/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

revision 602 by ph10, Wed May 25 08:29:03 2011 UTC revision 1477 by ph10, Wed May 21 17:53:49 2014 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) 1997-2011 University of Cambridge             Copyright (c) 1997-2014 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 PCRE_DEBUG is defined, we need the pcre_printint() function, which is  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. */  library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 73  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 88  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123  #define COMPILE_WORK_SIZE (4096)  #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
128  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138    /* Repeated character flags. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  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
# Line 219  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 231  static const char posix_names[] = Line 286  static const char posix_names[] =
286    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
288    
289  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    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 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* 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
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 261  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
329    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
332    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXsp[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXwd[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXwd[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *substitutes[] = {
348      string_PNd,           /* \D */
349      string_pNd,           /* \d */
350      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352      string_PXwd,          /* \W */
353      string_pXwd           /* \w */
354  };  };
355    
356  static const uschar *posix_substitutes[] = {  /* The POSIX class substitutes must be in the order of the POSIX class names,
357    (uschar *)"\\p{L}",     /* alpha */  defined above, and there are both positive and negative cases. NULL means no
358    (uschar *)"\\p{Ll}",    /* lower */  general substitute of a Unicode property escape (\p or \P). However, for some
359    (uschar *)"\\p{Lu}",    /* upper */  POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    (uschar *)"\\p{Xan}",   /* alnum */  directly. */
361    NULL,                   /* ascii */  
362    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pL[] =   {
363    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    NULL,                   /* graph */  static const pcre_uchar string_pLl[] =  {
366    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367    NULL,                   /* punct */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_pLu[] =  {
369    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370    NULL,                   /* xdigit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pXan[] = {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_h[] =    {
375      CHAR_BACKSLASH, CHAR_h, '\0' };
376    static const pcre_uchar string_pXps[] = {
377      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379    static const pcre_uchar string_PL[] =   {
380      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382    static const pcre_uchar string_PLl[] =  {
383      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PLu[] =  {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PXan[] = {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_H[] =    {
392      CHAR_BACKSLASH, CHAR_H, '\0' };
393    static const pcre_uchar string_PXps[] = {
394      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396    
397    static const pcre_uchar *posix_substitutes[] = {
398      string_pL,            /* alpha */
399      string_pLl,           /* lower */
400      string_pLu,           /* upper */
401      string_pXan,          /* alnum */
402      NULL,                 /* ascii */
403      string_h,             /* blank */
404      NULL,                 /* cntrl */
405      string_pNd,           /* digit */
406      NULL,                 /* graph */
407      NULL,                 /* print */
408      NULL,                 /* punct */
409      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411      NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
414    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
415    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
416    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
417    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
418    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
419    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
420    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
421    NULL,                   /* ^graph */    NULL,                 /* ^graph */
422    NULL,                   /* ^print */    NULL,                 /* ^print */
423    NULL,                   /* ^punct */    NULL,                 /* ^punct */
424    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429  #endif  #endif
430    
431  #define STRING(a)  # a  #define STRING(a)  # a
# Line 365  static const char error_texts[] = Line 484  static const char error_texts[] =
484    /* 30 */    /* 30 */
485    "unknown POSIX class name\0"    "unknown POSIX class name\0"
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 388  static const char error_texts[] = Line 507  static const char error_texts[] =
507    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508    /* 50 */    /* 50 */
509    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
510    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
512    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
513    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
514    /* 55 */    /* 55 */
515    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
516    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
517    "\\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"
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 407  static const char error_texts[] = Line 526  static const char error_texts[] =
526    /* 65 */    /* 65 */
527    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
528    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
529    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with Unicode property support\0"
530    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
531      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532      /* 70 */
533      "internal error: unknown opcode in find_fixedlength()\0"
534      "\\N is not supported in a class\0"
535      "too many forward references\0"
536      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537      "invalid UTF-16 string\0"
538      /* 75 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552    ;    ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 427  For convenience, we use the same bit def Line 567  For convenience, we use the same bit def
567    
568  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
569    
570    /* Using a simple comparison for decimal numbers rather than a memory read
571    is much faster, and the resulting code is simpler (the compiler turns it
572    into a subtraction and unsigned comparison). */
573    
574    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
575    
576  #ifndef EBCDIC  #ifndef EBCDIC
577    
578  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
579  UTF-8 mode. */  UTF-8 mode. */
580    
581  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
582    {    {
583    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
584    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 471  static const unsigned char digitab[] = Line 617  static const unsigned char digitab[] =
617    
618  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
619    
620  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
621    {    {
622    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
623    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 506  static const unsigned char digitab[] = Line 652  static const unsigned char digitab[] =
652    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
653    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
654    
655  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
656    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
657    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
658    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 542  static const unsigned char ebcdic_charta Line 688  static const unsigned char ebcdic_charta
688  #endif  #endif
689    
690    
691  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
692    between adjacent character-type opcodes. The left-hand (repeated) opcode is
693    used to select the row, and the right-hand opcode is use to select the column.
694    A value of 1 means that auto-possessification is OK. For example, the second
695    value in the first row means that \D+\d can be turned into \D++\d.
696    
697    The Unicode property types (\P and \p) have to be present to fill out the table
698    because of what their opcode values are, but the table values should always be
699    zero because property types are handled separately in the code. The last four
700    columns apply to items that cannot be repeated, so there is no need to have
701    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
702    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
703    
704    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
705    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
706    
707    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
708    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
709      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
710      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
711      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
712      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
713      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
714      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
720      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
724      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
725      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
726    };
727    
728  static BOOL  
729    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  /* This table is used to check whether auto-possessification is possible
730      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
731    left-hand (repeated) opcode is used to select the row, and the right-hand
732    opcode is used to select the column. The values are as follows:
733    
734      0   Always return FALSE (never auto-possessify)
735      1   Character groups are distinct (possessify if both are OP_PROP)
736      2   Check character categories in the same group (general or particular)
737      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
738    
739      4   Check left general category vs right particular category
740      5   Check right general category vs left particular category
741    
742      6   Left alphanum vs right general category
743      7   Left space vs right general category
744      8   Left word vs right general category
745    
746      9   Right alphanum vs left general category
747     10   Right space vs left general category
748     11   Right word vs left general category
749    
750     12   Left alphanum vs right particular category
751     13   Left space vs right particular category
752     14   Left word vs right particular category
753    
754     15   Right alphanum vs left particular category
755     16   Right space vs left particular category
756     17   Right word vs left particular category
757    */
758    
759    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
760    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
761      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
762      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
763      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
764      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
765      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
766      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
767      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
769      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
770      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
772    };
773    
774    /* This table is used to check whether auto-possessification is possible
775    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
776    specifies a general category and the other specifies a particular category. The
777    row is selected by the general category and the column by the particular
778    category. The value is 1 if the particular category is not part of the general
779    category. */
780    
781    static const pcre_uint8 catposstab[7][30] = {
782    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
783      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
784      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
790    };
791    
792    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
793    a general or particular category. The properties in each row are those
794    that apply to the character set in question. Duplication means that a little
795    unnecessary work is done when checking, but this keeps things much simpler
796    because they can all use the same code. For more details see the comment where
797    this table is used.
798    
799    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
800    "space", but from Perl 5.18 it's included, so both categories are treated the
801    same here. */
802    
803    static const pcre_uint8 posspropstab[3][4] = {
804      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
805      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
806      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807    };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
# Line 569  find_error_text(int n) Line 885  find_error_text(int n)
885  const char *s = error_texts;  const char *s = error_texts;
886  for (; n > 0; n--)  for (; n > 0; n--)
887    {    {
888    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
889    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
890    }    }
891  return s;  return s;
892  }  }
893    
894    
895    
896    /*************************************************
897    *           Expand the workspace                 *
898    *************************************************/
899    
900    /* This function is called during the second compiling phase, if the number of
901    forward references fills the existing workspace, which is originally a block on
902    the stack. A larger block is obtained from malloc() unless the ultimate limit
903    has been reached or the increase will be rather small.
904    
905    Argument: pointer to the compile data block
906    Returns:  0 if all went well, else an error number
907    */
908    
909    static int
910    expand_workspace(compile_data *cd)
911    {
912    pcre_uchar *newspace;
913    int newsize = cd->workspace_size * 2;
914    
915    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
916    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
917        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
918     return ERR72;
919    
920    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
921    if (newspace == NULL) return ERR21;
922    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
923    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
924    if (cd->workspace_size > COMPILE_WORK_SIZE)
925      (PUBL(free))((void *)cd->start_workspace);
926    cd->start_workspace = newspace;
927    cd->workspace_size = newsize;
928    return 0;
929    }
930    
931    
932    
933    /*************************************************
934    *            Check for counted repeat            *
935    *************************************************/
936    
937    /* This function is called when a '{' is encountered in a place where it might
938    start a quantifier. It looks ahead to see if it really is a quantifier or not.
939    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
940    where the ddds are digits.
941    
942    Arguments:
943      p         pointer to the first char after '{'
944    
945    Returns:    TRUE or FALSE
946    */
947    
948    static BOOL
949    is_counted_repeat(const pcre_uchar *p)
950    {
951    if (!IS_DIGIT(*p)) return FALSE;
952    p++;
953    while (IS_DIGIT(*p)) p++;
954    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
955    
956    if (*p++ != CHAR_COMMA) return FALSE;
957    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
958    
959    if (!IS_DIGIT(*p)) return FALSE;
960    p++;
961    while (IS_DIGIT(*p)) p++;
962    
963    return (*p == CHAR_RIGHT_CURLY_BRACKET);
964    }
965    
966    
967    
968  /*************************************************  /*************************************************
969  *            Handle escapes                      *  *            Handle escapes                      *
970  *************************************************/  *************************************************/
971    
972  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
973  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
974  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
975  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
976  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
977  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981      chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
985    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
986    
987  Returns:         zero or positive => a data character  Returns:         zero => a data character
988                   negative => a special escape sequence                   positive => a special escape sequence
989                     negative => a back reference
990                   on error, errorcodeptr is set                   on error, errorcodeptr is set
991  */  */
992    
993  static int  static int
994  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
995    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
996  {  {
997  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
998  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
999  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
1000    pcre_uint32 c;
1001    int escape = 0;
1002    int i;
1003    
1004  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1005  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1006    
1007  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1008    
1009  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1010    
1011  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1012  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.
1013  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1014    
1015  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1016  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
1017  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1018    else if ((i = escapes[c - CHAR_0]) != 0)
1019      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1020    
1021  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1022  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
1023  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1024    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1025  #endif  #endif
1026    
1027  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1028    
1029  else  else
1030    {    {
1031    const uschar *oldptr;    const pcre_uchar *oldptr;
1032    BOOL braced, negated;    BOOL braced, negated, overflow;
1033      int s;
1034    
1035    switch (c)    switch (c)
1036      {      {
# Line 642  else Line 1039  else
1039    
1040      case CHAR_l:      case CHAR_l:
1041      case CHAR_L:      case CHAR_L:
1042        *errorcodeptr = ERR37;
1043        break;
1044    
1045      case CHAR_u:      case CHAR_u:
1046        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1047          {
1048          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1049          Otherwise it is a lowercase u letter. */
1050          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1051            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1052            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1053            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1054            {
1055            c = 0;
1056            for (i = 0; i < 4; ++i)
1057              {
1058              register pcre_uint32 cc = *(++ptr);
1059    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1060              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1061              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1062    #else           /* EBCDIC coding */
1063              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1064              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1065    #endif
1066              }
1067    
1068    #if defined COMPILE_PCRE8
1069            if (c > (utf ? 0x10ffffU : 0xffU))
1070    #elif defined COMPILE_PCRE16
1071            if (c > (utf ? 0x10ffffU : 0xffffU))
1072    #elif defined COMPILE_PCRE32
1073            if (utf && c > 0x10ffffU)
1074    #endif
1075              {
1076              *errorcodeptr = ERR76;
1077              }
1078            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1079            }
1080          }
1081        else
1082          *errorcodeptr = ERR37;
1083        break;
1084    
1085      case CHAR_U:      case CHAR_U:
1086      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
1087        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1088      break;      break;
1089    
1090      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
1091        class, \g must be followed by one of a number of specific things:
1092    
1093      (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
1094      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 661  else Line 1102  else
1102      (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
1103      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1104      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1105      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1106    
1107      case CHAR_g:      case CHAR_g:
1108        if (isclass) break;
1109      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1110        {        {
1111        c = -ESC_g;        escape = ESC_g;
1112        break;        break;
1113        }        }
1114    
# Line 674  else Line 1116  else
1116    
1117      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1118        {        {
1119        const uschar *p;        const pcre_uchar *p;
1120        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1121          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1122        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1123          {          {
1124          c = -ESC_k;          escape = ESC_k;
1125          break;          break;
1126          }          }
1127        braced = TRUE;        braced = TRUE;
# Line 694  else Line 1136  else
1136        }        }
1137      else negated = FALSE;      else negated = FALSE;
1138    
1139      c = 0;      /* The integer range is limited by the machine's int representation. */
1140      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1141        c = c * 10 + *(++ptr) - CHAR_0;      overflow = FALSE;
1142        while (IS_DIGIT(ptr[1]))
1143      if (c < 0)   /* Integer overflow */        {
1144          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1145            {
1146            overflow = TRUE;
1147            break;
1148            }
1149          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1150          }
1151        if (overflow) /* Integer overflow */
1152        {        {
1153          while (IS_DIGIT(ptr[1]))
1154            ptr++;
1155        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1156        break;        break;
1157        }        }
# Line 710  else Line 1162  else
1162        break;        break;
1163        }        }
1164    
1165      if (c == 0)      if (s == 0)
1166        {        {
1167        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1168        break;        break;
# Line 718  else Line 1170  else
1170    
1171      if (negated)      if (negated)
1172        {        {
1173        if (c > bracount)        if (s > bracount)
1174          {          {
1175          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1176          break;          break;
1177          }          }
1178        c = bracount - (c - 1);        s = bracount - (s - 1);
1179        }        }
1180    
1181      c = -(ESC_REF + c);      escape = -s;
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1190      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1191      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1192      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1193      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1194      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 747  else Line 1203  else
1203      if (!isclass)      if (!isclass)
1204        {        {
1205        oldptr = ptr;        oldptr = ptr;
1206        c -= CHAR_0;        /* The integer range is limited by the machine's int representation. */
1207        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1208          c = c * 10 + *(++ptr) - CHAR_0;        overflow = FALSE;
1209        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1210            {
1211            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1212              {
1213              overflow = TRUE;
1214              break;
1215              }
1216            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1217            }
1218          if (overflow) /* Integer overflow */
1219          {          {
1220            while (IS_DIGIT(ptr[1]))
1221              ptr++;
1222          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1223          break;          break;
1224          }          }
1225        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          c = -(ESC_REF + c);          escape = -s;
1228          break;          break;
1229          }          }
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1234      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1235      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1243      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1244      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1245      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1246      than 3 octal digits. */      but no more than 3 octal digits. */
1247    
1248      case CHAR_0:      case CHAR_0:
1249      c -= CHAR_0;      c -= CHAR_0;
1250      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1251          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1252      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1253        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1254    #endif
1255      break;      break;
1256    
1257      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1258      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1259    
1260      case CHAR_x:      case CHAR_o:
1261      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262        {        {
1263        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1264        c = 0;        c = 0;
1265        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267          {          {
1268          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1269          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270          count++;  #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272    #endif
1273            c = (c << 3) + cc - CHAR_0 ;
1274    #if defined COMPILE_PCRE8
1275            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE16
1277            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280    #endif
1281            }
1282          if (overflow)
1283            {
1284            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285            *errorcodeptr = ERR34;
1286            }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288            {
1289            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292          }
1293        break;
1294    
1295        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298        case CHAR_x:
1299        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300          {
1301          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303            {
1304            c = 0;
1305            for (i = 0; i < 2; ++i)
1306              {
1307              register pcre_uint32 cc = *(++ptr);
1308  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1309          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1310          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1311  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1312          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1313          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1314  #endif  #endif
1315              }
1316          }          }
1317          }    /* End JavaScript handling */
1318    
1319        if (*pt == CHAR_RIGHT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        else
1327          {
1328          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329          {          {
1330          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          ptr += 2;
1331          ptr = pt;          c = 0;
1332          break;          overflow = FALSE;
1333          }          while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338        /* If the sequence of hex digits does not end with '}', then we don't  #ifdef COMPILE_PCRE32
1339        recognize this construct; fall through to the normal \x handling. */            if (c >= 0x10000000l) { overflow = TRUE; break; }
1340        }  #endif
1341    
1342    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345    #else           /* EBCDIC coding */
1346              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1347              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348    #endif
1349    
1350      /* Read just a single-byte hex-defined char */  #if defined COMPILE_PCRE8
1351              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352    #elif defined COMPILE_PCRE16
1353              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354    #elif defined COMPILE_PCRE32
1355              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356    #endif
1357              }
1358    
1359      c = 0;          if (overflow)
1360      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)            {
1361        {            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362        int cc;                                  /* Some compilers don't like */            *errorcodeptr = ERR34;
1363        cc = *(++ptr);                           /* ++ in initializers */            }
1364    
1365            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366              {
1367              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368              }
1369    
1370            /* If the sequence of hex digits does not end with '}', give an error.
1371            We used just to recognize this construct and fall through to the normal
1372            \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375            else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379    
1380          else
1381            {
1382            c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1391        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393  #endif  #endif
1394        }            }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* 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.
# Line 848  else Line 1403  else
1403    
1404      case CHAR_c:      case CHAR_c:
1405      c = *(++ptr);      c = *(++ptr);
1406      if (c == 0)      if (c == CHAR_NULL)
1407        {        {
1408        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1409        break;        break;
# Line 885  else Line 1440  else
1440    }    }
1441    
1442  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
1443  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1444    quantification such as \N{2,3}. */
1445    
1446  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1447         !is_counted_repeat(ptr+2))
1448    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1449    
1450  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1451    
1452  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1453    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1454    
1455  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1456    
1457  *ptrptr = ptr;  *ptrptr = ptr;
1458  return c;  *chptr = c;
1459    return escape;
1460  }  }
1461    
1462    
# Line 916  escape sequence. Line 1474  escape sequence.
1474  Argument:  Argument:
1475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1476    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1477    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1478      pdataptr       points to an unsigned int that is set to the detailed property value
1479    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1480    
1481  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1482  */  */
1483    
1484  static int  static BOOL
1485  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1486      unsigned int *pdataptr, int *errorcodeptr)
1487  {  {
1488  int c, i, bot, top;  pcre_uchar c;
1489  const uschar *ptr = *ptrptr;  int i, bot, top;
1490  char name[32];  const pcre_uchar *ptr = *ptrptr;
1491    pcre_uchar name[32];
1492    
1493  c = *(++ptr);  c = *(++ptr);
1494  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1495    
1496  *negptr = FALSE;  *negptr = FALSE;
1497    
# Line 944  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1505  if (c == CHAR_LEFT_CURLY_BRACKET)
1505      *negptr = TRUE;      *negptr = TRUE;
1506      ptr++;      ptr++;
1507      }      }
1508    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1509      {      {
1510      c = *(++ptr);      c = *(++ptr);
1511      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1512      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1513      name[i] = c;      name[i] = c;
1514      }      }
# Line 968  else Line 1529  else
1529  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1530    
1531  bot = 0;  bot = 0;
1532  top = _pcre_utt_size;  top = PRIV(utt_size);
1533    
1534  while (bot < top)  while (bot < top)
1535    {    {
1536      int r;
1537    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1538    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1539    if (c == 0)    if (r == 0)
1540      {      {
1541      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1542      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1543        return TRUE;
1544      }      }
1545    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1546    }    }
1547    
1548  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1549  *ptrptr = ptr;  *ptrptr = ptr;
1550  return -1;  return FALSE;
1551    
1552  ERROR_RETURN:  ERROR_RETURN:
1553  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1554  *ptrptr = ptr;  *ptrptr = ptr;
1555  return -1;  return FALSE;
1556  }  }
1557  #endif  #endif
1558    
1559    
1560    
   
 /*************************************************  
 *            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 == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
1561  /*************************************************  /*************************************************
1562  *         Read repeat counts                     *  *         Read repeat counts                     *
1563  *************************************************/  *************************************************/
# Line 1048  Returns:         pointer to '}' on succe Line 1577  Returns:         pointer to '}' on succe
1577                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1578  */  */
1579    
1580  static const uschar *  static const pcre_uchar *
1581  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)
1582  {  {
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
1585    
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  
 if (min < 0 || min > 65535)  
1587    {    {
1588    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1589    return p;    if (min > 65535)
1590    }      {
1591        *errorcodeptr = ERR5;
1592  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1593  Also, max must not be less than min. */      }
1594      }
1595    
1596  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1597    {    {
1598    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1599      {      {
1600      max = 0;      max = 0;
1601      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1602        {        {
1603        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1604        return p;        if (max > 65535)
1605        }          {
1606            *errorcodeptr = ERR5;
1607            return p;
1608            }
1609          }
1610      if (max < min)      if (max < min)
1611        {        {
1612        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1086  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1615  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1615      }      }
1616    }    }
1617    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1618  *minp = min;  *minp = min;
1619  *maxp = max;  *maxp = max;
1620  return p;  return p;
# Line 1097  return p; Line 1623  return p;
1623    
1624    
1625  /*************************************************  /*************************************************
1626  *  Subroutine for finding forward reference      *  *      Find first significant op code            *
1627  *************************************************/  *************************************************/
1628    
1629  /* This recursive function is called only from find_parens() below. The  /* This is called by several functions that scan a compiled expression looking
1630  top-level call starts at the beginning of the pattern. All other calls must  for a fixed first character, or an anchoring op code etc. It skips over things
1631  start at a parenthesis. It scans along a pattern's text looking for capturing  that do not influence this. For some calls, it makes sense to skip negative
1632  subpatterns, and counting them. If it finds a named pattern that matches the  forward and all backward assertions, and also the \b assertion; for others it
1633  name it is given, it returns its number. Alternatively, if the name is NULL, it  does not.
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
1634    
1635  Arguments:  Arguments:
1636    ptrptr       address of the current character pointer (updated)    code         pointer to the start of the group
1637    cd           compile background data    skipassert   TRUE if certain assertions are to be skipped
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf8         TRUE if we are in UTF-8 mode  
   count        pointer to the current capturing subpattern number (updated)  
1638    
1639  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1640  */  */
1641    
1642  static int  static const pcre_uchar*
1643  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode, BOOL utf8, int *count)  
1644  {  {
1645  uschar *ptr = *ptrptr;  for (;;)
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
1646    {    {
1647    /* Handle specials such as (*SKIP) or (*UTF8) etc. */    switch ((int)*code)
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
1648      {      {
1649      *count += 1;      case OP_ASSERT_NOT:
1650      if (name == NULL && *count == lorn) return *count;      case OP_ASSERTBACK:
1651      ptr++;      case OP_ASSERTBACK_NOT:
1652      }      if (!skipassert) return code;
1653        do code += GET(code, 1); while (*code == OP_ALT);
1654    /* All cases now have (? at the start. Remember when we are in a group      code += PRIV(OP_lengths)[*code];
1655    where the parenthesis numbers are duplicated. */      break;
1656    
1657    else if (ptr[2] == CHAR_VERTICAL_LINE)      case OP_WORD_BOUNDARY:
1658      {      case OP_NOT_WORD_BOUNDARY:
1659      ptr += 3;      if (!skipassert) return code;
1660      dup_parens = TRUE;      /* Fall through */
     }  
1661    
1662    /* Handle comments; all characters are allowed until a ket is reached. */      case OP_CALLOUT:
1663        case OP_CREF:
1664        case OP_DNCREF:
1665        case OP_RREF:
1666        case OP_DNRREF:
1667        case OP_DEF:
1668        code += PRIV(OP_lengths)[*code];
1669        break;
1670    
1671    else if (ptr[2] == CHAR_NUMBER_SIGN)      default:
1672      {      return code;
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
1673      }      }
1674      }
1675    /* Control never reaches here */
1676    }
1677    
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
1678    
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
1679    
1680    /* Start with (? but not a condition. */  /*************************************************
1681    *        Find the fixed length of a branch       *
1682    else  *************************************************/
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       int term;  
       const uschar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == ptr - thisname &&  
           strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (strncmp((const char *)ptr+2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF8  
       if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf8         TRUE if we are in UTF-8 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,  
   BOOL utf8)  
 {  
 uschar *ptr = (uschar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, it makes sense to skip negative  
 forward and all backward assertions, and also the \b assertion; for others it  
 does not.  
   
 Arguments:  
   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  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const uschar*  
 first_significant_code(const uschar *code, int *options, int optbit,  
   BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
     {  
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     case OP_WORD_BOUNDARY:  
     case OP_NOT_WORD_BOUNDARY:  
     if (!skipassert) return code;  
     /* Fall through */  
   
     case OP_CALLOUT:  
     case OP_CREF:  
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
     case OP_DEF:  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     default:  
     return code;  
     }  
   }  
 /* Control never reaches here */  
 }  
   
   
   
   
 /*************************************************  
 *        Find the fixed length of a branch       *  
 *************************************************/  
1683    
1684  /* Scan a branch 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,
1685  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.
# Line 1468  and doing the check at the end; a flag s Line 1694  and doing the check at the end; a flag s
1694    
1695  Arguments:  Arguments:
1696    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1697    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1698    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1699    cd       the "compile data" structure    cd       the "compile data" structure
1700    
1701  Returns:   the fixed length,  Returns:   the fixed length,
1702               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1703               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1704               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1705                 or -4 if an unknown opcode was encountered (internal error)
1706  */  */
1707    
1708  static int  static int
1709  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1710  {  {
1711  int length = -1;  int length = -1;
1712    
1713  register int branchlength = 0;  register int branchlength = 0;
1714  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1715    
1716  /* 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
1717  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1492  branch, check the length against that of Line 1719  branch, check the length against that of
1719  for (;;)  for (;;)
1720    {    {
1721    int d;    int d;
1722    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1723    register int op = *cc;    register pcre_uchar op = *cc;
1724    
1725    switch (op)    switch (op)
1726      {      {
1727        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1728        OP_BRA (normal non-capturing bracket) because the other variants of these
1729        opcodes are all concerned with unlimited repeated groups, which of course
1730        are not of fixed length. */
1731    
1732      case OP_CBRA:      case OP_CBRA:
1733      case OP_BRA:      case OP_BRA:
1734      case OP_ONCE:      case OP_ONCE:
1735        case OP_ONCE_NC:
1736      case OP_COND:      case OP_COND:
1737      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1738      if (d < 0) return d;      if (d < 0) return d;
1739      branchlength += d;      branchlength += d;
1740      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1741      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1742      break;      break;
1743    
1744      /* 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.
1745      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
1746      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
1747        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1748        because they all imply an unlimited repeat. */
1749    
1750      case OP_ALT:      case OP_ALT:
1751      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1752      case OP_END:      case OP_END:
1753        case OP_ACCEPT:
1754        case OP_ASSERT_ACCEPT:
1755      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1756        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1757      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1529  for (;;) Line 1765  for (;;)
1765    
1766      case OP_RECURSE:      case OP_RECURSE:
1767      if (!atend) return -3;      if (!atend) return -3;
1768      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1769      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1770      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1771      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1772      if (d < 0) return d;      if (d < 0) return d;
1773      branchlength += d;      branchlength += d;
1774      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1545  for (;;) Line 1781  for (;;)
1781      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1782      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1783      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1784      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1785        break;
1786    
1787      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1788    
1789      case OP_REVERSE:      case OP_MARK:
1790      case OP_CREF:      case OP_PRUNE_ARG:
1791      case OP_NCREF:      case OP_SKIP_ARG:
1792      case OP_RREF:      case OP_THEN_ARG:
1793      case OP_NRREF:      cc += cc[1] + PRIV(OP_lengths)[*cc];
1794      case OP_DEF:      break;
1795    
1796      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1797      case OP_CIRC:      case OP_CIRC:
1798      case OP_CIRCM:      case OP_CIRCM:
1799        case OP_CLOSE:
1800        case OP_COMMIT:
1801        case OP_CREF:
1802        case OP_DEF:
1803        case OP_DNCREF:
1804        case OP_DNRREF:
1805      case OP_DOLL:      case OP_DOLL:
1806      case OP_DOLLM:      case OP_DOLLM:
1807        case OP_EOD:
1808        case OP_EODN:
1809        case OP_FAIL:
1810      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1811        case OP_PRUNE:
1812        case OP_REVERSE:
1813        case OP_RREF:
1814        case OP_SET_SOM:
1815        case OP_SKIP:
1816        case OP_SOD:
1817        case OP_SOM:
1818        case OP_THEN:
1819      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1820      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1821      break;      break;
1822    
1823      /* Handle literal characters */      /* Handle literal characters */
# Line 1575  for (;;) Line 1825  for (;;)
1825      case OP_CHAR:      case OP_CHAR:
1826      case OP_CHARI:      case OP_CHARI:
1827      case OP_NOT:      case OP_NOT:
1828      case OP_NOTI:      case OP_NOTI:
1829      branchlength++;      branchlength++;
1830      cc += 2;      cc += 2;
1831  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1832      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1833  #endif  #endif
1834      break;      break;
1835    
# Line 1588  for (;;) Line 1837  for (;;)
1837      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1838    
1839      case OP_EXACT:      case OP_EXACT:
1840      branchlength += GET2(cc,1);      case OP_EXACTI:
1841      cc += 4;      case OP_NOTEXACT:
1842  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1843      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      branchlength += (int)GET2(cc,1);
1844        cc += _pcre_utf8_table4[cc[-1] & 0x3f];      cc += 2 + IMM2_SIZE;
1845    #ifdef SUPPORT_UTF
1846        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1847  #endif  #endif
1848      break;      break;
1849    
1850      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1851      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1852      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1853      cc += 4;        cc += 2;
1854        cc += 1 + IMM2_SIZE + 1;
1855      break;      break;
1856    
1857      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1609  for (;;) Line 1861  for (;;)
1861      cc += 2;      cc += 2;
1862      /* Fall through */      /* Fall through */
1863    
1864        case OP_HSPACE:
1865        case OP_VSPACE:
1866        case OP_NOT_HSPACE:
1867        case OP_NOT_VSPACE:
1868      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1869      case OP_DIGIT:      case OP_DIGIT:
1870      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1621  for (;;) Line 1877  for (;;)
1877      cc++;      cc++;
1878      break;      break;
1879    
1880      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1881        otherwise \C is coded as OP_ALLANY. */
1882    
1883      case OP_ANYBYTE:      case OP_ANYBYTE:
1884      return -2;      return -2;
1885    
1886      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1887    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1888      case OP_CLASS:      case OP_CLASS:
1889      case OP_NCLASS:      case OP_NCLASS:
1890      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1891        case OP_XCLASS:
1892        /* The original code caused an unsigned overflow in 64 bit systems,
1893        so now we use a conditional statement. */
1894        if (op == OP_XCLASS)
1895          cc += GET(cc, 1);
1896        else
1897          cc += PRIV(OP_lengths)[OP_CLASS];
1898    #else
1899        cc += PRIV(OP_lengths)[OP_CLASS];
1900    #endif
1901    
1902      switch (*cc)      switch (*cc)
1903        {        {
1904        case OP_CRSTAR:        case OP_CRSTAR:
1905        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1906          case OP_CRPLUS:
1907          case OP_CRMINPLUS:
1908        case OP_CRQUERY:        case OP_CRQUERY:
1909        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1910          case OP_CRPOSSTAR:
1911          case OP_CRPOSPLUS:
1912          case OP_CRPOSQUERY:
1913        return -1;        return -1;
1914    
1915        case OP_CRRANGE:        case OP_CRRANGE:
1916        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1917        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1918        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1919        cc += 5;        branchlength += (int)GET2(cc,1);
1920          cc += 1 + 2 * IMM2_SIZE;
1921        break;        break;
1922    
1923        default:        default:
# Line 1660  for (;;) Line 1927  for (;;)
1927    
1928      /* Anything else is variable length */      /* Anything else is variable length */
1929    
1930      default:      case OP_ANYNL:
1931        case OP_BRAMINZERO:
1932        case OP_BRAPOS:
1933        case OP_BRAPOSZERO:
1934        case OP_BRAZERO:
1935        case OP_CBRAPOS:
1936        case OP_EXTUNI:
1937        case OP_KETRMAX:
1938        case OP_KETRMIN:
1939        case OP_KETRPOS:
1940        case OP_MINPLUS:
1941        case OP_MINPLUSI:
1942        case OP_MINQUERY:
1943        case OP_MINQUERYI:
1944        case OP_MINSTAR:
1945        case OP_MINSTARI:
1946        case OP_MINUPTO:
1947        case OP_MINUPTOI:
1948        case OP_NOTMINPLUS:
1949        case OP_NOTMINPLUSI:
1950        case OP_NOTMINQUERY:
1951        case OP_NOTMINQUERYI:
1952        case OP_NOTMINSTAR:
1953        case OP_NOTMINSTARI:
1954        case OP_NOTMINUPTO:
1955        case OP_NOTMINUPTOI:
1956        case OP_NOTPLUS:
1957        case OP_NOTPLUSI:
1958        case OP_NOTPOSPLUS:
1959        case OP_NOTPOSPLUSI:
1960        case OP_NOTPOSQUERY:
1961        case OP_NOTPOSQUERYI:
1962        case OP_NOTPOSSTAR:
1963        case OP_NOTPOSSTARI:
1964        case OP_NOTPOSUPTO:
1965        case OP_NOTPOSUPTOI:
1966        case OP_NOTQUERY:
1967        case OP_NOTQUERYI:
1968        case OP_NOTSTAR:
1969        case OP_NOTSTARI:
1970        case OP_NOTUPTO:
1971        case OP_NOTUPTOI:
1972        case OP_PLUS:
1973        case OP_PLUSI:
1974        case OP_POSPLUS:
1975        case OP_POSPLUSI:
1976        case OP_POSQUERY:
1977        case OP_POSQUERYI:
1978        case OP_POSSTAR:
1979        case OP_POSSTARI:
1980        case OP_POSUPTO:
1981        case OP_POSUPTOI:
1982        case OP_QUERY:
1983        case OP_QUERYI:
1984        case OP_REF:
1985        case OP_REFI:
1986        case OP_DNREF:
1987        case OP_DNREFI:
1988        case OP_SBRA:
1989        case OP_SBRAPOS:
1990        case OP_SCBRA:
1991        case OP_SCBRAPOS:
1992        case OP_SCOND:
1993        case OP_SKIPZERO:
1994        case OP_STAR:
1995        case OP_STARI:
1996        case OP_TYPEMINPLUS:
1997        case OP_TYPEMINQUERY:
1998        case OP_TYPEMINSTAR:
1999        case OP_TYPEMINUPTO:
2000        case OP_TYPEPLUS:
2001        case OP_TYPEPOSPLUS:
2002        case OP_TYPEPOSQUERY:
2003        case OP_TYPEPOSSTAR:
2004        case OP_TYPEPOSUPTO:
2005        case OP_TYPEQUERY:
2006        case OP_TYPESTAR:
2007        case OP_TYPEUPTO:
2008        case OP_UPTO:
2009        case OP_UPTOI:
2010      return -1;      return -1;
2011    
2012        /* Catch unrecognized opcodes so that when new ones are added they
2013        are not forgotten, as has happened in the past. */
2014    
2015        default:
2016        return -4;
2017      }      }
2018    }    }
2019  /* Control never gets here */  /* Control never gets here */
# Line 1669  for (;;) Line 2021  for (;;)
2021    
2022    
2023    
   
2024  /*************************************************  /*************************************************
2025  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2026  *************************************************/  *************************************************/
# Line 1682  length. Line 2033  length.
2033    
2034  Arguments:  Arguments:
2035    code        points to start of expression    code        points to start of expression
2036    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2037    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2038    
2039  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
2040  */  */
2041    
2042  const uschar *  const pcre_uchar *
2043  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2044  {  {
2045  for (;;)  for (;;)
2046    {    {
2047    register int c = *code;    register pcre_uchar c = *code;
2048    
2049    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2050    
2051    /* 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 1706  for (;;) Line 2058  for (;;)
2058    
2059    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2060      {      {
2061      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2062      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2063      }      }
2064    
2065    /* Handle capturing bracket */    /* Handle capturing bracket */
2066    
2067    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2068               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2069      {      {
2070      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2071      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2072      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2073      }      }
2074    
2075    /* 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
# Line 1744  for (;;) Line 2097  for (;;)
2097        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2098        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2099        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2100        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2101            code += 2;
2102        break;        break;
2103    
2104        case OP_MARK:        case OP_MARK:
2105        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2106        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2107        case OP_THEN_ARG:        case OP_THEN_ARG:
2108        code += code[1+LINK_SIZE];        code += code[1];
2109        break;        break;
2110        }        }
2111    
2112      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2113    
2114      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2115    
2116    /* 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
2117    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
2118    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2119    
2120  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2121      if (utf8) switch(c)      if (utf) switch(c)
2122        {        {
2123        case OP_CHAR:        case OP_CHAR:
2124        case OP_CHARI:        case OP_CHARI:
# Line 1797  for (;;) Line 2148  for (;;)
2148        case OP_MINQUERYI:        case OP_MINQUERYI:
2149        case OP_POSQUERY:        case OP_POSQUERY:
2150        case OP_POSQUERYI:        case OP_POSQUERYI:
2151        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2152        break;        break;
2153        }        }
2154  #else  #else
2155      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2156  #endif  #endif
2157      }      }
2158    }    }
# Line 1818  instance of OP_RECURSE. Line 2169  instance of OP_RECURSE.
2169    
2170  Arguments:  Arguments:
2171    code        points to start of expression    code        points to start of expression
2172    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2173    
2174  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
2175  */  */
2176    
2177  static const uschar *  static const pcre_uchar *
2178  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2179  {  {
2180  for (;;)  for (;;)
2181    {    {
2182    register int c = *code;    register pcre_uchar c = *code;
2183    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2184    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2185    
# Line 1863  for (;;) Line 2214  for (;;)
2214        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2215        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2216        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2217        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2218            code += 2;
2219        break;        break;
2220    
2221        case OP_MARK:        case OP_MARK:
2222        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2223        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2224        case OP_THEN_ARG:        case OP_THEN_ARG:
2225        code += code[1+LINK_SIZE];        code += code[1];
2226        break;        break;
2227        }        }
2228    
2229      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2230    
2231      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2232    
2233      /* 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
2234      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
2235      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2236    
2237  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2238      if (utf8) switch(c)      if (utf) switch(c)
2239        {        {
2240        case OP_CHAR:        case OP_CHAR:
2241        case OP_CHARI:        case OP_CHARI:
2242          case OP_NOT:
2243          case OP_NOTI:
2244        case OP_EXACT:        case OP_EXACT:
2245        case OP_EXACTI:        case OP_EXACTI:
2246          case OP_NOTEXACT:
2247          case OP_NOTEXACTI:
2248        case OP_UPTO:        case OP_UPTO:
2249        case OP_UPTOI:        case OP_UPTOI:
2250          case OP_NOTUPTO:
2251          case OP_NOTUPTOI:
2252        case OP_MINUPTO:        case OP_MINUPTO:
2253        case OP_MINUPTOI:        case OP_MINUPTOI:
2254          case OP_NOTMINUPTO:
2255          case OP_NOTMINUPTOI:
2256        case OP_POSUPTO:        case OP_POSUPTO:
2257        case OP_POSUPTOI:        case OP_POSUPTOI:
2258          case OP_NOTPOSUPTO:
2259          case OP_NOTPOSUPTOI:
2260        case OP_STAR:        case OP_STAR:
2261        case OP_STARI:        case OP_STARI:
2262          case OP_NOTSTAR:
2263          case OP_NOTSTARI:
2264        case OP_MINSTAR:        case OP_MINSTAR:
2265        case OP_MINSTARI:        case OP_MINSTARI:
2266          case OP_NOTMINSTAR:
2267          case OP_NOTMINSTARI:
2268        case OP_POSSTAR:        case OP_POSSTAR:
2269        case OP_POSSTARI:        case OP_POSSTARI:
2270          case OP_NOTPOSSTAR:
2271          case OP_NOTPOSSTARI:
2272        case OP_PLUS:        case OP_PLUS:
2273        case OP_PLUSI:        case OP_PLUSI:
2274          case OP_NOTPLUS:
2275          case OP_NOTPLUSI:
2276        case OP_MINPLUS:        case OP_MINPLUS:
2277        case OP_MINPLUSI:        case OP_MINPLUSI:
2278          case OP_NOTMINPLUS:
2279          case OP_NOTMINPLUSI:
2280        case OP_POSPLUS:        case OP_POSPLUS:
2281        case OP_POSPLUSI:        case OP_POSPLUSI:
2282          case OP_NOTPOSPLUS:
2283          case OP_NOTPOSPLUSI:
2284        case OP_QUERY:        case OP_QUERY:
2285        case OP_QUERYI:        case OP_QUERYI:
2286          case OP_NOTQUERY:
2287          case OP_NOTQUERYI:
2288        case OP_MINQUERY:        case OP_MINQUERY:
2289        case OP_MINQUERYI:        case OP_MINQUERYI:
2290          case OP_NOTMINQUERY:
2291          case OP_NOTMINQUERYI:
2292        case OP_POSQUERY:        case OP_POSQUERY:
2293        case OP_POSQUERYI:        case OP_POSQUERYI:
2294        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_NOTPOSQUERY:
2295          case OP_NOTPOSQUERYI:
2296          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2297        break;        break;
2298        }        }
2299  #else  #else
2300      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2301  #endif  #endif
2302      }      }
2303    }    }
# Line 1943  bracket whose current branch will alread Line 2320  bracket whose current branch will alread
2320  Arguments:  Arguments:
2321    code        points to start of search    code        points to start of search
2322    endcode     points to where to stop    endcode     points to where to stop
2323    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2324    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2325      recurses    chain of recurse_check to catch mutual recursion
2326    
2327  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2328  */  */
2329    
2330    typedef struct recurse_check {
2331      struct recurse_check *prev;
2332      const pcre_uchar *group;
2333    } recurse_check;
2334    
2335  static BOOL  static BOOL
2336  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2337    compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2338  {  {
2339  register int c;  register pcre_uchar c;
2340  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2341    
2342    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2343       code < endcode;       code < endcode;
2344       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2345    {    {
2346    const uschar *ccode;    const pcre_uchar *ccode;
2347    
2348    c = *code;    c = *code;
2349    
# Line 1972  for (code = first_significant_code(code Line 2357  for (code = first_significant_code(code
2357      continue;      continue;
2358      }      }
2359    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2360    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2361    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2362      forward reference subroutine call, we can't. To detect forward reference
2363      we have to scan up the list that is kept in the workspace. This function is
2364      called only when doing the real compile, not during the pre-compile that
2365      measures the size of the compiled pattern. */
2366    
2367    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2368      {      {
2369      BOOL empty_branch = FALSE;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2370      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2371      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */  
2372      do      /* Test for forward reference or uncompleted reference. This is disabled
2373        {      when called to scan a completed pattern by setting cd->start_workspace to
2374        if (could_be_empty_branch(scode, endcode, utf8, cd))      NULL. */
2375          {  
2376          empty_branch = TRUE;      if (cd->start_workspace != NULL)
2377          {
2378          const pcre_uchar *tcode;
2379          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2380            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2381          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2382          }
2383    
2384        /* If we are scanning a completed pattern, there are no forward references
2385        and all groups are complete. We need to detect whether this is a recursive
2386        call, as otherwise there will be an infinite loop. If it is a recursion,
2387        just skip over it. Simple recursions are easily detected. For mutual
2388        recursions we keep a chain on the stack. */
2389    
2390        else
2391          {
2392          recurse_check *r = recurses;
2393          const pcre_uchar *endgroup = scode;
2394    
2395          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2396          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2397    
2398          for (r = recurses; r != NULL; r = r->prev)
2399            if (r->group == scode) break;
2400          if (r != NULL) continue;   /* Mutual recursion */
2401          }
2402    
2403        /* Completed reference; scan the referenced group, remembering it on the
2404        stack chain to detect mutual recursions. */
2405    
2406        empty_branch = FALSE;
2407        this_recurse.prev = recurses;
2408        this_recurse.group = scode;
2409    
2410        do
2411          {
2412          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2413            {
2414            empty_branch = TRUE;
2415          break;          break;
2416          }          }
2417        scode += GET(scode, 1);        scode += GET(scode, 1);
2418        }        }
2419      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2420    
2421      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2422      continue;      continue;
2423      }      }
2424    
2425      /* Groups with zero repeats can of course be empty; skip them. */
2426    
2427      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2428          c == OP_BRAPOSZERO)
2429        {
2430        code += PRIV(OP_lengths)[c];
2431        do code += GET(code, 1); while (*code == OP_ALT);
2432        c = *code;
2433        continue;
2434        }
2435    
2436      /* A nested group that is already marked as "could be empty" can just be
2437      skipped. */
2438    
2439      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2440          c == OP_SCBRA || c == OP_SCBRAPOS)
2441        {
2442        do code += GET(code, 1); while (*code == OP_ALT);
2443        c = *code;
2444        continue;
2445        }
2446    
2447    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2448    
2449    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2450          c == OP_CBRA || c == OP_CBRAPOS ||
2451          c == OP_ONCE || c == OP_ONCE_NC ||
2452          c == OP_COND)
2453      {      {
2454      BOOL empty_branch;      BOOL empty_branch;
2455      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2022  for (code = first_significant_code(code Line 2465  for (code = first_significant_code(code
2465        empty_branch = FALSE;        empty_branch = FALSE;
2466        do        do
2467          {          {
2468          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2469            empty_branch = TRUE;            empty_branch = TRUE;
2470          code += GET(code, 1);          code += GET(code, 1);
2471          }          }
# Line 2040  for (code = first_significant_code(code Line 2483  for (code = first_significant_code(code
2483      {      {
2484      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2485      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2486      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2487      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"
2488      here. */      here. */
2489    
2490  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2491      case OP_XCLASS:      case OP_XCLASS:
2492      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2493      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2052  for (code = first_significant_code(code Line 2495  for (code = first_significant_code(code
2495    
2496      case OP_CLASS:      case OP_CLASS:
2497      case OP_NCLASS:      case OP_NCLASS:
2498      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2499    
2500  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2501      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2502  #endif  #endif
2503    
# Line 2064  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2508        case OP_CRQUERY:        case OP_CRQUERY:
2509        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2510          case OP_CRPOSSTAR:
2511          case OP_CRPOSQUERY:
2512        break;        break;
2513    
2514        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2515        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2516        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2517          case OP_CRPOSPLUS:
2518        return FALSE;        return FALSE;
2519    
2520        case OP_CRRANGE:        case OP_CRRANGE:
2521        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2522          case OP_CRPOSRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2524        break;        break;
2525        }        }
# Line 2080  for (code = first_significant_code(code Line 2527  for (code = first_significant_code(code
2527    
2528      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2529    
2530        case OP_ANY:
2531        case OP_ALLANY:
2532        case OP_ANYBYTE:
2533    
2534      case OP_PROP:      case OP_PROP:
2535      case OP_NOTPROP:      case OP_NOTPROP:
2536        case OP_ANYNL:
2537    
2538        case OP_NOT_HSPACE:
2539        case OP_HSPACE:
2540        case OP_NOT_VSPACE:
2541        case OP_VSPACE:
2542      case OP_EXTUNI:      case OP_EXTUNI:
2543    
2544      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2545      case OP_DIGIT:      case OP_DIGIT:
2546      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2547      case OP_WHITESPACE:      case OP_WHITESPACE:
2548      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2549      case OP_WORDCHAR:      case OP_WORDCHAR:
2550      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2551      case OP_CHAR:      case OP_CHAR:
2552      case OP_CHARI:      case OP_CHARI:
2553      case OP_NOT:      case OP_NOT:
2554      case OP_NOTI:      case OP_NOTI:
2555    
2556      case OP_PLUS:      case OP_PLUS:
2557        case OP_PLUSI:
2558      case OP_MINPLUS:      case OP_MINPLUS:
2559      case OP_POSPLUS:      case OP_MINPLUSI:
2560      case OP_EXACT:  
2561      case OP_NOTPLUS:      case OP_NOTPLUS:
2562        case OP_NOTPLUSI:
2563      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2564        case OP_NOTMINPLUSI:
2565    
2566        case OP_POSPLUS:
2567        case OP_POSPLUSI:
2568      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2569        case OP_NOTPOSPLUSI:
2570    
2571        case OP_EXACT:
2572        case OP_EXACTI:
2573      case OP_NOTEXACT:      case OP_NOTEXACT:
2574        case OP_NOTEXACTI:
2575    
2576      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2577      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2578      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2579      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2580    
2581      return FALSE;      return FALSE;
2582    
2583      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2127  for (code = first_significant_code(code Line 2597  for (code = first_significant_code(code
2597      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2598      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2599      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2600      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2601          code += 2;
2602      break;      break;
2603    
2604      /* End of branch */      /* End of branch */
# Line 2135  for (code = first_significant_code(code Line 2606  for (code = first_significant_code(code
2606      case OP_KET:      case OP_KET:
2607      case OP_KETRMAX:      case OP_KETRMAX:
2608      case OP_KETRMIN:      case OP_KETRMIN:
2609        case OP_KETRPOS:
2610      case OP_ALT:      case OP_ALT:
2611      return TRUE;      return TRUE;
2612    
2613      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2614      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2615        followed by a multibyte character. */
2616    
2617  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2618      case OP_STAR:      case OP_STAR:
2619      case OP_STARI:      case OP_STARI:
2620        case OP_NOTSTAR:
2621        case OP_NOTSTARI:
2622    
2623      case OP_MINSTAR:      case OP_MINSTAR:
2624      case OP_MINSTARI:      case OP_MINSTARI:
2625        case OP_NOTMINSTAR:
2626        case OP_NOTMINSTARI:
2627    
2628      case OP_POSSTAR:      case OP_POSSTAR:
2629      case OP_POSSTARI:      case OP_POSSTARI:
2630        case OP_NOTPOSSTAR:
2631        case OP_NOTPOSSTARI:
2632    
2633      case OP_QUERY:      case OP_QUERY:
2634      case OP_QUERYI:      case OP_QUERYI:
2635        case OP_NOTQUERY:
2636        case OP_NOTQUERYI:
2637    
2638      case OP_MINQUERY:      case OP_MINQUERY:
2639      case OP_MINQUERYI:      case OP_MINQUERYI:
2640        case OP_NOTMINQUERY:
2641        case OP_NOTMINQUERYI:
2642    
2643      case OP_POSQUERY:      case OP_POSQUERY:
2644      case OP_POSQUERYI:      case OP_POSQUERYI:
2645      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_NOTPOSQUERY:
2646        case OP_NOTPOSQUERYI:
2647    
2648        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2649      break;      break;
2650    
2651      case OP_UPTO:      case OP_UPTO:
2652      case OP_UPTOI:      case OP_UPTOI:
2653        case OP_NOTUPTO:
2654        case OP_NOTUPTOI:
2655    
2656      case OP_MINUPTO:      case OP_MINUPTO:
2657      case OP_MINUPTOI:      case OP_MINUPTOI:
2658        case OP_NOTMINUPTO:
2659        case OP_NOTMINUPTOI:
2660    
2661      case OP_POSUPTO:      case OP_POSUPTO:
2662      case OP_POSUPTOI:      case OP_POSUPTOI:
2663      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_NOTPOSUPTO:
2664        case OP_NOTPOSUPTOI:
2665    
2666        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2667      break;      break;
2668  #endif  #endif
2669    
# Line 2173  for (code = first_significant_code(code Line 2673  for (code = first_significant_code(code
2673      case OP_MARK:      case OP_MARK:
2674      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2675      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2676      case OP_THEN_ARG:      case OP_THEN_ARG:
2677      code += code[1+LINK_SIZE];      code += code[1];
2678      break;      break;
2679    
2680      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2200  return TRUE; Line 2697  return TRUE;
2697  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
2698  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,
2699  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.
2700    This function is called only during the real compile, not during the
2701    pre-compile.
2702    
2703  Arguments:  Arguments:
2704    code        points to start of the recursion    code        points to start of the recursion
2705    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2706    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2707    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2708    cd          pointers to tables etc    cd          pointers to tables etc
2709    
2710  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2711  */  */
2712    
2713  static BOOL  static BOOL
2714  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2715    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2716  {  {
2717  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2718    {    {
2719    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2720      return FALSE;      return FALSE;
2721    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2722    }    }
# Line 2227  return TRUE; Line 2726  return TRUE;
2726    
2727    
2728  /*************************************************  /*************************************************
2729  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2730  *************************************************/  *************************************************/
2731    
2732  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2733  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2734    
2735  Returns:   TRUE or FALSE  Arguments:  c opcode
2736    Returns:    base opcode for the type
2737  */  */
2738    
2739  static BOOL  static pcre_uchar
2740  check_posix_syntax(const uschar *ptr, const uschar **endptr)  get_repeat_base(pcre_uchar c)
2741  {  {
2742  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2743  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2744  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2745    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2746    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else         (c >= OP_STARI)?      OP_STARI :
2747      {                               OP_STAR;
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     }  
   }  
 return FALSE;  
2748  }  }
2749    
2750    
2751    
2752    #ifdef SUPPORT_UCP
2753  /*************************************************  /*************************************************
2754  *          Check POSIX class name                *  *        Check a character and a property        *
2755  *************************************************/  *************************************************/
2756    
2757  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2758  such as [:alnum:].  is adjacent to a fixed character.
2759    
2760  Arguments:  Arguments:
2761    ptr        points to the first letter    c            the character
2762    len        the length of the name    ptype        the property type
2763      pdata        the data for the type
2764      negated      TRUE if it's a negated property (\P or \p{^)
2765    
2766  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2767  */  */
2768    
2769  static int  static BOOL
2770  check_posix_name(const uschar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2771      BOOL negated)
2772  {  {
2773  const char *pn = posix_names;  const pcre_uint32 *p;
2774  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     strncmp((const char *)ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2775    
2776  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2777  that is referenced. This means that groups can be replicated for fixed    {
2778  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2779  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2780  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2781  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2782    
2783  This function has been extended with the possibility of forward references for    case PT_GC:
2784  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2785    
2786  Arguments:    case PT_PC:
2787    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf8       TRUE in UTF-8 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2788    
2789  Returns:     nothing    case PT_SC:
2790  */    return (pdata == prop->script) == negated;
2791    
2792  static void    /* These are specials */
 adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  
   uschar *save_hwm)  
 {  
 uschar *ptr = group;  
2793    
2794  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)    case PT_ALNUM:
2795    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2796    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   uschar *hc;  
2797    
2798    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2799    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2800      at release 8.34. */
2801    
2802    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2803      case PT_PXSPACE:  /* POSIX space */
2804      switch(c)
2805      {      {
2806      offset = GET(hc, 0);      HSPACE_CASES:
2807      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2808        {      return negated;
2809        PUT(hc, 0, offset + adjust);  
2810        break;      default:
2811        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2812      }      }
2813      break;  /* Control never reaches here */
2814    
2815    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2816    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2817              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2818              c == CHAR_UNDERSCORE) == negated;
2819    
2820    if (hc >= cd->hwm)    case PT_CLIST:
2821      p = PRIV(ucd_caseless_sets) + prop->caseset;
2822      for (;;)
2823      {      {
2824      offset = GET(ptr, 1);      if (c < *p) return !negated;
2825      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2826      }      }
2827      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2828    }    }
2829    
2830    return FALSE;
2831  }  }
2832    #endif  /* SUPPORT_UCP */
2833    
2834    
2835    
2836  /*************************************************  /*************************************************
2837  *        Insert an automatic callout point       *  *        Fill the character property list        *
2838  *************************************************/  *************************************************/
2839    
2840  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2841  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2842    
2843  Arguments:  Arguments:
2844    code           current code pointer    code        points to start of expression
2845    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2846    cd             pointers to tables etc    fcc         points to case-flipping table
2847      list        points to output list
2848                  list[0] will be filled with the opcode
2849                  list[1] will be non-zero if this opcode
2850                    can match an empty character string
2851                  list[2..7] depends on the opcode
2852    
2853  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2854                  NULL if *code is not accepted
2855  */  */
2856    
2857  static uschar *  static const pcre_uchar *
2858  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2859      const pcre_uint8 *fcc, pcre_uint32 *list)
2860  {  {
2861  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2862  *code++ = 255;  pcre_uchar base;
2863  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uchar *end;
2864  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 chr;
 return code + 2*LINK_SIZE;  
 }  
   
2865    
2866    #ifdef SUPPORT_UCP
2867    pcre_uint32 *clist_dest;
2868    const pcre_uint32 *clist_src;
2869    #else
2870    utf = utf;  /* Suppress "unused parameter" compiler warning */
2871    #endif
2872    
2873  /*************************************************  list[0] = c;
2874  *         Complete a callout item                *  list[1] = FALSE;
2875  *************************************************/  code++;
   
 /* A callout item contains the length of the next item in the pattern, which  
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
2876    
2877  Arguments:  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2878    previous_callout   points to previous callout item    {
2879    ptr                current pattern pointer    base = get_repeat_base(c);
2880    cd                 pointers to tables etc    c -= (base - OP_STAR);
2881    
2882  Returns:             nothing    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2883  */      code += IMM2_SIZE;
2884    
2885  static void    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  
 {  
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2886    
2887      switch(base)
2888        {
2889        case OP_STAR:
2890        list[0] = OP_CHAR;
2891        break;
2892    
2893        case OP_STARI:
2894        list[0] = OP_CHARI;
2895        break;
2896    
2897  #ifdef SUPPORT_UCP      case OP_NOTSTAR:
2898  /*************************************************      list[0] = OP_NOT;
2899  *           Get othercase range                  *      break;
 *************************************************/  
2900    
2901  /* This function is passed the start and end of a class range, in UTF-8 mode      case OP_NOTSTARI:
2902  with UCP support. It searches up the characters, looking for internal ranges of      list[0] = OP_NOTI;
2903  characters in the "other" case. Each call returns the next one, updating the      break;
 start address.  
2904    
2905  Arguments:      case OP_TYPESTAR:
2906    cptr        points to starting character value; updated      list[0] = *code;
2907    d           end value      code++;
2908    ocptr       where to put start of othercase range      break;
2909    odptr       where to put end of othercase range      }
2910      c = list[0];
2911      }
2912    
2913  Yield:        TRUE when range returned; FALSE when no more  switch(c)
2914      {
2915      case OP_NOT_DIGIT:
2916      case OP_DIGIT:
2917      case OP_NOT_WHITESPACE:
2918      case OP_WHITESPACE:
2919      case OP_NOT_WORDCHAR:
2920      case OP_WORDCHAR:
2921      case OP_ANY:
2922      case OP_ALLANY:
2923      case OP_ANYNL:
2924      case OP_NOT_HSPACE:
2925      case OP_HSPACE:
2926      case OP_NOT_VSPACE:
2927      case OP_VSPACE:
2928      case OP_EXTUNI:
2929      case OP_EODN:
2930      case OP_EOD:
2931      case OP_DOLL:
2932      case OP_DOLLM:
2933      return code;
2934    
2935      case OP_CHAR:
2936      case OP_NOT:
2937      GETCHARINCTEST(chr, code);
2938      list[2] = chr;
2939      list[3] = NOTACHAR;
2940      return code;
2941    
2942      case OP_CHARI:
2943      case OP_NOTI:
2944      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2945      GETCHARINCTEST(chr, code);
2946      list[2] = chr;
2947    
2948    #ifdef SUPPORT_UCP
2949      if (chr < 128 || (chr < 256 && !utf))
2950        list[3] = fcc[chr];
2951      else
2952        list[3] = UCD_OTHERCASE(chr);
2953    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2954      list[3] = (chr < 256) ? fcc[chr] : chr;
2955    #else
2956      list[3] = fcc[chr];
2957    #endif
2958    
2959      /* The othercase might be the same value. */
2960    
2961      if (chr == list[3])
2962        list[3] = NOTACHAR;
2963      else
2964        list[4] = NOTACHAR;
2965      return code;
2966    
2967    #ifdef SUPPORT_UCP
2968      case OP_PROP:
2969      case OP_NOTPROP:
2970      if (code[0] != PT_CLIST)
2971        {
2972        list[2] = code[0];
2973        list[3] = code[1];
2974        return code + 2;
2975        }
2976    
2977      /* Convert only if we have enough space. */
2978    
2979      clist_src = PRIV(ucd_caseless_sets) + code[1];
2980      clist_dest = list + 2;
2981      code += 2;
2982    
2983      do {
2984         if (clist_dest >= list + 8)
2985           {
2986           /* Early return if there is not enough space. This should never
2987           happen, since all clists are shorter than 5 character now. */
2988           list[2] = code[0];
2989           list[3] = code[1];
2990           return code;
2991           }
2992         *clist_dest++ = *clist_src;
2993         }
2994      while(*clist_src++ != NOTACHAR);
2995    
2996      /* All characters are stored. The terminating NOTACHAR
2997      is copied form the clist itself. */
2998    
2999      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3000      return code;
3001    #endif
3002    
3003      case OP_NCLASS:
3004      case OP_CLASS:
3005    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3006      case OP_XCLASS:
3007      if (c == OP_XCLASS)
3008        end = code + GET(code, 0) - 1;
3009      else
3010    #endif
3011        end = code + 32 / sizeof(pcre_uchar);
3012    
3013      switch(*end)
3014        {
3015        case OP_CRSTAR:
3016        case OP_CRMINSTAR:
3017        case OP_CRQUERY:
3018        case OP_CRMINQUERY:
3019        case OP_CRPOSSTAR:
3020        case OP_CRPOSQUERY:
3021        list[1] = TRUE;
3022        end++;
3023        break;
3024    
3025        case OP_CRPLUS:
3026        case OP_CRMINPLUS:
3027        case OP_CRPOSPLUS:
3028        end++;
3029        break;
3030    
3031        case OP_CRRANGE:
3032        case OP_CRMINRANGE:
3033        case OP_CRPOSRANGE:
3034        list[1] = (GET2(end, 1) == 0);
3035        end += 1 + 2 * IMM2_SIZE;
3036        break;
3037        }
3038      list[2] = (pcre_uint32)(end - code);
3039      return end;
3040      }
3041    return NULL;    /* Opcode not accepted */
3042    }
3043    
3044    
3045    
3046    /*************************************************
3047    *    Scan further character sets for match       *
3048    *************************************************/
3049    
3050    /* Checks whether the base and the current opcode have a common character, in
3051    which case the base cannot be possessified.
3052    
3053    Arguments:
3054      code        points to the byte code
3055      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3056      cd          static compile data
3057      base_list   the data list of the base opcode
3058    
3059    Returns:      TRUE if the auto-possessification is possible
3060    */
3061    
3062    static BOOL
3063    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3064      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3065    {
3066    pcre_uchar c;
3067    pcre_uint32 list[8];
3068    const pcre_uint32 *chr_ptr;
3069    const pcre_uint32 *ochr_ptr;
3070    const pcre_uint32 *list_ptr;
3071    const pcre_uchar *next_code;
3072    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3073    const pcre_uchar *xclass_flags;
3074    #endif
3075    const pcre_uint8 *class_bitset;
3076    const pcre_uint8 *set1, *set2, *set_end;
3077    pcre_uint32 chr;
3078    BOOL accepted, invert_bits;
3079    
3080    /* Note: the base_list[1] contains whether the current opcode has greedy
3081    (represented by a non-zero value) quantifier. This is a different from
3082    other character type lists, which stores here that the character iterator
3083    matches to an empty string (also represented by a non-zero value). */
3084    
3085    for(;;)
3086      {
3087      /* All operations move the code pointer forward.
3088      Therefore infinite recursions are not possible. */
3089    
3090      c = *code;
3091    
3092      /* Skip over callouts */
3093    
3094      if (c == OP_CALLOUT)
3095        {
3096        code += PRIV(OP_lengths)[c];
3097        continue;
3098        }
3099    
3100      if (c == OP_ALT)
3101        {
3102        do code += GET(code, 1); while (*code == OP_ALT);
3103        c = *code;
3104        }
3105    
3106      switch(c)
3107        {
3108        case OP_END:
3109        case OP_KETRPOS:
3110        /* TRUE only in greedy case. The non-greedy case could be replaced by
3111        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3112        uses more memory, which we cannot get at this stage.) */
3113    
3114        return base_list[1] != 0;
3115    
3116        case OP_KET:
3117        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3118        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3119        cannot be converted to a possessive form. */
3120    
3121        if (base_list[1] == 0) return FALSE;
3122    
3123        switch(*(code - GET(code, 1)))
3124          {
3125          case OP_ASSERT:
3126          case OP_ASSERT_NOT:
3127          case OP_ASSERTBACK:
3128          case OP_ASSERTBACK_NOT:
3129          case OP_ONCE:
3130          case OP_ONCE_NC:
3131          /* Atomic sub-patterns and assertions can always auto-possessify their
3132          last iterator. */
3133          return TRUE;
3134          }
3135    
3136        code += PRIV(OP_lengths)[c];
3137        continue;
3138    
3139        case OP_ONCE:
3140        case OP_ONCE_NC:
3141        case OP_BRA:
3142        case OP_CBRA:
3143        next_code = code + GET(code, 1);
3144        code += PRIV(OP_lengths)[c];
3145    
3146        while (*next_code == OP_ALT)
3147          {
3148          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3149          code = next_code + 1 + LINK_SIZE;
3150          next_code += GET(next_code, 1);
3151          }
3152        continue;
3153    
3154        case OP_BRAZERO:
3155        case OP_BRAMINZERO:
3156    
3157        next_code = code + 1;
3158        if (*next_code != OP_BRA && *next_code != OP_CBRA
3159            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3160    
3161        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3162    
3163        /* The bracket content will be checked by the
3164        OP_BRA/OP_CBRA case above. */
3165        next_code += 1 + LINK_SIZE;
3166        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3167          return FALSE;
3168    
3169        code += PRIV(OP_lengths)[c];
3170        continue;
3171        }
3172    
3173      /* Check for a supported opcode, and load its properties. */
3174    
3175      code = get_chr_property_list(code, utf, cd->fcc, list);
3176      if (code == NULL) return FALSE;    /* Unsupported */
3177    
3178      /* If either opcode is a small character list, set pointers for comparing
3179      characters from that list with another list, or with a property. */
3180    
3181      if (base_list[0] == OP_CHAR)
3182        {
3183        chr_ptr = base_list + 2;
3184        list_ptr = list;
3185        }
3186      else if (list[0] == OP_CHAR)
3187        {
3188        chr_ptr = list + 2;
3189        list_ptr = base_list;
3190        }
3191    
3192      /* Character bitsets can also be compared to certain opcodes. */
3193    
3194      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3195    #ifdef COMPILE_PCRE8
3196          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3197          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3198    #endif
3199          )
3200        {
3201    #ifdef COMPILE_PCRE8
3202        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3203    #else
3204        if (base_list[0] == OP_CLASS)
3205    #endif
3206          {
3207          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3208          list_ptr = list;
3209          }
3210        else
3211          {
3212          set1 = (pcre_uint8 *)(code - list[2]);
3213          list_ptr = base_list;
3214          }
3215    
3216        invert_bits = FALSE;
3217        switch(list_ptr[0])
3218          {
3219          case OP_CLASS:
3220          case OP_NCLASS:
3221          set2 = (pcre_uint8 *)
3222            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3223          break;
3224    
3225    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3226          case OP_XCLASS:
3227          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3228          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3229          if ((*xclass_flags & XCL_MAP) == 0)
3230            {
3231            /* No bits are set for characters < 256. */
3232            if (list[1] == 0) return TRUE;
3233            /* Might be an empty repeat. */
3234            continue;
3235            }
3236          set2 = (pcre_uint8 *)(xclass_flags + 1);
3237          break;
3238    #endif
3239    
3240          case OP_NOT_DIGIT:
3241          invert_bits = TRUE;
3242          /* Fall through */
3243          case OP_DIGIT:
3244          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3245          break;
3246    
3247          case OP_NOT_WHITESPACE:
3248          invert_bits = TRUE;
3249          /* Fall through */
3250          case OP_WHITESPACE:
3251          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3252          break;
3253    
3254          case OP_NOT_WORDCHAR:
3255          invert_bits = TRUE;
3256          /* Fall through */
3257          case OP_WORDCHAR:
3258          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3259          break;
3260    
3261          default:
3262          return FALSE;
3263          }
3264    
3265        /* Because the sets are unaligned, we need
3266        to perform byte comparison here. */
3267        set_end = set1 + 32;
3268        if (invert_bits)
3269          {
3270          do
3271            {
3272            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3273            }
3274          while (set1 < set_end);
3275          }
3276        else
3277          {
3278          do
3279            {
3280            if ((*set1++ & *set2++) != 0) return FALSE;
3281            }
3282          while (set1 < set_end);
3283          }
3284    
3285        if (list[1] == 0) return TRUE;
3286        /* Might be an empty repeat. */
3287        continue;
3288        }
3289    
3290      /* Some property combinations also acceptable. Unicode property opcodes are
3291      processed specially; the rest can be handled with a lookup table. */
3292    
3293      else
3294        {
3295        pcre_uint32 leftop, rightop;
3296    
3297        leftop = base_list[0];
3298        rightop = list[0];
3299    
3300    #ifdef SUPPORT_UCP
3301        accepted = FALSE; /* Always set in non-unicode case. */
3302        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3303          {
3304          if (rightop == OP_EOD)
3305            accepted = TRUE;
3306          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3307            {
3308            int n;
3309            const pcre_uint8 *p;
3310            BOOL same = leftop == rightop;
3311            BOOL lisprop = leftop == OP_PROP;
3312            BOOL risprop = rightop == OP_PROP;
3313            BOOL bothprop = lisprop && risprop;
3314    
3315            /* There's a table that specifies how each combination is to be
3316            processed:
3317              0   Always return FALSE (never auto-possessify)
3318              1   Character groups are distinct (possessify if both are OP_PROP)
3319              2   Check character categories in the same group (general or particular)
3320              3   Return TRUE if the two opcodes are not the same
3321              ... see comments below
3322            */
3323    
3324            n = propposstab[base_list[2]][list[2]];
3325            switch(n)
3326              {
3327              case 0: break;
3328              case 1: accepted = bothprop; break;
3329              case 2: accepted = (base_list[3] == list[3]) != same; break;
3330              case 3: accepted = !same; break;
3331    
3332              case 4:  /* Left general category, right particular category */
3333              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3334              break;
3335    
3336              case 5:  /* Right general category, left particular category */
3337              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3338              break;
3339    
3340              /* This code is logically tricky. Think hard before fiddling with it.
3341              The posspropstab table has four entries per row. Each row relates to
3342              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3343              Only WORD actually needs all four entries, but using repeats for the
3344              others means they can all use the same code below.
3345    
3346              The first two entries in each row are Unicode general categories, and
3347              apply always, because all the characters they include are part of the
3348              PCRE character set. The third and fourth entries are a general and a
3349              particular category, respectively, that include one or more relevant
3350              characters. One or the other is used, depending on whether the check
3351              is for a general or a particular category. However, in both cases the
3352              category contains more characters than the specials that are defined
3353              for the property being tested against. Therefore, it cannot be used
3354              in a NOTPROP case.
3355    
3356              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3357              Underscore is covered by ucp_P or ucp_Po. */
3358    
3359              case 6:  /* Left alphanum vs right general category */
3360              case 7:  /* Left space vs right general category */
3361              case 8:  /* Left word vs right general category */
3362              p = posspropstab[n-6];
3363              accepted = risprop && lisprop ==
3364                (list[3] != p[0] &&
3365                 list[3] != p[1] &&
3366                (list[3] != p[2] || !lisprop));
3367              break;
3368    
3369              case 9:   /* Right alphanum vs left general category */
3370              case 10:  /* Right space vs left general category */
3371              case 11:  /* Right word vs left general category */
3372              p = posspropstab[n-9];
3373              accepted = lisprop && risprop ==
3374                (base_list[3] != p[0] &&
3375                 base_list[3] != p[1] &&
3376                (base_list[3] != p[2] || !risprop));
3377              break;
3378    
3379              case 12:  /* Left alphanum vs right particular category */
3380              case 13:  /* Left space vs right particular category */
3381              case 14:  /* Left word vs right particular category */
3382              p = posspropstab[n-12];
3383              accepted = risprop && lisprop ==
3384                (catposstab[p[0]][list[3]] &&
3385                 catposstab[p[1]][list[3]] &&
3386                (list[3] != p[3] || !lisprop));
3387              break;
3388    
3389              case 15:  /* Right alphanum vs left particular category */
3390              case 16:  /* Right space vs left particular category */
3391              case 17:  /* Right word vs left particular category */
3392              p = posspropstab[n-15];
3393              accepted = lisprop && risprop ==
3394                (catposstab[p[0]][base_list[3]] &&
3395                 catposstab[p[1]][base_list[3]] &&
3396                (base_list[3] != p[3] || !risprop));
3397              break;
3398              }
3399            }
3400          }
3401    
3402        else
3403    #endif  /* SUPPORT_UCP */
3404    
3405        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3406               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3407               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3408    
3409        if (!accepted)
3410          return FALSE;
3411    
3412        if (list[1] == 0) return TRUE;
3413        /* Might be an empty repeat. */
3414        continue;
3415        }
3416    
3417      /* Control reaches here only if one of the items is a small character list.
3418      All characters are checked against the other side. */
3419    
3420      do
3421        {
3422        chr = *chr_ptr;
3423    
3424        switch(list_ptr[0])
3425          {
3426          case OP_CHAR:
3427          ochr_ptr = list_ptr + 2;
3428          do
3429            {
3430            if (chr == *ochr_ptr) return FALSE;
3431            ochr_ptr++;
3432            }
3433          while(*ochr_ptr != NOTACHAR);
3434          break;
3435    
3436          case OP_NOT:
3437          ochr_ptr = list_ptr + 2;
3438          do
3439            {
3440            if (chr == *ochr_ptr)
3441              break;
3442            ochr_ptr++;
3443            }
3444          while(*ochr_ptr != NOTACHAR);
3445          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3446          break;
3447    
3448          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3449          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3450    
3451          case OP_DIGIT:
3452          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3453          break;
3454    
3455          case OP_NOT_DIGIT:
3456          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3457          break;
3458    
3459          case OP_WHITESPACE:
3460          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3461          break;
3462    
3463          case OP_NOT_WHITESPACE:
3464          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3465          break;
3466    
3467          case OP_WORDCHAR:
3468          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3469          break;
3470    
3471          case OP_NOT_WORDCHAR:
3472          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3473          break;
3474    
3475          case OP_HSPACE:
3476          switch(chr)
3477            {
3478            HSPACE_CASES: return FALSE;
3479            default: break;
3480            }
3481          break;
3482    
3483          case OP_NOT_HSPACE:
3484          switch(chr)
3485            {
3486            HSPACE_CASES: break;
3487            default: return FALSE;
3488            }
3489          break;
3490    
3491          case OP_ANYNL:
3492          case OP_VSPACE:
3493          switch(chr)
3494            {
3495            VSPACE_CASES: return FALSE;
3496            default: break;
3497            }
3498          break;
3499    
3500          case OP_NOT_VSPACE:
3501          switch(chr)
3502            {
3503            VSPACE_CASES: break;
3504            default: return FALSE;
3505            }
3506          break;
3507    
3508          case OP_DOLL:
3509          case OP_EODN:
3510          switch (chr)
3511            {
3512            case CHAR_CR:
3513            case CHAR_LF:
3514            case CHAR_VT:
3515            case CHAR_FF:
3516            case CHAR_NEL:
3517    #ifndef EBCDIC
3518            case 0x2028:
3519            case 0x2029:
3520    #endif  /* Not EBCDIC */
3521            return FALSE;
3522            }
3523          break;
3524    
3525          case OP_EOD:    /* Can always possessify before \z */
3526          break;
3527    
3528    #ifdef SUPPORT_UCP
3529          case OP_PROP:
3530          case OP_NOTPROP:
3531          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3532                list_ptr[0] == OP_NOTPROP))
3533            return FALSE;
3534          break;
3535    #endif
3536    
3537          case OP_NCLASS:
3538          if (chr > 255) return FALSE;
3539          /* Fall through */
3540    
3541          case OP_CLASS:
3542          if (chr > 255) break;
3543          class_bitset = (pcre_uint8 *)
3544            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3545          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3546          break;
3547    
3548    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3549          case OP_XCLASS:
3550          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3551              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3552          break;
3553    #endif
3554    
3555          default:
3556          return FALSE;
3557          }
3558    
3559        chr_ptr++;
3560        }
3561      while(*chr_ptr != NOTACHAR);