:
2172        case OP_MINPLUS:        case OP_MINPLUS:
2173          case OP_MINPLUSI:
2174        case OP_POSPLUS:        case OP_POSPLUS:
2175          case OP_POSPLUSI:
2176        case OP_QUERY:        case OP_QUERY:
2177          case OP_QUERYI:
2178        case OP_MINQUERY:        case OP_MINQUERY:
2179          case OP_MINQUERYI:
2180        case OP_POSQUERY:        case OP_POSQUERY:
2181        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2182          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2183        break;        break;
2184        }        }
2185    #else
2186        (void)(utf);  /* Keep compiler happy by referencing function argument */
2187  #endif  #endif
2188      }      }
2189    }    }
# Line 1416  instance of OP_RECURSE. Line 2200  instance of OP_RECURSE.
2200    
2201  Arguments:  Arguments:
2202    code        points to start of expression    code        points to start of expression
2203    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2204    
2205  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
2206  */  */
2207    
2208  static const uschar *  static const pcre_uchar *
2209  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2210  {  {
2211  for (;;)  for (;;)
2212    {    {
2213    register int c = *code;    register pcre_uchar c = *code;
2214    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2215    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2216    
# Line 1438  for (;;) Line 2222  for (;;)
2222    
2223    /* 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
2224    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
2225    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2226      must add in its length. */
2227    
2228    else    else
2229      {      {
# Line 1460  for (;;) Line 2245  for (;;)
2245        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2246        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2247        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2248        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2249            code += 2;
2250          break;
2251    
2252          case OP_MARK:
2253          case OP_PRUNE_ARG:
2254          case OP_SKIP_ARG:
2255          case OP_THEN_ARG:
2256          code += code[1];
2257        break;        break;
2258        }        }
2259    
2260      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2261    
2262      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2263    
2264      /* 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
2265      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
2266      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2267    
2268  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2269      if (utf8) switch(c)      if (utf) switch(c)
2270        {        {
2271        case OP_CHAR:        case OP_CHAR:
2272        case OP_CHARNC:        case OP_CHARI:
2273          case OP_NOT:
2274          case OP_NOTI:
2275        case OP_EXACT:        case OP_EXACT:
2276          case OP_EXACTI:
2277          case OP_NOTEXACT:
2278          case OP_NOTEXACTI:
2279        case OP_UPTO:        case OP_UPTO:
2280          case OP_UPTOI:
2281          case OP_NOTUPTO:
2282          case OP_NOTUPTOI:
2283        case OP_MINUPTO:        case OP_MINUPTO:
2284          case OP_MINUPTOI:
2285          case OP_NOTMINUPTO:
2286          case OP_NOTMINUPTOI:
2287        case OP_POSUPTO:        case OP_POSUPTO:
2288          case OP_POSUPTOI:
2289          case OP_NOTPOSUPTO:
2290          case OP_NOTPOSUPTOI:
2291        case OP_STAR:        case OP_STAR:
2292          case OP_STARI:
2293          case OP_NOTSTAR:
2294          case OP_NOTSTARI:
2295        case OP_MINSTAR:        case OP_MINSTAR:
2296          case OP_MINSTARI:
2297          case OP_NOTMINSTAR:
2298          case OP_NOTMINSTARI:
2299        case OP_POSSTAR:        case OP_POSSTAR:
2300          case OP_POSSTARI:
2301          case OP_NOTPOSSTAR:
2302          case OP_NOTPOSSTARI:
2303        case OP_PLUS:        case OP_PLUS:
2304          case OP_PLUSI:
2305          case OP_NOTPLUS:
2306          case OP_NOTPLUSI:
2307        case OP_MINPLUS:        case OP_MINPLUS:
2308          case OP_MINPLUSI:
2309          case OP_NOTMINPLUS:
2310          case OP_NOTMINPLUSI:
2311        case OP_POSPLUS:        case OP_POSPLUS:
2312          case OP_POSPLUSI:
2313          case OP_NOTPOSPLUS:
2314          case OP_NOTPOSPLUSI:
2315        case OP_QUERY:        case OP_QUERY:
2316          case OP_QUERYI:
2317          case OP_NOTQUERY:
2318          case OP_NOTQUERYI:
2319        case OP_MINQUERY:        case OP_MINQUERY:
2320          case OP_MINQUERYI:
2321          case OP_NOTMINQUERY:
2322          case OP_NOTMINQUERYI:
2323        case OP_POSQUERY:        case OP_POSQUERY:
2324        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2325          case OP_NOTPOSQUERY:
2326          case OP_NOTPOSQUERYI:
2327          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2328        break;        break;
2329        }        }
2330    #else
2331        (void)(utf);  /* Keep compiler happy by referencing function argument */
2332  #endif  #endif
2333      }      }
2334    }    }
# Line 1515  bracket whose current branch will alread Line 2351  bracket whose current branch will alread
2351  Arguments:  Arguments:
2352    code        points to start of search    code        points to start of search
2353    endcode     points to where to stop    endcode     points to where to stop
2354    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2355      cd          contains pointers to tables etc.
2356      recurses    chain of recurse_check to catch mutual recursion
2357    
2358  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2359  */  */
2360    
2361    typedef struct recurse_check {
2362      struct recurse_check *prev;
2363      const pcre_uchar *group;
2364    } recurse_check;
2365    
2366  static BOOL  static BOOL
2367  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2368      BOOL utf, compile_data *cd, recurse_check *recurses)
2369  {  {
2370  register int c;  register pcre_uchar c;
2371  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2372    
2373    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2374       code < endcode;       code < endcode;
2375       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2376    {    {
2377    const uschar *ccode;    const pcre_uchar *ccode;
2378    
2379    c = *code;    c = *code;
2380    
2381    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2382    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2383    
# Line 1542  for (code = first_significant_code(code Line 2388  for (code = first_significant_code(code
2388      continue;      continue;
2389      }      }
2390    
2391      /* For a recursion/subroutine call, if its end has been reached, which
2392      implies a backward reference subroutine call, we can scan it. If it's a
2393      forward reference subroutine call, we can't. To detect forward reference
2394      we have to scan up the list that is kept in the workspace. This function is
2395      called only when doing the real compile, not during the pre-compile that
2396      measures the size of the compiled pattern. */
2397    
2398      if (c == OP_RECURSE)
2399        {
2400        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2401        BOOL empty_branch;
2402    
2403        /* Test for forward reference or uncompleted reference. This is disabled
2404        when called to scan a completed pattern by setting cd->start_workspace to
2405        NULL. */
2406    
2407        if (cd->start_workspace != NULL)
2408          {
2409          const pcre_uchar *tcode;
2410          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2411            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2412          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2413          }
2414    
2415        /* If we are scanning a completed pattern, there are no forward references
2416        and all groups are complete. We need to detect whether this is a recursive
2417        call, as otherwise there will be an infinite loop. If it is a recursion,
2418        just skip over it. Simple recursions are easily detected. For mutual
2419        recursions we keep a chain on the stack. */
2420    
2421        else
2422          {
2423          recurse_check *r = recurses;
2424          const pcre_uchar *endgroup = scode;
2425    
2426          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2427          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2428    
2429          for (r = recurses; r != NULL; r = r->prev)
2430            if (r->group == scode) break;
2431          if (r != NULL) continue;   /* Mutual recursion */
2432          }
2433    
2434        /* Completed reference; scan the referenced group, remembering it on the
2435        stack chain to detect mutual recursions. */
2436    
2437        empty_branch = FALSE;
2438        this_recurse.prev = recurses;
2439        this_recurse.group = scode;
2440    
2441        do
2442          {
2443          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2444            {
2445            empty_branch = TRUE;
2446            break;
2447            }
2448          scode += GET(scode, 1);
2449          }
2450        while (*scode == OP_ALT);
2451    
2452        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2453        continue;
2454        }
2455    
2456    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2457    
2458    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2459          c == OP_BRAPOSZERO)
2460        {
2461        code += PRIV(OP_lengths)[c];
2462        do code += GET(code, 1); while (*code == OP_ALT);
2463        c = *code;
2464        continue;
2465        }
2466    
2467      /* A nested group that is already marked as "could be empty" can just be
2468      skipped. */
2469    
2470      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2471          c == OP_SCBRA || c == OP_SCBRAPOS)
2472      {      {
     code += _pcre_OP_lengths[c];  
2473      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2474      c = *code;      c = *code;
2475      continue;      continue;
# Line 1554  for (code = first_significant_code(code Line 2477  for (code = first_significant_code(code
2477    
2478    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2479    
2480    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2481          c == OP_CBRA || c == OP_CBRAPOS ||
2482          c == OP_ONCE || c == OP_ONCE_NC ||
2483          c == OP_COND)
2484      {      {
2485      BOOL empty_branch;      BOOL empty_branch;
2486      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2487    
2488      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2489        empty branch, so just skip over the conditional, because it could be empty.
2490        Otherwise, scan the individual branches of the group. */
2491    
2492      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;  
2493        code += GET(code, 1);        code += GET(code, 1);
2494        else
2495          {
2496          empty_branch = FALSE;
2497          do
2498            {
2499            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2500              empty_branch = TRUE;
2501            code += GET(code, 1);
2502            }
2503          while (*code == OP_ALT);
2504          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2505        }        }
2506      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2507      c = *code;      c = *code;
2508      continue;      continue;
2509      }      }
# Line 1580  for (code = first_significant_code(code Line 2514  for (code = first_significant_code(code
2514      {      {
2515      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2516      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2517      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2518      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"
2519      here. */      here. */
2520    
2521  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2522      case OP_XCLASS:      case OP_XCLASS:
2523      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2524      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1592  for (code = first_significant_code(code Line 2526  for (code = first_significant_code(code
2526    
2527      case OP_CLASS:      case OP_CLASS:
2528      case OP_NCLASS:      case OP_NCLASS:
2529      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2530    
2531  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2532      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2533  #endif  #endif
2534    
# Line 1620  for (code = first_significant_code(code Line 2554  for (code = first_significant_code(code
2554    
2555      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2556    
2557        case OP_ANY:
2558        case OP_ALLANY:
2559        case OP_ANYBYTE:
2560    
2561      case OP_PROP:      case OP_PROP:
2562      case OP_NOTPROP:      case OP_NOTPROP:
2563        case OP_ANYNL:
2564    
2565        case OP_NOT_HSPACE:
2566        case OP_HSPACE:
2567        case OP_NOT_VSPACE:
2568        case OP_VSPACE:
2569      case OP_EXTUNI:      case OP_EXTUNI:
2570    
2571      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2572      case OP_DIGIT:      case OP_DIGIT:
2573      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2574      case OP_WHITESPACE:      case OP_WHITESPACE:
2575      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2576      case OP_WORDCHAR:      case OP_WORDCHAR:
2577      case OP_ANY:  
     case OP_ANYBYTE:  
2578      case OP_CHAR:      case OP_CHAR:
2579      case OP_CHARNC:      case OP_CHARI:
2580      case OP_NOT:      case OP_NOT:
2581        case OP_NOTI:
2582    
2583      case OP_PLUS:      case OP_PLUS:
2584        case OP_PLUSI:
2585      case OP_MINPLUS:      case OP_MINPLUS:
2586      case OP_POSPLUS:      case OP_MINPLUSI:
2587      case OP_EXACT:  
2588      case OP_NOTPLUS:      case OP_NOTPLUS:
2589        case OP_NOTPLUSI:
2590      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2591        case OP_NOTMINPLUSI:
2592    
2593        case OP_POSPLUS:
2594        case OP_POSPLUSI:
2595      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2596        case OP_NOTPOSPLUSI:
2597    
2598        case OP_EXACT:
2599        case OP_EXACTI:
2600      case OP_NOTEXACT:      case OP_NOTEXACT:
2601        case OP_NOTEXACTI:
2602    
2603      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2604      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2605      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2606      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2607    
2608      return FALSE;      return FALSE;
2609    
2610      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 1665  for (code = first_significant_code(code Line 2624  for (code = first_significant_code(code
2624      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2625      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2626      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2627      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2628          code += 2;
2629      break;      break;
2630    
2631      /* End of branch */      /* End of branch */
# Line 1673  for (code = first_significant_code(code Line 2633  for (code = first_significant_code(code
2633      case OP_KET:      case OP_KET:
2634      case OP_KETRMAX:      case OP_KETRMAX:
2635      case OP_KETRMIN:      case OP_KETRMIN:
2636        case OP_KETRPOS:
2637      case OP_ALT:      case OP_ALT:
2638      return TRUE;      return TRUE;
2639    
2640      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2641      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2642        followed by a multibyte character. */
2643    
2644  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2645      case OP_STAR:      case OP_STAR:
2646        case OP_STARI:
2647        case OP_NOTSTAR:
2648        case OP_NOTSTARI:
2649    
2650      case OP_MINSTAR:      case OP_MINSTAR:
2651        case OP_MINSTARI:
2652        case OP_NOTMINSTAR:
2653        case OP_NOTMINSTARI:
2654    
2655      case OP_POSSTAR:      case OP_POSSTAR:
2656        case OP_POSSTARI:
2657        case OP_NOTPOSSTAR:
2658        case OP_NOTPOSSTARI:
2659    
2660      case OP_QUERY:      case OP_QUERY:
2661        case OP_QUERYI:
2662        case OP_NOTQUERY:
2663        case OP_NOTQUERYI:
2664    
2665      case OP_MINQUERY:      case OP_MINQUERY:
2666        case OP_MINQUERYI:
2667        case OP_NOTMINQUERY:
2668        case OP_NOTMINQUERYI:
2669    
2670      case OP_POSQUERY:      case OP_POSQUERY:
2671        case OP_POSQUERYI:
2672        case OP_NOTPOSQUERY:
2673        case OP_NOTPOSQUERYI:
2674    
2675        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2676        break;
2677    
2678      case OP_UPTO:      case OP_UPTO:
2679        case OP_UPTOI:
2680        case OP_NOTUPTO:
2681        case OP_NOTUPTOI:
2682    
2683      case OP_MINUPTO:      case OP_MINUPTO:
2684        case OP_MINUPTOI:
2685        case OP_NOTMINUPTO:
2686        case OP_NOTMINUPTOI:
2687    
2688      case OP_POSUPTO:      case OP_POSUPTO:
2689      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2690        case OP_NOTPOSUPTO:
2691        case OP_NOTPOSUPTOI:
2692    
2693        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2694      break;      break;
2695  #endif  #endif
2696    
2697        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2698        string. */
2699    
2700        case OP_MARK:
2701        case OP_PRUNE_ARG:
2702        case OP_SKIP_ARG:
2703        case OP_THEN_ARG:
2704        code += code[1];
2705        break;
2706    
2707        /* None of the remaining opcodes are required to match a character. */
2708    
2709        default:
2710        break;
2711      }      }
2712    }    }
2713    
# Line 1708  return TRUE; Line 2724  return TRUE;
2724  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
2725  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,
2726  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.
2727    This function is called only during the real compile, not during the
2728    pre-compile.
2729    
2730  Arguments:  Arguments:
2731    code        points to start of the recursion    code        points to start of the recursion
2732    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2733    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2734    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2735      cd          pointers to tables etc
2736    
2737  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2738  */  */
2739    
2740  static BOOL  static BOOL
2741  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2742    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2743  {  {
2744  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2745    {    {
2746    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2747        return FALSE;
2748    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2749    }    }
2750  return TRUE;  return TRUE;
# Line 1756  where Perl recognizes it as the POSIX cl Line 2776  where Perl recognizes it as the POSIX cl
2776  "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,
2777  I think.  I think.
2778    
2779    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2780    It seems that the appearance of a nested POSIX class supersedes an apparent
2781    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2782    a digit.
2783    
2784    In Perl, unescaped square brackets may also appear as part of class names. For
2785    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2786    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2787    seem right at all. PCRE does not allow closing square brackets in POSIX class
2788    names.
2789    
2790  Arguments:  Arguments:
2791    ptr      pointer to the initial [    ptr      pointer to the initial [
2792    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1764  Returns:   TRUE or FALSE Line 2795  Returns:   TRUE or FALSE
2795  */  */
2796    
2797  static BOOL  static BOOL
2798  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2799  {  {
2800  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
2801  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2802  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != CHAR_NULL; ptr++)
2803    {    {
2804    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2805        ptr++;
2806      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2807      else
2808      {      {
2809      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2810        {        {
2811        *endptr = ptr;        *endptr = ptr;
2812        return TRUE;        return TRUE;
2813        }        }
2814        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2815             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2816              ptr[1] == CHAR_EQUALS_SIGN) &&
2817            check_posix_syntax(ptr, endptr))
2818          return FALSE;
2819      }      }
2820    }    }
2821  return FALSE;  return FALSE;
# Line 1801  Returns:     a value representing the na Line 2839  Returns:     a value representing the na
2839  */  */
2840    
2841  static int  static int
2842  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2843  {  {
2844  const char *pn = posix_names;  const char *pn = posix_names;
2845  register int yield = 0;  register int yield = 0;
2846  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2847    {    {
2848    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2849      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
2850    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2851    yield++;    yield++;
2852    }    }
# Line 1824  return -1; Line 2862  return -1;
2862  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2863  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2864  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2865  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2866  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2867  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2868  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2869  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2870    OP_END.
2871    
2872  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2873  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1839  value in the reference (which is a group Line 2878  value in the reference (which is a group
2878  Arguments:  Arguments:
2879    group      points to the start of the group    group      points to the start of the group
2880    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2881    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2882    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2883    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
2884    
# Line 1847  Returns:     nothing Line 2886  Returns:     nothing
2886  */  */
2887    
2888  static void  static void
2889  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2890    uschar *save_hwm)    pcre_uchar *save_hwm)
2891  {  {
2892  uschar *ptr = group;  pcre_uchar *ptr = group;
2893    
2894  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2895    {    {
2896    int offset;    int offset;
2897    uschar *hc;    pcre_uchar *hc;
2898    
2899    /* 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
2900    reference. */    reference. */
2901    
2902    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2903      {      {
2904      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
2905      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
2906        {        {
2907        PUT(hc, 0, offset + adjust);        PUT(hc, 0, offset + adjust);
# Line 1875  while ((ptr = (uschar *)find_recurse(ptr Line 2914  while ((ptr = (uschar *)find_recurse(ptr
2914    
2915    if (hc >= cd->hwm)    if (hc >= cd->hwm)
2916      {      {
2917      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
2918      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2919      }      }
2920    
# Line 1900  Arguments: Line 2939  Arguments:
2939  Returns:         new code pointer  Returns:         new code pointer
2940  */  */
2941    
2942  static uschar *  static pcre_uchar *
2943  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2944  {  {
2945  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2946  *code++ = 255;  *code++ = 255;
2947  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2948  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2949  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2950  }  }
2951    
2952    
# Line 1929  Returns:             nothing Line 2968  Returns:             nothing
2968  */  */
2969    
2970  static void  static void
2971  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2972  {  {
2973  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2974  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2975  }  }
2976    
# Line 1943  PUT(previous_callout, 2 + LINK_SIZE, len Line 2982  PUT(previous_callout, 2 + LINK_SIZE, len
2982  *************************************************/  *************************************************/
2983    
2984  /* 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
2985  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
2986  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
2987  start address.  start address. A character with multiple other cases is returned on its own
2988    with a special return value.
2989    
2990  Arguments:  Arguments:
2991    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1953  Arguments: Line 2993  Arguments:
2993    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2994    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2995    
2996  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2997                   0 when a range is returned
2998                  >0 the CASESET offset for char with multiple other cases
2999                    in this case, ocptr contains the original
3000  */  */
3001    
3002  static BOOL  static int
3003  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3004    unsigned int *odptr)    pcre_uint32 *odptr)
3005  {  {
3006  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
3007    unsigned int co;
3008    
3009    /* Find the first character that has an other case. If it has multiple other
3010    cases, return its case offset value. */
3011    
3012  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
3013    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
3014      if ((co = UCD_CASESET(c)) != 0)
3015        {
3016        *ocptr = c++;   /* Character that has the set */
3017        *cptr = c;      /* Rest of input range */
3018        return (int)co;
3019        }
3020      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3021      }
3022    
3023  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
3024    
3025  *ocptr = othercase;  *ocptr = othercase;
3026  next = othercase + 1;  next = othercase + 1;
3027    
3028  for (++c; c <= d; c++)  for (++c; c <= d; c++)
3029    {    {
3030    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
3031    next++;    next++;
3032    }    }
3033    
3034  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
3035  *cptr = c;  *cptr = c;             /* Rest of input range */
3036    return 0;
3037    }
3038    
3039    
3040  return TRUE;  
3041    /*************************************************
3042    *        Check a character and a property        *
3043    *************************************************/
3044    
3045    /* This function is called by check_auto_possessive() when a property item
3046    is adjacent to a fixed character.
3047    
3048    Arguments:
3049      c            the character
3050      ptype        the property type
3051      pdata        the data for the type
3052      negated      TRUE if it's a negated property (\P or \p{^)
3053    
3054    Returns:       TRUE if auto-possessifying is OK
3055    */
3056    
3057    static BOOL
3058    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)
3059    {
3060    #ifdef SUPPORT_UCP
3061    const pcre_uint32 *p;
3062    #endif
3063    
3064    const ucd_record *prop = GET_UCD(c);
3065    
3066    switch(ptype)
3067      {
3068      case PT_LAMP:
3069      return (prop->chartype == ucp_Lu ||
3070              prop->chartype == ucp_Ll ||
3071              prop->chartype == ucp_Lt) == negated;
3072    
3073      case PT_GC:
3074      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
3075    
3076      case PT_PC:
3077      return (pdata == prop->chartype) == negated;
3078    
3079      case PT_SC:
3080      return (pdata == prop->script) == negated;
3081    
3082      /* These are specials */
3083    
3084      case PT_ALNUM:
3085      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3086              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
3087    
3088      case PT_SPACE:    /* Perl space */
3089      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3090              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
3091              == negated;
3092    
3093      case PT_PXSPACE:  /* POSIX space */
3094      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3095              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3096              c == CHAR_FF || c == CHAR_CR)
3097              == negated;
3098    
3099      case PT_WORD:
3100      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3101              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3102              c == CHAR_UNDERSCORE) == negated;
3103    
3104    #ifdef SUPPORT_UCP
3105      case PT_CLIST:
3106      p = PRIV(ucd_caseless_sets) + prop->caseset;
3107      for (;;)
3108        {
3109        if (c < *p) return !negated;
3110        if (c == *p++) return negated;
3111        }
3112      break;  /* Control never reaches here */
3113    #endif
3114      }
3115    
3116    return FALSE;
3117  }  }
3118  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3119    
# Line 1994  whether the next thing could possibly ma Line 3128  whether the next thing could possibly ma
3128  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3129    
3130  Arguments:  Arguments:
3131    op_code       the repeated op code    previous      pointer to the repeated opcode
3132    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  
3133    ptr           next character in pattern    ptr           next character in pattern
3134    options       options bits    options       options bits
3135    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns:        TRUE if possessifying is Line 3138  Returns:        TRUE if possessifying is
3138  */  */
3139    
3140  static BOOL  static BOOL
3141  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3142    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3143  {  {
3144  int next;  pcre_uint32 c = NOTACHAR;
3145    pcre_uint32 next;
3146    int escape;
3147    pcre_uchar op_code = *previous++;
3148    
3149  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3150    
# Line 2017  if ((options & PCRE_EXTENDED) != 0) Line 3152  if ((options & PCRE_EXTENDED) != 0)
3152    {    {
3153    for (;;)    for (;;)
3154      {      {
3155      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3156      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3157        {        {
3158        while (*(++ptr) != 0)        ptr++;
3159          while (*ptr != CHAR_NULL)
3160            {
3161          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3162            ptr++;
3163    #ifdef SUPPORT_UTF
3164            if (utf) FORWARDCHAR(ptr);
3165    #endif
3166            }
3167        }        }
3168      else break;      else break;
3169      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 3172  if ((options & PCRE_EXTENDED) != 0)
3172  /* 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
3173  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3174    
3175  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3176    {    {
3177    int temperrorcode = 0;    int temperrorcode = 0;
3178    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,
3179        FALSE);
3180    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3181    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3182    }    }
3183    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3184    {    {
3185  #ifdef SUPPORT_UTF8    escape = 0;
3186    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
3187      if (utf) { GETCHARINC(next, ptr); } else
3188  #endif  #endif
3189    next = *ptr++;    next = *ptr++;
3190    }    }
   
3191  else return FALSE;  else return FALSE;
3192    
3193  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2054  if ((options & PCRE_EXTENDED) != 0) Line 3196  if ((options & PCRE_EXTENDED) != 0)
3196    {    {
3197    for (;;)    for (;;)
3198      {      {
3199      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3200      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3201        {        {
3202        while (*(++ptr) != 0)        ptr++;
3203          while (*ptr != CHAR_NULL)
3204            {
3205          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3206            ptr++;
3207    #ifdef SUPPORT_UTF
3208            if (utf) FORWARDCHAR(ptr);
3209    #endif
3210            }
3211        }        }
3212      else break;      else break;
3213      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 3215  if ((options & PCRE_EXTENDED) != 0)
3215    
3216  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3217    
3218  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3219    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3220        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. */  
   
3221    
3222  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3223    
3224  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3225        op_code == OP_NOT || op_code == OP_NOTI)
3226    {    {
3227    case OP_CHAR:  #ifdef SUPPORT_UTF
3228  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3229    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3230      c = *previous;
3231  #endif  #endif
3232    return item != next;    }
3233    
3234    /* 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
3235    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3236    
3237    case OP_CHARNC:  if (escape == 0)
3238  #ifdef SUPPORT_UTF8    {
3239    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
3240      case, which maps to the special PT_CLIST property. Check this first. */
3241    
3242    #ifdef SUPPORT_UCP
3243      if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3244        {
3245        unsigned int ocs = UCD_CASESET(next);
3246        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3247        }
3248  #endif  #endif
3249    if (item == next) return FALSE;  
3250  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3251      {      {
3252      unsigned int othercase;      case OP_CHAR:
3253      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3254    
3255        /* For CHARI (caseless character) we must check the other case. If we have
3256        Unicode property support, we can use it to test the other case of
3257        high-valued characters. We know that next can have only one other case,
3258        because multi-other-case characters are dealt with above. */
3259    
3260        case OP_CHARI:
3261        if (c == next) return FALSE;
3262    #ifdef SUPPORT_UTF
3263        if (utf)
3264          {
3265          pcre_uint32 othercase;
3266          if (next < 128) othercase = cd->fcc[next]; else
3267  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3268      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE(next);
3269  #else  #else
3270      othercase = NOTACHAR;        othercase = NOTACHAR;
3271  #endif  #endif
3272      return (unsigned int)item != othercase;        return c != othercase;
3273      }        }
3274    else      else
3275  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3276    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
   
   /* For OP_NOT, "item" must be a single-byte character. */  
3277    
3278    case OP_NOT:      case OP_NOT:
3279    if (item == next) return TRUE;      return c == next;
3280    if ((options & PCRE_CASELESS) == 0) return FALSE;  
3281  #ifdef SUPPORT_UTF8      case OP_NOTI:
3282    if (utf8)      if (c == next) return TRUE;
3283      {  #ifdef SUPPORT_UTF
3284      unsigned int othercase;      if (utf)
3285      if (next < 128) othercase = cd->fcc[next]; else        {
3286          pcre_uint32 othercase;
3287          if (next < 128) othercase = cd->fcc[next]; else
3288  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3289      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE(next);
3290  #else  #else
3291      othercase = NOTACHAR;        othercase = NOTACHAR;
3292  #endif  #endif
3293      return (unsigned int)item == othercase;        return c == othercase;
3294      }        }
3295    else      else
3296  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3297    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3298    
3299    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3300    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3301    
3302    case OP_NOT_DIGIT:      case OP_DIGIT:
3303    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3304    
3305    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3306    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3307    
3308    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3309    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3310    
3311    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3312    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3313    
3314    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3315    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3316    
3317    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3318    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3319    switch(next)  
3320      {      case OP_HSPACE:
3321      case 0x09:      case OP_NOT_HSPACE:
3322      case 0x20:      switch(next)
3323      case 0xa0:        {
3324      case 0x1680:        HSPACE_CASES:
3325      case 0x180e:        return op_code == OP_NOT_HSPACE;
3326      case 0x2000:  
3327      case 0x2001:        default:
3328      case 0x2002:        return op_code != OP_NOT_HSPACE;
3329      case 0x2003:        }
3330      case 0x2004:  
3331      case 0x2005:      case OP_ANYNL:
3332      case 0x2006:      case OP_VSPACE:
3333      case 0x2007:      case OP_NOT_VSPACE:
3334      case 0x2008:      switch(next)
3335      case 0x2009:        {
3336      case 0x200A:        VSPACE_CASES:
3337      case 0x202f:        return op_code == OP_NOT_VSPACE;
3338      case 0x205f:  
3339      case 0x3000:        default:
3340      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3341      default:        }
3342      return op_code == OP_HSPACE;  
3343      }  #ifdef SUPPORT_UCP
3344        case OP_PROP:
3345        return check_char_prop(next, previous[0], previous[1], FALSE);
3346    
3347        case OP_NOTPROP:
3348        return check_char_prop(next, previous[0], previous[1], TRUE);
3349    #endif
3350    
   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;  
3351      default:      default:
3352      return op_code == OP_VSPACE;      return FALSE;
3353      }      }
   
   default:  
   return FALSE;  
3354    }    }
3355    
3356    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3357  /* 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
3358    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3359    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3360    replaced by OP_PROP codes when PCRE_UCP is set. */
3361    
3362  switch(op_code)  switch(op_code)
3363    {    {
3364    case OP_CHAR:    case OP_CHAR:
3365    case OP_CHARNC:    case OP_CHARI:
3366  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3367      {      {
3368      case ESC_d:      case ESC_d:
3369      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3370    
3371      case ESC_D:      case ESC_D:
3372      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3373    
3374      case ESC_s:      case ESC_s:
3375      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3376    
3377      case ESC_S:      case ESC_S:
3378      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3379    
3380      case ESC_w:      case ESC_w:
3381      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3382    
3383      case ESC_W:      case ESC_W:
3384      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3385    
3386      case ESC_h:      case ESC_h:
3387      case ESC_H:      case ESC_H:
3388      switch(item)      switch(c)
3389        {        {
3390        case 0x09:        HSPACE_CASES:
3391        case 0x20:        return escape != ESC_h;
3392        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;  
3393        default:        default:
3394        return -next == ESC_h;        return escape == ESC_h;
3395        }        }
3396    
3397      case ESC_v:      case ESC_v:
3398      case ESC_V:      case ESC_V:
3399      switch(item)      switch(c)
3400        {        {
3401        case 0x0a:        VSPACE_CASES:
3402        case 0x0b:        return escape != ESC_v;
3403        case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
3404        default:        default:
3405        return -next == ESC_v;        return escape == ESC_v;
3406          }
3407    
3408        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3409        their substitutions and process them. The result will always be either
3410        ESC_p or ESC_P. Then fall through to process those values. */
3411    
3412    #ifdef SUPPORT_UCP
3413        case ESC_du:
3414        case ESC_DU:
3415        case ESC_wu:
3416        case ESC_WU:
3417        case ESC_su:
3418        case ESC_SU:
3419          {
3420          int temperrorcode = 0;
3421          ptr = substitutes[escape - ESC_DU];
3422          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3423          if (temperrorcode != 0) return FALSE;
3424          ptr++;    /* For compatibility */
3425        }        }
3426        /* Fall through */
3427    
3428        case ESC_p:
3429        case ESC_P:
3430          {
3431          unsigned int ptype = 0, pdata = 0;
3432          int errorcodeptr;
3433          BOOL negated;
3434    
3435          ptr--;      /* Make ptr point at the p or P */
3436          if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))
3437            return FALSE;
3438          ptr++;      /* Point past the final curly ket */
3439    
3440          /* If the property item is optional, we have to give up. (When generated
3441          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3442          to the original \d etc. At this point, ptr will point to a zero byte. */
3443    
3444          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3445            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3446              return FALSE;
3447    
3448          /* Do the property check. */
3449    
3450          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3451          }
3452    #endif
3453    
3454      default:      default:
3455      return FALSE;      return FALSE;
3456      }      }
3457    
3458      /* In principle, support for Unicode properties should be integrated here as
3459      well. It means re-organizing the above code so as to get hold of the property
3460      values before switching on the op-code. However, I wonder how many patterns
3461      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3462      these op-codes are never generated.) */
3463    
3464    case OP_DIGIT:    case OP_DIGIT:
3465    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3466           next == -ESC_h || next == -ESC_v;           escape == ESC_h || escape == ESC_v || escape == ESC_R;
3467    
3468    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3469    return next == -ESC_d;    return escape == ESC_d;
3470    
3471    case OP_WHITESPACE:    case OP_WHITESPACE:
3472    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3473    
3474    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3475    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3476    
3477    case OP_HSPACE:    case OP_HSPACE:
3478    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3479             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3480    
3481    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3482    return next == -ESC_h;    return escape == ESC_h;
3483    
3484    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3485      case OP_ANYNL:
3486    case OP_VSPACE:    case OP_VSPACE:
3487    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3488    
3489    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3490    return next == -ESC_v;    return escape == ESC_v || escape == ESC_R;
3491    
3492    case OP_WORDCHAR:    case OP_WORDCHAR:
3493    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3494             escape == ESC_v || escape == ESC_R;
3495    
3496    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3497    return next == -ESC_w || next == -ESC_d;    return escape == ESC_w || escape == ESC_d;
3498    
3499    default:    default:
3500    return FALSE;    return FALSE;
# Line 2317  switch(op_code) Line 3506  switch(op_code)
3506    
3507    
3508  /*************************************************  /*************************************************
3509    *        Add a character or range to a class     *
3510    *************************************************/
3511    
3512    /* This function packages up the logic of adding a character or range of
3513    characters to a class. The character values in the arguments will be within the
3514    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3515    mutually recursive with the function immediately below.
3516    
3517    Arguments:
3518      classbits     the bit map for characters < 256
3519      uchardptr     points to the pointer for extra data
3520      options       the options word
3521      cd            contains pointers to tables etc.
3522      start         start of range character
3523      end           end of range character
3524    
3525    Returns:        the number of < 256 characters added
3526                    the pointer to extra data is updated
3527    */
3528    
3529    static int
3530    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3531      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3532    {
3533    pcre_uint32 c;
3534    int n8 = 0;
3535    
3536    /* If caseless matching is required, scan the range and process alternate
3537    cases. In Unicode, there are 8-bit characters that have alternate cases that
3538    are greater than 255 and vice-versa. Sometimes we can just extend the original
3539    range. */
3540    
3541    if ((options & PCRE_CASELESS) != 0)
3542      {
3543    #ifdef SUPPORT_UCP
3544      if ((options & PCRE_UTF8) != 0)
3545        {
3546        int rc;
3547        pcre_uint32 oc, od;
3548    
3549        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3550        c = start;
3551    
3552        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3553          {
3554          /* Handle a single character that has more than one other case. */
3555    
3556          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3557            PRIV(ucd_caseless_sets) + rc, oc);
3558    
3559          /* Do nothing if the other case range is within the original range. */
3560    
3561          else if (oc >= start && od <= end) continue;
3562    
3563          /* Extend the original range if there is overlap, noting that if oc < c, we
3564          can't have od > end because a subrange is always shorter than the basic
3565          range. Otherwise, use a recursive call to add the additional range. */
3566    
3567          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3568          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3569          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3570          }
3571        }
3572      else
3573    #endif  /* SUPPORT_UCP */
3574    
3575      /* Not UTF-mode, or no UCP */
3576    
3577      for (c = start; c <= end && c < 256; c++)
3578        {
3579        SETBIT(classbits, cd->fcc[c]);
3580        n8++;
3581        }
3582      }
3583    
3584    /* Now handle the original range. Adjust the final value according to the bit
3585    length - this means that the same lists of (e.g.) horizontal spaces can be used
3586    in all cases. */
3587    
3588    #if defined COMPILE_PCRE8
3589    #ifdef SUPPORT_UTF
3590      if ((options & PCRE_UTF8) == 0)
3591    #endif
3592      if (end > 0xff) end = 0xff;
3593    
3594    #elif defined COMPILE_PCRE16
3595    #ifdef SUPPORT_UTF
3596      if ((options & PCRE_UTF16) == 0)
3597    #endif
3598      if (end > 0xffff) end = 0xffff;
3599    
3600    #endif /* COMPILE_PCRE[8|16] */
3601    
3602    /* If all characters are less than 256, use the bit map. Otherwise use extra
3603    data. */
3604    
3605    if (end < 0x100)
3606      {
3607      for (c = start; c <= end; c++)
3608        {
3609        n8++;
3610        SETBIT(classbits, c);
3611        }
3612      }
3613    
3614    else
3615      {
3616      pcre_uchar *uchardata = *uchardptr;
3617    
3618    #ifdef SUPPORT_UTF
3619      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3620        {
3621        if (start < end)
3622          {
3623          *uchardata++ = XCL_RANGE;
3624          uchardata += PRIV(ord2utf)(start, uchardata);
3625          uchardata += PRIV(ord2utf)(end, uchardata);
3626          }
3627        else if (start == end)
3628          {
3629          *uchardata++ = XCL_SINGLE;
3630          uchardata += PRIV(ord2utf)(start, uchardata);
3631          }
3632        }
3633      else
3634    #endif  /* SUPPORT_UTF */
3635    
3636      /* Without UTF support, character values are constrained by the bit length,
3637      and can only be > 256 for 16-bit and 32-bit libraries. */
3638    
3639    #ifdef COMPILE_PCRE8
3640        {}
3641    #else
3642      if (start < end)
3643        {
3644        *uchardata++ = XCL_RANGE;
3645        *uchardata++ = start;
3646        *uchardata++ = end;
3647        }
3648      else if (start == end)
3649        {
3650        *uchardata++ = XCL_SINGLE;
3651        *uchardata++ = start;
3652        }
3653    #endif
3654    
3655      *uchardptr = uchardata;   /* Updata extra data pointer */
3656      }
3657    
3658    return n8;    /* Number of 8-bit characters */
3659    }
3660    
3661    
3662    
3663    
3664    /*************************************************
3665    *        Add a list of characters to a class     *
3666    *************************************************/
3667    
3668    /* This function is used for adding a list of case-equivalent characters to a
3669    class, and also for adding a list of horizontal or vertical whitespace. If the
3670    list is in order (which it should be), ranges of characters are detected and
3671    handled appropriately. This function is mutually recursive with the function
3672    above.
3673    
3674    Arguments:
3675      classbits     the bit map for characters < 256
3676      uchardptr     points to the pointer for extra data
3677      options       the options word
3678      cd            contains pointers to tables etc.
3679      p             points to row of 32-bit values, terminated by NOTACHAR
3680      except        character to omit; this is used when adding lists of
3681                      case-equivalent characters to avoid including the one we
3682                      already know about
3683    
3684    Returns:        the number of < 256 characters added
3685                    the pointer to extra data is updated
3686    */
3687    
3688    static int
3689    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3690      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3691    {
3692    int n8 = 0;
3693    while (p[0] < NOTACHAR)
3694      {
3695      int n = 0;
3696      if (p[0] != except)
3697        {
3698        while(p[n+1] == p[0] + n + 1) n++;
3699        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3700        }
3701      p += n + 1;
3702      }
3703    return n8;
3704    }
3705    
3706    
3707    
3708    /*************************************************
3709    *    Add characters not in a list to a class     *
3710    *************************************************/
3711    
3712    /* This function is used for adding the complement of a list of horizontal or
3713    vertical whitespace to a class. The list must be in order.
3714    
3715    Arguments:
3716      classbits     the bit map for characters < 256
3717      uchardptr     points to the pointer for extra data
3718      options       the options word
3719      cd            contains pointers to tables etc.
3720      p             points to row of 32-bit values, terminated by NOTACHAR
3721    
3722    Returns:        the number of < 256 characters added
3723                    the pointer to extra data is updated
3724    */
3725    
3726    static int
3727    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3728      int options, compile_data *cd, const pcre_uint32 *p)
3729    {
3730    BOOL utf = (options & PCRE_UTF8) != 0;
3731    int n8 = 0;
3732    if (p[0] > 0)
3733      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3734    while (p[0] < NOTACHAR)
3735      {
3736      while (p[1] == p[0] + 1) p++;
3737      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3738        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3739      p++;
3740      }
3741    return n8;
3742    }
3743    
3744    
3745    
3746    /*************************************************
3747  *           Compile one branch                   *  *           Compile one branch                   *
3748  *************************************************/  *************************************************/
3749    
# Line 2331  Arguments: Line 3758  Arguments:
3758    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3759    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3760    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3761    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr    place to put the first required character
3762    reqbyteptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
3763      reqcharptr     place to put the last required character
3764      reqcharflagsptr place to put the last required character flags, or a negative number
3765    bcptr          points to current branch chain    bcptr          points to current branch chain
3766      cond_depth     conditional nesting depth
3767    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3768    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3769                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2343  Returns:         TRUE on success Line 3773  Returns:         TRUE on success
3773  */  */
3774    
3775  static BOOL  static BOOL
3776  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3777    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
3778      pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
3779      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
3780      branch_chain *bcptr, int cond_depth,
3781    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3782  {  {
3783  int repeat_type, op_type;  int repeat_type, op_type;
3784  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3785  int bravalue = 0;  int bravalue = 0;
3786  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3787  int firstbyte, reqbyte;  pcre_uint32 firstchar, reqchar;
3788  int zeroreqbyte, zerofirstbyte;  pcre_int32 firstcharflags, reqcharflags;
3789  int req_caseopt, reqvary, tempreqvary;  pcre_uint32 zeroreqchar, zerofirstchar;
3790  int options = *optionsptr;  pcre_int32 zeroreqcharflags, zerofirstcharflags;
3791    pcre_int32 req_caseopt, reqvary, tempreqvary;
3792    int options = *optionsptr;               /* May change dynamically */
3793  int after_manual_callout = 0;  int after_manual_callout = 0;
3794  int length_prevgroup = 0;  int length_prevgroup = 0;
3795  register int c;  register pcre_uint32 c;
3796  register uschar *code = *codeptr;  int escape;
3797  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3798  uschar *orig_code = code;  pcre_uchar *last_code = code;
3799  uschar *tempcode;  pcre_uchar *orig_code = code;
3800    pcre_uchar *tempcode;
3801  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3802  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3803  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3804  const uschar *tempptr;  const pcre_uchar *tempptr;
3805  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3806  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3807  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3808  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3809    pcre_uint8 classbits[32];
3810  #ifdef SUPPORT_UTF8  
3811  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3812  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3813  uschar *class_utf8data;  dynamically as we process the pattern. */
3814  uschar *class_utf8data_base;  
3815  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3816    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3817    BOOL utf = (options & PCRE_UTF8) != 0;
3818    #ifndef COMPILE_PCRE32
3819    pcre_uchar utf_chars[6];
3820    #endif
3821  #else  #else
3822  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3823  uschar *utf8_char = NULL;  #endif
3824    
3825    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3826    class_uchardata always so that it can be passed to add_to_class() always,
3827    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3828    alternative calls for the different cases. */