/[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 360 by ph10, Wed Jul 9 20:00:28 2008 UTC revision 1048 by ph10, Fri Sep 28 16:03:14 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,
88        int, int, int *, int *, branch_chain *, compile_data *, int *);
89    
90    
91    
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 103  so this number is very generous.
103  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
104  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
105  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
106  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
107    filled up by repetitions of forward references, for example patterns like
108    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
109    that the workspace is expanded using malloc() in this situation. The value
110    below is therefore a minimum, and we put a maximum on it for safety. The
111    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
112    kicks in at the same number of forward references in all cases. */
113    
114    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
115    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
116    
117  #define COMPILE_WORK_SIZE (4096)  /* The overrun tests check for a slightly smaller size so that they detect the
118    overrun before it actually does run off the end of the data block. */
119    
120    #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    
127    /* Repeated character flags. */
128    
129    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
130    
131  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
132  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
133  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
134  is invalid. */  is invalid. */
135    
136  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
137    
138    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
139    in UTF-8 mode. */
140    
141  static const short int escapes[] = {  static const short int escapes[] = {
142       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
143       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
144     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
145  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
146  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
147  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
148     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
149  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
150  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
151       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
152         -ESC_D,                  -ESC_E,
153         0,                       -ESC_G,
154         -ESC_H,                  0,
155         0,                       -ESC_K,
156         0,                       0,
157         -ESC_N,                  0,
158         -ESC_P,                  -ESC_Q,
159         -ESC_R,                  -ESC_S,
160         0,                       0,
161         -ESC_V,                  -ESC_W,
162         -ESC_X,                  0,
163         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
164         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
165         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
166         CHAR_GRAVE_ACCENT,       7,
167         -ESC_b,                  0,
168         -ESC_d,                  ESC_e,
169         ESC_f,                   0,
170         -ESC_h,                  0,
171         0,                       -ESC_k,
172         0,                       0,
173         ESC_n,                   0,
174         -ESC_p,                  0,
175         ESC_r,                   -ESC_s,
176         ESC_tee,                 0,
177         -ESC_v,                  -ESC_w,
178         0,                       0,
179         -ESC_z
180  };  };
181    
182  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
183    
184    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
185    
186  static const short int escapes[] = {  static const short int escapes[] = {
187  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
188  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 201  static const short int escapes[] = {
201  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
202  /*  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,
203  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
204  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
205  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
206  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
207  /*  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 213  static const short int escapes[] = {
213    
214  /* 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
215  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
216  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
217    string is built from string macros so that it works in UTF-8 mode on EBCDIC
218    platforms. */
219    
220  typedef struct verbitem {  typedef struct verbitem {
221    int   len;    int   len;                 /* Length of verb name */
222    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
223      int   op_arg;              /* Op when arg present, or -1 if not allowed */
224  } verbitem;  } verbitem;
225    
226  static const char verbnames[] =  static const char verbnames[] =
227    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
228    "COMMIT\0"    STRING_MARK0
229    "F\0"    STRING_ACCEPT0
230    "FAIL\0"    STRING_COMMIT0
231    "PRUNE\0"    STRING_F0
232    "SKIP\0"    STRING_FAIL0
233    "THEN";    STRING_PRUNE0
234      STRING_SKIP0
235      STRING_THEN;
236    
237  static const verbitem verbs[] = {  static const verbitem verbs[] = {
238    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
239    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
240    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
241    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
242    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
243    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
244    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
245      { 4, OP_SKIP,   OP_SKIP_ARG  },
246      { 4, OP_THEN,   OP_THEN_ARG  }
247  };  };
248    
249  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 256  length entry. The first three must be al
256  for handling case independence. */  for handling case independence. */
257    
258  static const char posix_names[] =  static const char posix_names[] =
259    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
260    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
261    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
262      STRING_word0  STRING_xdigit;
263    
264  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
265    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 };
266    
267  /* 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 291  static const int posix_class_maps[] = {
291    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
292  };  };
293    
294    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
295    substitutes must be in the order of the names, defined above, and there are
296    both positive and negative cases. NULL means no substitute. */
297    
298    #ifdef SUPPORT_UCP
299    static const pcre_uchar string_PNd[]  = {
300      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
301      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
302    static const pcre_uchar string_pNd[]  = {
303      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
304      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
305    static const pcre_uchar string_PXsp[] = {
306      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
307      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
308    static const pcre_uchar string_pXsp[] = {
309      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
310      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
311    static const pcre_uchar string_PXwd[] = {
312      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_pXwd[] = {
315      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    
318    static const pcre_uchar *substitutes[] = {
319      string_PNd,           /* \D */
320      string_pNd,           /* \d */
321      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
322      string_pXsp,          /* \s */
323      string_PXwd,          /* \W */
324      string_pXwd           /* \w */
325    };
326    
327    static const pcre_uchar string_pL[] =   {
328      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_pLl[] =  {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_pLu[] =  {
334      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_pXan[] = {
337      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_h[] =    {
340      CHAR_BACKSLASH, CHAR_h, '\0' };
341    static const pcre_uchar string_pXps[] = {
342      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
343      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
344    static const pcre_uchar string_PL[] =   {
345      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
346      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347    static const pcre_uchar string_PLl[] =  {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    static const pcre_uchar string_PLu[] =  {
351      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
352      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
353    static const pcre_uchar string_PXan[] = {
354      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
355      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
356    static const pcre_uchar string_H[] =    {
357      CHAR_BACKSLASH, CHAR_H, '\0' };
358    static const pcre_uchar string_PXps[] = {
359      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    
362    static const pcre_uchar *posix_substitutes[] = {
363      string_pL,            /* alpha */
364      string_pLl,           /* lower */
365      string_pLu,           /* upper */
366      string_pXan,          /* alnum */
367      NULL,                 /* ascii */
368      string_h,             /* blank */
369      NULL,                 /* cntrl */
370      string_pNd,           /* digit */
371      NULL,                 /* graph */
372      NULL,                 /* print */
373      NULL,                 /* punct */
374      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
375      string_pXwd,          /* word */
376      NULL,                 /* xdigit */
377      /* Negated cases */
378      string_PL,            /* ^alpha */
379      string_PLl,           /* ^lower */
380      string_PLu,           /* ^upper */
381      string_PXan,          /* ^alnum */
382      NULL,                 /* ^ascii */
383      string_H,             /* ^blank */
384      NULL,                 /* ^cntrl */
385      string_PNd,           /* ^digit */
386      NULL,                 /* ^graph */
387      NULL,                 /* ^print */
388      NULL,                 /* ^punct */
389      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
390      string_PXwd,          /* ^word */
391      NULL                  /* ^xdigit */
392    };
393    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
394    #endif
395    
396  #define STRING(a)  # a  #define STRING(a)  # a
397  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 404  the number of relocations needed when a
404  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
405  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
406  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
407  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
408    
409    Each substring ends with \0 to insert a null character. This includes the final
410    substring, so that the whole string ends with \0\0, which can be detected when
411    counting through. */
412    
413  static const char error_texts[] =  static const char error_texts[] =
414    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 449  static const char error_texts[] =
449    /* 30 */    /* 30 */
450    "unknown POSIX class name\0"    "unknown POSIX class name\0"
451    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
452    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
453    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
454    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
455    /* 35 */    /* 35 */
456    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
457    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
458    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
459    "number after (?C is > 255\0"    "number after (?C is > 255\0"
460    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
461    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 472  static const char error_texts[] =
472    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
473    /* 50 */    /* 50 */
474    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
475    "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"
476    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
477    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
478    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
479    /* 55 */    /* 55 */
480    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
481    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
482    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
483    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
484    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
485    /* 60 */    /* 60 */
486    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
487    "number is too big\0"    "number is too big\0"
488    "subpattern name expected\0"    "subpattern name expected\0"
489    "digit expected after (?+\0"    "digit expected after (?+\0"
490    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
491      /* 65 */
492      "different names for subpatterns of the same number are not allowed\0"
493      "(*MARK) must have an argument\0"
494      "this version of PCRE is not compiled with Unicode property support\0"
495      "\\c must be followed by an ASCII character\0"
496      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
497      /* 70 */
498      "internal error: unknown opcode in find_fixedlength()\0"
499      "\\N is not supported in a class\0"
500      "too many forward references\0"
501      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
502      "invalid UTF-16 string\0"
503      /* 75 */
504      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
505      "character value in \\u.... sequence is too large\0"
506      ;
507    
508  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
509  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 322  For convenience, we use the same bit def Line 521  For convenience, we use the same bit def
521    
522  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
523    
524  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
525  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
526    into a subtraction and unsigned comparison). */
527    
528    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
529    
530    #ifndef EBCDIC
531    
532    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
533    UTF-8 mode. */
534    
535    static const pcre_uint8 digitab[] =
536    {    {
537    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
538    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 358  static const unsigned char digitab[] = Line 567  static const unsigned char digitab[] =
567    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
568    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
569    
570  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
571  static const unsigned char digitab[] =  
572    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
573    
574    static const pcre_uint8 digitab[] =
575    {    {
576    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
577    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 394  static const unsigned char digitab[] = Line 606  static const unsigned char digitab[] =
606    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
607    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
608    
609  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
610    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
611    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
612    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 430  static const unsigned char ebcdic_charta Line 642  static const unsigned char ebcdic_charta
642  #endif  #endif
643    
644    
 /* 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 *);  
   
645    
646    
647  /*************************************************  /*************************************************
# Line 455  static const char * Line 661  static const char *
661  find_error_text(int n)  find_error_text(int n)
662  {  {
663  const char *s = error_texts;  const char *s = error_texts;
664  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
665      {
666      while (*s++ != 0) {};
667      if (*s == 0) return "Error text not found (please report)";
668      }
669  return s;  return s;
670  }  }
671    
672    
673  /*************************************************  /*************************************************
674    *           Expand the workspace                 *
675    *************************************************/
676    
677    /* This function is called during the second compiling phase, if the number of
678    forward references fills the existing workspace, which is originally a block on
679    the stack. A larger block is obtained from malloc() unless the ultimate limit
680    has been reached or the increase will be rather small.
681    
682    Argument: pointer to the compile data block
683    Returns:  0 if all went well, else an error number
684    */
685    
686    static int
687    expand_workspace(compile_data *cd)
688    {
689    pcre_uchar *newspace;
690    int newsize = cd->workspace_size * 2;
691    
692    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
693    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
694        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
695     return ERR72;
696    
697    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
698    if (newspace == NULL) return ERR21;
699    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
700    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
701    if (cd->workspace_size > COMPILE_WORK_SIZE)
702      (PUBL(free))((void *)cd->start_workspace);
703    cd->start_workspace = newspace;
704    cd->workspace_size = newsize;
705    return 0;
706    }
707    
708    
709    
710    /*************************************************
711    *            Check for counted repeat            *
712    *************************************************/
713    
714    /* This function is called when a '{' is encountered in a place where it might
715    start a quantifier. It looks ahead to see if it really is a quantifier or not.
716    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
717    where the ddds are digits.
718    
719    Arguments:
720      p         pointer to the first char after '{'
721    
722    Returns:    TRUE or FALSE
723    */
724    
725    static BOOL
726    is_counted_repeat(const pcre_uchar *p)
727    {
728    if (!IS_DIGIT(*p)) return FALSE;
729    p++;
730    while (IS_DIGIT(*p)) p++;
731    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
732    
733    if (*p++ != CHAR_COMMA) return FALSE;
734    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
735    
736    if (!IS_DIGIT(*p)) return FALSE;
737    p++;
738    while (IS_DIGIT(*p)) p++;
739    
740    return (*p == CHAR_RIGHT_CURLY_BRACKET);
741    }
742    
743    
744    
745    /*************************************************
746  *            Handle escapes                      *  *            Handle escapes                      *
747  *************************************************/  *************************************************/
748    
# Line 485  Returns:         zero or positive => a d Line 767  Returns:         zero or positive => a d
767  */  */
768    
769  static int  static int
770  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
771    int options, BOOL isclass)    int options, BOOL isclass)
772  {  {
773  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
774  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
775  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
776    pcre_int32 c;
777    int i;
778    
779  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
780  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 503  if (c == 0) *errorcodeptr = ERR1; Line 787  if (c == 0) *errorcodeptr = ERR1;
787  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.
788  Otherwise further processing may be required. */  Otherwise further processing may be required. */
789    
790  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
791  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
792  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
793    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
794    
795  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
796  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
797    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
798  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
799  #endif  #endif
800    
# Line 516  else if ((i = escapes[c - 0x48]) != 0) Line 802  else if ((i = escapes[c - 0x48]) != 0)
802    
803  else  else
804    {    {
805    const uschar *oldptr;    const pcre_uchar *oldptr;
806    BOOL braced, negated;    BOOL braced, negated;
807    
808    switch (c)    switch (c)
# Line 524  else Line 810  else
810      /* 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
811      error. */      error. */
812    
813      case 'l':      case CHAR_l:
814      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
815      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
816      break;      break;
817    
818      /* \g must be followed by one of a number of specific things:      case CHAR_u:
819        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
820          {
821          /* In JavaScript, \u must be followed by four hexadecimal numbers.
822          Otherwise it is a lowercase u letter. */
823          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
824            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
825            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
826            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
827            {
828            c = 0;
829            for (i = 0; i < 4; ++i)
830              {
831              register int cc = *(++ptr);
832    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
833              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
834              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
835    #else           /* EBCDIC coding */
836              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
837              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
838    #endif
839              }
840    
841    #ifdef COMPILE_PCRE8
842            if (c > (utf ? 0x10ffff : 0xff))
843    #else
844    #ifdef COMPILE_PCRE16
845            if (c > (utf ? 0x10ffff : 0xffff))
846    #endif
847    #endif
848              {
849              *errorcodeptr = ERR76;
850              }
851            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
852            }
853          }
854        else
855          *errorcodeptr = ERR37;
856        break;
857    
858        case CHAR_U:
859        /* In JavaScript, \U is an uppercase U letter. */
860        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
861        break;
862    
863        /* In a character class, \g is just a literal "g". Outside a character
864        class, \g must be followed by one of a number of specific things:
865    
866      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
867      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 548  else Line 877  else
877      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
878      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
879    
880      case 'g':      case CHAR_g:
881      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
882        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
883        {        {
884        c = -ESC_g;        c = -ESC_g;
885        break;        break;
# Line 557  else Line 887  else
887    
888      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
889    
890      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891        {        {
892        const uschar *p;        const pcre_uchar *p;
893        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
894          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
895        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
896          {          {
897          c = -ESC_k;          c = -ESC_k;
898          break;          break;
# Line 572  else Line 902  else
902        }        }
903      else braced = FALSE;      else braced = FALSE;
904    
905      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
906        {        {
907        negated = TRUE;        negated = TRUE;
908        ptr++;        ptr++;
909        }        }
910      else negated = FALSE;      else negated = FALSE;
911    
912        /* The integer range is limited by the machine's int representation. */
913      c = 0;      c = 0;
914      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
915        c = c * 10 + *(++ptr) - '0';        {
916          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
917      if (c < 0)   /* Integer overflow */          {
918            c = -1;
919            break;
920            }
921          c = c * 10 + *(++ptr) - CHAR_0;
922          }
923        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
924        {        {
925          while (IS_DIGIT(ptr[1]))
926            ptr++;
927        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
928        break;        break;
929        }        }
930    
931      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
932        {        {
933        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
934        break;        break;
# Line 626  else Line 965  else
965      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
966      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
967    
968      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:
969      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
970    
971      if (!isclass)      if (!isclass)
972        {        {
973        oldptr = ptr;        oldptr = ptr;
974        c -= '0';        /* The integer range is limited by the machine's int representation. */
975        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
976          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
       if (c < 0)    /* Integer overflow */  
977          {          {
978            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
979              {
980              c = -1;
981              break;
982              }
983            c = c * 10 + *(++ptr) - CHAR_0;
984            }
985          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
986            {
987            while (IS_DIGIT(ptr[1]))
988              ptr++;
989          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
990          break;          break;
991          }          }
# Line 652  else Line 1001  else
1001      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.
1002      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1003    
1004      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1005        {        {
1006        ptr--;        ptr--;
1007        c = 0;        c = 0;
# Line 662  else Line 1011  else
1011      /* \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
1012      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
1013      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
1014      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,
1015      than 3 octal digits. */      but no more than 3 octal digits. */
1016    
1017      case '0':      case CHAR_0:
1018      c -= '0';      c -= CHAR_0;
1019      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1020          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1021      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1022        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1023    #endif
1024      break;      break;
1025    
1026      /* \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
1027      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.
1028      treated as a data character. */      If not, { is treated as a data character. */
1029    
1030        case CHAR_x:
1031        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1032          {
1033          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1034          Otherwise it is a lowercase x letter. */
1035          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1037            {
1038            c = 0;
1039            for (i = 0; i < 2; ++i)
1040              {
1041              register int cc = *(++ptr);
1042    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1044              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1045    #else           /* EBCDIC coding */
1046              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1047              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048    #endif
1049              }
1050            }
1051          break;
1052          }
1053    
1054      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1055        {        {
1056        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1057    
1058        c = 0;        c = 0;
1059        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1060          {          {
1061          register int cc = *pt++;          register int cc = *pt++;
1062          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1063    
1064  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1065          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1066          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1067  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1068          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1069          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1070    #endif
1071    
1072    #ifdef COMPILE_PCRE8
1073            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1074    #else
1075    #ifdef COMPILE_PCRE16
1076            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1077  #endif  #endif
1078    #endif
1079            }
1080    
1081          if (c < 0)
1082            {
1083            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1084            *errorcodeptr = ERR34;
1085          }          }
1086    
1087        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1088          {          {
1089          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1090          ptr = pt;          ptr = pt;
1091          break;          break;
1092          }          }
# Line 712  else Line 1098  else
1098      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1099    
1100      c = 0;      c = 0;
1101      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1102        {        {
1103        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1104        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1105  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1106        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1107        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1108  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1109        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1110        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1111  #endif  #endif
1112        }        }
1113      break;      break;
1114    
1115      /* 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.
1116      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
1117        coding is ASCII-specific, but then the whole concept of \cx is
1118      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1119    
1120      case 'c':      case CHAR_c:
1121      c = *(++ptr);      c = *(++ptr);
1122      if (c == 0)      if (c == 0)
1123        {        {
1124        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1125        break;        break;
1126        }        }
1127    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1128  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1129      if (c >= 'a' && c <= 'z') c -= 32;        {
1130          *errorcodeptr = ERR68;
1131          break;
1132          }
1133        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1134      c ^= 0x40;      c ^= 0x40;
1135  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1136      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1137      c ^= 0xC0;      c ^= 0xC0;
1138  #endif  #endif
1139      break;      break;
# Line 764  else Line 1155  else
1155      }      }
1156    }    }
1157    
1158    /* Perl supports \N{name} for character names, as well as plain \N for "not
1159    newline". PCRE does not support \N{name}. However, it does support
1160    quantification such as \N{2,3}. */
1161    
1162    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1163         !is_counted_repeat(ptr+2))
1164      *errorcodeptr = ERR37;
1165    
1166    /* If PCRE_UCP is set, we change the values for \d etc. */
1167    
1168    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1169      c -= (ESC_DU - ESC_D);
1170    
1171    /* Set the pointer to the final character before returning. */
1172    
1173  *ptrptr = ptr;  *ptrptr = ptr;
1174  return c;  return c;
1175  }  }
# Line 790  Returns:         type value from ucp_typ Line 1196  Returns:         type value from ucp_typ
1196  */  */
1197    
1198  static int  static int
1199  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1200  {  {
1201  int c, i, bot, top;  int c, i, bot, top;
1202  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1203  char name[32];  pcre_uchar name[32];
1204    
1205  c = *(++ptr);  c = *(++ptr);
1206  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 804  if (c == 0) goto ERROR_RETURN; Line 1210  if (c == 0) goto ERROR_RETURN;
1210  /* \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
1211  negation. */  negation. */
1212    
1213  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1214    {    {
1215    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1216      {      {
1217      *negptr = TRUE;      *negptr = TRUE;
1218      ptr++;      ptr++;
1219      }      }
1220    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1221      {      {
1222      c = *(++ptr);      c = *(++ptr);
1223      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1224      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1225      name[i] = c;      name[i] = c;
1226      }      }
1227    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1228    name[i] = 0;    name[i] = 0;
1229    }    }
1230    
# Line 835  else Line 1241  else
1241  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1242    
1243  bot = 0;  bot = 0;
1244  top = _pcre_utt_size;  top = PRIV(utt_size);
1245    
1246  while (bot < top)  while (bot < top)
1247    {    {
1248    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1249    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1250    if (c == 0)    if (c == 0)
1251      {      {
1252      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1253      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1254      }      }
1255    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1256    }    }
# Line 864  return -1; Line 1270  return -1;
1270    
1271    
1272  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1273  *         Read repeat counts                     *  *         Read repeat counts                     *
1274  *************************************************/  *************************************************/
1275    
# Line 915  Returns:         pointer to '}' on succe Line 1288  Returns:         pointer to '}' on succe
1288                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1289  */  */
1290    
1291  static const uschar *  static const pcre_uchar *
1292  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)
1293  {  {
1294  int min = 0;  int min = 0;
1295  int max = -1;  int max = -1;
# Line 924  int max = -1; Line 1297  int max = -1;
1297  /* 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
1298  an integer overflow. */  an integer overflow. */
1299    
1300  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1301  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1302    {    {
1303    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 934  if (min < 0 || min > 65535) Line 1307  if (min < 0 || min > 65535)
1307  /* 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.
1308  Also, max must not be less than min. */  Also, max must not be less than min. */
1309    
1310  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1311    {    {
1312    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1313      {      {
1314      max = 0;      max = 0;
1315      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1316      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1317        {        {
1318        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 964  return p; Line 1337  return p;
1337    
1338    
1339  /*************************************************  /*************************************************
1340  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1341  *************************************************/  *************************************************/
1342    
1343  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1344    top-level call starts at the beginning of the pattern. All other calls must
1345    start at a parenthesis. It scans along a pattern's text looking for capturing
1346  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
1347  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
1348  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
1349  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1350  be terminated by '>' because that is checked in the first pass.  
1351    This function was originally called only from the second pass, in which we know
1352    that if (?< or (?' or (?P< is encountered, the name will be correctly
1353    terminated because that is checked in the first pass. There is now one call to
1354    this function in the first pass, to check for a recursive back reference by
1355    name (so that we can make the whole group atomic). In this case, we need check
1356    only up to the current position in the pattern, and that is still OK because
1357    and previous occurrences will have been checked. To make this work, the test
1358    for "end of pattern" is a check against cd->end_pattern in the main loop,
1359    instead of looking for a binary zero. This means that the special first-pass
1360    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1361    processing items within the loop are OK, because afterwards the main loop will
1362    terminate.)
1363    
1364  Arguments:  Arguments:
1365    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1366    cd           compile background data    cd           compile background data
1367    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1368    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1369    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1370      utf          TRUE if we are in UTF-8 / UTF-16 mode
1371      count        pointer to the current capturing subpattern number (updated)
1372    
1373  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1374  */  */
1375    
1376  static int  static int
1377  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1378    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1379  {  {
1380  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1381  int count = cd->bracount;  int start_count = *count;
1382    int hwm_count = start_count;
1383    BOOL dup_parens = FALSE;
1384    
1385  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1386    dealing with. The very first call may not start with a parenthesis. */
1387    
1388    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1389    {    {
1390    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1391    
1392      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1393    
1394      /* Handle a normal, unnamed capturing parenthesis. */
1395    
1396      else if (ptr[1] != CHAR_QUESTION_MARK)
1397        {
1398        *count += 1;
1399        if (name == NULL && *count == lorn) return *count;
1400        ptr++;
1401        }
1402    
1403      /* All cases now have (? at the start. Remember when we are in a group
1404      where the parenthesis numbers are duplicated. */
1405    
1406      else if (ptr[2] == CHAR_VERTICAL_LINE)
1407        {
1408        ptr += 3;
1409        dup_parens = TRUE;
1410        }
1411    
1412      /* Handle comments; all characters are allowed until a ket is reached. */
1413    
1414      else if (ptr[2] == CHAR_NUMBER_SIGN)
1415        {
1416        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1417        goto FAIL_EXIT;
1418        }
1419    
1420      /* Handle a condition. If it is an assertion, just carry on so that it
1421      is processed as normal. If not, skip to the closing parenthesis of the
1422      condition (there can't be any nested parens). */
1423    
1424      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1425        {
1426        ptr += 2;
1427        if (ptr[1] != CHAR_QUESTION_MARK)
1428          {
1429          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1430          if (*ptr != 0) ptr++;
1431          }
1432        }
1433    
1434      /* Start with (? but not a condition. */
1435    
1436      else
1437        {
1438        ptr += 2;
1439        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1440    
1441        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1442    
1443        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1444            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1445          {
1446          int term;
1447          const pcre_uchar *thisname;
1448          *count += 1;
1449          if (name == NULL && *count == lorn) return *count;
1450          term = *ptr++;
1451          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1452          thisname = ptr;
1453          while (*ptr != term) ptr++;
1454          if (name != NULL && lorn == ptr - thisname &&
1455              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1456            return *count;
1457          term++;
1458          }
1459        }
1460      }
1461    
1462    /* Past any initial parenthesis handling, scan for parentheses or vertical
1463    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1464    first-pass call when this value is temporarily adjusted to stop at the current
1465    position. So DO NOT change this to a test for binary zero. */
1466    
1467    for (; ptr < cd->end_pattern; ptr++)
1468      {
1469    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1470    
1471    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1472      {      {
1473      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1474      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1475        {        {
1476        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1477        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1478        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1479        }        }
1480      continue;      continue;
1481      }      }
# Line 1012  for (; *ptr != 0; ptr++) Line 1483  for (; *ptr != 0; ptr++)
1483    /* Skip over character classes; this logic must be similar to the way they    /* Skip over character classes; this logic must be similar to the way they
1484    are handled for real. If the first character is '^', skip it. Also, if the    are handled for real. If the first character is '^', skip it. Also, if the
1485    first few characters (either before or after ^) are \Q\E or \E we skip them    first few characters (either before or after ^) are \Q\E or \E we skip them
1486    too. This makes for compatibility with Perl. */    too. This makes for compatibility with Perl. Note the use of STR macros to
1487      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1488    
1489    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1490      {      {
1491      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1492      for (;;)      for (;;)
1493        {        {
1494        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == '\\')  
1495          {          {
1496          if (ptr[1] == 'E') ptr++;          if (ptr[2] == CHAR_E)
1497            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr+= 2;
1498              else break;          else if (STRNCMP_UC_C8(ptr + 2,
1499                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1500              ptr += 4;
1501            else
1502              break;
1503          }          }
1504        else if (!negate_class && c == '^')        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1505            {
1506          negate_class = TRUE;          negate_class = TRUE;
1507            ptr++;
1508            }
1509        else break;        else break;
1510        }        }
1511    
1512      /* If the next character is ']', it is a data character that must be      /* If the next character is ']', it is a data character that must be
1513      skipped, except in JavaScript compatibility mode. */      skipped, except in JavaScript compatibility mode. */
1514    
1515      if (ptr[1] == ']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1516            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1517        ptr++;        ptr++;
1518    
1519      while (*(++ptr) != ']')      while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1520        {        {
1521        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1522        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1523          {          {
1524          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1525          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1526            {            {
1527            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1528            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1529            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1530            }            }
1531          continue;          continue;
1532          }          }
# Line 1057  for (; *ptr != 0; ptr++) Line 1536  for (; *ptr != 0; ptr++)
1536    
1537    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1538    
1539    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1540      {      {
1541      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1542      if (*ptr == 0) return -1;      while (*ptr != 0)
1543          {
1544          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1545          ptr++;
1546    #ifdef SUPPORT_UTF
1547          if (utf) FORWARDCHAR(ptr);
1548    #endif
1549          }
1550        if (*ptr == 0) goto FAIL_EXIT;
1551      continue;      continue;
1552      }      }
1553    
1554    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1555    
1556    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1557      {      {
1558      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1559      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1560      continue;      if (*ptr == 0) goto FAIL_EXIT;
1561      }      }
1562    
1563    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1564    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1565        if (dup_parens && *count < hwm_count) *count = hwm_count;
1566        goto FAIL_EXIT;
1567        }
1568    
1569    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1570        {
1571        if (*count > hwm_count) hwm_count = *count;
1572        *count = start_count;
1573        }
1574      }
1575    
1576    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1577         *ptr != '\'')  *ptrptr = ptr;
1578      continue;  return -1;
1579    }
1580    
1581    
1582    
1583    
1584    /*************************************************
1585    *       Find forward referenced subpattern       *
1586    *************************************************/
1587    
1588    /* This function scans along a pattern's text looking for capturing
1589    subpatterns, and counting them. If it finds a named pattern that matches the
1590    name it is given, it returns its number. Alternatively, if the name is NULL, it
1591    returns when it reaches a given numbered subpattern. This is used for forward
1592    references to subpatterns. We used to be able to start this scan from the
1593    current compiling point, using the current count value from cd->bracount, and
1594    do it all in a single loop, but the addition of the possibility of duplicate
1595    subpattern numbers means that we have to scan from the very start, in order to
1596    take account of such duplicates, and to use a recursive function to keep track
1597    of the different types of group.
1598    
1599    Arguments:
1600      cd           compile background data
1601      name         name to seek, or NULL if seeking a numbered subpattern
1602      lorn         name length, or subpattern number if name is NULL
1603      xmode        TRUE if we are in /x mode
1604      utf          TRUE if we are in UTF-8 / UTF-16 mode
1605    
1606    count++;  Returns:       the number of the found subpattern, or -1 if not found
1607    */
1608    
1609    static int
1610    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1611      BOOL utf)
1612    {
1613    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1614    int count = 0;
1615    int rc;
1616    
1617    /* If the pattern does not start with an opening parenthesis, the first call
1618    to find_parens_sub() will scan right to the end (if necessary). However, if it
1619    does start with a parenthesis, find_parens_sub() will return when it hits the
1620    matching closing parens. That is why we have to have a loop. */
1621    
1622    if (name == NULL && count == lorn) return count;  for (;;)
1623    term = *ptr++;    {
1624    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1625    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1626    }    }
1627    
1628  return -1;  return rc;
1629  }  }
1630    
1631    
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *      Find first significant op code            *  *      Find first significant op code            *
1636  *************************************************/  *************************************************/
1637    
1638  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1639  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1640  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1641  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1642  assertions, and also the \b assertion; for others it does not.  does not.
1643    
1644  Arguments:  Arguments:
1645    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1646    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1647    
1648  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1649  */  */
1650    
1651  static const uschar*  static const pcre_uchar*
1652  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1653  {  {
1654  for (;;)  for (;;)
1655    {    {
1656    switch ((int)*code)    switch ((int)*code)
1657      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1658      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1659      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1660      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1661      if (!skipassert) return code;      if (!skipassert) return code;
1662      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1663      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1664      break;      break;
1665    
1666      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1149  for (;;) Line 1670  for (;;)
1670    
1671      case OP_CALLOUT:      case OP_CALLOUT:
1672      case OP_CREF:      case OP_CREF:
1673        case OP_NCREF:
1674      case OP_RREF:      case OP_RREF:
1675        case OP_NRREF:
1676      case OP_DEF:      case OP_DEF:
1677      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1678      break;      break;
1679    
1680      default:      default:
# Line 1165  for (;;) Line 1688  for (;;)
1688    
1689    
1690  /*************************************************  /*************************************************
1691  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1692  *************************************************/  *************************************************/
1693    
1694  /* 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,
1695  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.
1696  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
1697    temporarily terminated with OP_END when this function is called.
1698    
1699    This function is called when a backward assertion is encountered, so that if it
1700    fails, the error message can point to the correct place in the pattern.
1701    However, we cannot do this when the assertion contains subroutine calls,
1702    because they can be forward references. We solve this by remembering this case
1703    and doing the check at the end; a flag specifies which mode we are running in.
1704    
1705  Arguments:  Arguments:
1706    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1707    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1708      atend    TRUE if called when the pattern is complete
1709  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1710               or -2 if \C was encountered  
1711    Returns:   the fixed length,
1712                 or -1 if there is no fixed length,
1713                 or -2 if \C was encountered (in UTF-8 mode only)
1714                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1715                 or -4 if an unknown opcode was encountered (internal error)
1716  */  */
1717    
1718  static int  static int
1719  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1720  {  {
1721  int length = -1;  int length = -1;
1722    
1723  register int branchlength = 0;  register int branchlength = 0;
1724  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1725    
1726  /* 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
1727  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1194  branch, check the length against that of Line 1729  branch, check the length against that of
1729  for (;;)  for (;;)
1730    {    {
1731    int d;    int d;
1732      pcre_uchar *ce, *cs;
1733    register int op = *cc;    register int op = *cc;
1734    
1735    switch (op)    switch (op)
1736      {      {
1737        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1738        OP_BRA (normal non-capturing bracket) because the other variants of these
1739        opcodes are all concerned with unlimited repeated groups, which of course
1740        are not of fixed length. */
1741    
1742      case OP_CBRA:      case OP_CBRA:
1743      case OP_BRA:      case OP_BRA:
1744      case OP_ONCE:      case OP_ONCE:
1745        case OP_ONCE_NC:
1746      case OP_COND:      case OP_COND:
1747      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1748      if (d < 0) return d;      if (d < 0) return d;
1749      branchlength += d;      branchlength += d;
1750      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1751      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1752      break;      break;
1753    
1754      /* 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.
1755      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
1756      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
1757        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1758        because they all imply an unlimited repeat. */
1759    
1760      case OP_ALT:      case OP_ALT:
1761      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1762      case OP_END:      case OP_END:
1763        case OP_ACCEPT:
1764        case OP_ASSERT_ACCEPT:
1765      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1766        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1767      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1224  for (;;) Line 1769  for (;;)
1769      branchlength = 0;      branchlength = 0;
1770      break;      break;
1771    
1772        /* A true recursion implies not fixed length, but a subroutine call may
1773        be OK. If the subroutine is a forward reference, we can't deal with
1774        it until the end of the pattern, so return -3. */
1775    
1776        case OP_RECURSE:
1777        if (!atend) return -3;
1778        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1779        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1780        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1781        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1782        if (d < 0) return d;
1783        branchlength += d;
1784        cc += 1 + LINK_SIZE;
1785        break;
1786    
1787      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1788    
1789      case OP_ASSERT:      case OP_ASSERT:
# Line 1231  for (;;) Line 1791  for (;;)
1791      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1792      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1793      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1794      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1795        break;
1796    
1797      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1798    
1799      case OP_REVERSE:      case OP_MARK:
1800        case OP_PRUNE_ARG:
1801        case OP_SKIP_ARG:
1802        case OP_THEN_ARG:
1803        cc += cc[1] + PRIV(OP_lengths)[*cc];
1804        break;
1805    
1806        case OP_CALLOUT:
1807        case OP_CIRC:
1808        case OP_CIRCM:
1809        case OP_CLOSE:
1810        case OP_COMMIT:
1811      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1812      case OP_DEF:      case OP_DEF:
1813      case OP_OPT:      case OP_DOLL:
1814      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1815      case OP_EOD:      case OP_EOD:
1816      case OP_EODN:      case OP_EODN:
1817      case OP_CIRC:      case OP_FAIL:
1818      case OP_DOLL:      case OP_NCREF:
1819        case OP_NRREF:
1820      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1821        case OP_PRUNE:
1822        case OP_REVERSE:
1823        case OP_RREF:
1824        case OP_SET_SOM:
1825        case OP_SKIP:
1826        case OP_SOD:
1827        case OP_SOM:
1828        case OP_THEN:
1829      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1830      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1831      break;      break;
1832    
1833      /* Handle literal characters */      /* Handle literal characters */
1834    
1835      case OP_CHAR:      case OP_CHAR:
1836      case OP_CHARNC:      case OP_CHARI:
1837      case OP_NOT:      case OP_NOT:
1838        case OP_NOTI:
1839      branchlength++;      branchlength++;
1840      cc += 2;      cc += 2;
1841  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1842      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1843  #endif  #endif
1844      break;      break;
1845    
# Line 1271  for (;;) Line 1847  for (;;)
1847      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1848    
1849      case OP_EXACT:      case OP_EXACT:
1850        case OP_EXACTI:
1851        case OP_NOTEXACT:
1852        case OP_NOTEXACTI:
1853      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1854      cc += 4;      cc += 2 + IMM2_SIZE;
1855  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1856      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1857  #endif  #endif
1858      break;      break;
1859    
1860      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1861      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1862      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1863      cc += 4;        cc += 2;
1864        cc += 1 + IMM2_SIZE + 1;
1865      break;      break;
1866    
1867      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1294  for (;;) Line 1871  for (;;)
1871      cc += 2;      cc += 2;
1872      /* Fall through */      /* Fall through */
1873    
1874        case OP_HSPACE:
1875        case OP_VSPACE:
1876        case OP_NOT_HSPACE:
1877        case OP_NOT_VSPACE:
1878      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1879      case OP_DIGIT:      case OP_DIGIT:
1880      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1306  for (;;) Line 1887  for (;;)
1887      cc++;      cc++;
1888      break;      break;
1889    
1890      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1891        otherwise \C is coded as OP_ALLANY. */
1892    
1893      case OP_ANYBYTE:      case OP_ANYBYTE:
1894      return -2;      return -2;
1895    
1896      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1897    
1898  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1899      case OP_XCLASS:      case OP_XCLASS:
1900      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1901      /* Fall through */      /* Fall through */
1902  #endif  #endif
1903    
1904      case OP_CLASS:      case OP_CLASS:
1905      case OP_NCLASS:      case OP_NCLASS:
1906      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1907    
1908      switch (*cc)      switch (*cc)
1909        {        {
1910          case OP_CRPLUS:
1911          case OP_CRMINPLUS:
1912        case OP_CRSTAR:        case OP_CRSTAR:
1913        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1914        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1333  for (;;) Line 1917  for (;;)
1917    
1918        case OP_CRRANGE:        case OP_CRRANGE:
1919        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1920        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1921        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1922        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1923        break;        break;
1924    
1925        default:        default:
# Line 1345  for (;;) Line 1929  for (;;)
1929    
1930      /* Anything else is variable length */      /* Anything else is variable length */
1931    
1932      default:      case OP_ANYNL:
1933        case OP_BRAMINZERO:
1934        case OP_BRAPOS:
1935        case OP_BRAPOSZERO:
1936        case OP_BRAZERO:
1937        case OP_CBRAPOS:
1938        case OP_EXTUNI:
1939        case OP_KETRMAX:
1940        case OP_KETRMIN:
1941        case OP_KETRPOS:
1942        case OP_MINPLUS:
1943        case OP_MINPLUSI:
1944        case OP_MINQUERY:
1945        case OP_MINQUERYI:
1946        case OP_MINSTAR:
1947        case OP_MINSTARI:
1948        case OP_MINUPTO:
1949        case OP_MINUPTOI:
1950        case OP_NOTMINPLUS:
1951        case OP_NOTMINPLUSI:
1952        case OP_NOTMINQUERY:
1953        case OP_NOTMINQUERYI:
1954        case OP_NOTMINSTAR:
1955        case OP_NOTMINSTARI:
1956        case OP_NOTMINUPTO:
1957        case OP_NOTMINUPTOI:
1958        case OP_NOTPLUS:
1959        case OP_NOTPLUSI:
1960        case OP_NOTPOSPLUS:
1961        case OP_NOTPOSPLUSI:
1962        case OP_NOTPOSQUERY:
1963        case OP_NOTPOSQUERYI:
1964        case OP_NOTPOSSTAR:
1965        case OP_NOTPOSSTARI:
1966        case OP_NOTPOSUPTO:
1967        case OP_NOTPOSUPTOI:
1968        case OP_NOTQUERY:
1969        case OP_NOTQUERYI:
1970        case OP_NOTSTAR:
1971        case OP_NOTSTARI:
1972        case OP_NOTUPTO:
1973        case OP_NOTUPTOI:
1974        case OP_PLUS:
1975        case OP_PLUSI:
1976        case OP_POSPLUS:
1977        case OP_POSPLUSI:
1978        case OP_POSQUERY:
1979        case OP_POSQUERYI:
1980        case OP_POSSTAR:
1981        case OP_POSSTARI:
1982        case OP_POSUPTO:
1983        case OP_POSUPTOI:
1984        case OP_QUERY:
1985        case OP_QUERYI:
1986        case OP_REF:
1987        case OP_REFI:
1988        case OP_SBRA:
1989        case OP_SBRAPOS:
1990        case OP_SCBRA:
1991        case OP_SCBRAPOS:
1992        case OP_SCOND:
1993        case OP_SKIPZERO:
1994        case OP_STAR:
1995        case OP_STARI:
1996        case OP_TYPEMINPLUS:
1997        case OP_TYPEMINQUERY:
1998        case OP_TYPEMINSTAR:
1999        case OP_TYPEMINUPTO:
2000        case OP_TYPEPLUS:
2001        case OP_TYPEPOSPLUS:
2002        case OP_TYPEPOSQUERY:
2003        case OP_TYPEPOSSTAR:
2004        case OP_TYPEPOSUPTO:
2005        case OP_TYPEQUERY:
2006        case OP_TYPESTAR:
2007        case OP_TYPEUPTO:
2008        case OP_UPTO:
2009        case OP_UPTOI:
2010      return -1;      return -1;
2011    
2012        /* Catch unrecognized opcodes so that when new ones are added they
2013        are not forgotten, as has happened in the past. */
2014    
2015        default:
2016        return -4;
2017      }      }
2018    }    }
2019  /* Control never gets here */  /* Control never gets here */
# Line 1356  for (;;) Line 2023  for (;;)
2023    
2024    
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
2028    
2029  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2030  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2031    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032    so that it can be called from pcre_study() when finding the minimum matching
2033    length.
2034    
2035  Arguments:  Arguments:
2036    code        points to start of expression    code        points to start of expression
2037    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2038    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2039    
2040  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
2041  */  */
2042    
2043  static const uschar *  const pcre_uchar *
2044  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045  {  {
2046  for (;;)  for (;;)
2047    {    {
2048    register int c = *code;    register int c = *code;
2049    
2050    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2051    
2052    /* 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 1384  for (;;) Line 2055  for (;;)
2055    
2056    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2057    
2058      /* Handle recursion */
2059    
2060      else if (c == OP_REVERSE)
2061        {
2062        if (number < 0) return (pcre_uchar *)code;
2063        code += PRIV(OP_lengths)[c];
2064        }
2065    
2066    /* Handle capturing bracket */    /* Handle capturing bracket */
2067    
2068    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2069               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070      {      {
2071      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2072      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2073      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2074      }      }
2075    
2076    /* 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
2077    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
2078    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079      must add in its length. */
2080    
2081    else    else
2082      {      {
# Line 1417  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102            code += 2;
2103          break;
2104    
2105          case OP_MARK:
2106          case OP_PRUNE_ARG:
2107          case OP_SKIP_ARG:
2108          code += code[1];
2109          break;
2110    
2111          case OP_THEN_ARG:
2112          code += code[1];
2113        break;        break;
2114        }        }
2115    
2116      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2117    
2118      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2119    
2120    /* 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
2121    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
2122    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2123    
2124  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2125      if (utf8) switch(c)      if (utf) switch(c)
2126        {        {
2127        case OP_CHAR:        case OP_CHAR:
2128        case OP_CHARNC:        case OP_CHARI:
2129        case OP_EXACT:        case OP_EXACT:
2130          case OP_EXACTI:
2131        case OP_UPTO:        case OP_UPTO:
2132          case OP_UPTOI:
2133        case OP_MINUPTO:        case OP_MINUPTO:
2134          case OP_MINUPTOI:
2135        case OP_POSUPTO:        case OP_POSUPTO:
2136          case OP_POSUPTOI:
2137        case OP_STAR:        case OP_STAR:
2138          case OP_STARI:
2139        case OP_MINSTAR:        case OP_MINSTAR:
2140          case OP_MINSTARI:
2141        case OP_POSSTAR:        case OP_POSSTAR:
2142          case OP_POSSTARI:
2143        case OP_PLUS:        case OP_PLUS:
2144          case OP_PLUSI:
2145        case OP_MINPLUS:        case OP_MINPLUS:
2146          case OP_MINPLUSI:
2147        case OP_POSPLUS:        case OP_POSPLUS:
2148          case OP_POSPLUSI:
2149        case OP_QUERY:        case OP_QUERY:
2150          case OP_QUERYI:
2151        case OP_MINQUERY:        case OP_MINQUERY:
2152          case OP_MINQUERYI:
2153        case OP_POSQUERY:        case OP_POSQUERY:
2154        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2155          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2156        break;        break;
2157        }        }
2158    #else
2159        (void)(utf);  /* Keep compiler happy by referencing function argument */
2160  #endif  #endif
2161      }      }
2162    }    }
# Line 1466  instance of OP_RECURSE. Line 2173  instance of OP_RECURSE.
2173    
2174  Arguments:  Arguments:
2175    code        points to start of expression    code        points to start of expression
2176    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2177    
2178  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
2179  */  */
2180    
2181  static const uschar *  static const pcre_uchar *
2182  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2183  {  {
2184  for (;;)  for (;;)
2185    {    {
# Line 1488  for (;;) Line 2195  for (;;)
2195    
2196    /* 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
2197    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
2198    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2199      must add in its length. */
2200    
2201    else    else
2202      {      {
# Line 1510  for (;;) Line 2218  for (;;)
2218        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2219        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2220        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2221        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2222            code += 2;
2223          break;
2224    
2225          case OP_MARK:
2226          case OP_PRUNE_ARG:
2227          case OP_SKIP_ARG:
2228          code += code[1];
2229          break;
2230    
2231          case OP_THEN_ARG:
2232          code += code[1];
2233        break;        break;
2234        }        }
2235    
2236      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2237    
2238      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2239    
2240      /* 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
2241      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
2242      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2243    
2244  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2245      if (utf8) switch(c)      if (utf) switch(c)
2246        {        {
2247        case OP_CHAR:        case OP_CHAR:
2248        case OP_CHARNC:        case OP_CHARI:
2249          case OP_NOT:
2250          case OP_NOTI:
2251        case OP_EXACT:        case OP_EXACT:
2252          case OP_EXACTI:
2253          case OP_NOTEXACT:
2254          case OP_NOTEXACTI:
2255        case OP_UPTO:        case OP_UPTO:
2256          case OP_UPTOI:
2257          case OP_NOTUPTO:
2258          case OP_NOTUPTOI:
2259        case OP_MINUPTO:        case OP_MINUPTO:
2260          case OP_MINUPTOI:
2261          case OP_NOTMINUPTO:
2262          case OP_NOTMINUPTOI:
2263        case OP_POSUPTO:        case OP_POSUPTO:
2264          case OP_POSUPTOI:
2265          case OP_NOTPOSUPTO:
2266          case OP_NOTPOSUPTOI:
2267        case OP_STAR:        case OP_STAR:
2268          case OP_STARI:
2269          case OP_NOTSTAR:
2270          case OP_NOTSTARI:
2271        case OP_MINSTAR:        case OP_MINSTAR:
2272          case OP_MINSTARI:
2273          case OP_NOTMINSTAR:
2274          case OP_NOTMINSTARI:
2275        case OP_POSSTAR:        case OP_POSSTAR:
2276          case OP_POSSTARI:
2277          case OP_NOTPOSSTAR:
2278          case OP_NOTPOSSTARI:
2279        case OP_PLUS:        case OP_PLUS:
2280          case OP_PLUSI:
2281          case OP_NOTPLUS:
2282          case OP_NOTPLUSI:
2283        case OP_MINPLUS:        case OP_MINPLUS:
2284          case OP_MINPLUSI:
2285          case OP_NOTMINPLUS:
2286          case OP_NOTMINPLUSI:
2287        case OP_POSPLUS:        case OP_POSPLUS:
2288          case OP_POSPLUSI:
2289          case OP_NOTPOSPLUS:
2290          case OP_NOTPOSPLUSI:
2291        case OP_QUERY:        case OP_QUERY:
2292          case OP_QUERYI:
2293          case OP_NOTQUERY:
2294          case OP_NOTQUERYI:
2295        case OP_MINQUERY:        case OP_MINQUERY:
2296          case OP_MINQUERYI:
2297          case OP_NOTMINQUERY:
2298          case OP_NOTMINQUERYI:
2299        case OP_POSQUERY:        case OP_POSQUERY:
2300        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2301          case OP_NOTPOSQUERY:
2302          case OP_NOTPOSQUERYI:
2303          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2304        break;        break;
2305        }        }
2306    #else
2307        (void)(utf);  /* Keep compiler happy by referencing function argument */
2308  #endif  #endif
2309      }      }
2310    }    }
# Line 1565  bracket whose current branch will alread Line 2327  bracket whose current branch will alread
2327  Arguments:  Arguments:
2328    code        points to start of search    code        points to start of search
2329    endcode     points to where to stop    endcode     points to where to stop
2330    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2331      cd          contains pointers to tables etc.
2332    
2333  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2334  */  */
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338      BOOL utf, compile_data *cd)
2339  {  {
2340  register int c;  register int c;
2341  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2342       code < endcode;       code < endcode;
2343       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2344    {    {
2345    const uschar *ccode;    const pcre_uchar *ccode;
2346    
2347    c = *code;    c = *code;
2348    
# Line 1592  for (code = first_significant_code(code Line 2356  for (code = first_significant_code(code
2356      continue;      continue;
2357      }      }
2358    
2359      /* For a recursion/subroutine call, if its end has been reached, which
2360      implies a backward reference subroutine call, we can scan it. If it's a
2361      forward reference subroutine call, we can't. To detect forward reference
2362      we have to scan up the list that is kept in the workspace. This function is
2363      called only when doing the real compile, not during the pre-compile that
2364      measures the size of the compiled pattern. */
2365    
2366      if (c == OP_RECURSE)
2367        {
2368        const pcre_uchar *scode;
2369        BOOL empty_branch;
2370    
2371        /* Test for forward reference */
2372    
2373        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2374          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2375    
2376        /* Not a forward reference, test for completed backward reference */
2377    
2378        empty_branch = FALSE;
2379        scode = cd->start_code + GET(code, 1);
2380        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2381    
2382        /* Completed backwards reference */
2383    
2384        do
2385          {
2386          if (could_be_empty_branch(scode, endcode, utf, cd))
2387            {
2388            empty_branch = TRUE;
2389            break;
2390            }
2391          scode += GET(scode, 1);
2392          }
2393        while (*scode == OP_ALT);
2394    
2395        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2396        continue;
2397        }
2398    
2399    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2400    
2401    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2402          c == OP_BRAPOSZERO)
2403        {
2404        code += PRIV(OP_lengths)[c];
2405        do code += GET(code, 1); while (*code == OP_ALT);
2406        c = *code;
2407        continue;
2408        }
2409    
2410      /* A nested group that is already marked as "could be empty" can just be
2411      skipped. */
2412    
2413      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2414          c == OP_SCBRA || c == OP_SCBRAPOS)
2415      {      {
     code += _pcre_OP_lengths[c];  
2416      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2417      c = *code;      c = *code;
2418      continue;      continue;
# Line 1604  for (code = first_significant_code(code Line 2420  for (code = first_significant_code(code
2420    
2421    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2422    
2423    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2424          c == OP_CBRA || c == OP_CBRAPOS ||
2425          c == OP_ONCE || c == OP_ONCE_NC ||
2426          c == OP_COND)
2427      {      {
2428      BOOL empty_branch;      BOOL empty_branch;
2429      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2430    
2431      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2432        empty branch, so just skip over the conditional, because it could be empty.
2433        Otherwise, scan the individual branches of the group. */
2434    
2435      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;  
2436        code += GET(code, 1);        code += GET(code, 1);
2437        else
2438          {
2439          empty_branch = FALSE;
2440          do
2441            {
2442            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2443              empty_branch = TRUE;
2444            code += GET(code, 1);
2445            }
2446          while (*code == OP_ALT);
2447          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2448        }        }
2449      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2450      c = *code;      c = *code;
2451      continue;      continue;
2452      }      }
# Line 1630  for (code = first_significant_code(code Line 2457  for (code = first_significant_code(code
2457      {      {
2458      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2459      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2460      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2461      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"
2462      here. */      here. */
2463    
2464  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2465      case OP_XCLASS:      case OP_XCLASS:
2466      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2467      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1642  for (code = first_significant_code(code Line 2469  for (code = first_significant_code(code
2469    
2470      case OP_CLASS:      case OP_CLASS:
2471      case OP_NCLASS:      case OP_NCLASS:
2472      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2473    
2474  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2475      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2476  #endif  #endif
2477    
# Line 1683  for (code = first_significant_code(code Line 2510  for (code = first_significant_code(code
2510      case OP_ALLANY:      case OP_ALLANY:
2511      case OP_ANYBYTE:      case OP_ANYBYTE:
2512      case OP_CHAR:      case OP_CHAR:
2513      case OP_CHARNC:      case OP_CHARI:
2514      case OP_NOT:      case OP_NOT:
2515        case OP_NOTI:
2516      case OP_PLUS:      case OP_PLUS:
2517      case OP_MINPLUS:      case OP_MINPLUS:
2518      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1716  for (code = first_significant_code(code Line 2544  for (code = first_significant_code(code
2544      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2545      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2546      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2547      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2548          code += 2;
2549      break;      break;
2550    
2551      /* End of branch */      /* End of branch */
# Line 1724  for (code = first_significant_code(code Line 2553  for (code = first_significant_code(code
2553      case OP_KET:      case OP_KET:
2554      case OP_KETRMAX:      case OP_KETRMAX:
2555      case OP_KETRMIN:      case OP_KETRMIN:
2556        case OP_KETRPOS:
2557      case OP_ALT:      case OP_ALT:
2558      return TRUE;      return TRUE;
2559    
2560      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2561      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2562    
2563  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2564      case OP_STAR:      case OP_STAR:
2565        case OP_STARI:
2566      case OP_MINSTAR:      case OP_MINSTAR:
2567        case OP_MINSTARI:
2568      case OP_POSSTAR:      case OP_POSSTAR:
2569        case OP_POSSTARI:
2570      case OP_QUERY:      case OP_QUERY:
2571        case OP_QUERYI:
2572      case OP_MINQUERY:      case OP_MINQUERY:
2573        case OP_MINQUERYI:
2574      case OP_POSQUERY:      case OP_POSQUERY:
2575        case OP_POSQUERYI:
2576        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2577        break;
2578    
2579      case OP_UPTO:      case OP_UPTO:
2580        case OP_UPTOI:
2581      case OP_MINUPTO:      case OP_MINUPTO:
2582        case OP_MINUPTOI:
2583      case OP_POSUPTO:      case OP_POSUPTO:
2584      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2585        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2586      break;      break;
2587  #endif  #endif
2588    
2589        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2590        string. */
2591    
2592        case OP_MARK:
2593        case OP_PRUNE_ARG:
2594        case OP_SKIP_ARG:
2595        code += code[1];
2596        break;
2597    
2598        case OP_THEN_ARG:
2599        code += code[1];
2600        break;
2601    
2602        /* None of the remaining opcodes are required to match a character. */
2603    
2604        default:
2605        break;
2606      }      }
2607    }    }
2608    
# Line 1759  return TRUE; Line 2619  return TRUE;
2619  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
2620  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,
2621  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.
2622    This function is called only during the real compile, not during the
2623    pre-compile.
2624    
2625  Arguments:  Arguments:
2626    code        points to start of the recursion    code        points to start of the recursion
2627    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2628    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2629    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2630      cd          pointers to tables etc
2631    
2632  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2633  */  */
2634    
2635  static BOOL  static BOOL
2636  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2637    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2638  {  {
2639  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2640    {    {
2641    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2642        return FALSE;
2643    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2644    }    }
2645  return TRUE;  return TRUE;
# Line 1807  where Perl recognizes it as the POSIX cl Line 2671  where Perl recognizes it as the POSIX cl
2671  "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,
2672  I think.  I think.
2673    
2674    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2675    It seems that the appearance of a nested POSIX class supersedes an apparent
2676    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2677    a digit.
2678    
2679    In Perl, unescaped square brackets may also appear as part of class names. For
2680    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2681    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2682    seem right at all. PCRE does not allow closing square brackets in POSIX class
2683    names.
2684    
2685  Arguments:  Arguments:
2686    ptr      pointer to the initial [    ptr      pointer to the initial [
2687    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1815  Returns:   TRUE or FALSE Line 2690  Returns:   TRUE or FALSE
2690  */  */
2691    
2692  static BOOL  static BOOL
2693  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2694  {  {
2695  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2696  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2697  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2698    {    {
2699    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2700        ptr++;
2701      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2702      else
2703      {      {
2704      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2705        {        {
2706        *endptr = ptr;        *endptr = ptr;
2707        return TRUE;        return TRUE;
2708        }        }
2709        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2710             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2711              ptr[1] == CHAR_EQUALS_SIGN) &&
2712            check_posix_syntax(ptr, endptr))
2713          return FALSE;
2714      }      }
2715    }    }
2716  return FALSE;  return FALSE;
# Line 1852  Returns:     a value representing the na Line 2734  Returns:     a value representing the na
2734  */  */
2735    
2736  static int  static int
2737  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2738  {  {
2739  const char *pn = posix_names;  const char *pn = posix_names;
2740  register int yield = 0;  register int yield = 0;
2741  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2742    {    {
2743    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2744      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2745    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2746    yield++;    yield++;
2747    }    }
# Line 1891  value in the reference (which is a group Line 2773  value in the reference (which is a group
2773  Arguments:  Arguments:
2774    group      points to the start of the group    group      points to the start of the group
2775    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2776    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2777    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2778    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
2779    
# Line 1899  Returns:     nothing Line 2781  Returns:     nothing
2781  */  */
2782    
2783  static void  static void
2784  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2785    uschar *save_hwm)    pcre_uchar *save_hwm)
2786  {  {
2787  uschar *ptr = group;  pcre_uchar *ptr = group;
2788    
2789  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2790    {    {
2791    int offset;    int offset;
2792    uschar *hc;    pcre_uchar *hc;
2793    
2794    /* 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
2795    reference. */    reference. */
# Line 1952  Arguments: Line 2834  Arguments:
2834  Returns:         new code pointer  Returns:         new code pointer
2835  */  */
2836    
2837  static uschar *  static pcre_uchar *
2838  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2839  {  {
2840  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2841  *code++ = 255;  *code++ = 255;
2842  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2843  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2844  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2845  }  }
2846    
2847    
# Line 1981  Returns:             nothing Line 2863  Returns:             nothing
2863  */  */
2864    
2865  static void  static void
2866  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2867  {  {
2868  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2869  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2870  }  }
2871    
# Line 1995  PUT(previous_callout, 2 + LINK_SIZE, len Line 2877  PUT(previous_callout, 2 + LINK_SIZE, len
2877  *************************************************/  *************************************************/
2878    
2879  /* 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
2880  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
2881  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
2882  start address.  start address. A character with multiple other cases is returned on its own
2883    with a special return value.
2884    
2885  Arguments:  Arguments:
2886    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2005  Arguments: Line 2888  Arguments:
2888    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2889    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2890    
2891  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2892                   0 when a range is returned
2893                  >0 the CASESET offset for char with multiple other cases
2894                    in this case, ocptr contains the original
2895  */  */
2896    
2897  static BOOL  static int
2898  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2899    unsigned int *odptr)    unsigned int *odptr)
2900  {  {
2901  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2902    int co;
2903    
2904    /* Find the first character that has an other case. If it has multiple other
2905    cases, return its case offset value. */
2906    
2907  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2908    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
2909      if ((co = UCD_CASESET(c)) != 0)
2910        {
2911        *ocptr = c++;   /* Character that has the set */
2912        *cptr = c;      /* Rest of input range */
2913        return co;
2914        }
2915      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2916      }
2917    
2918  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2919    
2920  *ocptr = othercase;  *ocptr = othercase;
2921  next = othercase + 1;  next = othercase + 1;
# Line 2028  for (++c; c <= d; c++) Line 2926  for (++c; c <= d; c++)
2926    next++;    next++;
2927    }    }
2928    
2929  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2930  *cptr = c;  *cptr = c;             /* Rest of input range */
2931    return 0;
2932    }
2933    
2934  return TRUE;  
2935    
2936    /*************************************************
2937    *        Check a character and a property        *
2938    *************************************************/
2939    
2940    /* This function is called by check_auto_possessive() when a property item
2941    is adjacent to a fixed character.
2942    
2943    Arguments:
2944      c            the character
2945      ptype        the property type
2946      pdata        the data for the type
2947      negated      TRUE if it's a negated property (\P or \p{^)
2948    
2949    Returns:       TRUE if auto-possessifying is OK
2950    */
2951    
2952    static BOOL
2953    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2954    {
2955    #ifdef SUPPORT_UCP
2956    const pcre_uint32 *p;
2957    #endif
2958    
2959    const ucd_record *prop = GET_UCD(c);
2960    
2961    switch(ptype)
2962      {
2963      case PT_LAMP:
2964      return (prop->chartype == ucp_Lu ||
2965              prop->chartype == ucp_Ll ||
2966              prop->chartype == ucp_Lt) == negated;
2967    
2968      case PT_GC:
2969      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2970    
2971      case PT_PC:
2972      return (pdata == prop->chartype) == negated;
2973    
2974      case PT_SC:
2975      return (pdata == prop->script) == negated;
2976    
2977      /* These are specials */
2978    
2979      case PT_ALNUM:
2980      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2981              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2982    
2983      case PT_SPACE:    /* Perl space */
2984      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2985              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2986              == negated;
2987    
2988      case PT_PXSPACE:  /* POSIX space */
2989      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2990              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2991              c == CHAR_FF || c == CHAR_CR)
2992              == negated;
2993    
2994      case PT_WORD:
2995      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2996              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2997              c == CHAR_UNDERSCORE) == negated;
2998    
2999    #ifdef SUPPORT_UCP
3000      case PT_CLIST:
3001      p = PRIV(ucd_caseless_sets) + prop->caseset;
3002      for (;;)
3003        {
3004        if ((unsigned int)c < *p) return !negated;
3005        if ((unsigned int)c == *p++) return negated;
3006        }
3007      break;  /* Control never reaches here */
3008    #endif
3009      }
3010    
3011    return FALSE;
3012  }  }
3013  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3014    
# Line 2046  whether the next thing could possibly ma Line 3023  whether the next thing could possibly ma
3023  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
3024    
3025  Arguments:  Arguments:
3026    op_code       the repeated op code    previous      pointer to the repeated opcode
3027    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
3028    ptr           next character in pattern    ptr           next character in pattern
3029    options       options bits    options       options bits
3030    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2058  Returns:        TRUE if possessifying is Line 3033  Returns:        TRUE if possessifying is
3033  */  */
3034    
3035  static BOOL  static BOOL
3036  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3037    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3038  {  {
3039  int next;  pcre_int32 c = NOTACHAR;
3040    pcre_int32 next;
3041    int op_code = *previous++;
3042    
3043  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
3044    
# Line 2069  if ((options & PCRE_EXTENDED) != 0) Line 3046  if ((options & PCRE_EXTENDED) != 0)
3046    {    {
3047    for (;;)    for (;;)
3048      {      {
3049      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3050      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3051        {        {
3052        while (*(++ptr) != 0)        ptr++;
3053          while (*ptr != 0)
3054            {
3055          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3056            ptr++;
3057    #ifdef SUPPORT_UTF
3058            if (utf) FORWARDCHAR(ptr);
3059    #endif
3060            }
3061        }        }
3062      else break;      else break;
3063      }      }
# Line 2082  if ((options & PCRE_EXTENDED) != 0) Line 3066  if ((options & PCRE_EXTENDED) != 0)
3066  /* 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
3067  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
3068    
3069  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
3070    {    {
3071    int temperrorcode = 0;    int temperrorcode = 0;
3072    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
3073    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3074    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3075    }    }
3076    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3077    {    {
3078  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3079    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
3080  #endif  #endif
3081    next = *ptr++;    next = *ptr++;
3082    }    }
   
3083  else return FALSE;  else return FALSE;
3084    
3085  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2106  if ((options & PCRE_EXTENDED) != 0) Line 3088  if ((options & PCRE_EXTENDED) != 0)
3088    {    {
3089    for (;;)    for (;;)
3090      {      {
3091      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3092      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3093        {        {
3094        while (*(++ptr) != 0)        ptr++;
3095          while (*ptr != 0)
3096            {
3097          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3098            ptr++;
3099    #ifdef SUPPORT_UTF
3100            if (utf) FORWARDCHAR(ptr);
3101    #endif
3102            }
3103        }        }
3104      else break;      else break;
3105      }      }
# Line 2118  if ((options & PCRE_EXTENDED) != 0) Line 3107  if ((options & PCRE_EXTENDED) != 0)
3107    
3108  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3109    
3110  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3111    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3112        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. */  
   
3113    
3114  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
3115    
3116  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3117        op_code == OP_NOT || op_code == OP_NOTI)
3118    {    {
3119    case OP_CHAR:  #ifdef SUPPORT_UTF
3120  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
3121    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
3122      c = *previous;
3123  #endif  #endif
3124    return item != next;    }
3125    
3126    /* 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
3127    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3128    
3129    case OP_CHARNC:  if (next >= 0)
3130  #ifdef SUPPORT_UTF8    {
3131    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
3132      case, which maps to the special PT_CLIST property. Check this first. */
3133    
3134    #ifdef SUPPORT_UCP
3135      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3136        {
3137        int ocs = UCD_CASESET(next);
3138        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3139        }
3140  #endif  #endif
3141    if (item == next) return FALSE;  
3142  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3143      {      {
3144      unsigned int othercase;      case OP_CHAR:
3145      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3146    
3147        /* For CHARI (caseless character) we must check the other case. If we have
3148        Unicode property support, we can use it to test the other case of
3149        high-valued characters. We know that next can have only one other case,
3150        because multi-other-case characters are dealt with above. */
3151    
3152        case OP_CHARI:
3153        if (c == next) return FALSE;
3154    #ifdef SUPPORT_UTF
3155        if (utf)
3156          {
3157          unsigned int othercase;
3158          if (next < 128) othercase = cd->fcc[next]; else
3159  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3160      othercase = UCD_OTHERCASE((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3161  #else  #else
3162      othercase = NOTACHAR;        othercase = NOTACHAR;
3163  #endif  #endif
3164      return (unsigned int)item != othercase;        return (unsigned int)c != othercase;
3165      }        }
3166    else      else
3167  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3168    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3169    
3170    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
3171        return c == next;
3172    case OP_NOT:  
3173    if (item == next) return TRUE;      case OP_NOTI:
3174    if ((options & PCRE_CASELESS) == 0) return FALSE;      if (c == next) return TRUE;
3175  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3176    if (utf8)      if (utf)
3177      {        {
3178      unsigned int othercase;        unsigned int othercase;
3179      if (next < 128) othercase = cd->fcc[next]; else        if (next < 128) othercase = cd->fcc[next]; else
3180  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3181      othercase = UCD_OTHERCASE(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3182  #else  #else
3183      othercase = NOTACHAR;        othercase = NOTACHAR;
3184  #endif  #endif
3185      return (unsigned int)item == othercase;        return (unsigned int)c == othercase;
3186      }        }
3187    else      else
3188  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3189    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3190    
3191    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3192    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3193    
3194    case OP_NOT_DIGIT:      case OP_DIGIT:
3195    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3196    
3197    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3198    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3199    
3200    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3201    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3202    
3203    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3204    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3205    
3206    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3207    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3208    
3209    case OP_HSPACE:      case OP_NOT_WORDCHAR:
3210    case OP_NOT_HSPACE:      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3211    switch(next)  
3212      {      case OP_HSPACE:
3213      case 0x09:      case OP_NOT_HSPACE:
3214      case 0x20:      switch(next)
3215      case 0xa0:        {
3216      case 0x1680:        HSPACE_CASES:
3217      case 0x180e:        return op_code == OP_NOT_HSPACE;
3218      case 0x2000:  
3219      case 0x2001:        default:
3220      case 0x2002:        return op_code != OP_NOT_HSPACE;
3221      case 0x2003:        }
3222      case 0x2004:  
3223      case 0x2005:      case OP_ANYNL:
3224      case 0x2006:      case OP_VSPACE:
3225      case 0x2007:      case OP_NOT_VSPACE:
3226      case 0x2008:      switch(next)
3227      case 0x2009:        {
3228      case 0x200A:        VSPACE_CASES:
3229      case 0x202f:        return op_code == OP_NOT_VSPACE;
3230      case 0x205f:  
3231      case 0x3000:        default:
3232      return op_code != OP_HSPACE;        return op_code != OP_NOT_VSPACE;
3233      default:        }
3234      return op_code == OP_HSPACE;  
3235      }  #ifdef SUPPORT_UCP
3236        case OP_PROP:
3237        return check_char_prop(next, previous[0], previous[1], FALSE);
3238    
3239        case OP_NOTPROP:
3240        return check_char_prop(next, previous[0], previous[1], TRUE);
3241    #endif
3242    
   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;  
3243      default:      default:
3244      return op_code == OP_VSPACE;      return FALSE;
3245      }      }
   
   default:  
   return FALSE;  
3246    }    }
3247    
3248    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3249  /* 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
3250    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3251    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3252    replaced by OP_PROP codes when PCRE_UCP is set. */
3253    
3254  switch(op_code)  switch(op_code)
3255    {    {
3256    case OP_CHAR:    case OP_CHAR:
3257    case OP_CHARNC:    case OP_CHARI:
 #ifdef SUPPORT_UTF8  
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
3258    switch(-next)    switch(-next)
3259      {      {
3260      case ESC_d:      case ESC_d:
3261      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3262    
3263      case ESC_D:      case ESC_D:
3264      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3265    
3266      case ESC_s:      case ESC_s:
3267      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3268    
3269      case ESC_S:      case ESC_S:
3270      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3271    
3272      case ESC_w:      case ESC_w:
3273      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3274    
3275      case ESC_W:      case ESC_W:
3276      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3277    
3278      case ESC_h:      case ESC_h:
3279      case ESC_H:      case ESC_H:
3280      switch(item)      switch(c)
3281        {        {
3282        case 0x09:        HSPACE_CASES:
       case 0x20:  
       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:  
3283        return -next != ESC_h;        return -next != ESC_h;
3284    
3285        default:        default:
3286        return -next == ESC_h;        return -next == ESC_h;
3287        }        }
3288    
3289      case ESC_v:      case ESC_v:
3290      case ESC_V:      case ESC_V:
3291      switch(item)      switch(c)
3292        {        {
3293        case 0x0a:        VSPACE_CASES:
       case 0x0b:  
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
3294        return -next != ESC_v;        return -next != ESC_v;
3295    
3296        default:        default:
3297        return -next == ESC_v;        return -next == ESC_v;
3298        }        }
3299    
3300        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3301        their substitutions and process them. The result will always be either
3302        -ESC_p or -ESC_P. Then fall through to process those values. */
3303    
3304    #ifdef SUPPORT_UCP
3305        case ESC_du:
3306        case ESC_DU:
3307        case ESC_wu:
3308        case ESC_WU:
3309        case ESC_su:
3310        case ESC_SU:
3311          {
3312          int temperrorcode = 0;
3313          ptr = substitutes[-next - ESC_DU];
3314          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3315          if (temperrorcode != 0) return FALSE;
3316          ptr++;    /* For compatibility */
3317          }
3318        /* Fall through */
3319    
3320        case ESC_p:
3321        case ESC_P:
3322          {
3323          int ptype, pdata, errorcodeptr;
3324          BOOL negated;
3325    
3326          ptr--;      /* Make ptr point at the p or P */
3327          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3328          if (ptype < 0) return FALSE;
3329          ptr++;      /* Point past the final curly ket */
3330    
3331          /* If the property item is optional, we have to give up. (When generated
3332          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3333          to the original \d etc. At this point, ptr will point to a zero byte. */
3334    
3335          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3336            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3337              return FALSE;
3338    
3339          /* Do the property check. */
3340    
3341          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3342          }
3343    #endif
3344    
3345      default:      default:
3346      return FALSE;      return FALSE;
3347      }      }
3348    
3349      /* In principle, support for Unicode properties should be integrated here as
3350      well. It means re-organizing the above code so as to get hold of the property
3351      values before switching on the op-code. However, I wonder how many patterns
3352      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3353      these op-codes are never generated.) */
3354    
3355    case OP_DIGIT:    case OP_DIGIT:
3356    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3357           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3358    
3359    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3360    return next == -ESC_d;    return next == -ESC_d;
# Line 2338  switch(op_code) Line 3363  switch(op_code)
3363    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
3364    
3365    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3366    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;
3367    
3368    case OP_HSPACE:    case OP_HSPACE:
3369    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3370             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3371    
3372    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3373    return next == -ESC_h;    return next == -ESC_h;
3374    
3375    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3376      case OP_ANYNL:
3377    case OP_VSPACE:    case OP_VSPACE:
3378    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3379    
3380    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3381    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3382    
3383    case OP_WORDCHAR:    case OP_WORDCHAR:
3384    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3385             next == -ESC_v || next == -ESC_R;
3386    
3387    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3388    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2369  switch(op_code) Line 3397  switch(op_code)
3397    
3398    
3399  /*************************************************  /*************************************************
3400    *        Add a character or range to a class     *
3401    *************************************************/
3402    
3403    /* This function packages up the logic of adding a character or range of
3404    characters to a class. The character values in the arguments will be within the
3405    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3406    mutually recursive with the function immediately below.
3407    
3408    Arguments:
3409      classbits     the bit map for characters < 256
3410      uchardptr     points to the pointer for extra data
3411      options       the options word
3412      cd            contains pointers to tables etc.
3413      start         start of range character
3414      end           end of range character
3415    
3416    Returns:        the number of < 256 characters added
3417                    the pointer to extra data is updated
3418    */
3419    
3420    static int
3421    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3422      compile_data *cd, unsigned int start, unsigned int end)
3423    {
3424    unsigned int c;
3425    int n8 = 0;
3426    
3427    /* If caseless matching is required, scan the range and process alternate
3428    cases. In Unicode, there are 8-bit characters that have alternate cases that
3429    are greater than 255 and vice-versa. Sometimes we can just extend the original
3430    range. */
3431    
3432    if ((options & PCRE_CASELESS) != 0)
3433      {
3434    #ifdef SUPPORT_UCP
3435      if ((options & PCRE_UTF8) != 0)
3436        {
3437        int rc;
3438        unsigned int oc, od;
3439    
3440        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3441        c = start;
3442    
3443        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3444          {
3445          /* Handle a single character that has more than one other case. */
3446    
3447          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3448            PRIV(ucd_caseless_sets) + rc, oc);
3449    
3450          /* Do nothing if the other case range is within the original range. */
3451    
3452          else if (oc >= start && od <= end) continue;
3453    
3454          /* Extend the original range if there is overlap, noting that if oc < c, we
3455          can't have od > end because a subrange is always shorter than the basic
3456          range. Otherwise, use a recursive call to add the additional range. */
3457    
3458          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3459          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3460          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3461          }
3462        }
3463      else
3464    #endif  /* SUPPORT_UCP */
3465    
3466      /* Not UTF-mode, or no UCP */
3467    
3468      for (c = start; c <= end && c < 256; c++)
3469        {
3470        SETBIT(classbits, cd->fcc[c]);
3471        n8++;
3472        }
3473      }
3474    
3475    /* Now handle the original range. Adjust the final value according to the bit
3476    length - this means that the same lists of (e.g.) horizontal spaces can be used
3477    in all cases. */
3478    
3479    #ifdef COMPILE_PCRE8
3480    #ifdef SUPPORT_UTF
3481      if ((options & PCRE_UTF8) == 0)
3482    #endif
3483      if (end > 0xff) end = 0xff;
3484    #endif
3485    
3486    #ifdef COMPILE_PCRE16
3487    #ifdef SUPPORT_UTF
3488      if ((options & PCRE_UTF16) == 0)
3489    #endif
3490      if (end > 0xffff) end = 0xffff;
3491    #endif
3492    
3493    /* If all characters are less than 256, use the bit map. Otherwise use extra
3494    data. */
3495    
3496    if (end < 0x100)
3497      {
3498      for (c = start; c <= end; c++)
3499        {
3500        n8++;
3501        SETBIT(classbits, c);
3502        }
3503      }
3504    
3505    else
3506      {
3507      pcre_uchar *uchardata = *uchardptr;
3508    
3509    #ifdef SUPPORT_UTF
3510      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3511        {
3512        if (start < end)
3513          {
3514          *uchardata++ = XCL_RANGE;
3515          uchardata += PRIV(ord2utf)(start, uchardata);
3516          uchardata += PRIV(ord2utf)(end, uchardata);
3517          }
3518        else if (start == end)
3519          {
3520          *uchardata++ = XCL_SINGLE;
3521          uchardata += PRIV(ord2utf)(start, uchardata);
3522          }
3523        }
3524      else
3525    #endif  /* SUPPORT_UTF */
3526    
3527      /* Without UTF support, character values are constrained by the bit length,
3528      and can only be > 256 for 16-bit and 32-bit libraries. */
3529    
3530    #ifdef COMPILE_PCRE8
3531        {}
3532    #else
3533      if (start < end)
3534        {
3535        *uchardata++ = XCL_RANGE;
3536        *uchardata++ = start;
3537        *uchardata++ = end;
3538        }
3539      else if (start == end)
3540        {
3541        *uchardata++ = XCL_SINGLE;
3542        *uchardata++ = start;
3543        }
3544    #endif
3545    
3546      *uchardptr = uchardata;   /* Updata extra data pointer */
3547      }
3548    
3549    return n8;    /* Number of 8-bit characters */
3550    }
3551    
3552    
3553    
3554    
3555    /*************************************************
3556    *        Add a list of characters to a class     *
3557    *************************************************/
3558    
3559    /* This function is used for adding a list of case-equivalent characters to a
3560    class, and also for adding a list of horizontal or vertical whitespace. If the
3561    list is in order (which it should be), ranges of characters are detected and
3562    handled appropriately. This function is mutually recursive with the function
3563    above.
3564    
3565    Arguments:
3566      classbits     the bit map for characters < 256
3567      uchardptr     points to the pointer for extra data
3568      options       the options word
3569      cd            contains pointers to tables etc.
3570      p             points to row of 32-bit values, terminated by NOTACHAR
3571      except        character to omit; this is used when adding lists of
3572                      case-equivalent characters to avoid including the one we
3573                      already know about
3574    
3575    Returns:        the number of < 256 characters added
3576                    the pointer to extra data is updated
3577    */
3578    
3579    static int
3580    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3581      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3582    {
3583    int n8 = 0;
3584    while (p[0] < NOTACHAR)
3585      {
3586      int n = 0;
3587      if (p[0] != except)
3588        {
3589        while(p[n+1] == p[0] + n + 1) n++;
3590        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3591        }
3592      p += n + 1;
3593      }
3594    return n8;
3595    }
3596    
3597    
3598    
3599    /*************************************************
3600    *    Add characters not in a list to a class     *
3601    *************************************************/
3602    
3603    /* This function is used for adding the complement of a list of horizontal or
3604    vertical whitespace to a class. The list must be in order.
3605    
3606    Arguments:
3607      classbits     the bit map for characters < 256
3608      uchardptr     points to the pointer for extra data
3609      options       the options word
3610      cd            contains pointers to tables etc.
3611      p             points to row of 32-bit values, terminated by NOTACHAR
3612    
3613    Returns:        the number of < 256 characters added
3614                    the pointer to extra data is updated
3615    */
3616    
3617    static int
3618    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3619      int options, compile_data *cd, const pcre_uint32 *p)
3620    {
3621    int n8 = 0;
3622    if (p[0] > 0)
3623      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3624    while (p[0] < NOTACHAR)
3625      {
3626      while (p[1] == p[0] + 1) p++;
3627      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3628        (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);
3629      p++;
3630      }
3631    return n8;
3632    }
3633    
3634    
3635    
3636    /*************************************************
3637  *           Compile one branch                   *  *           Compile one branch                   *
3638  *************************************************/  *************************************************/
3639    
# Line 2383  Arguments: Line 3648  Arguments:
3648    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3649    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3650    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3651    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3652    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3653    bcptr          points to current branch chain    bcptr          points to current branch chain
3654      cond_depth     conditional nesting depth
3655    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3656    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3657                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2395  Returns:         TRUE on success Line 3661  Returns:         TRUE on success
3661  */  */
3662    
3663  static BOOL  static BOOL
3664  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3665    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3666      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3667    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3668  {  {
3669  int repeat_type, op_type;  int repeat_type, op_type;
3670  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3671  int bravalue = 0;  int bravalue = 0;
3672  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3673  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3674  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3675  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3676  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3677  int after_manual_callout = 0;  int after_manual_callout = 0;
3678  int length_prevgroup = 0;  int length_prevgroup = 0;
3679  register int c;  register int c;
3680  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3681  uschar *last_code = code;  pcre_uchar *last_code = code;
3682  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3683  uschar *tempcode;  pcre_uchar *tempcode;
3684  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3685  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3686  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3687  const uschar *tempptr;  const pcre_uchar *tempptr;
3688  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3689  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3690  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3691  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3692    pcre_uint8 classbits[32];
3693  #ifdef SUPPORT_UTF8  
3694  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3695  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3696  uschar *class_utf8data;  dynamically as we process the pattern. */
3697  uschar *class_utf8data_base;  
3698  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3699    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3700    BOOL utf = (options & PCRE_UTF8) != 0;
3701    pcre_uchar utf_chars[6];
3702  #else  #else
3703  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3704  uschar *utf8_char = NULL;  #endif
3705    
3706    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3707    class_uchardata always so that it can be passed to add_to_class() always,
3708    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3709    alternative calls for the different cases. */
3710    
3711    pcre_uchar *class_uchardata;
3712    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3713    BOOL xclass;
3714    pcre_uchar *class_uchardata_base;
3715  #endif  #endif
3716    
3717  #ifdef DEBUG  #ifdef PCRE_DEBUG
3718  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3719  #endif  #endif
3720    
# Line 2445  greedy_non_default = greedy_default ^ 1; Line 3725  greedy_non_default = greedy_default ^ 1;
3725    
3726  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3727  matching encountered yet". It gets changed to REQ_NONE if we hit something that  matching encountered yet". It gets changed to REQ_NONE if we hit something that
3728  matches a non-fixed char first char; reqbyte just remains unset if we never  matches a non-fixed char first char; reqchar just remains unset if we never
3729  find one.  find one.
3730    
3731  When we hit a repeat whose minimum is zero, we may have to adjust these values  When we hit a repeat whose minimum is zero, we may have to adjust these values
3732  to take the zero repeat into account. This is implemented by setting them to  to take the zero repeat into account. This is implemented by setting them to
3733  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3734  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3735    
3736  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3737    
3738  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3739  according to the current setting of the caseless flag. REQ_CASELESS is a bit  or zero, according to the current setting of the caseless flag. The
3740  value > 255. It is added into the firstbyte or reqbyte variables to record the  REQ_CASELESS leaves the lower 28 bit empty. It is added into the
3741  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3742    value. This is used only for ASCII characters. */
3743    
3744  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3745    
3746  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3747    
# Line 2472  for (;; ptr++) Line 3753  for (;; ptr++)
3753    BOOL is_quantifier;    BOOL is_quantifier;
3754    BOOL is_recurse;    BOOL is_recurse;
3755    BOOL reset_bracount;    BOOL reset_bracount;
3756    int class_charcount;    int class_has_8bitchar;
3757    int class_lastchar;    int class_one_char;
3758    int newoptions;    int newoptions;
3759    int recno;    int recno;
3760    int refsign;    int refsign;
3761    int skipbytes;    int skipbytes;
3762    int subreqbyte;    int subreqchar;
3763    int subfirstbyte;    int subfirstchar;
3764    int terminator;    int terminator;
3765    int mclength;    int mclength;
3766    uschar mcbuffer[8];    int tempbracount;
3767      pcre_uchar mcbuffer[8];
3768    
3769    /* Get next byte in the pattern */    /* Get next character in the pattern */
3770    
3771    c = *ptr;    c = *ptr;
3772    
3773      /* If we are at the end of a nested substitution, revert to the outer level
3774      string. Nesting only happens one level deep. */
3775    
3776      if (c == 0 && nestptr != NULL)
3777        {
3778        ptr = nestptr;
3779        nestptr = NULL;
3780        c = *ptr;
3781        }
3782    
3783    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3784    previous cycle of this loop. */    previous cycle of this loop. */
3785    
3786    if (lengthptr != NULL)    if (lengthptr != NULL)
3787      {      {
3788  #ifdef DEBUG  #ifdef PCRE_DEBUG
3789      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3790  #endif  #endif
3791      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3792            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3793        {        {
3794        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3795        goto FAILED;        goto FAILED;
# Line 2518  for (;; ptr++) Line 3811  for (;; ptr++)
3811        goto FAILED;        goto FAILED;
3812        }        }
3813    
3814      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3815      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3816          (int)(code - last_code), c, c));
3817    
3818      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3819      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2529  for (;; ptr++) Line 3823  for (;; ptr++)
3823        {        {
3824        if (previous > orig_code)        if (previous > orig_code)
3825          {          {
3826          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3827          code -= previous - orig_code;          code -= previous - orig_code;
3828          previous = orig_code;          previous = orig_code;
3829          }          }
# Line 2545  for (;; ptr++) Line 3839  for (;; ptr++)
3839    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3840    reference list. */    reference list. */
3841    
3842    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3843               WORK_SIZE_SAFETY_MARGIN)
3844      {      {
3845      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3846      goto FAILED;      goto FAILED;
# Line 2555  for (;; ptr++) Line 3850  for (;; ptr++)
3850    
3851    if (inescq && c != 0)    if (inescq && c != 0)
3852      {      {
3853      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3854        {        {
3855        inescq = FALSE;        inescq = FALSE;
3856        ptr++;        ptr++;
# Line 2581  for (;; ptr++) Line 3876  for (;; ptr++)
3876    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3877    a quantifier. */    a quantifier. */
3878    
3879    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3880      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3881        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3882    
3883    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3884         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2592  for (;; ptr++) Line 3888  for (;; ptr++)
3888      previous_callout = NULL;      previous_callout = NULL;
3889      }      }
3890    
3891    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3892    
3893    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3894      {      {
3895      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3896      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3897        {        {
3898        while (*(++ptr) != 0)        ptr++;
3899          while (*ptr != 0)
3900          {          {
3901          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3902            ptr++;
3903    #ifdef SUPPORT_UTF
3904            if (utf) FORWARDCHAR(ptr);
3905    #endif
3906          }          }
3907        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3908    
# Line 2622  for (;; ptr++) Line 3923  for (;; ptr++)
3923      {      {
3924      /* ===================================================================*/      /* ===================================================================*/
3925      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3926      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3927      case ')':      case CHAR_RIGHT_PARENTHESIS:
3928      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3929      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3930      *codeptr = code;      *codeptr = code;
3931      *ptrptr = ptr;      *ptrptr = ptr;
3932      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2635  for (;; ptr++) Line 3936  for (;; ptr++)
3936          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3937          goto FAILED;          goto FAILED;
3938          }          }
3939        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3940        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3941        }        }
3942      return TRUE;      return TRUE;
# Line 2645  for (;; ptr++) Line 3946  for (;; ptr++)
3946      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3947      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3948    
3949      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3950        previous = NULL;
3951      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3952        {        {
3953        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3954          *code++ = OP_CIRCM;
3955        }        }
3956      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3957      break;      break;
3958    
3959      case '$':      case CHAR_DOLLAR_SIGN:
3960      previous = NULL;      previous = NULL;
3961      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3962      break;      break;
3963    
3964      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3965      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3966    
3967      case '.':      case CHAR_DOT:
3968      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3969      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3970      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3971      previous = code;      previous = code;
3972      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3973      break;      break;
# Line 2686  for (;; ptr++) Line 3988  for (;; ptr++)
3988      In JavaScript compatibility mode, an isolated ']' causes an error. In      In JavaScript compatibility mode, an isolated ']' causes an error. In
3989      default (Perl) mode, it is treated as a data character. */      default (Perl) mode, it is treated as a data character. */
3990    
3991      case ']':      case CHAR_RIGHT_SQUARE_BRACKET:
3992      if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3993        {        {
3994        *errorcodeptr = ERR64;        *errorcodeptr = ERR64;
# Line 2694  for (;; ptr++) Line 3996  for (;; ptr++)
3996        }        }
3997      goto NORMAL_CHAR;      goto NORMAL_CHAR;
3998    
3999      case '[':      case CHAR_LEFT_SQUARE_BRACKET:
4000      previous = code;      previous = code;
4001    
4002      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4003      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
4004    
4005      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4006             ptr[1] == CHAR_EQUALS_SIGN) &&
4007          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
4008        {        {
4009        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4010        goto FAILED;        goto FAILED;
4011        }        }
4012    
# Line 2715  for (;; ptr++) Line 4018  for (;; ptr++)
4018      for (;;)      for (;;)
4019        {        {
4020        c = *(++ptr);        c = *(++ptr);
4021        if (c == '\\')        if (c == CHAR_BACKSLASH)
4022          {          {
4023          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
4024            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
4025              else break;          else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4026              ptr += 3;
4027            else
4028              break;
4029          }          }
4030        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4031          negate_class = TRUE;          negate_class = TRUE;
4032        else break;        else break;
4033        }        }
# Line 2731  for (;; ptr++) Line 4037  for (;; ptr++)
4037      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4038      [^] must match any character, so generate OP_ALLANY. */      [^] must match any character, so generate OP_ALLANY. */
4039    
4040      if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)      if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4041            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4042        {        {
4043        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4044        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4045        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
4046        break;        break;
4047        }        }
4048    
# Line 2745  for (;; ptr++) Line 4052  for (;; ptr++)
4052    
4053      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4054    
4055      /* Keep a count of chars with values < 256 so that we can optimize the case      /* For optimization purposes, we track some properties of the class:
4056      of just a single character (as long as it's < 256). However, For higher      class_has_8bitchar will be non-zero if the class contains at least one <