/[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 637 by ph10, Sun Jul 24 17:44:12 2011 UTC revision 1382 by zherczeg, Fri Oct 18 07:55:07 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2011 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When 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 231  static const char posix_names[] = Line 272  static const char posix_names[] =
272    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
273    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
274    
275  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
276    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 };
277    
278  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 266  substitutes must be in the order of the Line 307  substitutes must be in the order of the
307  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
308    
309  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
310  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
311    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
312    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
313    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
314    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
315    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
316    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
317      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pXsp[] = {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_PXwd[] = {
323      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_pXwd[] = {
326      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    
329    static const pcre_uchar *substitutes[] = {
330      string_PNd,           /* \D */
331      string_pNd,           /* \d */
332      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
333      string_pXsp,          /* \s */
334      string_PXwd,          /* \W */
335      string_pXwd           /* \w */
336  };  };
337    
338  static const uschar *posix_substitutes[] = {  static const pcre_uchar string_pL[] =   {
339    (uschar *)"\\p{L}",     /* alpha */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
340    (uschar *)"\\p{Ll}",    /* lower */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
341    (uschar *)"\\p{Lu}",    /* upper */  static const pcre_uchar string_pLl[] =  {
342    (uschar *)"\\p{Xan}",   /* alnum */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
343    NULL,                   /* ascii */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
344    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pLu[] =  {
345    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
346    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347    NULL,                   /* graph */  static const pcre_uchar string_pXan[] = {
348    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
349    NULL,                   /* punct */    CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_h[] =    {
351    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_h, '\0' };
352    NULL,                   /* xdigit */  static const pcre_uchar string_pXps[] = {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_PL[] =   {
356      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_PLl[] =  {
359      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_PLu[] =  {
362      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
363      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
364    static const pcre_uchar string_PXan[] = {
365      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
366      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
367    static const pcre_uchar string_H[] =    {
368      CHAR_BACKSLASH, CHAR_H, '\0' };
369    static const pcre_uchar string_PXps[] = {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    
373    static const pcre_uchar *posix_substitutes[] = {
374      string_pL,            /* alpha */
375      string_pLl,           /* lower */
376      string_pLu,           /* upper */
377      string_pXan,          /* alnum */
378      NULL,                 /* ascii */
379      string_h,             /* blank */
380      NULL,                 /* cntrl */
381      string_pNd,           /* digit */
382      NULL,                 /* graph */
383      NULL,                 /* print */
384      NULL,                 /* punct */
385      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
386      string_pXwd,          /* word */
387      NULL,                 /* xdigit */
388    /* Negated cases */    /* Negated cases */
389    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
390    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
391    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
392    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
393    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
394    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
395    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
396    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
397    NULL,                   /* ^graph */    NULL,                 /* ^graph */
398    NULL,                   /* ^print */    NULL,                 /* ^print */
399    NULL,                   /* ^punct */    NULL,                 /* ^punct */
400    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
401    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */
402    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
403  };  };
404  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
405  #endif  #endif
406    
407  #define STRING(a)  # a  #define STRING(a)  # a
# Line 365  static const char error_texts[] = Line 460  static const char error_texts[] =
460    /* 30 */    /* 30 */
461    "unknown POSIX class name\0"    "unknown POSIX class name\0"
462    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
463    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
464    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
465    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
466    /* 35 */    /* 35 */
467    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
468    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 388  static const char error_texts[] = Line 483  static const char error_texts[] =
483    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
484    /* 50 */    /* 50 */
485    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
486    "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"
487    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
488    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
489    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
# Line 399  static const char error_texts[] = Line 494  static const char error_texts[] =
494    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
495    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496    /* 60 */    /* 60 */
497    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
498    "number is too big\0"    "number is too big\0"
499    "subpattern name expected\0"    "subpattern name expected\0"
500    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 407  static const char error_texts[] = Line 502  static const char error_texts[] =
502    /* 65 */    /* 65 */
503    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
504    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
505    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with Unicode property support\0"
506    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
507    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
508      /* 70 */
509      "internal error: unknown opcode in find_fixedlength()\0"
510      "\\N is not supported in a class\0"
511      "too many forward references\0"
512      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
513      "invalid UTF-16 string\0"
514      /* 75 */
515      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516      "character value in \\u.... sequence is too large\0"
517      "invalid UTF-32 string\0"
518      "setting UTF is disabled by the application\0"
519      "non-hex character in \\x{} (closing brace missing?)\0"
520      /* 80 */
521      "non-octal character in \\o{} (closing brace missing?)\0"
522      "missing opening brace after \\o\0"
523    ;    ;
524    
525  /* 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 428  For convenience, we use the same bit def Line 538  For convenience, we use the same bit def
538    
539  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
540    
541    /* Using a simple comparison for decimal numbers rather than a memory read
542    is much faster, and the resulting code is simpler (the compiler turns it
543    into a subtraction and unsigned comparison). */
544    
545    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
546    
547  #ifndef EBCDIC  #ifndef EBCDIC
548    
549  /* 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
550  UTF-8 mode. */  UTF-8 mode. */
551    
552  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
553    {    {
554    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
555    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 472  static const unsigned char digitab[] = Line 588  static const unsigned char digitab[] =
588    
589  /* 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. */
590    
591  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
592    {    {
593    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
594    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 507  static const unsigned char digitab[] = Line 623  static const unsigned char digitab[] =
623    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
624    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
625    
626  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
627    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
628    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
629    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 543  static const unsigned char ebcdic_charta Line 659  static const unsigned char ebcdic_charta
659  #endif  #endif
660    
661    
662  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
663    between adjacent character-type opcodes. The left-hand (repeated) opcode is
664    used to select the row, and the right-hand opcode is use to select the column.
665    A value of 1 means that auto-possessification is OK. For example, the second
666    value in the first row means that \D+\d can be turned into \D++\d.
667    
668    The Unicode property types (\P and \p) have to be present to fill out the table
669    because of what their opcode values are, but the table values should always be
670    zero because property types are handled separately in the code. The last four
671    columns apply to items that cannot be repeated, so there is no need to have
672    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674    
675    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677    
678    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
680      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
681      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
682      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
683      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
684      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
685      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
687      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
689      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
691      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
693      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
694      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
696      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
697    };
698    
699    
700  static BOOL  /* This table is used to check whether auto-possessification is possible
701    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702      int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
703    opcode is used to select the column. The values are as follows:
704    
705      0   Always return FALSE (never auto-possessify)
706      1   Character groups are distinct (possessify if both are OP_PROP)
707      2   Check character categories in the same group (general or particular)
708      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709    
710      4   Check left general category vs right particular category
711      5   Check right general category vs left particular category
712    
713      6   Left alphanum vs right general category
714      7   Left space vs right general category
715      8   Left word vs right general category
716    
717      9   Right alphanum vs left general category
718     10   Right space vs left general category
719     11   Right word vs left general category
720    
721     12   Left alphanum vs right particular category
722     13   Left space vs right particular category
723     14   Left word vs right particular category
724    
725     15   Right alphanum vs left particular category
726     16   Right space vs left particular category
727     17   Right word vs left particular category
728    */
729    
730    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
733      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
734      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
735      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
736      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
737      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
738      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
739      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
740      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
741      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
742      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
743    };
744    
745    /* This table is used to check whether auto-possessification is possible
746    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747    specifies a general category and the other specifies a particular category. The
748    row is selected by the general category and the column by the particular
749    category. The value is 1 if the particular category is not part of the general
750    category. */
751    
752    static const pcre_uint8 catposstab[7][30] = {
753    /* 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 */
754      { 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 */
755      { 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 */
756      { 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 */
757      { 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 */
758      { 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 */
759      { 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 */
760      { 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 */
761    };
762    
763    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764    a general or particular category. The properties in each row are those
765    that apply to the character set in question. Duplication means that a little
766    unnecessary work is done when checking, but this keeps things much simpler
767    because they can all use the same code. For more details see the comment where
768    this table is used.
769    
770    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771    "space", but from Perl 5.18 it's included, so both categories are treated the
772    same here. */
773    
774    static const pcre_uint8 posspropstab[3][4] = {
775      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
776      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
777      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
778    };
779    
780    /* This table is used when converting repeating opcodes into possessified
781    versions as a result of an explicit possessive quantifier such as ++. A zero
782    value means there is no possessified version - in those cases the item in
783    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
784    because all relevant opcodes are less than that. */
785    
786    static const pcre_uint8 opcode_possessify[] = {
787      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
788      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
789    
790      0,                       /* NOTI */
791      OP_POSSTAR, 0,           /* STAR, MINSTAR */
792      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
793      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
794      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
795      0,                       /* EXACT */
796      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
797    
798      OP_POSSTARI, 0,          /* STARI, MINSTARI */
799      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
800      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
801      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
802      0,                       /* EXACTI */
803      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
804    
805      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
806      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
807      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
808      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
809      0,                       /* NOTEXACT */
810      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
813      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
814      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
815      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
816      0,                       /* NOTEXACTI */
817      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
820      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
821      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
822      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
823      0,                       /* TYPEEXACT */
824      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
827      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
828      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
829      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
830      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
831    
832      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
833      0, 0,                    /* REF, REFI */
834      0, 0,                    /* DNREF, DNREFI */
835      0, 0                     /* RECURSE, CALLOUT */
836    };
837    
838    
839    
# Line 570  find_error_text(int n) Line 856  find_error_text(int n)
856  const char *s = error_texts;  const char *s = error_texts;
857  for (; n > 0; n--)  for (; n > 0; n--)
858    {    {
859    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
860    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
861    }    }
862  return s;  return s;
863  }  }
864    
865    
866    
867    /*************************************************
868    *           Expand the workspace                 *
869    *************************************************/
870    
871    /* This function is called during the second compiling phase, if the number of
872    forward references fills the existing workspace, which is originally a block on
873    the stack. A larger block is obtained from malloc() unless the ultimate limit
874    has been reached or the increase will be rather small.
875    
876    Argument: pointer to the compile data block
877    Returns:  0 if all went well, else an error number
878    */
879    
880    static int
881    expand_workspace(compile_data *cd)
882    {
883    pcre_uchar *newspace;
884    int newsize = cd->workspace_size * 2;
885    
886    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
887    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
888        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
889     return ERR72;
890    
891    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
892    if (newspace == NULL) return ERR21;
893    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
894    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
895    if (cd->workspace_size > COMPILE_WORK_SIZE)
896      (PUBL(free))((void *)cd->start_workspace);
897    cd->start_workspace = newspace;
898    cd->workspace_size = newsize;
899    return 0;
900    }
901    
902    
903    
904    /*************************************************
905    *            Check for counted repeat            *
906    *************************************************/
907    
908    /* This function is called when a '{' is encountered in a place where it might
909    start a quantifier. It looks ahead to see if it really is a quantifier or not.
910    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
911    where the ddds are digits.
912    
913    Arguments:
914      p         pointer to the first char after '{'
915    
916    Returns:    TRUE or FALSE
917    */
918    
919    static BOOL
920    is_counted_repeat(const pcre_uchar *p)
921    {
922    if (!IS_DIGIT(*p)) return FALSE;
923    p++;
924    while (IS_DIGIT(*p)) p++;
925    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
926    
927    if (*p++ != CHAR_COMMA) return FALSE;
928    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
929    
930    if (!IS_DIGIT(*p)) return FALSE;
931    p++;
932    while (IS_DIGIT(*p)) p++;
933    
934    return (*p == CHAR_RIGHT_CURLY_BRACKET);
935    }
936    
937    
938    
939  /*************************************************  /*************************************************
940  *            Handle escapes                      *  *            Handle escapes                      *
941  *************************************************/  *************************************************/
942    
943  /* 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
944  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
945  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.
946  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
947  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
948  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
949    
950  Arguments:  Arguments:
951    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
952      chptr          points to a returned data character
953    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
954    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
955    options        the options bits    options        the options bits
956    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
957    
958  Returns:         zero or positive => a data character  Returns:         zero => a data character
959                   negative => a special escape sequence                   positive => a special escape sequence
960                     negative => a back reference
961                   on error, errorcodeptr is set                   on error, errorcodeptr is set
962  */  */
963    
964  static int  static int
965  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
966    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
967  {  {
968  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
969  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
970  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
971    pcre_uint32 c;
972    int escape = 0;
973    int i;
974    
975  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
976  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
977    
978  /* 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. */
979    
980  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
981    
982  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
983  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.
984  Otherwise further processing may be required. */  Otherwise further processing may be required. */
985    
986  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
987  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
988  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
989    else if ((i = escapes[c - CHAR_0]) != 0)
990      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
991    
992  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
993  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
994  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
995    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
996  #endif  #endif
997    
998  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
999    
1000  else  else
1001    {    {
1002    const uschar *oldptr;    const pcre_uchar *oldptr;
1003    BOOL braced, negated;    BOOL braced, negated, overflow;
1004      int s;
1005    
1006    switch (c)    switch (c)
1007      {      {
# Line 643  else Line 1010  else
1010    
1011      case CHAR_l:      case CHAR_l:
1012      case CHAR_L:      case CHAR_L:
1013        *errorcodeptr = ERR37;
1014        break;
1015    
1016      case CHAR_u:      case CHAR_u:
1017        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1018          {
1019          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1020          Otherwise it is a lowercase u letter. */
1021          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1022            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1023            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1024            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1025            {
1026            c = 0;
1027            for (i = 0; i < 4; ++i)
1028              {
1029              register pcre_uint32 cc = *(++ptr);
1030    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1031              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1032              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1033    #else           /* EBCDIC coding */
1034              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1035              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1036    #endif
1037              }
1038    
1039    #if defined COMPILE_PCRE8
1040            if (c > (utf ? 0x10ffffU : 0xffU))
1041    #elif defined COMPILE_PCRE16
1042            if (c > (utf ? 0x10ffffU : 0xffffU))
1043    #elif defined COMPILE_PCRE32
1044            if (utf && c > 0x10ffffU)
1045    #endif
1046              {
1047              *errorcodeptr = ERR76;
1048              }
1049            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1050            }
1051          }
1052        else
1053          *errorcodeptr = ERR37;
1054        break;
1055    
1056      case CHAR_U:      case CHAR_U:
1057      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
1058        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1059      break;      break;
1060    
1061      /* \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
1062        class, \g must be followed by one of a number of specific things:
1063    
1064      (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
1065      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 662  else Line 1073  else
1073      (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
1074      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1075      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1076      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1077    
1078      case CHAR_g:      case CHAR_g:
1079        if (isclass) break;
1080      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1081        {        {
1082        c = -ESC_g;        escape = ESC_g;
1083        break;        break;
1084        }        }
1085    
# Line 675  else Line 1087  else
1087    
1088      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1089        {        {
1090        const uschar *p;        const pcre_uchar *p;
1091        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++)
1092          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1093        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1094          {          {
1095          c = -ESC_k;          escape = ESC_k;
1096          break;          break;
1097          }          }
1098        braced = TRUE;        braced = TRUE;
# Line 695  else Line 1107  else
1107        }        }
1108      else negated = FALSE;      else negated = FALSE;
1109    
1110      c = 0;      /* The integer range is limited by the machine's int representation. */
1111      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1112        c = c * 10 + *(++ptr) - CHAR_0;      overflow = FALSE;
1113        while (IS_DIGIT(ptr[1]))
1114      if (c < 0)   /* Integer overflow */        {
1115          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1116            {
1117            overflow = TRUE;
1118            break;
1119            }
1120          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1121          }
1122        if (overflow) /* Integer overflow */
1123        {        {
1124          while (IS_DIGIT(ptr[1]))
1125            ptr++;
1126        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1127        break;        break;
1128        }        }
# Line 711  else Line 1133  else
1133        break;        break;
1134        }        }
1135    
1136      if (c == 0)      if (s == 0)
1137        {        {
1138        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1139        break;        break;
# Line 719  else Line 1141  else
1141    
1142      if (negated)      if (negated)
1143        {        {
1144        if (c > bracount)        if (s > bracount)
1145          {          {
1146          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1147          break;          break;
1148          }          }
1149        c = bracount - (c - 1);        s = bracount - (s - 1);
1150        }        }
1151    
1152      c = -(ESC_REF + c);      escape = -s;
1153      break;      break;
1154    
1155      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1156      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
1157      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1158        recommended to avoid the ambiguities in the old syntax.
1159    
1160      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
1161      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
1162      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
1163      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
1164      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
1165      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
1166      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1167    
1168        Inside a character class, \ followed by a digit is always either a literal
1169        8 or 9 or an octal number. */
1170    
1171      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:
1172      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 748  else Line 1174  else
1174      if (!isclass)      if (!isclass)
1175        {        {
1176        oldptr = ptr;        oldptr = ptr;
1177        c -= CHAR_0;        /* The integer range is limited by the machine's int representation. */
1178        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1179          c = c * 10 + *(++ptr) - CHAR_0;        overflow = FALSE;
1180        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1181            {
1182            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1183              {
1184              overflow = TRUE;
1185              break;
1186              }
1187            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1188            }
1189          if (overflow) /* Integer overflow */
1190          {          {
1191            while (IS_DIGIT(ptr[1]))
1192              ptr++;
1193          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1194          break;          break;
1195          }          }
1196        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1197          {          {
1198          c = -(ESC_REF + c);          escape = -s;
1199          break;          break;
1200          }          }
1201        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1202        }        }
1203    
1204      /* 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
1205      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
1206      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
1207        changed so as not to insert the binary zero. */
1208    
1209      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1210        {  
1211        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1212    
1213      /* \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
1214      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
1215      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
1216      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,
1217      than 3 octal digits. */      but no more than 3 octal digits. */
1218    
1219      case CHAR_0:      case CHAR_0:
1220      c -= CHAR_0;      c -= CHAR_0;
1221      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1222          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1223      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1224        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1225    #endif
1226      break;      break;
1227    
1228      /* \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
1229      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. */  
1230    
1231      case CHAR_x:      case CHAR_o:
1232      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1233        {        {
1234        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1235        c = 0;        c = 0;
1236        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1237          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1238          {          {
1239          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1240          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1241          count++;  #ifdef COMPILE_PCRE32
1242            if (c >= 0x20000000l) { overflow = TRUE; break; }
1243    #endif
1244            c = (c << 3) + cc - CHAR_0 ;
1245    #if defined COMPILE_PCRE8
1246            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1247    #elif defined COMPILE_PCRE16
1248            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1249    #elif defined COMPILE_PCRE32
1250            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1251    #endif
1252            }
1253          if (overflow)
1254            {
1255            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1256            *errorcodeptr = ERR34;
1257            }
1258          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1259            {
1260            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1261            }
1262          else *errorcodeptr = ERR80;
1263          }
1264        break;
1265    
1266        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1267        numbers. Otherwise it is a lowercase x letter. */
1268    
1269        case CHAR_x:
1270        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1271          {
1272          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1273            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1274            {
1275            c = 0;
1276            for (i = 0; i < 2; ++i)
1277              {
1278              register pcre_uint32 cc = *(++ptr);
1279  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1280          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1281          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1282  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1283          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 */
1284          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1285  #endif  #endif
1286              }
1287          }          }
1288          }    /* End JavaScript handling */
1289    
1290        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1291        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1292        digits. If not, { used to be treated as a data character. However, Perl
1293        seems to read hex digits up to the first non-such, and ignore the rest, so
1294        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1295        now gives an error. */
1296    
1297        if (*pt == CHAR_RIGHT_CURLY_BRACKET)      else
1298          {
1299          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1300          {          {
1301          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          ptr += 2;
1302          ptr = pt;          c = 0;
1303          break;          overflow = FALSE;
1304          }          while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1305              {
1306              register pcre_uint32 cc = *ptr++;
1307              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1308    
1309        /* If the sequence of hex digits does not end with '}', then we don't  #ifdef COMPILE_PCRE32
1310        recognize this construct; fall through to the normal \x handling. */            if (c >= 0x10000000l) { overflow = TRUE; break; }
1311        }  #endif
1312    
1313      /* Read just a single-byte hex-defined char */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1314              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1315              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1316    #else           /* EBCDIC coding */
1317              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1318              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1319    #endif
1320    
1321      c = 0;  #if defined COMPILE_PCRE8
1322      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1323        {  #elif defined COMPILE_PCRE16
1324        int cc;                                  /* Some compilers don't like */            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1325        cc = *(++ptr);                           /* ++ in initializers */  #elif defined COMPILE_PCRE32
1326              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1327    #endif
1328              }
1329    
1330            if (overflow)
1331              {
1332              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1333              *errorcodeptr = ERR34;
1334              }
1335    
1336            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1337              {
1338              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1339              }
1340    
1341            /* If the sequence of hex digits does not end with '}', give an error.
1342            We used just to recognize this construct and fall through to the normal
1343            \x handling, but nowadays Perl gives an error, which seems much more
1344            sensible, so we do too. */
1345    
1346            else *errorcodeptr = ERR79;
1347            }   /* End of \x{} processing */
1348    
1349          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1350    
1351          else
1352            {
1353            c = 0;
1354            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1355              {
1356              pcre_uint32 cc;                          /* Some compilers don't like */
1357              cc = *(++ptr);                           /* ++ in initializers */
1358  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1359        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1360        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1361  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1362        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1363        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1364  #endif  #endif
1365        }            }
1366            }     /* End of \xdd handling */
1367          }       /* End of Perl-style \x handling */
1368      break;      break;
1369    
1370      /* 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 849  else Line 1374  else
1374    
1375      case CHAR_c:      case CHAR_c:
1376      c = *(++ptr);      c = *(++ptr);
1377      if (c == 0)      if (c == CHAR_NULL)
1378        {        {
1379        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1380        break;        break;
# Line 886  else Line 1411  else
1411    }    }
1412    
1413  /* 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
1414  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1415    quantification such as \N{2,3}. */
1416    
1417  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1418         !is_counted_repeat(ptr+2))
1419    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1420    
1421  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1422    
1423  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1424    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1425    
1426  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1427    
1428  *ptrptr = ptr;  *ptrptr = ptr;
1429  return c;  *chptr = c;
1430    return escape;
1431  }  }
1432    
1433    
# Line 917  escape sequence. Line 1445  escape sequence.
1445  Argument:  Argument:
1446    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1447    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
1448    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
1449      pdataptr       points to an unsigned int that is set to the detailed property value
1450    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1451    
1452  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
1453  */  */
1454    
1455  static int  static BOOL
1456  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1457      unsigned int *pdataptr, int *errorcodeptr)
1458  {  {
1459  int c, i, bot, top;  pcre_uchar c;
1460  const uschar *ptr = *ptrptr;  int i, bot, top;
1461  char name[32];  const pcre_uchar *ptr = *ptrptr;
1462    pcre_uchar name[32];
1463    
1464  c = *(++ptr);  c = *(++ptr);
1465  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1466    
1467  *negptr = FALSE;  *negptr = FALSE;
1468    
# Line 945  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1476  if (c == CHAR_LEFT_CURLY_BRACKET)
1476      *negptr = TRUE;      *negptr = TRUE;
1477      ptr++;      ptr++;
1478      }      }
1479    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1480      {      {
1481      c = *(++ptr);      c = *(++ptr);
1482      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1483      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1484      name[i] = c;      name[i] = c;
1485      }      }
# Line 969  else Line 1500  else
1500  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1501    
1502  bot = 0;  bot = 0;
1503  top = _pcre_utt_size;  top = PRIV(utt_size);
1504    
1505  while (bot < top)  while (bot < top)
1506    {    {
1507      int r;
1508    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1509    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1510    if (c == 0)    if (r == 0)
1511      {      {
1512      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1513      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1514        return TRUE;
1515      }      }
1516    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1517    }    }
1518    
1519  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1520  *ptrptr = ptr;  *ptrptr = ptr;
1521  return -1;  return FALSE;
1522    
1523  ERROR_RETURN:  ERROR_RETURN:
1524  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1525  *ptrptr = ptr;  *ptrptr = ptr;
1526  return -1;  return FALSE;
1527  }  }
1528  #endif  #endif
1529    
1530    
1531    
   
 /*************************************************  
 *            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);  
 }  
   
   
   
1532  /*************************************************  /*************************************************
1533  *         Read repeat counts                     *  *         Read repeat counts                     *
1534  *************************************************/  *************************************************/
# Line 1049  Returns:         pointer to '}' on succe Line 1548  Returns:         pointer to '}' on succe
1548                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1549  */  */
1550    
1551  static const uschar *  static const pcre_uchar *
1552  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)
1553  {  {
1554  int min = 0;  int min = 0;
1555  int max = -1;  int max = -1;
# Line 1058  int max = -1; Line 1557  int max = -1;
1557  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1558  an integer overflow. */  an integer overflow. */
1559    
1560  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1561  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1562    {    {
1563    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1073  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1572  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1572    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1573      {      {
1574      max = 0;      max = 0;
1575      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1576      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1577        {        {
1578        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1098  return p; Line 1597  return p;
1597    
1598    
1599  /*************************************************  /*************************************************
1600  *  Subroutine for finding forward reference      *  *      Find first significant op code            *
1601  *************************************************/  *************************************************/
1602    
1603  /* This recursive function is called only from find_parens() below. The  /* This is called by several functions that scan a compiled expression looking
1604  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
1605  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
1606  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
1607  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.)  
1608    
1609  Arguments:  Arguments:
1610    ptrptr       address of the current character pointer (updated)    code         pointer to the start of the group
1611    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)  
1612    
1613  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1614  */  */
1615    
1616  static int  static const pcre_uchar*
1617  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)  
1618  {  {
1619  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)  
1620    {    {
1621    /* Handle specials such as (*SKIP) or (*UTF8) etc. */    switch ((int)*code)
1622        {
1623        case OP_ASSERT_NOT:
1624        case OP_ASSERTBACK:
1625        case OP_ASSERTBACK_NOT:
1626        if (!skipassert) return code;
1627        do code += GET(code, 1); while (*code == OP_ALT);
1628        code += PRIV(OP_lengths)[*code];
1629        break;
1630    
1631    if (ptr[1] == CHAR_ASTERISK) ptr += 2;      case OP_WORD_BOUNDARY:
1632        case OP_NOT_WORD_BOUNDARY:
1633        if (!skipassert) return code;
1634        /* Fall through */
1635    
1636    /* Handle a normal, unnamed capturing parenthesis. */      case OP_CALLOUT:
1637        case OP_CREF:
1638        case OP_DNCREF:
1639        case OP_RREF:
1640        case OP_DNRREF:
1641        case OP_DEF:
1642        code += PRIV(OP_lengths)[*code];
1643        break;
1644    
1645    else if (ptr[1] != CHAR_QUESTION_MARK)      default:
1646      {      return code;
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
1647      }      }
1648      }
1649    /* Control never reaches here */
1650    }
1651    
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
1652    
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
1653    
1654    /* Handle comments; all characters are allowed until a ket is reached. */  /*************************************************
1655    *        Find the fixed length of a branch       *
1656    *************************************************/
1657    
1658    else if (ptr[2] == CHAR_NUMBER_SIGN)  /* Scan a branch and compute the fixed length of subject that will match it,
1659      {  if the length is fixed. This is needed for dealing with backward assertions.
1660      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  In UTF8 mode, the result is in characters rather than bytes. The branch is
1661      goto FAIL_EXIT;  temporarily terminated with OP_END when this function is called.
     }  
1662    
1663    /* Handle a condition. If it is an assertion, just carry on so that it  This function is called when a backward assertion is encountered, so that if it
1664    is processed as normal. If not, skip to the closing parenthesis of the  fails, the error message can point to the correct place in the pattern.
1665    condition (there can't be any nested parens). */  However, we cannot do this when the assertion contains subroutine calls,
1666    because they can be forward references. We solve this by remembering this case
1667    and doing the check at the end; a flag specifies which mode we are running in.
1668    
1669    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  Arguments:
1670      {    code     points to the start of the pattern (the bracket)
1671      ptr += 2;    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1672      if (ptr[1] != CHAR_QUESTION_MARK)    atend    TRUE if called when the pattern is complete
1673        {    cd       the "compile data" structure
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
1674    
1675    /* Start with (? but not a condition. */  Returns:   the fixed length,
1676                 or -1 if there is no fixed length,
1677    else               or -2 if \C was encountered (in UTF-8 mode only)
1678      {               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1679      ptr += 2;               or -4 if an unknown opcode was encountered (internal error)
1680      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  
   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, 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       *  
 *************************************************/  
   
 /* Scan a branch and compute the fixed length of subject that will match it,  
 if the length is fixed. This is needed for dealing with backward assertions.  
 In UTF8 mode, the result is in characters rather than bytes. The branch is  
 temporarily terminated with OP_END when this function is called.  
   
 This function is called when a backward assertion is encountered, so that if it  
 fails, the error message can point to the correct place in the pattern.  
 However, we cannot do this when the assertion contains subroutine calls,  
 because they can be forward references. We solve this by remembering this case  
 and doing the check at the end; a flag specifies which mode we are running in.  
   
 Arguments:  
   code     points to the start of the pattern (the bracket)  
   utf8     TRUE in UTF-8 mode  
   atend    TRUE if called when the pattern is complete  
   cd       the "compile data" structure  
   
 Returns:   the fixed length,  
              or -1 if there is no fixed length,  
              or -2 if \C was encountered  
              or -3 if an OP_RECURSE item was encountered and atend is FALSE  
 */  
1681    
1682  static int  static int
1683  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1684  {  {
1685  int length = -1;  int length = -1;
1686    
1687  register int branchlength = 0;  register int branchlength = 0;
1688  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1689    
1690  /* 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
1691  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1489  branch, check the length against that of Line 1693  branch, check the length against that of
1693  for (;;)  for (;;)
1694    {    {
1695    int d;    int d;
1696    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1697    register int op = *cc;    register pcre_uchar op = *cc;
1698    
1699    switch (op)    switch (op)
1700      {      {
1701      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1702      OP_BRA (normal non-capturing bracket) because the other variants of these      OP_BRA (normal non-capturing bracket) because the other variants of these
1703      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1704      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1705    
1706      case OP_CBRA:      case OP_CBRA:
1707      case OP_BRA:      case OP_BRA:
1708      case OP_ONCE:      case OP_ONCE:
1709        case OP_ONCE_NC:
1710      case OP_COND:      case OP_COND:
1711      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1712      if (d < 0) return d;      if (d < 0) return d;
1713      branchlength += d;      branchlength += d;
1714      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1715      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1716      break;      break;
1717    
1718      /* 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.
1719      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
1720      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
1721      Note that we must not include the OP_KETRxxx opcodes here, because they      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1722      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1723    
1724      case OP_ALT:      case OP_ALT:
1725      case OP_KET:      case OP_KET:
1726      case OP_END:      case OP_END:
1727        case OP_ACCEPT:
1728        case OP_ASSERT_ACCEPT:
1729      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1730        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1731      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1532  for (;;) Line 1739  for (;;)
1739    
1740      case OP_RECURSE:      case OP_RECURSE:
1741      if (!atend) return -3;      if (!atend) return -3;
1742      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1743      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1744      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1745      d = find_fixedlength(cs + 2, utf8, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1746      if (d < 0) return d;      if (d < 0) return d;
1747      branchlength += d;      branchlength += d;
1748      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1548  for (;;) Line 1755  for (;;)
1755      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1756      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1757      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1758      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1759        break;
1760    
1761      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1762    
1763      case OP_REVERSE:      case OP_MARK:
1764      case OP_CREF:      case OP_PRUNE_ARG:
1765      case OP_NCREF:      case OP_SKIP_ARG:
1766      case OP_RREF:      case OP_THEN_ARG:
1767      case OP_NRREF:      cc += cc[1] + PRIV(OP_lengths)[*cc];
1768      case OP_DEF:      break;
1769    
1770      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1771      case OP_CIRC:      case OP_CIRC:
1772      case OP_CIRCM:      case OP_CIRCM:
1773        case OP_CLOSE:
1774        case OP_COMMIT:
1775        case OP_CREF:
1776        case OP_DEF:
1777        case OP_DNCREF:
1778        case OP_DNRREF:
1779      case OP_DOLL:      case OP_DOLL:
1780      case OP_DOLLM:      case OP_DOLLM:
1781        case OP_EOD:
1782        case OP_EODN:
1783        case OP_FAIL:
1784      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1785        case OP_PRUNE:
1786        case OP_REVERSE:
1787        case OP_RREF:
1788        case OP_SET_SOM:
1789        case OP_SKIP:
1790        case OP_SOD:
1791        case OP_SOM:
1792        case OP_THEN:
1793      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1794      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1795      break;      break;
1796    
1797      /* Handle literal characters */      /* Handle literal characters */
# Line 1581  for (;;) Line 1802  for (;;)
1802      case OP_NOTI:      case OP_NOTI:
1803      branchlength++;      branchlength++;
1804      cc += 2;      cc += 2;
1805  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1806      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1807  #endif  #endif
1808      break;      break;
1809    
# Line 1590  for (;;) Line 1811  for (;;)
1811      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1812    
1813      case OP_EXACT:      case OP_EXACT:
1814      branchlength += GET2(cc,1);      case OP_EXACTI:
1815      cc += 4;      case OP_NOTEXACT:
1816  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1817      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      branchlength += (int)GET2(cc,1);
1818        cc += 2 + IMM2_SIZE;
1819    #ifdef SUPPORT_UTF
1820        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1821  #endif  #endif
1822      break;      break;
1823    
1824      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1825      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1826      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1827      cc += 4;        cc += 2;
1828        cc += 1 + IMM2_SIZE + 1;
1829      break;      break;
1830    
1831      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1610  for (;;) Line 1835  for (;;)
1835      cc += 2;      cc += 2;
1836      /* Fall through */      /* Fall through */
1837    
1838        case OP_HSPACE:
1839        case OP_VSPACE:
1840        case OP_NOT_HSPACE:
1841        case OP_NOT_VSPACE:
1842      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1843      case OP_DIGIT:      case OP_DIGIT:
1844      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1622  for (;;) Line 1851  for (;;)
1851      cc++;      cc++;
1852      break;      break;
1853    
1854      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1855        otherwise \C is coded as OP_ALLANY. */
1856    
1857      case OP_ANYBYTE:      case OP_ANYBYTE:
1858      return -2;      return -2;
1859    
1860      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1861    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1862      case OP_CLASS:      case OP_CLASS:
1863      case OP_NCLASS:      case OP_NCLASS:
1864      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1865        case OP_XCLASS:
1866        /* The original code caused an unsigned overflow in 64 bit systems,
1867        so now we use a conditional statement. */
1868        if (op == OP_XCLASS)
1869          cc += GET(cc, 1);
1870        else
1871          cc += PRIV(OP_lengths)[OP_CLASS];
1872    #else
1873        cc += PRIV(OP_lengths)[OP_CLASS];
1874    #endif
1875    
1876      switch (*cc)      switch (*cc)
1877        {        {
1878        case OP_CRSTAR:        case OP_CRSTAR:
1879        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1880          case OP_CRPLUS:
1881          case OP_CRMINPLUS:
1882        case OP_CRQUERY:        case OP_CRQUERY:
1883        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1884          case OP_CRPOSSTAR:
1885          case OP_CRPOSPLUS:
1886          case OP_CRPOSQUERY:
1887        return -1;        return -1;
1888    
1889        case OP_CRRANGE:        case OP_CRRANGE:
1890        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1891        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1892        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1893        cc += 5;        branchlength += (int)GET2(cc,1);
1894          cc += 1 + 2 * IMM2_SIZE;
1895        break;        break;
1896    
1897        default:        default:
# Line 1661  for (;;) Line 1901  for (;;)
1901    
1902      /* Anything else is variable length */      /* Anything else is variable length */
1903    
1904      default:      case OP_ANYNL:
1905        case OP_BRAMINZERO:
1906        case OP_BRAPOS:
1907        case OP_BRAPOSZERO:
1908        case OP_BRAZERO:
1909        case OP_CBRAPOS:
1910        case OP_EXTUNI:
1911        case OP_KETRMAX:
1912        case OP_KETRMIN:
1913        case OP_KETRPOS:
1914        case OP_MINPLUS:
1915        case OP_MINPLUSI:
1916        case OP_MINQUERY:
1917        case OP_MINQUERYI:
1918        case OP_MINSTAR:
1919        case OP_MINSTARI:
1920        case OP_MINUPTO:
1921        case OP_MINUPTOI:
1922        case OP_NOTMINPLUS:
1923        case OP_NOTMINPLUSI:
1924        case OP_NOTMINQUERY:
1925        case OP_NOTMINQUERYI:
1926        case OP_NOTMINSTAR:
1927        case OP_NOTMINSTARI:
1928        case OP_NOTMINUPTO:
1929        case OP_NOTMINUPTOI:
1930        case OP_NOTPLUS:
1931        case OP_NOTPLUSI:
1932        case OP_NOTPOSPLUS:
1933        case OP_NOTPOSPLUSI:
1934        case OP_NOTPOSQUERY:
1935        case OP_NOTPOSQUERYI:
1936        case OP_NOTPOSSTAR:
1937        case OP_NOTPOSSTARI:
1938        case OP_NOTPOSUPTO:
1939        case OP_NOTPOSUPTOI:
1940        case OP_NOTQUERY:
1941        case OP_NOTQUERYI:
1942        case OP_NOTSTAR:
1943        case OP_NOTSTARI:
1944        case OP_NOTUPTO:
1945        case OP_NOTUPTOI:
1946        case OP_PLUS:
1947        case OP_PLUSI:
1948        case OP_POSPLUS:
1949        case OP_POSPLUSI:
1950        case OP_POSQUERY:
1951        case OP_POSQUERYI:
1952        case OP_POSSTAR:
1953        case OP_POSSTARI:
1954        case OP_POSUPTO:
1955        case OP_POSUPTOI:
1956        case OP_QUERY:
1957        case OP_QUERYI:
1958        case OP_REF:
1959        case OP_REFI:
1960        case OP_DNREF:
1961        case OP_DNREFI:
1962        case OP_SBRA:
1963        case OP_SBRAPOS:
1964        case OP_SCBRA:
1965        case OP_SCBRAPOS:
1966        case OP_SCOND:
1967        case OP_SKIPZERO:
1968        case OP_STAR:
1969        case OP_STARI:
1970        case OP_TYPEMINPLUS:
1971        case OP_TYPEMINQUERY:
1972        case OP_TYPEMINSTAR:
1973        case OP_TYPEMINUPTO:
1974        case OP_TYPEPLUS:
1975        case OP_TYPEPOSPLUS:
1976        case OP_TYPEPOSQUERY:
1977        case OP_TYPEPOSSTAR:
1978        case OP_TYPEPOSUPTO:
1979        case OP_TYPEQUERY:
1980        case OP_TYPESTAR:
1981        case OP_TYPEUPTO:
1982        case OP_UPTO:
1983        case OP_UPTOI:
1984      return -1;      return -1;
1985    
1986        /* Catch unrecognized opcodes so that when new ones are added they
1987        are not forgotten, as has happened in the past. */
1988    
1989        default:
1990        return -4;
1991      }      }
1992    }    }
1993  /* Control never gets here */  /* Control never gets here */
# Line 1670  for (;;) Line 1995  for (;;)
1995    
1996    
1997    
   
1998  /*************************************************  /*************************************************
1999  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2000  *************************************************/  *************************************************/
# Line 1683  length. Line 2007  length.
2007    
2008  Arguments:  Arguments:
2009    code        points to start of expression    code        points to start of expression
2010    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2011    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2012    
2013  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
2014  */  */
2015    
2016  const uschar *  const pcre_uchar *
2017  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2018  {  {
2019  for (;;)  for (;;)
2020    {    {
2021    register int c = *code;    register pcre_uchar c = *code;
2022    
2023    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2024    
# Line 1708  for (;;) Line 2032  for (;;)
2032    
2033    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2034      {      {
2035      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2036      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2037      }      }
2038    
2039    /* Handle capturing bracket */    /* Handle capturing bracket */
# Line 1717  for (;;) Line 2041  for (;;)
2041    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2042             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2043      {      {
2044      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2045      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2046      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2047      }      }
2048    
2049    /* 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 1747  for (;;) Line 2071  for (;;)
2071        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2072        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2073        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2074        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2075            code += 2;
2076        break;        break;
2077    
2078        case OP_MARK:        case OP_MARK:
2079        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2080        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2081        case OP_THEN_ARG:        case OP_THEN_ARG:
2082        code += code[1+LINK_SIZE];        code += code[1];
2083        break;        break;
2084        }        }
2085    
2086      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2087    
2088      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2089    
2090    /* 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
2091    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
2092    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2093    
2094  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2095      if (utf8) switch(c)      if (utf) switch(c)
2096        {        {
2097        case OP_CHAR:        case OP_CHAR:
2098        case OP_CHARI:        case OP_CHARI:
# Line 1800  for (;;) Line 2122  for (;;)
2122        case OP_MINQUERYI:        case OP_MINQUERYI:
2123        case OP_POSQUERY:        case OP_POSQUERY:
2124        case OP_POSQUERYI:        case OP_POSQUERYI:
2125        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2126        break;        break;
2127        }        }
2128  #else  #else
2129      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2130  #endif  #endif
2131      }      }
2132    }    }
# Line 1821  instance of OP_RECURSE. Line 2143  instance of OP_RECURSE.
2143    
2144  Arguments:  Arguments:
2145    code        points to start of expression    code        points to start of expression
2146    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2147    
2148  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
2149  */  */
2150    
2151  static const uschar *  static const pcre_uchar *
2152  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2153  {  {
2154  for (;;)  for (;;)
2155    {    {
2156    register int c = *code;    register pcre_uchar c = *code;
2157    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2158    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2159    
# Line 1866  for (;;) Line 2188  for (;;)
2188        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2189        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2190        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2191        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2192            code += 2;
2193        break;        break;
2194    
2195        case OP_MARK:        case OP_MARK:
2196        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2197        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2198          case OP_THEN_ARG:
2199        code += code[1];        code += code[1];
2200        break;        break;
   
       case OP_THEN_ARG:  
       code += code[1+LINK_SIZE];  
       break;  
2201        }        }
2202    
2203      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2204    
2205      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2206    
2207      /* 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
2208      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
2209      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2210    
2211  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2212      if (utf8) switch(c)      if (utf) switch(c)
2213        {        {
2214        case OP_CHAR:        case OP_CHAR:
2215        case OP_CHARI:        case OP_CHARI:
2216          case OP_NOT:
2217          case OP_NOTI:
2218        case OP_EXACT:        case OP_EXACT:
2219        case OP_EXACTI:        case OP_EXACTI:
2220          case OP_NOTEXACT:
2221          case OP_NOTEXACTI:
2222        case OP_UPTO:        case OP_UPTO:
2223        case OP_UPTOI:        case OP_UPTOI:
2224          case OP_NOTUPTO:
2225          case OP_NOTUPTOI:
2226        case OP_MINUPTO:        case OP_MINUPTO:
2227        case OP_MINUPTOI:        case OP_MINUPTOI:
2228          case OP_NOTMINUPTO:
2229          case OP_NOTMINUPTOI:
2230        case OP_POSUPTO:        case OP_POSUPTO:
2231        case OP_POSUPTOI:        case OP_POSUPTOI:
2232          case OP_NOTPOSUPTO:
2233          case OP_NOTPOSUPTOI:
2234        case OP_STAR:        case OP_STAR:
2235        case OP_STARI:        case OP_STARI:
2236          case OP_NOTSTAR:
2237          case OP_NOTSTARI:
2238        case OP_MINSTAR:        case OP_MINSTAR:
2239        case OP_MINSTARI:        case OP_MINSTARI:
2240          case OP_NOTMINSTAR:
2241          case OP_NOTMINSTARI:
2242        case OP_POSSTAR:        case OP_POSSTAR:
2243        case OP_POSSTARI:        case OP_POSSTARI:
2244          case OP_NOTPOSSTAR:
2245          case OP_NOTPOSSTARI:
2246        case OP_PLUS:        case OP_PLUS:
2247        case OP_PLUSI:        case OP_PLUSI:
2248          case OP_NOTPLUS:
2249          case OP_NOTPLUSI:
2250        case OP_MINPLUS:        case OP_MINPLUS:
2251        case OP_MINPLUSI:        case OP_MINPLUSI:
2252          case OP_NOTMINPLUS:
2253          case OP_NOTMINPLUSI:
2254        case OP_POSPLUS:        case OP_POSPLUS:
2255        case OP_POSPLUSI:        case OP_POSPLUSI:
2256          case OP_NOTPOSPLUS:
2257          case OP_NOTPOSPLUSI:
2258        case OP_QUERY:        case OP_QUERY:
2259        case OP_QUERYI:        case OP_QUERYI:
2260          case OP_NOTQUERY:
2261          case OP_NOTQUERYI:
2262        case OP_MINQUERY:        case OP_MINQUERY:
2263        case OP_MINQUERYI:        case OP_MINQUERYI:
2264          case OP_NOTMINQUERY:
2265          case OP_NOTMINQUERYI:
2266        case OP_POSQUERY:        case OP_POSQUERY:
2267        case OP_POSQUERYI:        case OP_POSQUERYI:
2268        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_NOTPOSQUERY:
2269          case OP_NOTPOSQUERYI:
2270          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2271        break;        break;
2272        }        }
2273  #else  #else
2274      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2275  #endif  #endif
2276      }      }
2277    }    }
# Line 1946  bracket whose current branch will alread Line 2294  bracket whose current branch will alread
2294  Arguments:  Arguments:
2295    code        points to start of search    code        points to start of search
2296    endcode     points to where to stop    endcode     points to where to stop
2297    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2298    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2299      recurses    chain of recurse_check to catch mutual recursion
2300    
2301  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2302  */  */
2303    
2304    typedef struct recurse_check {
2305      struct recurse_check *prev;
2306      const pcre_uchar *group;
2307    } recurse_check;
2308    
2309  static BOOL  static BOOL
2310  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2311    compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2312  {  {
2313  register int c;  register pcre_uchar c;
2314  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);  recurse_check this_recurse;
2315    
2316    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2317       code < endcode;       code < endcode;
2318       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2319    {    {
2320    const uschar *ccode;    const pcre_uchar *ccode;
2321    
2322    c = *code;    c = *code;
2323    
# Line 1978  for (code = first_significant_code(code Line 2334  for (code = first_significant_code(code
2334    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2335    implies a backward reference subroutine call, we can scan it. If it's a    implies a backward reference subroutine call, we can scan it. If it's a
2336    forward reference subroutine call, we can't. To detect forward reference    forward reference subroutine call, we can't. To detect forward reference
2337    we have to scan up the list that is kept in the workspace. This function is    we have to scan up the list that is kept in the workspace. This function is
2338    called only when doing the real compile, not during the pre-compile that    called only when doing the real compile, not during the pre-compile that
2339    measures the size of the compiled pattern. */    measures the size of the compiled pattern. */
2340    
2341    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2342      {      {
2343      const uschar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2344      BOOL empty_branch;      BOOL empty_branch;
   
     /* Test for forward reference */  
   
     for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)  
       if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;  
2345    
2346      /* Not a forward reference, test for completed backward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2347        when called to scan a completed pattern by setting cd->start_workspace to
2348        NULL. */
2349    
2350        if (cd->start_workspace != NULL)
2351          {
2352          const pcre_uchar *tcode;
2353          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2354            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2355          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2356          }
2357    
2358        /* If we are scanning a completed pattern, there are no forward references
2359        and all groups are complete. We need to detect whether this is a recursive
2360        call, as otherwise there will be an infinite loop. If it is a recursion,
2361        just skip over it. Simple recursions are easily detected. For mutual
2362        recursions we keep a chain on the stack. */
2363    
2364        else
2365          {
2366          recurse_check *r = recurses;
2367          const pcre_uchar *endgroup = scode;
2368    
2369          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2370          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2371    
2372          for (r = recurses; r != NULL; r = r->prev)
2373            if (r->group == scode) break;
2374          if (r != NULL) continue;   /* Mutual recursion */
2375          }
2376    
2377        /* Completed reference; scan the referenced group, remembering it on the
2378        stack chain to detect mutual recursions. */
2379    
2380      empty_branch = FALSE;      empty_branch = FALSE;
2381      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2382      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
2383    
     /* Completed backwards reference */  
   
2384      do      do
2385        {        {
2386        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2387          {          {
2388          empty_branch = TRUE;          empty_branch = TRUE;
2389          break;          break;
# Line 2010  for (code = first_significant_code(code Line 2391  for (code = first_significant_code(code
2391        scode += GET(scode, 1);        scode += GET(scode, 1);
2392        }        }
2393      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2394    
2395      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2396      continue;      continue;
2397      }      }
# Line 2020  for (code = first_significant_code(code Line 2401  for (code = first_significant_code(code
2401    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2402        c == OP_BRAPOSZERO)        c == OP_BRAPOSZERO)
2403      {      {
2404      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2405      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2406      c = *code;      c = *code;
2407      continue;      continue;
# Line 2041  for (code = first_significant_code(code Line 2422  for (code = first_significant_code(code
2422    
2423    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2424        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2425        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2426          c == OP_COND)
2427      {      {
2428      BOOL empty_branch;      BOOL empty_branch;
2429      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2057  for (code = first_significant_code(code Line 2439  for (code = first_significant_code(code
2439        empty_branch = FALSE;        empty_branch = FALSE;
2440        do        do
2441          {          {
2442          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2443            empty_branch = TRUE;            empty_branch = TRUE;
2444          code += GET(code, 1);          code += GET(code, 1);
2445          }          }
# Line 2075  for (code = first_significant_code(code Line 2457  for (code = first_significant_code(code
2457      {      {
2458      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2459      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2460      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2461      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2462      here. */      here. */
2463    
2464  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2465      case OP_XCLASS:      case OP_XCLASS:
2466      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2467      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2087  for (code = first_significant_code(code Line 2469  for (code = first_significant_code(code
2469    
2470      case OP_CLASS:      case OP_CLASS:
2471      case OP_NCLASS:      case OP_NCLASS:
2472      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2473    
2474  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2475      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2476  #endif  #endif
2477    
# Line 2099  for (code = first_significant_code(code Line 2481  for (code = first_significant_code(code
2481        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2482        case OP_CRQUERY:        case OP_CRQUERY:
2483        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2484          case OP_CRPOSSTAR:
2485          case OP_CRPOSQUERY:
2486        break;        break;
2487    
2488        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2489        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2490        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2491          case OP_CRPOSPLUS:
2492        return FALSE;        return FALSE;
2493    
2494        case OP_CRRANGE:        case OP_CRRANGE:
2495        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2496          case OP_CRPOSRANGE:
2497        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2498        break;        break;
2499        }        }
# Line 2115  for (code = first_significant_code(code Line 2501  for (code = first_significant_code(code
2501    
2502      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2503    
2504        case OP_ANY:
2505        case OP_ALLANY:
2506        case OP_ANYBYTE:
2507    
2508      case OP_PROP:      case OP_PROP:
2509      case OP_NOTPROP:      case OP_NOTPROP:
2510        case OP_ANYNL:
2511    
2512        case OP_NOT_HSPACE:
2513        case OP_HSPACE:
2514        case OP_NOT_VSPACE:
2515        case OP_VSPACE:
2516      case OP_EXTUNI:      case OP_EXTUNI:
2517    
2518      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2519      case OP_DIGIT:      case OP_DIGIT:
2520      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2521      case OP_WHITESPACE:      case OP_WHITESPACE:
2522      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2523      case OP_WORDCHAR:      case OP_WORDCHAR:
2524      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2525      case OP_CHAR:      case OP_CHAR:
2526      case OP_CHARI:      case OP_CHARI:
2527      case OP_NOT:      case OP_NOT:
2528      case OP_NOTI:      case OP_NOTI:
2529    
2530      case OP_PLUS:      case OP_PLUS:
2531        case OP_PLUSI:
2532      case OP_MINPLUS:      case OP_MINPLUS:
2533      case OP_POSPLUS:      case OP_MINPLUSI:
2534      case OP_EXACT:  
2535      case OP_NOTPLUS:      case OP_NOTPLUS:
2536        case OP_NOTPLUSI:
2537      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2538        case OP_NOTMINPLUSI:
2539    
2540        case OP_POSPLUS:
2541        case OP_POSPLUSI:
2542      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2543        case OP_NOTPOSPLUSI:
2544    
2545        case OP_EXACT:
2546        case OP_EXACTI:
2547      case OP_NOTEXACT:      case OP_NOTEXACT:
2548        case OP_NOTEXACTI:
2549    
2550      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2551      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2552      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2553      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2554    
2555      return FALSE;      return FALSE;
2556    
2557      /* 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 2162  for (code = first_significant_code(code Line 2571  for (code = first_significant_code(code
2571      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2572      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2573      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2574      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2575          code += 2;
2576      break;      break;
2577    
2578      /* End of branch */      /* End of branch */
# Line 2175  for (code = first_significant_code(code Line 2585  for (code = first_significant_code(code
2585      return TRUE;      return TRUE;
2586    
2587      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2588      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2589        followed by a multibyte character. */
2590    
2591  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2592      case OP_STAR:      case OP_STAR:
2593      case OP_STARI:      case OP_STARI:
2594        case OP_NOTSTAR:
2595        case OP_NOTSTARI:
2596    
2597      case OP_MINSTAR:      case OP_MINSTAR:
2598      case OP_MINSTARI:      case OP_MINSTARI:
2599        case OP_NOTMINSTAR:
2600        case OP_NOTMINSTARI:
2601    
2602      case OP_POSSTAR:      case OP_POSSTAR:
2603      case OP_POSSTARI:      case OP_POSSTARI:
2604        case OP_NOTPOSSTAR:
2605        case OP_NOTPOSSTARI:
2606    
2607      case OP_QUERY:      case OP_QUERY:
2608      case OP_QUERYI:      case OP_QUERYI:
2609        case OP_NOTQUERY:
2610        case OP_NOTQUERYI:
2611    
2612      case OP_MINQUERY:      case OP_MINQUERY:
2613      case OP_MINQUERYI:      case OP_MINQUERYI:
2614        case OP_NOTMINQUERY:
2615        case OP_NOTMINQUERYI:
2616    
2617      case OP_POSQUERY:      case OP_POSQUERY:
2618      case OP_POSQUERYI:      case OP_POSQUERYI:
2619      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_NOTPOSQUERY:
2620        case OP_NOTPOSQUERYI:
2621    
2622        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2623      break;      break;
2624    
2625      case OP_UPTO:      case OP_UPTO:
2626      case OP_UPTOI:      case OP_UPTOI:
2627        case OP_NOTUPTO:
2628        case OP_NOTUPTOI:
2629    
2630      case OP_MINUPTO:      case OP_MINUPTO:
2631      case OP_MINUPTOI:      case OP_MINUPTOI:
2632        case OP_NOTMINUPTO:
2633        case OP_NOTMINUPTOI:
2634    
2635      case OP_POSUPTO:      case OP_POSUPTO:
2636      case OP_POSUPTOI:      case OP_POSUPTOI:
2637      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_NOTPOSUPTO:
2638        case OP_NOTPOSUPTOI:
2639    
2640        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2641      break;      break;
2642  #endif  #endif
2643    
# Line 2209  for (code = first_significant_code(code Line 2647  for (code = first_significant_code(code
2647      case OP_MARK:      case OP_MARK:
2648      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2649      case OP_SKIP_ARG:      case OP_SKIP_ARG:
2650        case OP_THEN_ARG:
2651      code += code[1];      code += code[1];
2652      break;      break;
2653    
2654      case OP_THEN_ARG:      /* None of the remaining opcodes are required to match a character. */
2655      code += code[1+LINK_SIZE];  
2656        default:
2657      break;      break;
2658        }
2659      }
2660    
2661      /* None of the remaining opcodes are required to match a character. */  return TRUE;
2662    }
2663    
2664    
2665    
2666    /*************************************************
2667    *    Scan compiled regex for non-emptiness       *
2668    *************************************************/
2669    
2670    /* This function is called to check for left recursive calls. We want to check
2671    the current branch of the current pattern to see if it could match the empty
2672    string. If it could, we must look outwards for branches at other levels,
2673    stopping when we pass beyond the bracket which is the subject of the recursion.
2674    This function is called only during the real compile, not during the
2675    pre-compile.
2676    
2677    Arguments:
2678      code        points to start of the recursion
2679      endcode     points to where to stop (current RECURSE item)
2680      bcptr       points to the chain of current (unclosed) branch starts
2681      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2682      cd          pointers to tables etc
2683    
2684    Returns:      TRUE if what is matched could be empty
2685    */
2686    
2687    static BOOL
2688    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2689      branch_chain *bcptr, BOOL utf, compile_data *cd)
2690    {
2691    while (bcptr != NULL && bcptr->current_branch >= code)
2692      {
2693      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2694        return FALSE;
2695      bcptr = bcptr->outer;
2696      }
2697    return TRUE;
2698    }
2699    
2700    
2701    
2702    /*************************************************
2703    *        Base opcode of repeated opcodes         *
2704    *************************************************/
2705    
2706    /* Returns the base opcode for repeated single character type opcodes. If the
2707    opcode is not a repeated character type, it returns with the original value.
2708    
2709    Arguments:  c opcode
2710    Returns:    base opcode for the type
2711    */
2712    
2713    static pcre_uchar
2714    get_repeat_base(pcre_uchar c)
2715    {
2716    return (c > OP_TYPEPOSUPTO)? c :
2717           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2718           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2719           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2720           (c >= OP_STARI)?      OP_STARI :
2721                                 OP_STAR;
2722    }
2723    
2724    
2725    
2726    #ifdef SUPPORT_UCP
2727    /*************************************************
2728    *        Check a character and a property        *
2729    *************************************************/
2730    
2731    /* This function is called by check_auto_possessive() when a property item
2732    is adjacent to a fixed character.
2733    
2734    Arguments:
2735      c            the character
2736      ptype        the property type
2737      pdata        the data for the type
2738      negated      TRUE if it's a negated property (\P or \p{^)
2739    
2740    Returns:       TRUE if auto-possessifying is OK
2741    */
2742    
2743    static BOOL
2744    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2745      BOOL negated)
2746    {
2747    const pcre_uint32 *p;
2748    const ucd_record *prop = GET_UCD(c);
2749    
2750    switch(ptype)
2751      {
2752      case PT_LAMP:
2753      return (prop->chartype == ucp_Lu ||
2754              prop->chartype == ucp_Ll ||
2755              prop->chartype == ucp_Lt) == negated;
2756    
2757      case PT_GC:
2758      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2759    
2760      case PT_PC:
2761      return (pdata == prop->chartype) == negated;
2762    
2763      case PT_SC:
2764      return (pdata == prop->script) == negated;
2765    
2766      /* These are specials */
2767    
2768      case PT_ALNUM:
2769      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2770              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2771    
2772      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2773      means that Perl space and POSIX space are now identical. PCRE was changed
2774      at release 8.34. */
2775    
2776      case PT_SPACE:    /* Perl space */
2777      case PT_PXSPACE:  /* POSIX space */
2778      switch(c)
2779        {
2780        HSPACE_CASES:
2781        VSPACE_CASES:
2782        return negated;
2783    
2784      default:      default:
2785        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2786        }
2787      break;  /* Control never reaches here */
2788    
2789      case PT_WORD:
2790      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2791              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2792              c == CHAR_UNDERSCORE) == negated;
2793    
2794      case PT_CLIST:
2795      p = PRIV(ucd_caseless_sets) + prop->caseset;
2796      for (;;)
2797        {
2798        if (c < *p) return !negated;
2799        if (c == *p++) return negated;
2800        }
2801      break;  /* Control never reaches here */
2802      }
2803    
2804    return FALSE;
2805    }
2806    #endif  /* SUPPORT_UCP */
2807    
2808    
2809    
2810    /*************************************************
2811    *        Fill the character property list        *
2812    *************************************************/
2813    
2814    /* Checks whether the code points to an opcode that can take part in auto-
2815    possessification, and if so, fills a list with its properties.
2816    
2817    Arguments:
2818      code        points to start of expression
2819      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2820      fcc         points to case-flipping table
2821      list        points to output list
2822                  list[0] will be filled with the opcode
2823                  list[1] will be non-zero if this opcode
2824                    can match an empty character string
2825                  list[2..7] depends on the opcode
2826    
2827    Returns:      points to the start of the next opcode if *code is accepted
2828                  NULL if *code is not accepted
2829    */
2830    
2831    static const pcre_uchar *
2832    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2833      const pcre_uint8 *fcc, pcre_uint32 *list)
2834    {
2835    pcre_uchar c = *code;
2836    const pcre_uchar *end;
2837    const pcre_uint32 *clist_src;
2838    pcre_uint32 *clist_dest;
2839    pcre_uint32 chr;
2840    pcre_uchar base;
2841    
2842    list[0] = c;
2843    list[1] = FALSE;
2844    code++;
2845    
2846    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2847      {
2848      base = get_repeat_base(c);
2849      c -= (base - OP_STAR);
2850    
2851      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2852        code += IMM2_SIZE;
2853    
2854      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2855    
2856      switch(base)
2857        {
2858        case OP_STAR:
2859        list[0] = OP_CHAR;
2860        break;
2861    
2862        case OP_STARI:
2863        list[0] = OP_CHARI;
2864        break;
2865    
2866        case OP_NOTSTAR:
2867        list[0] = OP_NOT;
2868        break;
2869    
2870        case OP_NOTSTARI:
2871        list[0] = OP_NOTI;
2872        break;
2873    
2874        case OP_TYPESTAR:
2875        list[0] = *code;
2876        code++;
2877        break;
2878        }
2879      c = list[0];
2880      }
2881    
2882    switch(c)
2883      {
2884      case OP_NOT_DIGIT:
2885      case OP_DIGIT:
2886      case OP_NOT_WHITESPACE:
2887      case OP_WHITESPACE:
2888      case OP_NOT_WORDCHAR:
2889      case OP_WORDCHAR:
2890      case OP_ANY:
2891      case OP_ALLANY:
2892      case OP_ANYNL:
2893      case OP_NOT_HSPACE:
2894      case OP_HSPACE:
2895      case OP_NOT_VSPACE:
2896      case OP_VSPACE:
2897      case OP_EXTUNI:
2898      case OP_EODN:
2899      case OP_EOD:
2900      case OP_DOLL:
2901      case OP_DOLLM:
2902      return code;
2903    
2904      case OP_CHAR:
2905      case OP_NOT:
2906      GETCHARINCTEST(chr, code);
2907      list[2] = chr;
2908      list[3] = NOTACHAR;
2909      return code;
2910    
2911      case OP_CHARI:
2912      case OP_NOTI:
2913      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2914      GETCHARINCTEST(chr, code);
2915      list[2] = chr;
2916    
2917    #ifdef SUPPORT_UCP
2918      if (chr < 128 || (chr < 256 && !utf))
2919        list[3] = fcc[chr];
2920      else
2921        list[3] = UCD_OTHERCASE(chr);
2922    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2923      list[3] = (chr < 256) ? fcc[chr] : chr;
2924    #else
2925      list[3] = fcc[chr];
2926    #endif
2927    
2928      /* The othercase might be the same value. */
2929    
2930      if (chr == list[3])
2931        list[3] = NOTACHAR;
2932      else
2933        list[4] = NOTACHAR;
2934      return code;
2935    
2936    #ifdef SUPPORT_UCP
2937      case OP_PROP:
2938      case OP_NOTPROP:
2939      if (code[0] != PT_CLIST)
2940        {
2941        list[2] = code[0];
2942        list[3] = code[1];
2943        return code + 2;
2944        }
2945    
2946      /* Convert only if we have enough space. */
2947    
2948      clist_src = PRIV(ucd_caseless_sets) + code[1];
2949      clist_dest = list + 2;
2950      code += 2;
2951    
2952      do {
2953         if (clist_dest >= list + 8)
2954           {
2955           /* Early return if there is not enough space. This should never
2956           happen, since all clists are shorter than 5 character now. */
2957           list[2] = code[0];
2958           list[3] = code[1];
2959           return code;
2960           }
2961         *clist_dest++ = *clist_src;
2962         }
2963      while(*clist_src++ != NOTACHAR);
2964    
2965      /* All characters are stored. The terminating NOTACHAR
2966      is copied form the clist itself. */
2967    
2968      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2969      return code;
2970    #endif
2971    
2972      case OP_NCLASS:
2973      case OP_CLASS:
2974    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2975      case OP_XCLASS:
2976    
2977      if (c == OP_XCLASS)
2978        end = code + GET(code, 0) - 1;
2979      else
2980    #endif
2981        end = code + 32 / sizeof(pcre_uchar);
2982    
2983      switch(*end)
2984        {
2985        case OP_CRSTAR:
2986        case OP_CRMINSTAR:
2987        case OP_CRQUERY:
2988        case OP_CRMINQUERY:
2989        case OP_CRPOSSTAR:
2990        case OP_CRPOSQUERY:
2991        list[1] = TRUE;
2992        end++;
2993        break;
2994    
2995        case OP_CRPLUS:
2996        case OP_CRMINPLUS:
2997        case OP_CRPOSPLUS:
2998        end++;
2999        break;
3000    
3001        case OP_CRRANGE:
3002        case OP_CRMINRANGE:
3003        case OP_CRPOSRANGE:
3004        list[1] = (GET2(end, 1) == 0);
3005        end += 1 + 2 * IMM2_SIZE;
3006        break;
3007        }
3008      list[2] = end - code;
3009      return end;
3010      }
3011    return NULL;    /* Opcode not accepted */
3012    }
3013    
3014    
3015    
3016    /*************************************************
3017    *    Scan further character sets for match       *
3018    *************************************************/
3019    
3020    /* Checks whether the base and the current opcode have a common character, in
3021    which case the base cannot be possessified.
3022    
3023    Arguments:
3024      code        points to the byte code
3025      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3026      cd          static compile data
3027      base_list   the data list of the base opcode
3028    
3029    Returns:      TRUE if the auto-possessification is possible
3030    */
3031    
3032    static BOOL
3033    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3034      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3035    {
3036    pcre_uchar c;
3037    pcre_uint32 list[8];
3038    const pcre_uint32 *chr_ptr;
3039    const pcre_uint32 *ochr_ptr;
3040    const pcre_uint32 *list_ptr;
3041    const pcre_uchar *next_code;
3042    const pcre_uint8 *class_bitset;
3043    const pcre_uint32 *set1, *set2, *set_end;
3044    pcre_uint32 chr;
3045    BOOL accepted, invert_bits;
3046    
3047    /* Note: the base_list[1] contains whether the current opcode has greedy
3048    (represented by a non-zero value) quantifier. This is a different from
3049    other character type lists, which stores here that the character iterator
3050    matches to an empty string (also represented by a non-zero value). */
3051    
3052    for(;;)
3053      {
3054      c = *code;
3055    
3056      /* Skip over callouts */
3057    
3058      if (c == OP_CALLOUT)
3059        {
3060        code += PRIV(OP_lengths)[c];
3061        continue;
3062        }
3063    
3064      if (c == OP_ALT)
3065        {
3066        do code += GET(code, 1); while (*code == OP_ALT);
3067        c = *code;
3068        }
3069    
3070      switch(c)
3071        {
3072        case OP_END:
3073        case OP_KETRPOS:
3074        /* TRUE only in greedy case. The non-greedy case could be replaced by
3075        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3076        uses more memory, which we cannot get at this stage.) */
3077    
3078        return base_list[1] != 0;
3079    
3080        case OP_KET:
3081        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3082        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3083        cannot be converted to a possessive form. */
3084    
3085        if (base_list[1] == 0) return FALSE;
3086    
3087        switch(*(code - GET(code, 1)))
3088          {
3089          case OP_ASSERT:
3090          case OP_ASSERT_NOT:
3091          case OP_ASSERTBACK:
3092          case OP_ASSERTBACK_NOT:
3093          case OP_ONCE:
3094          case OP_ONCE_NC:
3095          /* Atomic sub-patterns and assertions can always auto-possessify their
3096          last iterator. */
3097          return TRUE;
3098          }
3099    
3100        code += PRIV(OP_lengths)[c];
3101        continue;
3102    
3103        case OP_ONCE:
3104        case OP_ONCE_NC:
3105        case OP_BRA:
3106        case OP_CBRA:
3107        next_code = code;
3108        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3109    
3110        /* We do not support repeated brackets, because they can lead to
3111        infinite recursion. */
3112    
3113        if (*next_code != OP_KET) return FALSE;
3114    
3115        next_code = code + GET(code, 1);
3116        code += PRIV(OP_lengths)[c];
3117    
3118        while (*next_code == OP_ALT)
3119          {
3120          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3121          code = next_code + 1 + LINK_SIZE;
3122          next_code += GET(next_code, 1);
3123          }
3124        continue;
3125    
3126        case OP_BRAZERO:
3127        case OP_BRAMINZERO:
3128    
3129        next_code = code + 1;
3130        if (*next_code != OP_BRA && *next_code != OP_CBRA)
3131          return FALSE;
3132    
3133        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3134    
3135        /* We do not support repeated brackets, because they can lead to
3136        infinite recursion. */
3137        if (*next_code != OP_KET) return FALSE;
3138    
3139        /* The bracket content will be checked by the
3140        OP_BRA/OP_CBRA case above. */
3141        next_code += 1 + LINK_SIZE;
3142        if (!compare_opcodes(next_code, utf, cd, base_list, base_end)) return FALSE;
3143    
3144        code += PRIV(OP_lengths)[c];
3145        continue;
3146        }
3147    
3148      /* Check for a supported opcode, and load its properties. */
3149    
3150      code = get_chr_property_list(code, utf, cd->fcc, list);
3151      if (code == NULL) return FALSE;    /* Unsupported */
3152    
3153      /* If either opcode is a small character list, set pointers for comparing
3154      characters from that list with another list, or with a property. */
3155    
3156      if (base_list[0] == OP_CHAR)
3157        {
3158        chr_ptr = base_list + 2;
3159        list_ptr = list;
3160        }
3161      else if (list[0] == OP_CHAR)
3162        {
3163        chr_ptr = list + 2;
3164        list_ptr = base_list;
3165        }
3166    
3167      /* Character bitsets can also be compared to certain opcodes. */
3168    
3169      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3170    #ifdef COMPILE_PCRE8
3171          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3172          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3173    #endif
3174          )
3175        {
3176    #ifdef COMPILE_PCRE8
3177        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3178    #else
3179        if (base_list[0] == OP_CLASS)
3180    #endif
3181          {
3182          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3183          list_ptr = list;
3184          }
3185        else
3186          {
3187          set1 = (pcre_uint32 *)(code - list[2]);
3188          list_ptr = base_list;
3189          }
3190    
3191        invert_bits = FALSE;
3192        switch(list_ptr[0])
3193          {
3194          case OP_CLASS:
3195          case OP_NCLASS:
3196          set2 = (pcre_uint32 *)
3197            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3198          break;
3199    
3200          /* OP_XCLASS cannot be supported here, because its bitset
3201          is not necessarily complete. E.g: [a-\0x{200}] is stored
3202          as a character range, and the appropriate bits are not set. */
3203    
3204          case OP_NOT_DIGIT:
3205            invert_bits = TRUE;
3206            /* Fall through */
3207          case OP_DIGIT:
3208            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3209            break;
3210    
3211          case OP_NOT_WHITESPACE:
3212            invert_bits = TRUE;
3213            /* Fall through */
3214          case OP_WHITESPACE:
3215            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3216            break;
3217    
3218          case OP_NOT_WORDCHAR:
3219            invert_bits = TRUE;
3220            /* Fall through */
3221          case OP_WORDCHAR:
3222            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3223            break;
3224    
3225          default:
3226          return FALSE;
3227          }
3228    
3229        /* Compare 4 bytes to improve speed. */
3230        set_end = set1 + (32 / 4);
3231        if (invert_bits)
3232          {
3233          do
3234            {
3235            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3236            }
3237          while (set1 < set_end);
3238          }
3239        else
3240          {
3241          do
3242            {
3243            if ((*set1++ & *set2++) != 0) return FALSE;
3244            }
3245          while (set1 < set_end);
3246          }
3247    
3248        if (list[1] == 0) return TRUE;
3249        /* Might be an empty repeat. */
3250        continue;
3251        }
3252    
3253      /* Some property combinations also acceptable. Unicode property opcodes are
3254      processed specially; the rest can be handled with a lookup table. */
3255    
3256      else
3257        {
3258        pcre_uint32 leftop, rightop;
3259    
3260        leftop = base_list[0];
3261        rightop = list[0];
3262    
3263    #ifdef SUPPORT_UCP
3264        accepted = FALSE; /* Always set in non-unicode case. */
3265        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3266          {
3267          if (rightop == OP_EOD)
3268            accepted = TRUE;
3269          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3270            {
3271            int n;
3272            const pcre_uint8 *p;
3273            BOOL same = leftop == rightop;
3274            BOOL lisprop = leftop == OP_PROP;
3275            BOOL risprop = rightop == OP_PROP;
3276            BOOL bothprop = lisprop && risprop;
3277    
3278            /* There's a table that specifies how each combination is to be
3279            processed:
3280              0   Always return FALSE (never auto-possessify)
3281              1   Character groups are distinct (possessify if both are OP_PROP)
3282              2   Check character categories in the same group (general or particular)
3283              3   Return TRUE if the two opcodes are not the same
3284              ... see comments below
3285            */
3286    
3287            n = propposstab[base_list[2]][list[2]];
3288            switch(n)
3289              {
3290              case 0: break;
3291              case 1: accepted = bothprop; break;
3292              case 2: accepted = (base_list[3] == list[3]) != same; break;
3293              case 3: accepted = !same; break;
3294    
3295              case 4:  /* Left general category, right particular category */
3296              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3297              break;
3298    
3299              case 5:  /* Right general category, left particular category */
3300              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3301              break;
3302    
3303              /* This code is logically tricky. Think hard before fiddling with it.
3304              The posspropstab table has four entries per row. Each row relates to
3305              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3306              Only WORD actually needs all four entries, but using repeats for the
3307              others means they can all use the same code below.
3308    
3309              The first two entries in each row are Unicode general categories, and
3310              apply always, because all the characters they include are part of the
3311              PCRE character set. The third and fourth entries are a general and a
3312              particular category, respectively, that include one or more relevant
3313              characters. One or the other is used, depending on whether the check
3314              is for a general or a particular category. However, in both cases the
3315              category contains more characters than the specials that are defined
3316              for the property being tested against. Therefore, it cannot be used
3317              in a NOTPROP case.
3318    
3319              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3320              Underscore is covered by ucp_P or ucp_Po. */
3321    
3322              case 6:  /* Left alphanum vs right general category */
3323              case 7:  /* Left space vs right general category */
3324              case 8:  /* Left word vs right general category */
3325              p = posspropstab[n-6];
3326              accepted = risprop && lisprop ==
3327                (list[3] != p[0] &&
3328                 list[3] != p[1] &&
3329                (list[3] != p[2] || !lisprop));
3330              break;
3331    
3332              case 9:   /* Right alphanum vs left general category */
3333              case 10:  /* Right space vs left general category */
3334              case 11:  /* Right word vs left general category */
3335              p = posspropstab[n-9];
3336              accepted = lisprop && risprop ==
3337                (base_list[3] != p[0] &&
3338                 base_list[3] != p[1] &&
3339                (base_list[3] != p[2] || !risprop));
3340              break;
3341    
3342              case 12:  /* Left alphanum vs right particular category */
3343              case 13:  /* Left space vs right particular category */
3344              case 14:  /* Left word vs right particular category */
3345              p = posspropstab[n-12];
3346              accepted = risprop && lisprop ==
3347                (catposstab[p[0]][list[3]] &&
3348                 catposstab[p[1]][list[3]] &&
3349                (list[3] != p[3] || !lisprop));
3350              break;
3351    
3352              case 15:  /* Right alphanum vs left particular category */
3353              case 16:  /* Right space vs left particular category */
3354              case 17:  /* Right word vs left particular category */
3355              p = posspropstab[n-15];
3356              accepted = lisprop && risprop ==
3357                (catposstab[p[0]][base_list[3]] &&
3358                 catposstab[p[1]][base_list[3]] &&
3359                (base_list[3] != p[3] || !risprop));
3360              break;
3361              }
3362            }
3363          }
3364    
3365        else
3366    #endif  /* SUPPORT_UCP */
3367    
3368        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3369               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3370               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3371    
3372        if (!accepted)
3373          return FALSE;
3374    
3375        if (list[1] == 0) return TRUE;
3376        /* Might be an empty repeat. */
3377        continue;
3378        }
3379    
3380      /* Control reaches here only if one of the items is a small character list.
3381      All characters are checked against the other side. */
3382    
3383      do
3384        {
3385        chr = *chr_ptr;
3386    
3387        switch(list_ptr[0])
3388          {
3389          case OP_CHAR:
3390          ochr_ptr = list_ptr + 2;
3391          do
3392            {
3393            if (chr == *ochr_ptr) return FALSE;
3394            ochr_ptr++;
3395            }
3396          while(*ochr_ptr != NOTACHAR);
3397          break;
3398    
3399          case OP_NOT:
3400          ochr_ptr = list_ptr + 2;
3401          do
3402            {
3403            if (chr == *ochr_ptr)
3404              break;
3405            ochr_ptr++;
3406            }
3407          while(*ochr_ptr != NOTACHAR);
3408          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3409          break;
3410    
3411          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3412          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3413    
3414          case OP_DIGIT:
3415          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3416          break;
3417    
3418          case OP_NOT_DIGIT:
3419          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3420          break;
3421    
3422          case OP_WHITESPACE:
3423          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3424          break;
3425    
3426          case OP_NOT_WHITESPACE:
3427          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3428          break;
3429    
3430          case OP_WORDCHAR:
3431          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3432          break;
3433    
3434          case OP_NOT_WORDCHAR:
3435          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3436          break;
3437    
3438          case OP_HSPACE:
3439          switch(chr)
3440            {
3441            HSPACE_CASES: return FALSE;
3442            default: break;
3443            }
3444          break;
3445    
3446          case OP_NOT_HSPACE:
3447          switch(chr)
3448            {
3449            HSPACE_CASES: break;
3450            default: return FALSE;
3451            }
3452          break;
3453    
3454          case OP_ANYNL:
3455          case OP_VSPACE:
3456          switch(chr)
3457            {
3458            VSPACE_CASES: return FALSE;
3459            default: break;
3460            }
3461          break;
3462    
3463          case OP_NOT_VSPACE:
3464          switch(chr)
3465            {
3466            VSPACE_CASES: break;
3467            default: return FALSE;
3468            }
3469          break;
3470    
3471          case OP_DOLL:
3472          case OP_EODN:
3473          switch (chr)
3474            {
3475            case CHAR_CR:
3476            case CHAR_LF:
3477            case CHAR_VT:
3478            case CHAR_FF:
3479            case CHAR_NEL:
3480    #ifndef EBCDIC
3481            case 0x2028:
3482            case 0x2029:
3483    #endif  /* Not EBCDIC */
3484            return FALSE;
3485            }
3486          break;
3487    
3488          case OP_EOD:    /* Can always possessify before \z */
3489          break;
3490    
3491          case OP_PROP:
3492          case OP_NOTPROP:
3493          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3494                list_ptr[0] == OP_NOTPROP))
3495            return FALSE;
3496          break;
3497    
3498          case OP_NCLASS:
3499          if (chr > 255) return FALSE;
3500          /* Fall through */
3501    
3502          case OP_CLASS:
3503          if (chr > 255) break;
3504          class_bitset = (pcre_uint8 *)
3505            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3506          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3507          break;
3508    
3509    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3510          case OP_XCLASS:
3511          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3512              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3513          break;
3514    #endif
3515    
3516          default:
3517          return FALSE;
3518          }
3519    
3520        chr_ptr++;
3521        }
3522      while(*chr_ptr != NOTACHAR);
3523    
3524      /* At least one character must be matched from this opcode. */
3525    
3526      if (list[1] == 0) return TRUE;
3527      }
3528    
3529    return FALSE;
3530    }
3531    
3532    
3533    
3534    /*************************************************
3535    *    Scan compiled regex for auto-possession     *
3536    *************************************************/
3537    
3538    /* Replaces single character iterations with their possessive alternatives
3539    if appropriate. This function modifies the compiled opcode!
3540    
3541    Arguments:
3542      code        points to start of the byte code
3543      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3544      cd          static compile data
3545    
3546    Returns:      nothing
3547    */
3548    
3549    static void
3550    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3551    {
3552    register pcre_uchar c;
3553    const pcre_uchar *end;
3554    pcre_uchar *repeat_opcode;
3555    pcre_uint32 list[8];
3556    
3557    for (;;)
3558      {
3559      c = *code;
3560    
3561      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3562        {
3563        c -= get_repeat_base(c) - OP_STAR;
3564        end = (c <= OP_MINUPTO) ?
3565          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3566        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3567    
3568        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3569          {
3570          switch(c)
3571            {
3572            case OP_STAR:
3573            *code += OP_POSSTAR - OP_STAR;
3574            break;
3575    
3576            case OP_MINSTAR:
3577            *code += OP_POSSTAR - OP_MINSTAR;
3578            break;
3579    
3580            case OP_PLUS:
3581            *code += OP_POSPLUS - OP_PLUS;
3582            break;
3583    
3584            case OP_MINPLUS:
3585            *code += OP_POSPLUS - OP_MINPLUS;
3586            break;
3587    
3588            case OP_QUERY:
3589            *code += OP_POSQUERY - OP_QUERY;
3590            break;
3591    
3592            case OP_MINQUERY:
3593            *code += OP_POSQUERY - OP_MINQUERY;
3594            break;
3595    
3596            case OP_UPTO:
3597            *code += OP_POSUPTO - OP_UPTO;
3598            break;
3599    
3600            case OP_MINUPTO:
3601            *code += OP_MINUPTO - OP_UPTO;
3602            break;
3603            }
3604          }
3605        c = *code;
3606        }
3607      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3608        {
3609    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3610        if (c == OP_XCLASS)
3611          repeat_opcode = code + GET(code, 1);
3612        else
3613    #endif
3614          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3615    
3616        c = *repeat_opcode;
3617        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3618          {
3619          /* end must not be NULL. */
3620          end = get_chr_property_list(code, utf, cd->fcc, list);
3621    
3622          list[1] = (c & 1) == 0;
3623    
3624          if (compare_opcodes(end, utf, cd, list, end))
3625            {
3626            switch (c)
3627              {
3628              case OP_CRSTAR:
3629              case OP_CRMINSTAR:
3630              *repeat_opcode = OP_CRPOSSTAR;
3631              break;
3632    
3633              case OP_CRPLUS:
3634              case OP_CRMINPLUS:
3635              *repeat_opcode = OP_CRPOSPLUS;
3636              break;
3637    
3638              case OP_CRQUERY:
3639              case OP_CRMINQUERY:
3640              *repeat_opcode = OP_CRPOSQUERY;
3641              break;
3642    
3643              case OP_CRRANGE:
3644              case OP_CRMINRANGE:
3645              *repeat_opcode = OP_CRPOSRANGE;
3646              break;
3647              }
3648            }
3649          }
3650        c = *code;
3651        }
3652    
3653      switch(c)
3654        {
3655        case OP_END:
3656        return;
3657    
3658        case OP_TYPESTAR:
3659        case OP_TYPEMINSTAR:
3660        case OP_TYPEPLUS:
3661        case OP_TYPEMINPLUS:
3662        case OP_TYPEQUERY:
3663        case OP_TYPEMINQUERY:
3664        case OP_TYPEPOSSTAR:
3665        case OP_TYPEPOSPLUS:
3666        case OP_TYPEPOSQUERY:
3667        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3668        break;
3669    
3670        case OP_TYPEUPTO:
3671        case OP_TYPEMINUPTO:
3672        case OP_TYPEEXACT:
3673        case OP_TYPEPOSUPTO:
3674        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3675          code += 2;
3676        break;
3677    
3678    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3679        case OP_XCLASS:
3680        code += GET(code, 1);
3681        break;
3682    #endif
3683    
3684        case OP_MARK:
3685        case OP_PRUNE_ARG:
3686        case OP_SKIP_ARG:
3687        case OP_THEN_ARG:
3688        code += code[1];
3689        break;
3690        }
3691    
3692      /* Add in the fixed length from the table */
3693    
3694      code += PRIV(OP_lengths)[c];
3695    
3696      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3697      a multi-byte character. The length in the table is a minimum, so we have to
3698      arrange to skip the extra bytes. */
3699    
3700    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3701      if (utf) switch(c)
3702        {
3703        case OP_CHAR:
3704        case OP_CHARI:
3705        case OP_NOT:
3706        case OP_NOTI:
3707        case OP_STAR:
3708        case OP_MINSTAR:
3709        case OP_PLUS:
3710        case OP_MINPLUS:
3711        case OP_QUERY:
3712        case OP_MINQUERY:
3713        case OP_UPTO:
3714        case OP_MINUPTO:
3715        case OP_EXACT:
3716        case OP_POSSTAR:
3717        case OP_POSPLUS:
3718        case OP_POSQUERY:
3719        case OP_POSUPTO:
3720        case OP_STARI:
3721        case OP_MINSTARI:
3722        case OP_PLUSI:
3723        case OP_MINPLUSI:
3724        case OP_QUERYI:
3725        case OP_MINQUERYI:
3726        case OP_UPTOI:
3727        case OP_MINUPTOI:
3728        case OP_EXACTI:
3729        case OP_POSSTARI:
3730        case OP_POSPLUSI:
3731        case OP_POSQUERYI:
3732        case OP_POSUPTOI:
3733        case OP_NOTSTAR:
3734        case OP_NOTMINSTAR:
3735        case OP_NOTPLUS:
3736        case OP_NOTMINPLUS:
3737        case OP_NOTQUERY:
3738        case OP_NOTMINQUERY:
3739        case OP_NOTUPTO:
3740        case OP_NOTMINUPTO:
3741        case OP_NOTEXACT:
3742        case OP_NOTPOSSTAR:
3743        case OP_NOTPOSPLUS:
3744        case OP_NOTPOSQUERY:
3745        case OP_NOTPOSUPTO:
3746        case OP_NOTSTARI:
3747        case OP_NOTMINSTARI:
3748        case OP_NOTPLUSI:
3749        case OP_NOTMINPLUSI:
3750        case OP_NOTQUERYI:
3751        case OP_NOTMINQUERYI:
3752        case OP_NOTUPTOI:
3753        case OP_NOTMINUPTOI:
3754        case OP_NOTEXACTI:
3755        case OP_NOTPOSSTARI:
3756        case OP_NOTPOSPLUSI:
3757        case OP_NOTPOSQUERYI:
3758        case OP_NOTPOSUPTOI:
3759        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3760      break;      break;
3761      }      }
3762    #else
3763      (void)(utf);  /* Keep compiler happy by referencing function argument */
3764    #endif
3765    }    }
   
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *    Scan compiled regex for non-emptiness       *  
 *************************************************/  
   
 /* This function is called to check for left recursive calls. We want to check  
 the current branch of the current pattern to see if it could match the empty  
 string. If it could, we must look outwards for branches at other levels,  
 stopping when we pass beyond the bracket which is the subject of the recursion.  
 This function is called only during the real compile, not during the  
 pre-compile.  
   
 Arguments:  
   code        points to start of the recursion  
   endcode     points to where to stop (current RECURSE item)  
   bcptr       points to the chain of current (unclosed) branch starts  
   utf8        TRUE if in UTF-8 mode  
   cd          pointers to tables etc  
   
 Returns:      TRUE if what is matched could be empty  
 */  
   
 static BOOL  
 could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  
   BOOL utf8, compile_data *cd)  
 {  
 while (bcptr != NULL && bcptr->current_branch >= code)  
   {  
   if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))  
     return FALSE;  
   bcptr = bcptr->outer;  
   }  
 return TRUE;  
3766  }  }
3767    
3768    
# Line 2288  where Perl recognizes it as the POSIX cl Line 3791  where Perl recognizes it as the POSIX cl
3791  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3792  I think.  I think.
3793    
3794    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3795    It seems that the appearance of a nested POSIX class supersedes an apparent
3796    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3797    a digit.
3798    
3799    In Perl, unescaped square brackets may also appear as part of class names. For
3800    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3801    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3802    seem right at all. PCRE does not allow closing square brackets in POSIX class
3803    names.
3804    
3805  Arguments:  Arguments:
3806    ptr      pointer to the initial [    ptr      pointer to the initial [
3807    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2296  Returns:   TRUE or FALSE Line 3810  Returns:   TRUE or FALSE
3810  */  */
3811    
3812  static BOOL  static BOOL
3813  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3814  {  {
3815  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3816  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3817  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != CHAR_NULL; ptr++)
3818    {    {
3819    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3820        ptr++;
3821      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3822      else
3823      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
3824      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3825        {        {
3826        *endptr = ptr;        *endptr = ptr;
3827        return TRUE;        return TRUE;
3828        }        }
3829        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3830             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3831              ptr[1] == CHAR_EQUALS_SIGN) &&
3832            check_posix_syntax(ptr, endptr))
3833          return FALSE;
3834      }      }
3835    }    }
3836  return FALSE;  return FALSE;
# Line 2333  Returns:     a value representing the na Line 3854  Returns:     a value representing the na
3854  */  */
3855    
3856  static int  static int
3857  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
3858  {  {
3859  const char *pn = posix_names;  const char *pn = posix_names;
3860  register int yield = 0;  register int yield = 0;
3861  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
3862    {    {
3863    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
3864      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3865    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
3866    yield++;    yield++;
3867    }    }
# Line 2372  value in the reference (which is a group Line 3893  value in the reference (which is a group
3893  Arguments:  Arguments:
3894    group      points to the start of the group    group      points to the start of the group
3895    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
3896    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3897    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
3898    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm   the hwm forward reference pointer at the start of the group
3899    
# Line 2380  Returns:     nothing Line 3901  Returns:     nothing
3901  */  */
3902    
3903  static void  static void
3904  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3905    uschar *save_hwm)    pcre_uchar *save_hwm)
3906  {  {
3907  uschar *ptr = group;  pcre_uchar *ptr = group;
3908    
3909  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3910    {    {
3911    int offset;    int offset;
3912    uschar *hc;    pcre_uchar *hc;
3913    
3914    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
3915    reference. */    reference. */
3916    
3917    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3918      {      {
3919      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
3920      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
3921        {        {
3922        PUT(hc, 0, offset + adjust);        PUT(hc, 0, offset + adjust);
# Line 2408  while ((ptr = (uschar *)find_recurse(ptr Line 3929  while ((ptr = (uschar *)find_recurse(ptr
3929    
3930    if (hc >= cd->hwm)    if (hc >= cd->hwm)
3931      {      {
3932      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
3933      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3934      }      }
3935    
# Line 2433  Arguments: Line 3954  Arguments:
3954  Returns:         new code pointer  Returns:         new code pointer
3955  */  */
3956    
3957  static uschar *  static pcre_uchar *
3958  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3959  {  {
3960  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
3961  *code++ = 255;  *code++ = 255;
3962  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3963  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
3964  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
3965  }  }
3966    
3967    
# Line 2462  Returns:             nothing Line 3983  Returns:             nothing
3983  */  */
3984    
3985  static void  static void
3986  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3987  {  {
3