/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

code/trunk/pcre_compile.c revision 333 by ph10, Thu Apr 10 19:55:57 2008 UTC code/branches/pcre16/pcre_compile.c revision 798 by zherczeg, Sun Dec 11 18:07:25 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2011 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_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    /* Private flags added to firstchar and reqchar. */
101    
102    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
103    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
104    
105    /* Repeated character flags. */
106    
107    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
108    
109  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
110  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
111  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
112  is invalid. */  is invalid. */
113    
114  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
115    
116    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
117    in UTF-8 mode. */
118    
119  static const short int escapes[] = {  static const short int escapes[] = {
120       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
121       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
122     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
123  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
124  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
125  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
126     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
127  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
128  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
129       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
130         -ESC_D,                  -ESC_E,
131         0,                       -ESC_G,
132         -ESC_H,                  0,
133         0,                       -ESC_K,
134         0,                       0,
135         -ESC_N,                  0,
136         -ESC_P,                  -ESC_Q,
137         -ESC_R,                  -ESC_S,
138         0,                       0,
139         -ESC_V,                  -ESC_W,
140         -ESC_X,                  0,
141         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
142         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
143         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
144         CHAR_GRAVE_ACCENT,       7,
145         -ESC_b,                  0,
146         -ESC_d,                  ESC_e,
147         ESC_f,                   0,
148         -ESC_h,                  0,
149         0,                       -ESC_k,
150         0,                       0,
151         ESC_n,                   0,
152         -ESC_p,                  0,
153         ESC_r,                   -ESC_s,
154         ESC_tee,                 0,
155         -ESC_v,                  -ESC_w,
156         0,                       0,
157         -ESC_z
158  };  };
159    
160  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
161    
162    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
163    
164  static const short int escapes[] = {  static const short int escapes[] = {
165  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
166  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 179  static const short int escapes[] = {
179  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
180  /*  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,
181  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
182  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
183  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
184  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
185  /*  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 191  static const short int escapes[] = {
191    
192  /* 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
193  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
194  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
195    string is built from string macros so that it works in UTF-8 mode on EBCDIC
196    platforms. */
197    
198  typedef struct verbitem {  typedef struct verbitem {
199    int   len;    int   len;                 /* Length of verb name */
200    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
201      int   op_arg;              /* Op when arg present, or -1 if not allowed */
202  } verbitem;  } verbitem;
203    
204  static const char verbnames[] =  static const char verbnames[] =
205    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
206    "COMMIT\0"    STRING_MARK0
207    "F\0"    STRING_ACCEPT0
208    "FAIL\0"    STRING_COMMIT0
209    "PRUNE\0"    STRING_F0
210    "SKIP\0"    STRING_FAIL0
211    "THEN";    STRING_PRUNE0
212      STRING_SKIP0
213      STRING_THEN;
214    
215  static const verbitem verbs[] = {  static const verbitem verbs[] = {
216    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
217    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
218    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
219    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
220    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
221    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
222    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
223      { 4, OP_SKIP,   OP_SKIP_ARG  },
224      { 4, OP_THEN,   OP_THEN_ARG  }
225  };  };
226    
227  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 234  length entry. The first three must be al
234  for handling case independence. */  for handling case independence. */
235    
236  static const char posix_names[] =  static const char posix_names[] =
237    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
238    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
239    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
240      STRING_word0  STRING_xdigit;
241    
242  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
243    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 };
244    
245  /* 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 269  static const int posix_class_maps[] = {
269    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
270  };  };
271    
272    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
273    substitutes must be in the order of the names, defined above, and there are
274    both positive and negative cases. NULL means no substitute. */
275    
276    #ifdef SUPPORT_UCP
277    static const pcre_uchar string_PNd[]  = {
278      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
279      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
280    static const pcre_uchar string_pNd[]  = {
281      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
282      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
283    static const pcre_uchar string_PXsp[] = {
284      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
285      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
286    static const pcre_uchar string_pXsp[] = {
287      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
288      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
289    static const pcre_uchar string_PXwd[] = {
290      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
291      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
292    static const pcre_uchar string_pXwd[] = {
293      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
294      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
295    
296    static const pcre_uchar *substitutes[] = {
297      string_PNd,           /* \D */
298      string_pNd,           /* \d */
299      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
300      string_pXsp,          /* \s */
301      string_PXwd,          /* \W */
302      string_pXwd           /* \w */
303    };
304    
305    static const pcre_uchar string_pL[] =   {
306      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
307      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
308    static const pcre_uchar string_pLl[] =  {
309      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
310      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
311    static const pcre_uchar string_pLu[] =  {
312      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_pXan[] = {
315      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    static const pcre_uchar string_h[] =    {
318      CHAR_BACKSLASH, CHAR_h, '\0' };
319    static const pcre_uchar string_pXps[] = {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_PL[] =   {
323      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_PLl[] =  {
326      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    static const pcre_uchar string_PLu[] =  {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_PXan[] = {
332      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_H[] =    {
335      CHAR_BACKSLASH, CHAR_H, '\0' };
336    static const pcre_uchar string_PXps[] = {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    
340    static const pcre_uchar *posix_substitutes[] = {
341      string_pL,            /* alpha */
342      string_pLl,           /* lower */
343      string_pLu,           /* upper */
344      string_pXan,          /* alnum */
345      NULL,                 /* ascii */
346      string_h,             /* blank */
347      NULL,                 /* cntrl */
348      string_pNd,           /* digit */
349      NULL,                 /* graph */
350      NULL,                 /* print */
351      NULL,                 /* punct */
352      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
353      string_pXwd,          /* word */
354      NULL,                 /* xdigit */
355      /* Negated cases */
356      string_PL,            /* ^alpha */
357      string_PLl,           /* ^lower */
358      string_PLu,           /* ^upper */
359      string_PXan,          /* ^alnum */
360      NULL,                 /* ^ascii */
361      string_H,             /* ^blank */
362      NULL,                 /* ^cntrl */
363      string_PNd,           /* ^digit */
364      NULL,                 /* ^graph */
365      NULL,                 /* ^print */
366      NULL,                 /* ^punct */
367      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
368      string_PXwd,          /* ^word */
369      NULL                  /* ^xdigit */
370    };
371    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
372    #endif
373    
374  #define STRING(a)  # a  #define STRING(a)  # a
375  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 382  the number of relocations needed when a
382  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
383  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
384  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
385  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
386    
387    Each substring ends with \0 to insert a null character. This includes the final
388    substring, so that the whole string ends with \0\0, which can be detected when
389    counting through. */
390    
391  static const char error_texts[] =  static const char error_texts[] =
392    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 433  static const char error_texts[] =
433    /* 35 */    /* 35 */
434    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
435    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
436    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
437    "number after (?C is > 255\0"    "number after (?C is > 255\0"
438    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
439    /* 40 */    /* 40 */
# Line 293  static const char error_texts[] = Line 455  static const char error_texts[] =
455    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
456    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
457    /* 55 */    /* 55 */
458    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
459    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
460    "\\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"
461    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
462    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
463    /* 60 */    /* 60 */
464    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
465    "number is too big\0"    "number is too big\0"
466    "subpattern name expected\0"    "subpattern name expected\0"
467    "digit expected after (?+";    "digit expected after (?+\0"
468      "] is an invalid data character in JavaScript compatibility mode\0"
469      /* 65 */
470      "different names for subpatterns of the same number are not allowed\0"
471      "(*MARK) must have an argument\0"
472      "this version of PCRE is not compiled with PCRE_UCP support\0"
473      "\\c must be followed by an ASCII character\0"
474      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
475      /* 70 */
476      "internal error: unknown opcode in find_fixedlength()\0"
477      "Not allowed UTF-8 / UTF-16 code point (>= 0xd800 && <= 0xdfff)\0"
478      ;
479    
480  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
481  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 493  For convenience, we use the same bit def
493    
494  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
495    
496  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
497  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
498    into a subtraction and unsigned comparison). */
499    
500    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
501    
502    #ifndef EBCDIC
503    
504    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
505    UTF-8 mode. */
506    
507    static const pcre_uint8 digitab[] =
508    {    {
509    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
510    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 357  static const unsigned char digitab[] = Line 539  static const unsigned char digitab[] =
539    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
540    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
541    
542  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
543  static const unsigned char digitab[] =  
544    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
545    
546    static const pcre_uint8 digitab[] =
547    {    {
548    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
549    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 393  static const unsigned char digitab[] = Line 578  static const unsigned char digitab[] =
578    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
579    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
580    
581  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
582    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
583    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
584    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 432  static const unsigned char ebcdic_charta Line 617  static const unsigned char ebcdic_charta
617  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
618    
619  static BOOL  static BOOL
620    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
621      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
622    
623    
# Line 454  static const char * Line 639  static const char *
639  find_error_text(int n)  find_error_text(int n)
640  {  {
641  const char *s = error_texts;  const char *s = error_texts;
642  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
643      {
644      while (*s++ != 0) {};
645      if (*s == 0) return "Error text not found (please report)";
646      }
647  return s;  return s;
648  }  }
649    
650    
651  /*************************************************  /*************************************************
652    *            Check for counted repeat            *
653    *************************************************/
654    
655    /* This function is called when a '{' is encountered in a place where it might
656    start a quantifier. It looks ahead to see if it really is a quantifier or not.
657    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
658    where the ddds are digits.
659    
660    Arguments:
661      p         pointer to the first char after '{'
662    
663    Returns:    TRUE or FALSE
664    */
665    
666    static BOOL
667    is_counted_repeat(const pcre_uchar *p)
668    {
669    if (!IS_DIGIT(*p)) return FALSE;
670    p++;
671    while (IS_DIGIT(*p)) p++;
672    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
673    
674    if (*p++ != CHAR_COMMA) return FALSE;
675    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
676    
677    if (!IS_DIGIT(*p)) return FALSE;
678    p++;
679    while (IS_DIGIT(*p)) p++;
680    
681    return (*p == CHAR_RIGHT_CURLY_BRACKET);
682    }
683    
684    
685    
686    /*************************************************
687  *            Handle escapes                      *  *            Handle escapes                      *
688  *************************************************/  *************************************************/
689    
# Line 484  Returns:         zero or positive => a d Line 708  Returns:         zero or positive => a d
708  */  */
709    
710  static int  static int
711  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
712    int options, BOOL isclass)    int options, BOOL isclass)
713  {  {
714  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
715  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
716  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
717    pcre_int32 c;
718    int i;
719    
720  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
721  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 728  if (c == 0) *errorcodeptr = ERR1;
728  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.
729  Otherwise further processing may be required. */  Otherwise further processing may be required. */
730    
731  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
732  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
733  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
734    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
735    
736  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
737  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
738    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
739  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
740  #endif  #endif
741    
# Line 515  else if ((i = escapes[c - 0x48]) != 0) Line 743  else if ((i = escapes[c - 0x48]) != 0)
743    
744  else  else
745    {    {
746    const uschar *oldptr;    const pcre_uchar *oldptr;
747    BOOL braced, negated;    BOOL braced, negated;
748    
749    switch (c)    switch (c)
# Line 523  else Line 751  else
751      /* 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
752      error. */      error. */
753    
754      case 'l':      case CHAR_l:
755      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
756      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
757      break;      break;
758    
759      /* \g must be followed by one of a number of specific things:      case CHAR_u:
760        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
761          {
762          /* In JavaScript, \u must be followed by four hexadecimal numbers.
763          Otherwise it is a lowercase u letter. */
764          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
765            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
766            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
767            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
768            {
769            c = 0;
770            for (i = 0; i < 4; ++i)
771              {
772              register int cc = *(++ptr);
773    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
774              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
775              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
776    #else           /* EBCDIC coding */
777              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
778              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
779    #endif
780              }
781            }
782          }
783        else
784          *errorcodeptr = ERR37;
785        break;
786    
787        case CHAR_U:
788        /* In JavaScript, \U is an uppercase U letter. */
789        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
790        break;
791    
792        /* In a character class, \g is just a literal "g". Outside a character
793        class, \g must be followed by one of a number of specific things:
794    
795      (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
796      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
797      5.10 feature.      5.10 feature.
798    
799      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
800      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
801      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
802      was \k.      was \k.
803    
804      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
805      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
806      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
807      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
808    
809      case 'g':      case CHAR_g:
810      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
811        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
812        {        {
813        c = -ESC_g;        c = -ESC_g;
814        break;        break;
815        }        }
816    
817      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
818    
819      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
820        {        {
821        const uschar *p;        const pcre_uchar *p;
822        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
823          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
824        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
825          {          {
826          c = -ESC_k;          c = -ESC_k;
827          break;          break;
# Line 571  else Line 831  else
831        }        }
832      else braced = FALSE;      else braced = FALSE;
833    
834      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
835        {        {
836        negated = TRUE;        negated = TRUE;
837        ptr++;        ptr++;
838        }        }
839      else negated = FALSE;      else negated = FALSE;
840    
841        /* The integer range is limited by the machine's int representation. */
842      c = 0;      c = 0;
843      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
       c = c * 10 + *(++ptr) - '0';  
   
     if (c < 0)   /* Integer overflow */  
844        {        {
845          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
846            {
847            c = -1;
848            break;
849            }
850          c = c * 10 + *(++ptr) - CHAR_0;
851          }
852        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
853          {
854          while (IS_DIGIT(ptr[1]))
855            ptr++;
856        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
857        break;        break;
858        }        }
859    
860      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
861        {        {
862        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
863        break;        break;
864        }        }
865    
866      if (c == 0)      if (c == 0)
867        {        {
868        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
869        break;        break;
870        }        }
871    
872      if (negated)      if (negated)
873        {        {
# Line 625  else Line 894  else
894      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
895      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
896    
897      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:
898      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
899    
900      if (!isclass)      if (!isclass)
901        {        {
902        oldptr = ptr;        oldptr = ptr;
903        c -= '0';        /* The integer range is limited by the machine's int representation. */
904        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
905          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
       if (c < 0)    /* Integer overflow */  
906          {          {
907            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
908              {
909              c = -1;
910              break;
911              }
912            c = c * 10 + *(++ptr) - CHAR_0;
913            }
914          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
915            {
916            while (IS_DIGIT(ptr[1]))
917              ptr++;
918          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
919          break;          break;
920          }          }
# Line 651  else Line 930  else
930      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.
931      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
932    
933      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
934        {        {
935        ptr--;        ptr--;
936        c = 0;        c = 0;
# Line 664  else Line 943  else
943      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
944      than 3 octal digits. */      than 3 octal digits. */
945    
946      case '0':      case CHAR_0:
947      c -= '0';      c -= CHAR_0;
948      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
949          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
950      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
951      break;      break;
952    
953      /* \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
954      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.
955      treated as a data character. */      If not, { is treated as a data character. */
956    
957      case 'x':      case CHAR_x:
958      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
959        {        {
960        const uschar *pt = ptr + 2;        /* In JavaScript, \x must be followed by two hexadecimal numbers.
961        int count = 0;        Otherwise it is a lowercase x letter. */
962          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
963            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
964            {
965            c = 0;
966            for (i = 0; i < 2; ++i)
967              {
968              register int cc = *(++ptr);
969    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
970              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
971              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
972    #else           /* EBCDIC coding */
973              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
974              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
975    #endif
976              }
977            }
978          break;
979          }
980    
981        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
982          {
983          const pcre_uchar *pt = ptr + 2;
984    
985        c = 0;        c = 0;
986        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
987          {          {
988          register int cc = *pt++;          register int cc = *pt++;
989          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
990    
991  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
992          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
993          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
994  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
995          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
996          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
997    #endif
998    
999    #ifdef COMPILE_PCRE8
1000            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1001    #else
1002    #ifdef COMPILE_PCRE16
1003            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1004  #endif  #endif
1005    #endif
1006            }
1007    
1008          if (c < 0)
1009            {
1010            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1011            *errorcodeptr = ERR34;
1012          }          }
1013    
1014        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1015          {          {
1016          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR71;
1017          ptr = pt;          ptr = pt;
1018          break;          break;
1019          }          }
# Line 711  else Line 1025  else
1025      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1026    
1027      c = 0;      c = 0;
1028      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1029        {        {
1030        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1031        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1032  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1033        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1034        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1035  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1036        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1037        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1038  #endif  #endif
1039        }        }
1040      break;      break;
1041    
1042      /* 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.
1043      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
1044        coding is ASCII-specific, but then the whole concept of \cx is
1045      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1046    
1047      case 'c':      case CHAR_c:
1048      c = *(++ptr);      c = *(++ptr);
1049      if (c == 0)      if (c == 0)
1050        {        {
1051        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1052        break;        break;
1053        }        }
1054    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1055  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1056      if (c >= 'a' && c <= 'z') c -= 32;        {
1057          *errorcodeptr = ERR68;
1058          break;
1059          }
1060        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1061      c ^= 0x40;      c ^= 0x40;
1062  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1063      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1064      c ^= 0xC0;      c ^= 0xC0;
1065  #endif  #endif
1066      break;      break;
# Line 763  else Line 1082  else
1082      }      }
1083    }    }
1084    
1085    /* Perl supports \N{name} for character names, as well as plain \N for "not
1086    newline". PCRE does not support \N{name}. However, it does support
1087    quantification such as \N{2,3}. */
1088    
1089    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1090         !is_counted_repeat(ptr+2))
1091      *errorcodeptr = ERR37;
1092    
1093    /* If PCRE_UCP is set, we change the values for \d etc. */
1094    
1095    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1096      c -= (ESC_DU - ESC_D);
1097    
1098    /* Set the pointer to the final character before returning. */
1099    
1100  *ptrptr = ptr;  *ptrptr = ptr;
1101  return c;  return c;
1102  }  }
# Line 789  Returns:         type value from ucp_typ Line 1123  Returns:         type value from ucp_typ
1123  */  */
1124    
1125  static int  static int
1126  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1127  {  {
1128  int c, i, bot, top;  int c, i, bot, top;
1129  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1130  char name[32];  pcre_uchar name[32];
1131    
1132  c = *(++ptr);  c = *(++ptr);
1133  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 803  if (c == 0) goto ERROR_RETURN; Line 1137  if (c == 0) goto ERROR_RETURN;
1137  /* \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
1138  negation. */  negation. */
1139    
1140  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1141    {    {
1142    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1143      {      {
1144      *negptr = TRUE;      *negptr = TRUE;
1145      ptr++;      ptr++;
# Line 814  if (c == '{') Line 1148  if (c == '{')
1148      {      {
1149      c = *(++ptr);      c = *(++ptr);
1150      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1151      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1152      name[i] = c;      name[i] = c;
1153      }      }
1154    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1155    name[i] = 0;    name[i] = 0;
1156    }    }
1157    
# Line 834  else Line 1168  else
1168  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1169    
1170  bot = 0;  bot = 0;
1171  top = _pcre_utt_size;  top = PRIV(utt_size);
1172    
1173  while (bot < top)  while (bot < top)
1174    {    {
1175    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1176    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1177    if (c == 0)    if (c == 0)
1178      {      {
1179      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1180      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1181      }      }
1182    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1183    }    }
# Line 863  return -1; Line 1197  return -1;
1197    
1198    
1199  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1200  *         Read repeat counts                     *  *         Read repeat counts                     *
1201  *************************************************/  *************************************************/
1202    
# Line 914  Returns:         pointer to '}' on succe Line 1215  Returns:         pointer to '}' on succe
1215                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1216  */  */
1217    
1218  static const uschar *  static const pcre_uchar *
1219  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)
1220  {  {
1221  int min = 0;  int min = 0;
1222  int max = -1;  int max = -1;
# Line 923  int max = -1; Line 1224  int max = -1;
1224  /* 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
1225  an integer overflow. */  an integer overflow. */
1226    
1227  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1228  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1229    {    {
1230    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1234  if (min < 0 || min > 65535)
1234  /* 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.
1235  Also, max must not be less than min. */  Also, max must not be less than min. */
1236    
1237  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1238    {    {
1239    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1240      {      {
1241      max = 0;      max = 0;
1242      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1243      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1244        {        {
1245        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1264  return p;
1264    
1265    
1266  /*************************************************  /*************************************************
1267  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1268  *************************************************/  *************************************************/
1269    
1270  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1271    top-level call starts at the beginning of the pattern. All other calls must
1272    start at a parenthesis. It scans along a pattern's text looking for capturing
1273  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
1274  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
1275  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
1276  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1277  be terminated by '>' because that is checked in the first pass.  
1278    This function was originally called only from the second pass, in which we know
1279    that if (?< or (?' or (?P< is encountered, the name will be correctly
1280    terminated because that is checked in the first pass. There is now one call to
1281    this function in the first pass, to check for a recursive back reference by
1282    name (so that we can make the whole group atomic). In this case, we need check
1283    only up to the current position in the pattern, and that is still OK because
1284    and previous occurrences will have been checked. To make this work, the test
1285    for "end of pattern" is a check against cd->end_pattern in the main loop,
1286    instead of looking for a binary zero. This means that the special first-pass
1287    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1288    processing items within the loop are OK, because afterwards the main loop will
1289    terminate.)
1290    
1291  Arguments:  Arguments:
1292    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1293    count        current count of capturing parens so far encountered    cd           compile background data
1294    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1295    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1296    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1297      utf          TRUE if we are in UTF-8 / UTF-16 mode
1298      count        pointer to the current capturing subpattern number (updated)
1299    
1300  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1301  */  */
1302    
1303  static int  static int
1304  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1305    BOOL xmode)    BOOL xmode, BOOL utf, int *count)
1306  {  {
1307  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1308    int start_count = *count;
1309    int hwm_count = start_count;
1310    BOOL dup_parens = FALSE;
1311    
1312  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1313    dealing with. The very first call may not start with a parenthesis. */
1314    
1315    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1316    {    {
1317    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1318    
1319      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1320    
1321      /* Handle a normal, unnamed capturing parenthesis. */
1322    
1323      else if (ptr[1] != CHAR_QUESTION_MARK)
1324        {
1325        *count += 1;
1326        if (name == NULL && *count == lorn) return *count;
1327        ptr++;
1328        }
1329    
1330      /* All cases now have (? at the start. Remember when we are in a group
1331      where the parenthesis numbers are duplicated. */
1332    
1333      else if (ptr[2] == CHAR_VERTICAL_LINE)
1334        {
1335        ptr += 3;
1336        dup_parens = TRUE;
1337        }
1338    
1339      /* Handle comments; all characters are allowed until a ket is reached. */
1340    
1341      else if (ptr[2] == CHAR_NUMBER_SIGN)
1342        {
1343        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1344        goto FAIL_EXIT;
1345        }
1346    
1347      /* Handle a condition. If it is an assertion, just carry on so that it
1348      is processed as normal. If not, skip to the closing parenthesis of the
1349      condition (there can't be any nested parens). */
1350    
1351      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1352        {
1353        ptr += 2;
1354        if (ptr[1] != CHAR_QUESTION_MARK)
1355          {
1356          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1357          if (*ptr != 0) ptr++;
1358          }
1359        }
1360    
1361      /* Start with (? but not a condition. */
1362    
1363      else
1364        {
1365        ptr += 2;
1366        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1367    
1368        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1369    
1370        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1371            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1372          {
1373          int term;
1374          const pcre_uchar *thisname;
1375          *count += 1;
1376          if (name == NULL && *count == lorn) return *count;
1377          term = *ptr++;
1378          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1379          thisname = ptr;
1380          while (*ptr != term) ptr++;
1381          if (name != NULL && lorn == ptr - thisname &&
1382              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1383            return *count;
1384          term++;
1385          }
1386        }
1387      }
1388    
1389    /* Past any initial parenthesis handling, scan for parentheses or vertical
1390    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1391    first-pass call when this value is temporarily adjusted to stop at the current
1392    position. So DO NOT change this to a test for binary zero. */
1393    
1394    for (; ptr < cd->end_pattern; ptr++)
1395      {
1396    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1397    
1398    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1399      {      {
1400      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1401      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1402        {        {
1403        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1404        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1405        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1406        }        }
1407      continue;      continue;
1408      }      }
1409    
1410    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1411      are handled for real. If the first character is '^', skip it. Also, if the
1412      first few characters (either before or after ^) are \Q\E or \E we skip them
1413      too. This makes for compatibility with Perl. Note the use of STR macros to
1414      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1415    
1416    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1417      {      {
1418      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1419        for (;;)
1420          {
1421          if (ptr[1] == CHAR_BACKSLASH)
1422            {
1423            if (ptr[2] == CHAR_E)
1424              ptr+= 2;
1425            else if (STRNCMP_UC_C8(ptr + 2,
1426                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1427              ptr += 4;
1428            else
1429              break;
1430            }
1431          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1432            {
1433            negate_class = TRUE;
1434            ptr++;
1435            }
1436          else break;
1437          }
1438    
1439        /* If the next character is ']', it is a data character that must be
1440        skipped, except in JavaScript compatibility mode. */
1441    
1442        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1443            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1444          ptr++;
1445    
1446        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1447        {        {
1448        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1449        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1450          {          {
1451          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1452          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1453            {            {
1454            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1455            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1456            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1457            }            }
1458          continue;          continue;
1459          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1463  for (; *ptr != 0; ptr++)
1463    
1464    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1465    
1466    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1467      {      {
1468      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1469      if (*ptr == 0) return -1;      while (*ptr != 0)
1470          {
1471          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1472          ptr++;
1473    #ifdef SUPPORT_UTF
1474          if (utf) FORWARDCHAR(ptr);
1475    #endif
1476          }
1477        if (*ptr == 0) goto FAIL_EXIT;
1478      continue;      continue;
1479      }      }
1480    
1481    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1482    
1483    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1484      {      {
1485      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1486      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1487      continue;      if (*ptr == 0) goto FAIL_EXIT;
1488      }      }
1489    
1490    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1491    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1492        if (dup_parens && *count < hwm_count) *count = hwm_count;
1493        goto FAIL_EXIT;
1494        }
1495    
1496      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1497        {
1498        if (*count > hwm_count) hwm_count = *count;
1499        *count = start_count;
1500        }
1501      }
1502    
1503    FAIL_EXIT:
1504    *ptrptr = ptr;
1505    return -1;
1506    }
1507    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1508    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1509    
   count++;  
1510    
1511    if (name == NULL && count == lorn) return count;  /*************************************************
1512    term = *ptr++;  *       Find forward referenced subpattern       *
1513    if (term == '<') term = '>';  *************************************************/
1514    thisname = ptr;  
1515    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1516    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1517        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1518      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1519    references to subpatterns. We used to be able to start this scan from the
1520    current compiling point, using the current count value from cd->bracount, and
1521    do it all in a single loop, but the addition of the possibility of duplicate
1522    subpattern numbers means that we have to scan from the very start, in order to
1523    take account of such duplicates, and to use a recursive function to keep track
1524    of the different types of group.
1525    
1526    Arguments:
1527      cd           compile background data
1528      name         name to seek, or NULL if seeking a numbered subpattern
1529      lorn         name length, or subpattern number if name is NULL
1530      xmode        TRUE if we are in /x mode
1531      utf          TRUE if we are in UTF-8 / UTF-16 mode
1532    
1533    Returns:       the number of the found subpattern, or -1 if not found
1534    */
1535    
1536    static int
1537    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1538      BOOL utf)
1539    {
1540    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1541    int count = 0;
1542    int rc;
1543    
1544    /* If the pattern does not start with an opening parenthesis, the first call
1545    to find_parens_sub() will scan right to the end (if necessary). However, if it
1546    does start with a parenthesis, find_parens_sub() will return when it hits the
1547    matching closing parens. That is why we have to have a loop. */
1548    
1549    for (;;)
1550      {
1551      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1552      if (rc > 0 || *ptr++ == 0) break;
1553    }    }
1554    
1555  return -1;  return rc;
1556  }  }
1557    
1558    
1559    
1560    
1561  /*************************************************  /*************************************************
1562  *      Find first significant op code            *  *      Find first significant op code            *
1563  *************************************************/  *************************************************/
1564    
1565  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1566  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
1567  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
1568  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
1569  assertions, and also the \b assertion; for others it does not.  does not.
1570    
1571  Arguments:  Arguments:
1572    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  
1573    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1574    
1575  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1576  */  */
1577    
1578  static const uschar*  static const pcre_uchar*
1579  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1580  {  {
1581  for (;;)  for (;;)
1582    {    {
1583    switch ((int)*code)    switch ((int)*code)
1584      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1585      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1586      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1587      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1588      if (!skipassert) return code;      if (!skipassert) return code;
1589      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1590      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1591      break;      break;
1592    
1593      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1123  for (;;) Line 1597  for (;;)
1597    
1598      case OP_CALLOUT:      case OP_CALLOUT:
1599      case OP_CREF:      case OP_CREF:
1600        case OP_NCREF:
1601      case OP_RREF:      case OP_RREF:
1602        case OP_NRREF:
1603      case OP_DEF:      case OP_DEF:
1604      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1605      break;      break;
1606    
1607      default:      default:
# Line 1139  for (;;) Line 1615  for (;;)
1615    
1616    
1617  /*************************************************  /*************************************************
1618  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1619  *************************************************/  *************************************************/
1620    
1621  /* 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,
1622  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.
1623  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
1624    temporarily terminated with OP_END when this function is called.
1625    
1626    This function is called when a backward assertion is encountered, so that if it
1627    fails, the error message can point to the correct place in the pattern.
1628    However, we cannot do this when the assertion contains subroutine calls,
1629    because they can be forward references. We solve this by remembering this case
1630    and doing the check at the end; a flag specifies which mode we are running in.
1631    
1632  Arguments:  Arguments:
1633    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1634    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1635      atend    TRUE if called when the pattern is complete
1636  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1637               or -2 if \C was encountered  
1638    Returns:   the fixed length,
1639                 or -1 if there is no fixed length,
1640                 or -2 if \C was encountered (in UTF-8 mode only)
1641                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1642                 or -4 if an unknown opcode was encountered (internal error)
1643  */  */
1644    
1645  static int  static int
1646  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1647  {  {
1648  int length = -1;  int length = -1;
1649    
1650  register int branchlength = 0;  register int branchlength = 0;
1651  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1652    
1653  /* 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
1654  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1168  branch, check the length against that of Line 1656  branch, check the length against that of
1656  for (;;)  for (;;)
1657    {    {
1658    int d;    int d;
1659      pcre_uchar *ce, *cs;
1660    register int op = *cc;    register int op = *cc;
1661    switch (op)    switch (op)
1662      {      {
1663        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1664        OP_BRA (normal non-capturing bracket) because the other variants of these
1665        opcodes are all concerned with unlimited repeated groups, which of course
1666        are not of fixed length. */
1667    
1668      case OP_CBRA:      case OP_CBRA:
1669      case OP_BRA:      case OP_BRA:
1670      case OP_ONCE:      case OP_ONCE:
1671        case OP_ONCE_NC:
1672      case OP_COND:      case OP_COND:
1673      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1674      if (d < 0) return d;      if (d < 0) return d;
1675      branchlength += d;      branchlength += d;
1676      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1677      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1678      break;      break;
1679    
1680      /* 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.
1681      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
1682      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
1683        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1684        because they all imply an unlimited repeat. */
1685    
1686      case OP_ALT:      case OP_ALT:
1687      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1688      case OP_END:      case OP_END:
1689        case OP_ACCEPT:
1690        case OP_ASSERT_ACCEPT:
1691      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1692        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1693      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1695  for (;;)
1695      branchlength = 0;      branchlength = 0;
1696      break;      break;
1697    
1698        /* A true recursion implies not fixed length, but a subroutine call may
1699        be OK. If the subroutine is a forward reference, we can't deal with
1700        it until the end of the pattern, so return -3. */
1701    
1702        case OP_RECURSE:
1703        if (!atend) return -3;
1704        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1705        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1706        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1707        d = find_fixedlength(cs + 2, utf, atend, cd);
1708        if (d < 0) return d;
1709        branchlength += d;
1710        cc += 1 + LINK_SIZE;
1711        break;
1712    
1713      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1714    
1715      case OP_ASSERT:      case OP_ASSERT:
# Line 1209  for (;;) Line 1721  for (;;)
1721    
1722      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1723    
1724      case OP_REVERSE:      case OP_MARK:
1725        case OP_PRUNE_ARG:
1726        case OP_SKIP_ARG:
1727        case OP_THEN_ARG:
1728        cc += cc[1] + PRIV(OP_lengths)[*cc];
1729        break;
1730    
1731        case OP_CALLOUT:
1732        case OP_CIRC:
1733        case OP_CIRCM:
1734        case OP_CLOSE:
1735        case OP_COMMIT:
1736      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1737      case OP_DEF:      case OP_DEF:
1738      case OP_OPT:      case OP_DOLL:
1739      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1740      case OP_EOD:      case OP_EOD:
1741      case OP_EODN:      case OP_EODN:
1742      case OP_CIRC:      case OP_FAIL:
1743      case OP_DOLL:      case OP_NCREF:
1744        case OP_NRREF:
1745      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1746        case OP_PRUNE:
1747        case OP_REVERSE:
1748        case OP_RREF:
1749        case OP_SET_SOM:
1750        case OP_SKIP:
1751        case OP_SOD:
1752        case OP_SOM:
1753        case OP_THEN:
1754      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1755      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1756      break;      break;
1757    
1758      /* Handle literal characters */      /* Handle literal characters */
1759    
1760      case OP_CHAR:      case OP_CHAR:
1761      case OP_CHARNC:      case OP_CHARI:
1762      case OP_NOT:      case OP_NOT:
1763        case OP_NOTI:
1764      branchlength++;      branchlength++;
1765      cc += 2;      cc += 2;
1766  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1767      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1768  #endif  #endif
1769      break;      break;
1770    
# Line 1245  for (;;) Line 1772  for (;;)
1772      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1773    
1774      case OP_EXACT:      case OP_EXACT:
1775        case OP_EXACTI:
1776        case OP_NOTEXACT:
1777        case OP_NOTEXACTI:
1778      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1779      cc += 4;      cc += 2 + IMM2_SIZE;
1780  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1781      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1782  #endif  #endif
1783      break;      break;
1784    
1785      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1786      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1787      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1788      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1789      break;      break;
1790    
1791      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1268  for (;;) Line 1795  for (;;)
1795      cc += 2;      cc += 2;
1796      /* Fall through */      /* Fall through */
1797    
1798        case OP_HSPACE:
1799        case OP_VSPACE:
1800        case OP_NOT_HSPACE:
1801        case OP_NOT_VSPACE:
1802      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1803      case OP_DIGIT:      case OP_DIGIT:
1804      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1806  for (;;)
1806      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1807      case OP_WORDCHAR:      case OP_WORDCHAR:
1808      case OP_ANY:      case OP_ANY:
1809        case OP_ALLANY:
1810      branchlength++;      branchlength++;
1811      cc++;      cc++;
1812      break;      break;
1813    
1814      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1815        otherwise \C is coded as OP_ALLANY. */
1816    
1817      case OP_ANYBYTE:      case OP_ANYBYTE:
1818      return -2;      return -2;
1819    
1820      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1821    
1822  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1823      case OP_XCLASS:      case OP_XCLASS:
1824      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1825      /* Fall through */      /* Fall through */
1826  #endif  #endif
1827    
1828      case OP_CLASS:      case OP_CLASS:
1829      case OP_NCLASS:      case OP_NCLASS:
1830      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1831    
1832      switch (*cc)      switch (*cc)
1833        {        {
1834          case OP_CRPLUS:
1835          case OP_CRMINPLUS:
1836        case OP_CRSTAR:        case OP_CRSTAR:
1837        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1838        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1306  for (;;) Line 1841  for (;;)
1841    
1842        case OP_CRRANGE:        case OP_CRRANGE:
1843        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1844        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1845        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1846        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1847        break;        break;
1848    
1849        default:        default:
# Line 1318  for (;;) Line 1853  for (;;)
1853    
1854      /* Anything else is variable length */      /* Anything else is variable length */
1855    
1856      default:      case OP_ANYNL:
1857        case OP_BRAMINZERO:
1858        case OP_BRAPOS:
1859        case OP_BRAPOSZERO:
1860        case OP_BRAZERO:
1861        case OP_CBRAPOS:
1862        case OP_EXTUNI:
1863        case OP_KETRMAX:
1864        case OP_KETRMIN:
1865        case OP_KETRPOS:
1866        case OP_MINPLUS:
1867        case OP_MINPLUSI:
1868        case OP_MINQUERY:
1869        case OP_MINQUERYI:
1870        case OP_MINSTAR:
1871        case OP_MINSTARI:
1872        case OP_MINUPTO:
1873        case OP_MINUPTOI:
1874        case OP_NOTMINPLUS:
1875        case OP_NOTMINPLUSI:
1876        case OP_NOTMINQUERY:
1877        case OP_NOTMINQUERYI:
1878        case OP_NOTMINSTAR:
1879        case OP_NOTMINSTARI:
1880        case OP_NOTMINUPTO:
1881        case OP_NOTMINUPTOI:
1882        case OP_NOTPLUS:
1883        case OP_NOTPLUSI:
1884        case OP_NOTPOSPLUS:
1885        case OP_NOTPOSPLUSI:
1886        case OP_NOTPOSQUERY:
1887        case OP_NOTPOSQUERYI:
1888        case OP_NOTPOSSTAR:
1889        case OP_NOTPOSSTARI:
1890        case OP_NOTPOSUPTO:
1891        case OP_NOTPOSUPTOI:
1892        case OP_NOTQUERY:
1893        case OP_NOTQUERYI:
1894        case OP_NOTSTAR:
1895        case OP_NOTSTARI:
1896        case OP_NOTUPTO:
1897        case OP_NOTUPTOI:
1898        case OP_PLUS:
1899        case OP_PLUSI:
1900        case OP_POSPLUS:
1901        case OP_POSPLUSI:
1902        case OP_POSQUERY:
1903        case OP_POSQUERYI:
1904        case OP_POSSTAR:
1905        case OP_POSSTARI:
1906        case OP_POSUPTO:
1907        case OP_POSUPTOI:
1908        case OP_QUERY:
1909        case OP_QUERYI:
1910        case OP_REF:
1911        case OP_REFI:
1912        case OP_SBRA:
1913        case OP_SBRAPOS:
1914        case OP_SCBRA:
1915        case OP_SCBRAPOS:
1916        case OP_SCOND:
1917        case OP_SKIPZERO:
1918        case OP_STAR:
1919        case OP_STARI:
1920        case OP_TYPEMINPLUS:
1921        case OP_TYPEMINQUERY:
1922        case OP_TYPEMINSTAR:
1923        case OP_TYPEMINUPTO:
1924        case OP_TYPEPLUS:
1925        case OP_TYPEPOSPLUS:
1926        case OP_TYPEPOSQUERY:
1927        case OP_TYPEPOSSTAR:
1928        case OP_TYPEPOSUPTO:
1929        case OP_TYPEQUERY:
1930        case OP_TYPESTAR:
1931        case OP_TYPEUPTO:
1932        case OP_UPTO:
1933        case OP_UPTOI:
1934      return -1;      return -1;
1935    
1936        /* Catch unrecognized opcodes so that when new ones are added they
1937        are not forgotten, as has happened in the past. */
1938    
1939        default:
1940        return -4;
1941      }      }
1942    }    }
1943  /* Control never gets here */  /* Control never gets here */
# Line 1329  for (;;) Line 1947  for (;;)
1947    
1948    
1949  /*************************************************  /*************************************************
1950  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1951  *************************************************/  *************************************************/
1952    
1953  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1954  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1955    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1956    so that it can be called from pcre_study() when finding the minimum matching
1957    length.
1958    
1959  Arguments:  Arguments:
1960    code        points to start of expression    code        points to start of expression
1961    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
1962    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1963    
1964  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
1965  */  */
1966    
1967  static const uschar *  const pcre_uchar *
1968  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
1969  {  {
1970  for (;;)  for (;;)
1971    {    {
1972    register int c = *code;    register int c = *code;
1973    
1974    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1975    
1976    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1357  for (;;) Line 1979  for (;;)
1979    
1980    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1981    
1982      /* Handle recursion */
1983    
1984      else if (c == OP_REVERSE)
1985        {
1986        if (number < 0) return (pcre_uchar *)code;
1987        code += PRIV(OP_lengths)[c];
1988        }
1989    
1990    /* Handle capturing bracket */    /* Handle capturing bracket */
1991    
1992    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1993               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1994      {      {
1995      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1996      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
1997      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1998      }      }
1999    
2000    /* 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
2001    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
2002    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2003      must add in its length. */
2004    
2005    else    else
2006      {      {
# Line 1390  for (;;) Line 2022  for (;;)
2022        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2023        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2024        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2025        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2026            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2027          break;
2028    
2029          case OP_MARK:
2030          case OP_PRUNE_ARG:
2031          case OP_SKIP_ARG:
2032          code += code[1];
2033          break;
2034    
2035          case OP_THEN_ARG:
2036          code += code[1];
2037        break;        break;
2038        }        }
2039    
2040      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2041    
2042      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2043    
2044    /* 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
2045    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
2046    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2047    
2048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2049      if (utf8) switch(c)      if (utf) switch(c)
2050        {        {
2051        case OP_CHAR:        case OP_CHAR:
2052        case OP_CHARNC:        case OP_CHARI:
2053        case OP_EXACT:        case OP_EXACT:
2054          case OP_EXACTI:
2055        case OP_UPTO:        case OP_UPTO:
2056          case OP_UPTOI:
2057        case OP_MINUPTO:        case OP_MINUPTO:
2058          case OP_MINUPTOI:
2059        case OP_POSUPTO:        case OP_POSUPTO:
2060          case OP_POSUPTOI:
2061        case OP_STAR:        case OP_STAR:
2062          case OP_STARI:
2063        case OP_MINSTAR:        case OP_MINSTAR:
2064          case OP_MINSTARI:
2065        case OP_POSSTAR:        case OP_POSSTAR:
2066          case OP_POSSTARI:
2067        case OP_PLUS:        case OP_PLUS:
2068          case OP_PLUSI:
2069        case OP_MINPLUS:        case OP_MINPLUS:
2070          case OP_MINPLUSI:
2071        case OP_POSPLUS:        case OP_POSPLUS:
2072          case OP_POSPLUSI:
2073        case OP_QUERY:        case OP_QUERY:
2074          case OP_QUERYI:
2075        case OP_MINQUERY:        case OP_MINQUERY:
2076          case OP_MINQUERYI:
2077        case OP_POSQUERY:        case OP_POSQUERY:
2078        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2079          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2080        break;        break;
2081        }        }
2082    #else
2083        (void)(utf);  /* Keep compiler happy by referencing function argument */
2084  #endif  #endif
2085      }      }
2086    }    }
# Line 1439  instance of OP_RECURSE. Line 2097  instance of OP_RECURSE.
2097    
2098  Arguments:  Arguments:
2099    code        points to start of expression    code        points to start of expression
2100    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2101    
2102  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
2103  */  */
2104    
2105  static const uschar *  static const pcre_uchar *
2106  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2107  {  {
2108  for (;;)  for (;;)
2109    {    {
# Line 1461  for (;;) Line 2119  for (;;)
2119    
2120    /* 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
2121    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
2122    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2123      must add in its length. */
2124    
2125    else    else
2126      {      {
# Line 1483  for (;;) Line 2142  for (;;)
2142        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2143        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2144        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2145        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2146            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2147          break;
2148    
2149          case OP_MARK:
2150          case OP_PRUNE_ARG:
2151          case OP_SKIP_ARG:
2152          code += code[1];
2153          break;
2154    
2155          case OP_THEN_ARG:
2156          code += code[1];
2157        break;        break;
2158        }        }
2159    
2160      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2161    
2162      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2163    
2164      /* 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
2165      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
2166      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2167    
2168  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2169      if (utf8) switch(c)      if (utf) switch(c)
2170        {        {
2171        case OP_CHAR:        case OP_CHAR:
2172        case OP_CHARNC:        case OP_CHARI:
2173        case OP_EXACT:        case OP_EXACT:
2174          case OP_EXACTI:
2175        case OP_UPTO:        case OP_UPTO:
2176          case OP_UPTOI:
2177        case OP_MINUPTO:        case OP_MINUPTO:
2178          case OP_MINUPTOI:
2179        case OP_POSUPTO:        case OP_POSUPTO:
2180          case OP_POSUPTOI:
2181        case OP_STAR:        case OP_STAR:
2182          case OP_STARI:
2183        case OP_MINSTAR:        case OP_MINSTAR:
2184          case OP_MINSTARI:
2185        case OP_POSSTAR:        case OP_POSSTAR:
2186          case OP_POSSTARI:
2187        case OP_PLUS:        case OP_PLUS:
2188          case OP_PLUSI:
2189        case OP_MINPLUS:        case OP_MINPLUS:
2190          case OP_MINPLUSI:
2191        case OP_POSPLUS:        case OP_POSPLUS:
2192          case OP_POSPLUSI:
2193        case OP_QUERY:        case OP_QUERY:
2194          case OP_QUERYI:
2195        case OP_MINQUERY:        case OP_MINQUERY:
2196          case OP_MINQUERYI:
2197        case OP_POSQUERY:        case OP_POSQUERY:
2198        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2199          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2200        break;        break;
2201        }        }
2202    #else
2203        (void)(utf);  /* Keep compiler happy by referencing function argument */
2204  #endif  #endif
2205      }      }
2206    }    }
# Line 1538  bracket whose current branch will alread Line 2223  bracket whose current branch will alread
2223  Arguments:  Arguments:
2224    code        points to start of search    code        points to start of search
2225    endcode     points to where to stop    endcode     points to where to stop
2226    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2227      cd          contains pointers to tables etc.
2228    
2229  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2230  */  */
2231    
2232  static BOOL  static BOOL
2233  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2234      BOOL utf, compile_data *cd)
2235  {  {
2236  register int c;  register int c;
2237  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2238       code < endcode;       code < endcode;
2239       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2240    {    {
2241    const uschar *ccode;    const pcre_uchar *ccode;
2242    
2243    c = *code;    c = *code;
2244    
# Line 1565  for (code = first_significant_code(code Line 2252  for (code = first_significant_code(code
2252      continue;      continue;
2253      }      }
2254    
2255      /* For a recursion/subroutine call, if its end has been reached, which
2256      implies a backward reference subroutine call, we can scan it. If it's a
2257      forward reference subroutine call, we can't. To detect forward reference
2258      we have to scan up the list that is kept in the workspace. This function is
2259      called only when doing the real compile, not during the pre-compile that
2260      measures the size of the compiled pattern. */
2261    
2262      if (c == OP_RECURSE)
2263        {
2264        const pcre_uchar *scode;
2265        BOOL empty_branch;
2266    
2267        /* Test for forward reference */
2268    
2269        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2270          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2271    
2272        /* Not a forward reference, test for completed backward reference */
2273    
2274        empty_branch = FALSE;
2275        scode = cd->start_code + GET(code, 1);
2276        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2277    
2278        /* Completed backwards reference */
2279    
2280        do
2281          {
2282          if (could_be_empty_branch(scode, endcode, utf, cd))
2283            {
2284            empty_branch = TRUE;
2285            break;
2286            }
2287          scode += GET(scode, 1);
2288          }
2289        while (*scode == OP_ALT);
2290    
2291        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2292        continue;
2293        }
2294    
2295    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2296    
2297    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2298          c == OP_BRAPOSZERO)
2299      {      {
2300      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2301      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2302      c = *code;      c = *code;
2303      continue;      continue;
2304      }      }
2305    
2306    /* For other groups, scan the branches. */    /* A nested group that is already marked as "could be empty" can just be
2307      skipped. */
2308    
2309    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_SBRA  || c == OP_SBRAPOS ||
2310          c == OP_SCBRA || c == OP_SCBRAPOS)
2311      {      {
2312      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
2313      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
2314        continue;
2315        }
2316    
2317      /* Scan a closed bracket */    /* For other groups, scan the branches. */
2318    
2319      empty_branch = FALSE;    if (c == OP_BRA  || c == OP_BRAPOS ||
2320      do        c == OP_CBRA || c == OP_CBRAPOS ||
2321        {        c == OP_ONCE || c == OP_ONCE_NC ||
2322        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        c == OP_COND)
2323          empty_branch = TRUE;      {
2324        BOOL empty_branch;
2325        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2326    
2327        /* If a conditional group has only one branch, there is a second, implied,
2328        empty branch, so just skip over the conditional, because it could be empty.
2329        Otherwise, scan the individual branches of the group. */
2330    
2331        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2332        code += GET(code, 1);        code += GET(code, 1);
2333        else
2334          {
2335          empty_branch = FALSE;
2336          do
2337            {
2338            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2339              empty_branch = TRUE;
2340            code += GET(code, 1);
2341            }
2342          while (*code == OP_ALT);
2343          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2344        }        }
2345      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2346      c = *code;      c = *code;
2347      continue;      continue;
2348      }      }
# Line 1603  for (code = first_significant_code(code Line 2353  for (code = first_significant_code(code
2353      {      {
2354      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2355      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2356      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2357      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"
2358      here. */      here. */
2359    
2360  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2361      case OP_XCLASS:      case OP_XCLASS:
2362      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2363      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1615  for (code = first_significant_code(code Line 2365  for (code = first_significant_code(code
2365    
2366      case OP_CLASS:      case OP_CLASS:
2367      case OP_NCLASS:      case OP_NCLASS:
2368      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2369    
2370  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2371      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2372  #endif  #endif
2373    
# Line 1653  for (code = first_significant_code(code Line 2403  for (code = first_significant_code(code
2403      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2404      case OP_WORDCHAR:      case OP_WORDCHAR:
2405      case OP_ANY:      case OP_ANY:
2406        case OP_ALLANY:
2407      case OP_ANYBYTE:      case OP_ANYBYTE:
2408      case OP_CHAR:      case OP_CHAR:
2409      case OP_CHARNC:      case OP_CHARI:
2410      case OP_NOT:      case OP_NOT:
2411        case OP_NOTI:
2412      case OP_PLUS:      case OP_PLUS:
2413      case OP_MINPLUS:      case OP_MINPLUS:
2414      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1688  for (code = first_significant_code(code Line 2440  for (code = first_significant_code(code
2440      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2441      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2442      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2443      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2444          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2445      break;      break;
2446    
2447      /* End of branch */      /* End of branch */
# Line 1696  for (code = first_significant_code(code Line 2449  for (code = first_significant_code(code
2449      case OP_KET:      case OP_KET:
2450      case OP_KETRMAX:      case OP_KETRMAX:
2451      case OP_KETRMIN:      case OP_KETRMIN:
2452        case OP_KETRPOS:
2453      case OP_ALT:      case OP_ALT:
2454      return TRUE;      return TRUE;
2455    
2456      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2457      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2458    
2459  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2460      case OP_STAR:      case OP_STAR:
2461        case OP_STARI:
2462      case OP_MINSTAR:      case OP_MINSTAR:
2463        case OP_MINSTARI:
2464      case OP_POSSTAR:      case OP_POSSTAR:
2465        case OP_POSSTARI:
2466      case OP_QUERY:      case OP_QUERY:
2467        case OP_QUERYI:
2468      case OP_MINQUERY:      case OP_MINQUERY:
2469        case OP_MINQUERYI:
2470      case OP_POSQUERY:      case OP_POSQUERY:
2471        case OP_POSQUERYI:
2472        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2473        break;
2474    
2475      case OP_UPTO:      case OP_UPTO:
2476        case OP_UPTOI:
2477      case OP_MINUPTO:      case OP_MINUPTO:
2478        case OP_MINUPTOI:
2479      case OP_POSUPTO:      case OP_POSUPTO:
2480      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2481        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2482      break;      break;
2483  #endif  #endif
2484    
2485        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2486        string. */
2487    
2488        case OP_MARK:
2489        case OP_PRUNE_ARG:
2490        case OP_SKIP_ARG:
2491        code += code[1];
2492        break;
2493    
2494        case OP_THEN_ARG:
2495        code += code[1];
2496        break;
2497    
2498        /* None of the remaining opcodes are required to match a character. */
2499    
2500        default:
2501        break;
2502      }      }
2503    }    }
2504    
# Line 1731  return TRUE; Line 2515  return TRUE;
2515  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
2516  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,
2517  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.
2518    This function is called only during the real compile, not during the
2519    pre-compile.
2520    
2521  Arguments:  Arguments:
2522    code        points to start of the recursion    code        points to start of the recursion
2523    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2524    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2525    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2526      cd          pointers to tables etc
2527    
2528  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2529  */  */
2530    
2531  static BOOL  static BOOL
2532  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2533    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2534  {  {
2535  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2536    {    {
2537    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2538        return FALSE;
2539    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2540    }    }
2541  return TRUE;  return TRUE;
# Line 1779  where Perl recognizes it as the POSIX cl Line 2567  where Perl recognizes it as the POSIX cl
2567  "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,
2568  I think.  I think.
2569    
2570    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2571    It seems that the appearance of a nested POSIX class supersedes an apparent
2572    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2573    a digit.
2574    
2575    In Perl, unescaped square brackets may also appear as part of class names. For
2576    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2577    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2578    seem right at all. PCRE does not allow closing square brackets in POSIX class
2579    names.
2580    
2581  Arguments:  Arguments:
2582    ptr      pointer to the initial [    ptr      pointer to the initial [
2583    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1787  Returns:   TRUE or FALSE Line 2586  Returns:   TRUE or FALSE
2586  */  */
2587    
2588  static BOOL  static BOOL
2589  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2590  {  {
2591  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2592  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2593  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2594    {    {
2595    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2596        ptr++;
2597      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2598      else
2599      {      {
2600      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2601        {        {
2602        *endptr = ptr;        *endptr = ptr;
2603        return TRUE;        return TRUE;
2604        }        }
2605        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2606             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2607              ptr[1] == CHAR_EQUALS_SIGN) &&
2608            check_posix_syntax(ptr, endptr))
2609          return FALSE;
2610      }      }
2611    }    }
2612  return FALSE;  return FALSE;
# Line 1824  Returns:     a value representing the na Line 2630  Returns:     a value representing the na
2630  */  */
2631    
2632  static int  static int
2633  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2634  {  {
2635  const char *pn = posix_names;  const char *pn = posix_names;
2636  register int yield = 0;  register int yield = 0;
2637  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2638    {    {
2639    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2640      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2641    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2642    yield++;    yield++;
2643    }    }
# Line 1847  return -1; Line 2653  return -1;
2653  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2654  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2655  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2656  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2657  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2658  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2659  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2660  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2661    OP_END.
2662    
2663  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2664  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1862  value in the reference (which is a group Line 2669  value in the reference (which is a group
2669  Arguments:  Arguments:
2670    group      points to the start of the group    group      points to the start of the group
2671    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2672    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2673    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2674    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
2675    
# Line 1870  Returns:     nothing Line 2677  Returns:     nothing
2677  */  */
2678    
2679  static void  static void
2680  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2681    uschar *save_hwm)    pcre_uchar *save_hwm)
2682  {  {
2683  uschar *ptr = group;  pcre_uchar *ptr = group;
2684    
2685  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2686    {    {
2687    int offset;    int offset;
2688    uschar *hc;    pcre_uchar *hc;
2689    
2690    /* 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
2691    reference. */    reference. */
# Line 1923  Arguments: Line 2730  Arguments:
2730  Returns:         new code pointer  Returns:         new code pointer
2731  */  */
2732    
2733  static uschar *  static pcre_uchar *
2734  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2735  {  {
2736  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2737  *code++ = 255;  *code++ = 255;
2738  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2739  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2740  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2741  }  }
2742    
2743    
# Line 1952  Returns:             nothing Line 2759  Returns:             nothing
2759  */  */
2760    
2761  static void  static void
2762  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2763  {  {
2764  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2765  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2766  }  }
2767    
# Line 1986  get_othercase_range(unsigned int *cptr, Line 2793  get_othercase_range(unsigned int *cptr,
2793  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2794    
2795  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2796    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2797    
2798  if (c > d) return FALSE;  if (c > d) return FALSE;
2799    
# Line 1995  next = othercase + 1; Line 2802  next = othercase + 1;
2802    
2803  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2804    {    {
2805    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2806    next++;    next++;
2807    }    }
2808    
# Line 2004  for (++c; c <= d; c++) Line 2811  for (++c; c <= d; c++)
2811    
2812  return TRUE;  return TRUE;
2813  }  }
2814    
2815    
2816    
2817    /*************************************************
2818    *        Check a character and a property        *
2819    *************************************************/
2820    
2821    /* This function is called by check_auto_possessive() when a property item
2822    is adjacent to a fixed character.
2823    
2824    Arguments:
2825      c            the character
2826      ptype        the property type
2827      pdata        the data for the type
2828      negated      TRUE if it's a negated property (\P or \p{^)
2829    
2830    Returns:       TRUE if auto-possessifying is OK
2831    */
2832    
2833    static BOOL
2834    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2835    {
2836    const ucd_record *prop = GET_UCD(c);
2837    switch(ptype)
2838      {
2839      case PT_LAMP:
2840      return (prop->chartype == ucp_Lu ||
2841              prop->chartype == ucp_Ll ||
2842              prop->chartype == ucp_Lt) == negated;
2843    
2844      case PT_GC:
2845      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2846    
2847      case PT_PC:
2848      return (pdata == prop->chartype) == negated;
2849    
2850      case PT_SC:
2851      return (pdata == prop->script) == negated;
2852    
2853      /* These are specials */
2854    
2855      case PT_ALNUM:
2856      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2857              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2858    
2859      case PT_SPACE:    /* Perl space */
2860      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2861              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2862              == negated;
2863    
2864      case PT_PXSPACE:  /* POSIX space */
2865      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2866              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2867              c == CHAR_FF || c == CHAR_CR)
2868              == negated;
2869    
2870      case PT_WORD:
2871      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2872              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2873              c == CHAR_UNDERSCORE) == negated;
2874      }
2875    return FALSE;
2876    }
2877  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2878    
2879    
# Line 2017  whether the next thing could possibly ma Line 2887  whether the next thing could possibly ma
2887  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2888    
2889  Arguments:  Arguments:
2890    op_code       the repeated op code    previous      pointer to the repeated opcode
2891    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  
2892    ptr           next character in pattern    ptr           next character in pattern
2893    options       options bits    options       options bits
2894    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2029  Returns:        TRUE if possessifying is Line 2897  Returns:        TRUE if possessifying is
2897  */  */
2898    
2899  static BOOL  static BOOL
2900  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2901    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2902  {  {
2903  int next;  pcre_int32 c, next;
2904    int op_code = *previous++;
2905    
2906  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2907    
# Line 2041  if ((options & PCRE_EXTENDED) != 0) Line 2910  if ((options & PCRE_EXTENDED) != 0)
2910    for (;;)    for (;;)
2911      {      {
2912      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2913      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2914        {        {
2915        while (*(++ptr) != 0)        ptr++;
2916          while (*ptr != 0)
2917            {
2918          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2919            ptr++;
2920    #ifdef SUPPORT_UTF
2921            if (utf) FORWARDCHAR(ptr);
2922    #endif
2923            }
2924        }        }
2925      else break;      else break;
2926      }      }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 2929  if ((options & PCRE_EXTENDED) != 0)
2929  /* 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
2930  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2931    
2932  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2933    {    {
2934    int temperrorcode = 0;    int temperrorcode = 0;
2935    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2936    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2937    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2938    }    }
2939    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2940    {    {
2941  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2942    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2943  #endif  #endif
2944    next = *ptr++;    next = *ptr++;
2945    }    }
   
2946  else return FALSE;  else return FALSE;
2947    
2948  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2078  if ((options & PCRE_EXTENDED) != 0) Line 2952  if ((options & PCRE_EXTENDED) != 0)
2952    for (;;)    for (;;)
2953      {      {
2954      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2955      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2956        {        {
2957        while (*(++ptr) != 0)        ptr++;
2958          while (*ptr != 0)
2959            {
2960          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2961            ptr++;
2962    #ifdef SUPPORT_UTF
2963            if (utf) FORWARDCHAR(ptr);
2964    #endif
2965            }
2966        }        }
2967      else break;      else break;
2968      }      }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 2970  if ((options & PCRE_EXTENDED) != 0)
2970    
2971  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2972    
2973  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2974    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2975        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. */  
   
2976    
2977  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2978    the next item is a character. */
2979    
2980  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2981    {    {
2982    case OP_CHAR:    case OP_CHAR:
2983  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2984    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2985    #else
2986      c = *previous;
2987  #endif  #endif
2988    return item != next;    return c != next;
2989    
2990    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2991    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2992    high-valued characters. */    high-valued characters. */
2993    
2994    case OP_CHARNC:    case OP_CHARI:
2995  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2996    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2997  #endif  #else
2998    if (item == next) return FALSE;    c = *previous;
2999  #ifdef SUPPORT_UTF8  #endif
3000    if (utf8)    if (c == next) return FALSE;
3001    #ifdef SUPPORT_UTF
3002      if (utf)
3003      {      {
3004      unsigned int othercase;      unsigned int othercase;
3005      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3006  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3007      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
3008  #else  #else
3009      othercase = NOTACHAR;      othercase = NOTACHAR;
3010  #endif  #endif
3011      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3012      }      }
3013    else    else
3014  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3015    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
3016    
3017    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3018      opcodes are not used for multi-byte characters, because they are coded using
3019      an XCLASS instead. */
3020    
3021    case OP_NOT:    case OP_NOT:
3022    if (item == next) return TRUE;    return (c = *previous) == next;
3023    if ((options & PCRE_CASELESS) == 0) return FALSE;  
3024  #ifdef SUPPORT_UTF8    case OP_NOTI:
3025    if (utf8)    if ((c = *previous) == next) return TRUE;
3026    #ifdef SUPPORT_UTF
3027      if (utf)
3028      {      {
3029      unsigned int othercase;      unsigned int othercase;
3030      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3031  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3032      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
3033  #else  #else
3034      othercase = NOTACHAR;      othercase = NOTACHAR;
3035  #endif  #endif
3036      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3037      }      }
3038    else    else
3039  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3040    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
3041    
3042      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3043      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3044    
3045    case OP_DIGIT:    case OP_DIGIT:
3046    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2195  if (next >= 0) switch(op_code) Line 3083  if (next >= 0) switch(op_code)
3083      case 0x202f:      case 0x202f:
3084      case 0x205f:      case 0x205f:
3085      case 0x3000:      case 0x3000:
3086      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3087      default:      default:
3088      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3089      }      }
3090    
3091      case OP_ANYNL:
3092    case OP_VSPACE:    case OP_VSPACE:
3093    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3094    switch(next)    switch(next)
# Line 2211  if (next >= 0) switch(op_code) Line 3100  if (next >= 0) switch(op_code)
3100      case 0x85:      case 0x85:
3101      case 0x2028:      case 0x2028:
3102      case 0x2029:      case 0x2029:
3103      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3104      default:      default:
3105      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3106      }      }
3107    
3108    #ifdef SUPPORT_UCP
3109      case OP_PROP:
3110      return check_char_prop(next, previous[0], previous[1], FALSE);
3111    
3112      case OP_NOTPROP:
3113      return check_char_prop(next, previous[0], previous[1], TRUE);
3114    #endif
3115    
3116    default:    default:
3117    return FALSE;    return FALSE;
3118    }    }
3119    
3120    
3121  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3122    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3123    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3124    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3125    replaced by OP_PROP codes when PCRE_UCP is set. */
3126    
3127  switch(op_code)  switch(op_code)
3128    {    {
3129    case OP_CHAR:    case OP_CHAR:
3130    case OP_CHARNC:    case OP_CHARI:
3131  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3132    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3133    #else
3134      c = *previous;
3135  #endif  #endif
3136    switch(-next)    switch(-next)
3137      {      {
3138      case ESC_d:      case ESC_d:
3139      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3140    
3141      case ESC_D:      case ESC_D:
3142      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3143    
3144      case ESC_s:      case ESC_s:
3145      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3146    
3147      case ESC_S:      case ESC_S:
3148      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3149    
3150      case ESC_w:      case ESC_w:
3151      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3152    
3153      case ESC_W:      case ESC_W:
3154      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3155    
3156      case ESC_h:      case ESC_h:
3157      case ESC_H:      case ESC_H:
3158      switch(item)      switch(c)
3159        {        {
3160        case 0x09:        case 0x09:
3161        case 0x20:        case 0x20:
# Line 2280  switch(op_code) Line 3183  switch(op_code)
3183    
3184      case ESC_v:      case ESC_v:
3185      case ESC_V:      case ESC_V:
3186      switch(item)      switch(c)
3187        {        {
3188        case 0x0a:        case 0x0a:
3189        case 0x0b:        case 0x0b:
# Line 2294  switch(op_code) Line 3197  switch(op_code)
3197        return -next == ESC_v;        return -next == ESC_v;
3198        }        }
3199    
3200        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3201        their substitutions and process them. The result will always be either
3202        -ESC_p or -ESC_P. Then fall through to process those values. */
3203    
3204    #ifdef SUPPORT_UCP
3205        case ESC_du:
3206        case ESC_DU:
3207        case ESC_wu:
3208        case ESC_WU:
3209        case ESC_su:
3210        case ESC_SU:
3211          {
3212          int temperrorcode = 0;
3213          ptr = substitutes[-next - ESC_DU];
3214          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3215          if (temperrorcode != 0) return FALSE;
3216          ptr++;    /* For compatibility */
3217          }
3218        /* Fall through */
3219    
3220        case ESC_p:
3221        case ESC_P:
3222          {
3223          int ptype, pdata, errorcodeptr;
3224          BOOL negated;
3225    
3226          ptr--;      /* Make ptr point at the p or P */
3227          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3228          if (ptype < 0) return FALSE;
3229          ptr++;      /* Point past the final curly ket */
3230    
3231          /* If the property item is optional, we have to give up. (When generated
3232          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3233          to the original \d etc. At this point, ptr will point to a zero byte. */
3234    
3235          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3236            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3237              return FALSE;
3238    
3239          /* Do the property check. */
3240    
3241          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3242          }
3243    #endif
3244    
3245      default:      default:
3246      return FALSE;      return FALSE;
3247      }      }
3248    
3249      /* In principle, support for Unicode properties should be integrated here as
3250      well. It means re-organizing the above code so as to get hold of the property
3251      values before switching on the op-code. However, I wonder how many patterns
3252      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3253      these op-codes are never generated.) */
3254    
3255    case OP_DIGIT:    case OP_DIGIT:
3256    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3257           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3258    
3259    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3260    return next == -ESC_d;    return next == -ESC_d;
3261    
3262    case OP_WHITESPACE:    case OP_WHITESPACE:
3263    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3264    
3265    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3266    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3267    
3268    case OP_HSPACE:    case OP_HSPACE:
3269    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3270             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3271    
3272    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3273    return next == -ESC_h;    return next == -ESC_h;
3274    
3275    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3276      case OP_ANYNL:
3277    case OP_VSPACE:    case OP_VSPACE:
3278    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3279    
3280    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3281    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3282    
3283    case OP_WORDCHAR:    case OP_WORDCHAR:
3284    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3285             next == -ESC_v || next == -ESC_R;
3286    
3287    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3288    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2354  Arguments: Line 3311  Arguments:
3311    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3312    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3313    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3314    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3315    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3316    bcptr          points to current branch chain    bcptr          points to current branch chain
3317      cond_depth     conditional nesting depth
3318    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3319    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3320                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2366  Returns:         TRUE on success Line 3324  Returns:         TRUE on success
3324  */  */
3325    
3326  static BOOL  static BOOL
3327  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3328    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3329      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3330    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3331  {  {
3332  int repeat_type, op_type;  int repeat_type, op_type;
3333  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3334  int bravalue = 0;  int bravalue = 0;
3335  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3336  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3337  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3338  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3339  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3340  int after_manual_callout = 0;  int after_manual_callout = 0;
3341  int length_prevgroup = 0;  int length_prevgroup = 0;
3342  register int c;  register int c;
3343  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3344  uschar *last_code = code;  pcre_uchar *last_code = code;
3345  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3346  uschar *tempcode;  pcre_uchar *tempcode;
3347  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3348  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3349  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3350  const uschar *tempptr;  const pcre_uchar *tempptr;
3351  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3352  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3353  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3354  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3355    pcre_uint8 classbits[32];
3356  #ifdef SUPPORT_UTF8  
3357  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3358  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3359  uschar *class_utf8data;  dynamically as we process the pattern. */
3360  uschar *class_utf8data_base;  
3361  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3362    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3363    BOOL utf = (options & PCRE_UTF8) != 0;
3364    pcre_uchar utf_chars[6];
3365  #else  #else
3366  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3367  uschar *utf8_char = NULL;  #endif
3368    
3369    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3370    
3371    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3372    BOOL xclass;
3373    pcre_uchar *class_uchardata;
3374    pcre_uchar *class_uchardata_base;
3375  #endif  #endif
3376    
3377  #ifdef DEBUG  #ifdef PCRE_DEBUG
3378  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3379  #endif  #endif
3380    
# Line 2416  greedy_non_default = greedy_default ^ 1; Line 3385  greedy_non_default = greedy_default ^ 1;
3385    
3386  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3387  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
3388  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
3389  find one.  find one.
3390    
3391  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
3392  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
3393  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3394  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3395    
3396  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3397    
3398  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3399  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
3400  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
3401  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3402    value. This is used only for ASCII characters. */
3403    
3404  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3405    
3406  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3407    
# Line 2443  for (;; ptr++) Line 3413  for (;; ptr++)
3413    BOOL is_quantifier;    BOOL is_quantifier;
3414    BOOL is_recurse;    BOOL is_recurse;
3415    BOOL reset_bracount;    BOOL reset_bracount;
3416    int class_charcount;    int class_has_8bitchar;
3417      int class_single_char;
3418    int class_lastchar;    int class_lastchar;
3419    int newoptions;    int newoptions;
3420    int recno;    int recno;
3421    int refsign;    int refsign;
3422    int skipbytes;    int skipbytes;
3423    int subreqbyte;    int subreqchar;
3424    int subfirstbyte;    int subfirstchar;
3425    int terminator;    int terminator;
3426    int mclength;    int mclength;
3427    uschar mcbuffer[8];    int tempbracount;
3428      pcre_uchar mcbuffer[8];
3429    
3430    /* Get next byte in the pattern */    /* Get next character in the pattern */
3431    
3432    c = *ptr;    c = *ptr;
3433    
3434      /* If we are at the end of a nested substitution, revert to the outer level
3435      string. Nesting only happens one level deep. */
3436    
3437      if (c == 0 && nestptr != NULL)
3438        {
3439        ptr = nestptr;
3440        nestptr = NULL;
3441        c = *ptr;
3442        }
3443    
3444    /* 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
3445    previous cycle of this loop. */    previous cycle of this loop. */
3446    
3447    if (lengthptr != NULL)    if (lengthptr != NULL)
3448      {      {
3449  #ifdef DEBUG  #ifdef PCRE_DEBUG
3450      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3451  #endif  #endif
3452      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3453        {        {
3454        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3455        goto FAILED;        goto FAILED;
# Line 2489  for (;; ptr++) Line 3471  for (;; ptr++)
3471        goto FAILED;        goto FAILED;
3472        }        }
3473    
3474      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3475      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3476          (int)(code - last_code), c, c));
3477    
3478      /* 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
3479      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 2500  for (;; ptr++) Line 3483  for (;; ptr++)
3483        {        {
3484        if (previous > orig_code)        if (previous > orig_code)
3485          {          {
3486          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3487          code -= previous - orig_code;          code -= previous - orig_code;
3488          previous = orig_code;          previous = orig_code;
3489          }          }
# Line 2516  for (;; ptr++) Line 3499  for (;; ptr++)
3499    /* 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
3500    reference list. */    reference list. */
3501    
3502    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3503      {      {
3504      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3505      goto FAILED;      goto FAILED;
# Line 2526  for (;; ptr++) Line 3509  for (;; ptr++)
3509    
3510    if (inescq && c != 0)    if (inescq && c != 0)
3511      {      {
3512      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3513        {        {
3514        inescq = FALSE;        inescq = FALSE;
3515        ptr++;        ptr++;
# Line 2552  for (;; ptr++) Line 3535  for (;; ptr++)
3535    /* 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
3536    a quantifier. */    a quantifier. */
3537    
3538    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3539      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3540        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3541    
3542    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3543         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2563  for (;; ptr++) Line 3547  for (;; ptr++)
3547      previous_callout = NULL;      previous_callout = NULL;
3548      }      }
3549    
3550    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3551    
3552    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3553      {      {
3554      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3555      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3556        {        {
3557        while (*(++ptr) != 0)        ptr++;
3558          while (*ptr != 0)
3559          {          {
3560          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3561            ptr++;
3562    #ifdef SUPPORT_UTF
3563            if (utf) FORWARDCHAR(ptr);
3564    #endif
3565          }          }
3566        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3567    
# Line 2593  for (;; ptr++) Line 3582  for (;; ptr++)
3582      {      {
3583      /* ===================================================================*/      /* ===================================================================*/
3584      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3585      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3586      case ')':      case CHAR_RIGHT_PARENTHESIS:
3587      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3588      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3589      *codeptr = code;      *codeptr = code;
3590      *ptrptr = ptr;      *ptrptr = ptr;
3591      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2606  for (;; ptr++) Line 3595  for (;; ptr++)
3595          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3596          goto FAILED;          goto FAILED;
3597          }          }
3598        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3599        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3600        }        }
3601      return TRUE;      return TRUE;
# Line 2616  for (;; ptr++) Line 3605  for (;; ptr++)
3605      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3606      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3607    
3608      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3609        previous = NULL;
3610      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3611        {        {
3612        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3613          *code++ = OP_CIRCM;
3614        }        }
3615      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3616      break;      break;
3617    
3618      case '$':      case CHAR_DOLLAR_SIGN:
3619      previous = NULL;      previous = NULL;
3620      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3621      break;      break;
3622    
3623      /* 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
3624      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3625    
3626      case '.':      case CHAR_DOT:
3627      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3628      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3629      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3630      previous = code;      previous = code;
3631      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3632      break;      break;
3633    
3634    
# Line 2653  for (;; ptr++) Line 3643  for (;; ptr++)
3643      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3644      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3645      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3646    
3647      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3648        default (Perl) mode, it is treated as a data character. */
3649    
3650        case CHAR_RIGHT_SQUARE_BRACKET:
3651        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3652          {
3653          *errorcodeptr = ERR64;
3654          goto FAILED;
3655          }
3656        goto NORMAL_CHAR;
3657    
3658        case CHAR_LEFT_SQUARE_BRACKET:
3659      previous = code;      previous = code;
3660    
3661      /* 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
3662      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. */
3663    
3664      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3665             ptr[1] == CHAR_EQUALS_SIGN) &&
3666          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3667        {        {
3668        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3669        goto FAILED;        goto FAILED;
3670        }        }
3671    
# Line 2676  for (;; ptr++) Line 3677  for (;; ptr++)
3677      for (;;)      for (;;)
3678        {        {
3679        c = *(++ptr);        c = *(++ptr);
3680        if (c == '\\')        if (c == CHAR_BACKSLASH)
3681          {          {
3682          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3683            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3684              else break;          else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
3685              ptr += 3;
3686            else
3687              break;
3688          }          }
3689        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3690          negate_class = TRUE;          negate_class = TRUE;
3691        else break;        else break;
3692        }        }
3693    
3694        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3695        an initial ']' is taken as a data character -- the code below handles
3696        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3697        [^] must match any character, so generate OP_ALLANY. */
3698    
3699        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3700            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3701          {
3702          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3703          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3704          zerofirstchar = firstchar;
3705          break;
3706          }
3707    
3708      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
3709      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3710      correctly (they are all included in the class). */      correctly (they are all included in the class). */
3711    
3712      should_flip_negation = FALSE;      should_flip_negation = FALSE;
3713    
3714      /* 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.
3715      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
3716      valued UTF-8 characters, we don't yet do any optimization. */      < 256 character. class_single_char will be 1, if the class only contains
3717        a single character. */
3718    
3719      class_charcount = 0;      class_has_8bitchar = 0;
3720        class_single_char = 0;
3721      class_lastchar = -1;      class_lastchar = -1;
3722    
3723      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
# Line 2705  for (;; ptr++) Line 3725  for (;; ptr++)
3725      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3726      */      */
3727    
3728      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3729    
3730  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3731      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3732      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3733      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3734  #endif  #endif
3735    
3736      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2719  for (;; ptr++) Line 3739  for (;; ptr++)
3739    
3740      if (c != 0) do      if (c != 0) do
3741        {        {
3742        const uschar *oldptr;        const pcre_uchar *oldptr;
3743    
3744  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3745        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3746          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3747          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3748          }          }
3749    #endif
3750    
3751        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3752          /* In the pre-compile phase, accumulate the length of any extra
3753        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3754        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
3755        (which is on the stack). */        (which is on the stack). */
3756    
3757        if (lengthptr != NULL)        if (lengthptr != NULL)
3758          {          {
3759          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_uchardata - class_uchardata_base;
3760          class_utf8data = class_utf8data_base;          class_uchardata = class_uchardata_base;
3761          }          }
   
3762  #endif  #endif
3763    
3764        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3765    
3766        if (inescq)        if (inescq)
3767          {          {
3768          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3769            {            {
3770            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3771            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2759  for (;; ptr++) Line 3780  for (;; ptr++)
3780        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3781        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3782    
3783        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3784            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3785            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3786          {          {
3787          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3788          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3789          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3790          uschar pbits[32];          pcre_uint8 pbits[32];
3791    
3792          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3793            {            {
3794            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3795            goto FAILED;            goto FAILED;
3796            }            }
3797    
3798          ptr += 2;          ptr += 2;
3799          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3800            {            {
3801            local_negate = TRUE;            local_negate = TRUE;
3802            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3803            ptr++;            ptr++;
3804            }            }
3805    
3806          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3807          if (posix_class < 0)          if (posix_class < 0)
3808            {            {
3809            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2796  for (;; ptr++) Line 3817  for (;; ptr++)
3817          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3818            posix_class = 0;            posix_class = 0;
3819    
3820          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3821          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3822          subtract bits that may be in the main map already. At the end we or the  
3823          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3824            if ((options & PCRE_UCP) != 0)
3825              {
3826              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3827              if (posix_substitutes[pc] != NULL)
3828                {
3829                nestptr = tempptr + 1;
3830                ptr = posix_substitutes[pc] - 1;
3831                continue;
3832                }
3833              }
3834    #endif
3835            /* In the non-UCP case, we build the bit map for the POSIX class in a
3836            chunk of local store because we may be adding and subtracting from it,
3837            and we don't want to subtract bits that may be in the main map already.
3838            At the end we or the result into the bit map that is being built. */
3839    
3840          posix_class *= 3;          posix_class *= 3;
3841    
3842          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3843    
3844          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3845            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3846    
3847          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3848    
# Line 2837  for (;; ptr++) Line 3873  for (;; ptr++)
3873            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3874    
3875          ptr = tempptr + 1;          ptr = tempptr + 1;
3876          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3877            class_has_8bitchar = 1;
3878            /* Every class contains at least two characters. */
3879            class_single_char = 2;
3880          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3881          }          }
3882    
3883        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3884        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3885        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3886        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so
3887        to 'or' into the one we are building. We assume they have more than one        speculatively set both class_has_8bitchar class_single_char bigger
3888        character in them, so set class_charcount bigger than one. */        than one. Unrecognized escapes fall through and are either treated
3889          as literal characters (by default), or are faulted if
3890          PCRE_EXTRA is set. */
3891    
3892        if (c == '\\')        if (c == CHAR_BACKSLASH)
3893          {          {
3894          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3895          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3896    
3897          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3898          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3899            {            {
3900            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3901              {              {
3902              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3903              }              }
# Line 2869  for (;; ptr++) Line 3908  for (;; ptr++)
3908    
3909          if (c < 0)          if (c < 0)
3910            {            {
3911            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3912            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
3913              class_has_8bitchar++;
3914            /* Save time by not doing this in the pre-compile phase. */            /* Every class contains at least two characters. */
3915              class_single_char += 2;
3916    
3917            if (lengthptr == NULL) switch (-c)            switch (-c)
3918              {              {
3919    #ifdef SUPPORT_UCP
3920                case ESC_du:     /* These are the values given for \d etc */
3921                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3922                case ESC_wu:     /* escape sequence with an appropriate \p */
3923                case ESC_WU:     /* or \P to test Unicode properties instead */
3924                case ESC_su:     /* of the default ASCII testing. */
3925                case ESC_SU:
3926                nestptr = ptr;
3927                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3928                class_has_8bitchar--;                /* Undo! */
3929                continue;
3930    #endif
3931              case ESC_d:              case ESC_d:
3932              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3933              continue;              continue;
# Line 2894  for (;; ptr++) Line 3946  for (;; ptr++)
3946              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3947              continue;              continue;
3948    
3949                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3950                if it was previously set by something earlier in the character
3951                class. */
3952    
3953              case ESC_s:              case ESC_s:
3954              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3955              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3956                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3957              continue;              continue;
3958    
3959              case ESC_S:              case ESC_S:
# Line 2905  for (;; ptr++) Line 3962  for (;; ptr++)
3962              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3963              continue;              continue;
3964    
3965              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3966              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3967              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3968              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
3969  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3970              if (utf8)              if (utf)
3971                {                {
3972                class_utf8 = TRUE;                xclass = TRUE;
3973                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3974                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
3975                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3976                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
3977                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
3978                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
3979                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200A, class_uchardata);
3980                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3981                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
3982                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3983                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
3984                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3985                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
3986                }                }
3987  #endif  #endif
3988              continue;              continue;
             }  
3989    
3990            if (-c == ESC_H)              case ESC_H:
             {  
3991              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3992                {                {
3993                int x = 0xff;                int x = 0xff;
# Line 2959  for (;; ptr++) Line 4001  for (;; ptr++)
4001                classbits[c] |= x;                classbits[c] |= x;
4002                }                }
4003    
4004  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4005              if (utf8)              if (utf)
4006                {                {
4007                class_utf8 = TRUE;                xclass = TRUE;
4008                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4009                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4010                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
4011                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4012                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
4013                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
4014                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4015                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
4016                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
4017                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4018                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200B, class_uchardata);
4019                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
4020                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4021                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
4022                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
4023                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4024                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
4025                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
4026                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4027                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
4028                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4029                }                }
4030  #endif  #endif
4031              continue;              continue;
             }  
4032    
4033            if (-c == ESC_v)              case ESC_v:
             {  
4034              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
4035              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
4036              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
4037              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
4038              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
4039  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4040              if (utf8)              if (utf)
4041                {                {
4042                class_utf8 = TRUE;                xclass = TRUE;
4043                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4044                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
4045                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4046                }                }
4047  #endif  #endif
4048              continue;              continue;
             }  
4049    
4050            if (-c == ESC_V)              case ESC_V:
             {  
4051              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
4052                {                {
4053                int x = 0xff;                int x = 0xff;
# Line 3026  for (;; ptr++) Line 4064  for (;; ptr++)
4064                classbits[c] |= x;                classbits[c] |= x;
4065                }                }
4066    
4067  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4068              if (utf8)              if (utf)
4069                {                {
4070                class_utf8 = TRUE;                xclass = TRUE;
4071                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4072                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4073                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
4074                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4075                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);               &nb