/[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 745 by ph10, Mon Nov 14 11:41:03 2011 UTC revision 1412 by ph10, Sun Dec 15 17:01:46 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 219  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 231  static const char posix_names[] = Line 286  static const char posix_names[] =
286    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
288    
289  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 261  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
329    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
332    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXsp[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXwd[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXwd[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *substitutes[] = {
348      string_PNd,           /* \D */
349      string_pNd,           /* \d */
350      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352      string_PXwd,          /* \W */
353      string_pXwd           /* \w */
354  };  };
355    
356  static const uschar *posix_substitutes[] = {  /* The POSIX class substitutes must be in the order of the POSIX class names,
357    (uschar *)"\\p{L}",     /* alpha */  defined above, and there are both positive and negative cases. NULL means no
358    (uschar *)"\\p{Ll}",    /* lower */  general substitute of a Unicode property escape (\p or \P). However, for some
359    (uschar *)"\\p{Lu}",    /* upper */  POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    (uschar *)"\\p{Xan}",   /* alnum */  directly. */
361    NULL,                   /* ascii */  
362    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pL[] =   {
363    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    NULL,                   /* graph */  static const pcre_uchar string_pLl[] =  {
366    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367    NULL,                   /* punct */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_pLu[] =  {
369    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370    NULL,                   /* xdigit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pXan[] = {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_h[] =    {
375      CHAR_BACKSLASH, CHAR_h, '\0' };
376    static const pcre_uchar string_pXps[] = {
377      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379    static const pcre_uchar string_PL[] =   {
380      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382    static const pcre_uchar string_PLl[] =  {
383      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PLu[] =  {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PXan[] = {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_H[] =    {
392      CHAR_BACKSLASH, CHAR_H, '\0' };
393    static const pcre_uchar string_PXps[] = {
394      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396    
397    static const pcre_uchar *posix_substitutes[] = {
398      string_pL,            /* alpha */
399      string_pLl,           /* lower */
400      string_pLu,           /* upper */
401      string_pXan,          /* alnum */
402      NULL,                 /* ascii */
403      string_h,             /* blank */
404      NULL,                 /* cntrl */
405      string_pNd,           /* digit */
406      NULL,                 /* graph */
407      NULL,                 /* print */
408      NULL,                 /* punct */
409      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411      NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
414    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
415    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
416    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
417    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
418    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
419    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
420    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
421    NULL,                   /* ^graph */    NULL,                 /* ^graph */
422    NULL,                   /* ^print */    NULL,                 /* ^print */
423    NULL,                   /* ^punct */    NULL,                 /* ^punct */
424    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429  #endif  #endif
430    
431  #define STRING(a)  # a  #define STRING(a)  # a
# Line 365  static const char error_texts[] = Line 484  static const char error_texts[] =
484    /* 30 */    /* 30 */
485    "unknown POSIX class name\0"    "unknown POSIX class name\0"
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 388  static const char error_texts[] = Line 507  static const char error_texts[] =
507    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508    /* 50 */    /* 50 */
509    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
510    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
512    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
513    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
# Line 399  static const char error_texts[] = Line 518  static const char error_texts[] =
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 407  static const char error_texts[] = Line 526  static const char error_texts[] =
526    /* 65 */    /* 65 */
527    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
528    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
529    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with Unicode property support\0"
530    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
531    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532      /* 70 */
533      "internal error: unknown opcode in find_fixedlength()\0"
534      "\\N is not supported in a class\0"
535      "too many forward references\0"
536      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537      "invalid UTF-16 string\0"
538      /* 75 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550    ;    ;
551    
552  /* 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 565  For convenience, we use the same bit def
565    
566  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
567    
568    /* Using a simple comparison for decimal numbers rather than a memory read
569    is much faster, and the resulting code is simpler (the compiler turns it
570    into a subtraction and unsigned comparison). */
571    
572    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
573    
574  #ifndef EBCDIC  #ifndef EBCDIC
575    
576  /* 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
577  UTF-8 mode. */  UTF-8 mode. */
578    
579  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
580    {    {
581    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
582    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 615  static const unsigned char digitab[] =
615    
616  /* 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. */
617    
618  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
619    {    {
620    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
621    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 650  static const unsigned char digitab[] =
650    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
651    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
652    
653  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
654    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
655    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
656    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 686  static const unsigned char ebcdic_charta
686  #endif  #endif
687    
688    
689  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726  static BOOL  
727    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
728      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729    left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* 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 */
781      { 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 */
782      { 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 */
783      { 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 */
784      { 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 */
785      { 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 */
786      { 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 */
787      { 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 */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
# Line 570  find_error_text(int n) Line 883  find_error_text(int n)
883  const char *s = error_texts;  const char *s = error_texts;
884  for (; n > 0; n--)  for (; n > 0; n--)
885    {    {
886    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
887    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
888    }    }
889  return s;  return s;
890  }  }
891    
892    
893    
894    /*************************************************
895    *           Expand the workspace                 *
896    *************************************************/
897    
898    /* This function is called during the second compiling phase, if the number of
899    forward references fills the existing workspace, which is originally a block on
900    the stack. A larger block is obtained from malloc() unless the ultimate limit
901    has been reached or the increase will be rather small.
902    
903    Argument: pointer to the compile data block
904    Returns:  0 if all went well, else an error number
905    */
906    
907    static int
908    expand_workspace(compile_data *cd)
909    {
910    pcre_uchar *newspace;
911    int newsize = cd->workspace_size * 2;
912    
913    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
914    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
915        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
916     return ERR72;
917    
918    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
919    if (newspace == NULL) return ERR21;
920    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
921    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
922    if (cd->workspace_size > COMPILE_WORK_SIZE)
923      (PUBL(free))((void *)cd->start_workspace);
924    cd->start_workspace = newspace;
925    cd->workspace_size = newsize;
926    return 0;
927    }
928    
929    
930    
931  /*************************************************  /*************************************************
932  *            Check for counted repeat            *  *            Check for counted repeat            *
933  *************************************************/  *************************************************/
# Line 593  Returns:    TRUE or FALSE Line 944  Returns:    TRUE or FALSE
944  */  */
945    
946  static BOOL  static BOOL
947  is_counted_repeat(const uschar *p)  is_counted_repeat(const pcre_uchar *p)
948  {  {
949  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if (!IS_DIGIT(*p)) return FALSE;
950  while ((digitab[*p] & ctype_digit) != 0) p++;  p++;
951    while (IS_DIGIT(*p)) p++;
952  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953    
954  if (*p++ != CHAR_COMMA) return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
955  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956    
957  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if (!IS_DIGIT(*p)) return FALSE;
958  while ((digitab[*p] & ctype_digit) != 0) p++;  p++;
959    while (IS_DIGIT(*p)) p++;
960    
961  return (*p == CHAR_RIGHT_CURLY_BRACKET);  return (*p == CHAR_RIGHT_CURLY_BRACKET);
962  }  }
# Line 615  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 968  return (*p == CHAR_RIGHT_CURLY_BRACKET);
968  *************************************************/  *************************************************/
969    
970  /* 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
971  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
972  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.
973  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
974  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
975  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979      chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
983    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
984    
985  Returns:         zero or positive => a data character  Returns:         zero => a data character
986                   negative => a special escape sequence                   positive => a special escape sequence
987                     negative => a back reference
988                   on error, errorcodeptr is set                   on error, errorcodeptr is set
989  */  */
990    
991  static int  static int
992  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
994  {  {
995  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
996  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
997  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
998    pcre_uint32 c;
999    int escape = 0;
1000    int i;
1001    
1002  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1003  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1004    
1005  /* 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. */
1006    
1007  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008    
1009  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010  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.
1011  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1012    
1013  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1014  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
1015  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1016    else if ((i = escapes[c - CHAR_0]) != 0)
1017      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018    
1019  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1020  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
1021  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1022    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1023  #endif  #endif
1024    
1025  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1026    
1027  else  else
1028    {    {
1029    const uschar *oldptr;    const pcre_uchar *oldptr;
1030    BOOL braced, negated;    BOOL braced, negated, overflow;
1031      int s;
1032    
1033    switch (c)    switch (c)
1034      {      {
# Line 684  else Line 1045  else
1045        {        {
1046        /* In JavaScript, \u must be followed by four hexadecimal numbers.        /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047        Otherwise it is a lowercase u letter. */        Otherwise it is a lowercase u letter. */
1048        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049             && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1050            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1051            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1052          {          {
1053          c = 0;          c = 0;
1054          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1055            {            {
1056            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1057  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1058            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1059            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 699  else Line 1062  else
1062            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063  #endif  #endif
1064            }            }
1065    
1066    #if defined COMPILE_PCRE8
1067            if (c > (utf ? 0x10ffffU : 0xffU))
1068    #elif defined COMPILE_PCRE16
1069            if (c > (utf ? 0x10ffffU : 0xffffU))
1070    #elif defined COMPILE_PCRE32
1071            if (utf && c > 0x10ffffU)
1072    #endif
1073              {
1074              *errorcodeptr = ERR76;
1075              }
1076            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1077          }          }
1078        }        }
1079      else      else
# Line 725  else Line 1100  else
1100      (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
1101      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1102      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1103      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1104    
1105      case CHAR_g:      case CHAR_g:
1106      if (isclass) break;      if (isclass) break;
1107      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1108        {        {
1109        c = -ESC_g;        escape = ESC_g;
1110        break;        break;
1111        }        }
1112    
# Line 739  else Line 1114  else
1114    
1115      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116        {        {
1117        const uschar *p;        const pcre_uchar *p;
1118        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++)
1119          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121          {          {
1122          c = -ESC_k;          escape = ESC_k;
1123          break;          break;
1124          }          }
1125        braced = TRUE;        braced = TRUE;
# Line 759  else Line 1134  else
1134        }        }
1135      else negated = FALSE;      else negated = FALSE;
1136    
1137      c = 0;      /* The integer range is limited by the machine's int representation. */
1138      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1139        c = c * 10 + *(++ptr) - CHAR_0;      overflow = FALSE;
1140        while (IS_DIGIT(ptr[1]))
1141      if (c < 0)   /* Integer overflow */        {
1142          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1143            {
1144            overflow = TRUE;
1145            break;
1146            }
1147          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1148          }
1149        if (overflow) /* Integer overflow */
1150        {        {
1151          while (IS_DIGIT(ptr[1]))
1152            ptr++;
1153        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1154        break;        break;
1155        }        }
# Line 775  else Line 1160  else
1160        break;        break;
1161        }        }
1162    
1163      if (c == 0)      if (s == 0)
1164        {        {
1165        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1166        break;        break;
# Line 783  else Line 1168  else
1168    
1169      if (negated)      if (negated)
1170        {        {
1171        if (c > bracount)        if (s > bracount)
1172          {          {
1173          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1174          break;          break;
1175          }          }
1176        c = bracount - (c - 1);        s = bracount - (s - 1);
1177        }        }
1178    
1179      c = -(ESC_REF + c);      escape = -s;
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      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
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      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
1188      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
1189      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
1190      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
1191      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
1192      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
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195        Inside a character class, \ followed by a digit is always either a literal
1196        8 or 9 or an octal number. */
1197    
1198      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:
1199      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 812  else Line 1201  else
1201      if (!isclass)      if (!isclass)
1202        {        {
1203        oldptr = ptr;        oldptr = ptr;
1204        c -= CHAR_0;        /* The integer range is limited by the machine's int representation. */
1205        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1206          c = c * 10 + *(++ptr) - CHAR_0;        overflow = FALSE;
1207        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1208            {
1209            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1210              {
1211              overflow = TRUE;
1212              break;
1213              }
1214            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1215            }
1216          if (overflow) /* Integer overflow */
1217          {          {
1218            while (IS_DIGIT(ptr[1]))
1219              ptr++;
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          c = -(ESC_REF + c);          escape = -s;
1226          break;          break;
1227          }          }
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* 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
1232      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
1233      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
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \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
1241      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
1242      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
1243      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,
1244      than 3 octal digits. */      but no more than 3 octal digits. */
1245    
1246      case CHAR_0:      case CHAR_0:
1247      c -= CHAR_0;      c -= CHAR_0;
1248      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1249          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1250      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1251        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1252    #endif
1253        break;
1254    
1255        /* \o is a relatively new Perl feature, supporting a more general way of
1256        specifying character codes in octal. The only supported form is \o{ddd}. */
1257    
1258        case CHAR_o:
1259        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260          {
1261          ptr += 2;
1262          c = 0;
1263          overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265            {
1266            register pcre_uint32 cc = *ptr++;
1267            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268    #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270    #endif
1271            c = (c << 3) + cc - CHAR_0 ;
1272    #if defined COMPILE_PCRE8
1273            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274    #elif defined COMPILE_PCRE16
1275            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278    #endif
1279            }
1280          if (overflow)
1281            {
1282            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283            *errorcodeptr = ERR34;
1284            }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286            {
1287            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290          }
1291      break;      break;
1292    
1293      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      numbers. Otherwise it is a lowercase x letter. */
     treated as a data character. */  
1295    
1296      case CHAR_x:      case CHAR_x:
1297      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
1299        /* In JavaScript, \x must be followed by two hexadecimal numbers.        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300        Otherwise it is a lowercase x letter. */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
       if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)  
1301          {          {
1302          c = 0;          c = 0;
1303          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1304            {            {
1305            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1306  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1307            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1308            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 876  else Line 1312  else
1312  #endif  #endif
1313            }            }
1314          }          }
1315        break;        }    /* End JavaScript handling */
       }  
1316    
1317      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        const uschar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1320        int count = 0;      seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        c = 0;      else
1325        while ((digitab[*pt] & ctype_xdigit) != 0)        {
1326          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          register int cc = *pt++;          ptr += 2;
1329          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1330          count++;          overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336    #ifdef COMPILE_PCRE32
1337              if (c >= 0x10000000l) { overflow = TRUE; break; }
1338    #endif
1339    
1340  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1344          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 */
1345          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346  #endif  #endif
         }  
1347    
1348        if (*pt == CHAR_RIGHT_CURLY_BRACKET)  #if defined COMPILE_PCRE8
1349          {            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;  #elif defined COMPILE_PCRE16
1351          ptr = pt;            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352          break;  #elif defined COMPILE_PCRE32
1353          }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354    #endif
1355              }
1356    
1357        /* If the sequence of hex digits does not end with '}', then we don't          if (overflow)
1358        recognize this construct; fall through to the normal \x handling. */            {
1359        }            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360              *errorcodeptr = ERR34;
1361              }
1362    
1363            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364              {
1365              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366              }
1367    
1368      /* Read just a single-byte hex-defined char */          /* If the sequence of hex digits does not end with '}', give an error.
1369            We used just to recognize this construct and fall through to the normal
1370            \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373      c = 0;          else *errorcodeptr = ERR79;
1374      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)          }   /* End of \x{} processing */
1375        {  
1376        int cc;                                  /* Some compilers don't like */        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377        cc = *(++ptr);                           /* ++ in initializers */  
1378          else
1379            {
1380            c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1389        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391  #endif  #endif
1392        }            }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* 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 935  else Line 1401  else
1401    
1402      case CHAR_c:      case CHAR_c:
1403      c = *(++ptr);      c = *(++ptr);
1404      if (c == 0)      if (c == CHAR_NULL)
1405        {        {
1406        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1407        break;        break;
# Line 975  else Line 1441  else
1441  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1442  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1443    
1444  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1445       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1446    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1447    
1448  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1449    
1450  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1451    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1452    
1453  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1454    
1455  *ptrptr = ptr;  *ptrptr = ptr;
1456  return c;  *chptr = c;
1457    return escape;
1458  }  }
1459    
1460    
# Line 1005  escape sequence. Line 1472  escape sequence.
1472  Argument:  Argument:
1473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1474    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
1475    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
1476      pdataptr       points to an unsigned int that is set to the detailed property value
1477    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1478    
1479  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
1480  */  */
1481    
1482  static int  static BOOL
1483  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484      unsigned int *pdataptr, int *errorcodeptr)
1485  {  {
1486  int c, i, bot, top;  pcre_uchar c;
1487  const uschar *ptr = *ptrptr;  int i, bot, top;
1488  char name[32];  const pcre_uchar *ptr = *ptrptr;
1489    pcre_uchar name[32];
1490    
1491  c = *(++ptr);  c = *(++ptr);
1492  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1493    
1494  *negptr = FALSE;  *negptr = FALSE;
1495    
# Line 1033  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1503  if (c == CHAR_LEFT_CURLY_BRACKET)
1503      *negptr = TRUE;      *negptr = TRUE;
1504      ptr++;      ptr++;
1505      }      }
1506    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507      {      {
1508      c = *(++ptr);      c = *(++ptr);
1509      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1510      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511      name[i] = c;      name[i] = c;
1512      }      }
# Line 1057  else Line 1527  else
1527  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1528    
1529  bot = 0;  bot = 0;
1530  top = _pcre_utt_size;  top = PRIV(utt_size);
1531    
1532  while (bot < top)  while (bot < top)
1533    {    {
1534      int r;
1535    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1536    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537    if (c == 0)    if (r == 0)
1538      {      {
1539      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1540      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1541        return TRUE;
1542      }      }
1543    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1544    }    }
1545    
1546  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1547  *ptrptr = ptr;  *ptrptr = ptr;
1548  return -1;  return FALSE;
1549    
1550  ERROR_RETURN:  ERROR_RETURN:
1551  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1552  *ptrptr = ptr;  *ptrptr = ptr;
1553  return -1;  return FALSE;
1554  }  }
1555  #endif  #endif
1556    
1557    
1558    
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1104  Returns:         pointer to '}' on succe Line 1575  Returns:         pointer to '}' on succe
1575                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1576  */  */
1577    
1578  static const uschar *  static const pcre_uchar *
1579  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)
1580  {  {
1581  int min = 0;  int min = 0;
1582  int max = -1;  int max = -1;
# Line 1113  int max = -1; Line 1584  int max = -1;
1584  /* 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
1585  an integer overflow. */  an integer overflow. */
1586    
1587  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1588  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1589    {    {
1590    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1128  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1599  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1599    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1600      {      {
1601      max = 0;      max = 0;
1602      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1603      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1604        {        {
1605        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1153  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
1627  *  Subroutine for finding forward reference      *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
1630  /* This recursive function is called only from find_parens() below. The  /* This is called by several functions that scan a compiled expression looking
1631  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
1632  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
1633  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
1634  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.)  
1635    
1636  Arguments:  Arguments:
1637    ptrptr       address of the current character pointer (updated)    code         pointer to the start of the group
1638    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)  
1639    
1640  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1641  */  */
1642    
1643  static int  static const pcre_uchar*
1644  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)  
1645  {  {
1646  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)  
1647    {    {
1648    /* Handle specials such as (*SKIP) or (*UTF8) etc. */    switch ((int)*code)
1649        {
1650        case OP_ASSERT_NOT:
1651        case OP_ASSERTBACK:
1652        case OP_ASSERTBACK_NOT:
1653        if (!skipassert) return code;
1654        do code += GET(code, 1); while (*code == OP_ALT);
1655        code += PRIV(OP_lengths)[*code];
1656        break;
1657    
1658    if (ptr[1] == CHAR_ASTERISK) ptr += 2;      case OP_WORD_BOUNDARY:
1659        case OP_NOT_WORD_BOUNDARY:
1660        if (!skipassert) return code;
1661        /* Fall through */
1662    
1663    /* Handle a normal, unnamed capturing parenthesis. */      case OP_CALLOUT:
1664        case OP_CREF:
1665        case OP_DNCREF:
1666        case OP_RREF:
1667        case OP_DNRREF:
1668        case OP_DEF:
1669        code += PRIV(OP_lengths)[*code];
1670        break;
1671    
1672    else if (ptr[1] != CHAR_QUESTION_MARK)      default:
1673      {      return code;
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
1674      }      }
1675      }
1676    /* Control never reaches here */
1677    }
1678    
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
1679    
   /* Handle comments; all characters are allowed until a ket is reached. */  
1680    
1681    else if (ptr[2] == CHAR_NUMBER_SIGN)  /*************************************************
1682      {  *        Find the fixed length of a branch       *
1683      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  *************************************************/
     goto FAIL_EXIT;  
     }  
1684    
1685    /* Handle a condition. If it is an assertion, just carry on so that it  /* Scan a branch and compute the fixed length of subject that will match it,
1686    is processed as normal. If not, skip to the closing parenthesis of the  if the length is fixed. This is needed for dealing with backward assertions.
1687    condition (there can't be any nested parens). */  In UTF8 mode, the result is in characters rather than bytes. The branch is
1688    temporarily terminated with OP_END when this function is called.
1689    
1690    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  This function is called when a backward assertion is encountered, so that if it
1691      {  fails, the error message can point to the correct place in the pattern.
1692      ptr += 2;  However, we cannot do this when the assertion contains subroutine calls,
1693      if (ptr[1] != CHAR_QUESTION_MARK)  because they can be forward references. We solve this by remembering this case
1694        {  and doing the check at the end; a flag specifies which mode we are running in.
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
1695    
1696    /* Start with (? but not a condition. */  Arguments:
1697      code     points to the start of the pattern (the bracket)
1698      utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699      atend    TRUE if called when the pattern is complete
1700      cd       the "compile data" structure
1701    
1702    else  Returns:   the fixed length,
1703      {               or -1 if there is no fixed length,
1704      ptr += 2;               or -2 if \C was encountered (in UTF-8 mode only)
1705      if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706                 or -4 if an unknown opcode was encountered (internal error)
1707    */
1708    
1709      /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  static int
1710    find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711    {
1712    int length = -1;
1713    
1714      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  register int branchlength = 0;
1715          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  register pcre_uchar *cc = code + 1 + LINK_SIZE;
       {  
       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++;  
       }  
     }  
   }  
1716    
1717  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Scan along the opcodes for this branch. If we get to the end of the
1718  bars. Stop if we get to cd->end_pattern. Note that this is important for the  branch, check the length against that of the other branches. */
 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. */  
1719    
1720  for (; ptr < cd->end_pattern; ptr++)  for (;;)
1721    {    {
1722    /* Skip over backslashed characters and also entire \Q...\E */    int d;
1723      pcre_uchar *ce, *cs;
1724      register pcre_uchar op = *cc;
1725    
1726    if (*ptr == CHAR_BACKSLASH)    switch (op)
1727      {      {
1728      if (*(++ptr) == 0) goto FAIL_EXIT;      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729      if (*ptr == CHAR_Q) for (;;)      OP_BRA (normal non-capturing bracket) because the other variants of these
1730        {      opcodes are all concerned with unlimited repeated groups, which of course
1731        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};      are not of fixed length. */
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
1732    
1733    /* Skip over character classes; this logic must be similar to the way they      case OP_CBRA:
1734    are handled for real. If the first character is '^', skip it. Also, if the      case OP_BRA:
1735    first few characters (either before or after ^) are \Q\E or \E we skip them      case OP_ONCE:
1736    too. This makes for compatibility with Perl. Note the use of STR macros to      case OP_ONCE_NC:
1737    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */      case OP_COND:
1738        d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739        if (d < 0) return d;
1740        branchlength += d;
1741        do cc += GET(cc, 1); while (*cc == OP_ALT);
1742        cc += 1 + LINK_SIZE;
1743        break;
1744    
1745    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1746      {      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1747      BOOL negate_class = FALSE;      an ALT. If it is END it's the end of the outer call. All can be handled by
1748      for (;;)      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749        {      because they all imply an unlimited repeat. */
       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;  
       }  
1750    
1751      /* If the next character is ']', it is a data character that must be      case OP_ALT:
1752      skipped, except in JavaScript compatibility mode. */      case OP_KET:
1753        case OP_END:
1754        case OP_ACCEPT:
1755        case OP_ASSERT_ACCEPT:
1756        if (length < 0) length = branchlength;
1757          else if (length != branchlength) return -1;
1758        if (*cc != OP_ALT) return length;
1759        cc += 1 + LINK_SIZE;
1760        branchlength = 0;
1761        break;
1762    
1763      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&      /* A true recursion implies not fixed length, but a subroutine call may
1764          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)      be OK. If the subroutine is a forward reference, we can't deal with
1765        ptr++;      it until the end of the pattern, so return -3. */
   
     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  
 */  
   
 static int  
 find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)  
 {  
 int length = -1;  
   
 register int branchlength = 0;  
 register uschar *cc = code + 1 + LINK_SIZE;  
   
 /* Scan along the opcodes for this branch. If we get to the end of the  
 branch, check the length against that of the other branches. */  
   
 for (;;)  
   {  
   int d;  
   uschar *ce, *cs;  
   register int op = *cc;  
   switch (op)  
     {  
     /* We only need to continue for OP_CBRA (normal capturing bracket) and  
     OP_BRA (normal non-capturing bracket) because the other variants of these  
     opcodes are all concerned with unlimited repeated groups, which of course  
     are not of fixed length. They will cause a -1 response from the default  
     case of this switch. */  
   
     case OP_CBRA:  
     case OP_BRA:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_COND:  
     d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);  
     if (d < 0) return d;  
     branchlength += d;  
     do cc += GET(cc, 1); while (*cc == OP_ALT);  
     cc += 1 + LINK_SIZE;  
     break;  
   
     /* Reached end of a branch; if it's a ket it is the end of a nested  
     call. If it's ALT it is an alternation in a nested call. If it is  
     END it's the end of the outer call. All can be handled by the same code.  
     Note that we must not include the OP_KETRxxx opcodes here, because they  
     all imply an unlimited repeat. */  
   
     case OP_ALT:  
     case OP_KET:  
     case OP_END:  
     if (length < 0) length = branchlength;  
       else if (length != branchlength) return -1;  
     if (*cc != OP_ALT) return length;  
     cc += 1 + LINK_SIZE;  
     branchlength = 0;  
     break;  
   
     /* A true recursion implies not fixed length, but a subroutine call may  
     be OK. If the subroutine is a forward reference, we can't deal with  
     it until the end of the pattern, so return -3. */  
1766    
1767      case OP_RECURSE:      case OP_RECURSE:
1768      if (!atend) return -3;      if (!atend) return -3;
1769      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1770      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1771      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1772      d = find_fixedlength(cs + 2, utf8, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1773      if (d < 0) return d;      if (d < 0) return d;
1774      branchlength += d;      branchlength += d;
1775      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1604  for (;;) Line 1782  for (;;)
1782      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1783      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1784      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1785      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1786        break;
1787    
1788      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1789    
1790      case OP_REVERSE:      case OP_MARK:
1791      case OP_CREF:      case OP_PRUNE_ARG:
1792      case OP_NCREF:      case OP_SKIP_ARG:
1793      case OP_RREF:      case OP_THEN_ARG:
1794      case OP_NRREF:      cc += cc[1] + PRIV(OP_lengths)[*cc];
1795      case OP_DEF:      break;
1796    
1797      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1798      case OP_CIRC:      case OP_CIRC:
1799      case OP_CIRCM:      case OP_CIRCM:
1800        case OP_CLOSE:
1801        case OP_COMMIT:
1802        case OP_CREF:
1803        case OP_DEF:
1804        case OP_DNCREF:
1805        case OP_DNRREF:
1806      case OP_DOLL:      case OP_DOLL:
1807      case OP_DOLLM:      case OP_DOLLM:
1808        case OP_EOD:
1809        case OP_EODN:
1810        case OP_FAIL:
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812        case OP_PRUNE:
1813        case OP_REVERSE:
1814        case OP_RREF:
1815        case OP_SET_SOM:
1816        case OP_SKIP:
1817        case OP_SOD:
1818        case OP_SOM:
1819        case OP_THEN:
1820      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1821      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1822      break;      break;
1823    
1824      /* Handle literal characters */      /* Handle literal characters */
# Line 1637  for (;;) Line 1829  for (;;)
1829      case OP_NOTI:      case OP_NOTI:
1830      branchlength++;      branchlength++;
1831      cc += 2;      cc += 2;
1832  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1833      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834  #endif  #endif
1835      break;      break;
1836    
# Line 1646  for (;;) Line 1838  for (;;)
1838      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1839    
1840      case OP_EXACT:      case OP_EXACT:
1841      branchlength += GET2(cc,1);      case OP_EXACTI:
1842      cc += 4;      case OP_NOTEXACT:
1843  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1844      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      branchlength += (int)GET2(cc,1);
1845        cc += 2 + IMM2_SIZE;
1846    #ifdef SUPPORT_UTF
1847        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848  #endif  #endif
1849      break;      break;
1850    
1851      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1852      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1853      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854      cc += 4;        cc += 2;
1855        cc += 1 + IMM2_SIZE + 1;
1856      break;      break;
1857    
1858      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1666  for (;;) Line 1862  for (;;)
1862      cc += 2;      cc += 2;
1863      /* Fall through */      /* Fall through */
1864    
1865        case OP_HSPACE:
1866        case OP_VSPACE:
1867        case OP_NOT_HSPACE:
1868        case OP_NOT_VSPACE:
1869      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1870      case OP_DIGIT:      case OP_DIGIT:
1871      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1678  for (;;) Line 1878  for (;;)
1878      cc++;      cc++;
1879      break;      break;
1880    
1881      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1882        otherwise \C is coded as OP_ALLANY. */
1883    
1884      case OP_ANYBYTE:      case OP_ANYBYTE:
1885      return -2;      return -2;
1886    
1887      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1888    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1889      case OP_CLASS:      case OP_CLASS:
1890      case OP_NCLASS:      case OP_NCLASS:
1891      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892        case OP_XCLASS:
1893        /* The original code caused an unsigned overflow in 64 bit systems,
1894        so now we use a conditional statement. */
1895        if (op == OP_XCLASS)
1896          cc += GET(cc, 1);
1897        else
1898          cc += PRIV(OP_lengths)[OP_CLASS];
1899    #else
1900        cc += PRIV(OP_lengths)[OP_CLASS];
1901    #endif
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1919        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        cc += 5;        branchlength += (int)GET2(cc,1);
1921          cc += 1 + 2 * IMM2_SIZE;
1922        break;        break;
1923    
1924        default:        default:
# Line 1717  for (;;) Line 1928  for (;;)
1928    
1929      /* Anything else is variable length */      /* Anything else is variable length */
1930    
1931      default:      case OP_ANYNL:
1932        case OP_BRAMINZERO:
1933        case OP_BRAPOS:
1934        case OP_BRAPOSZERO:
1935        case OP_BRAZERO:
1936        case OP_CBRAPOS:
1937        case OP_EXTUNI:
1938        case OP_KETRMAX:
1939        case OP_KETRMIN:
1940        case OP_KETRPOS:
1941        case OP_MINPLUS:
1942        case OP_MINPLUSI:
1943        case OP_MINQUERY:
1944        case OP_MINQUERYI:
1945        case OP_MINSTAR:
1946        case OP_MINSTARI:
1947        case OP_MINUPTO:
1948        case OP_MINUPTOI:
1949        case OP_NOTMINPLUS:
1950        case OP_NOTMINPLUSI:
1951        case OP_NOTMINQUERY:
1952        case OP_NOTMINQUERYI:
1953        case OP_NOTMINSTAR:
1954        case OP_NOTMINSTARI:
1955        case OP_NOTMINUPTO:
1956        case OP_NOTMINUPTOI:
1957        case OP_NOTPLUS:
1958        case OP_NOTPLUSI:
1959        case OP_NOTPOSPLUS:
1960        case OP_NOTPOSPLUSI:
1961        case OP_NOTPOSQUERY:
1962        case OP_NOTPOSQUERYI:
1963        case OP_NOTPOSSTAR:
1964        case OP_NOTPOSSTARI:
1965        case OP_NOTPOSUPTO:
1966        case OP_NOTPOSUPTOI:
1967        case OP_NOTQUERY:
1968        case OP_NOTQUERYI:
1969        case OP_NOTSTAR:
1970        case OP_NOTSTARI:
1971        case OP_NOTUPTO:
1972        case OP_NOTUPTOI:
1973        case OP_PLUS:
1974        case OP_PLUSI:
1975        case OP_POSPLUS:
1976        case OP_POSPLUSI:
1977        case OP_POSQUERY:
1978        case OP_POSQUERYI:
1979        case OP_POSSTAR:
1980        case OP_POSSTARI:
1981        case OP_POSUPTO:
1982        case OP_POSUPTOI:
1983        case OP_QUERY:
1984        case OP_QUERYI:
1985        case OP_REF:
1986        case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989        case OP_SBRA:
1990        case OP_SBRAPOS:
1991        case OP_SCBRA:
1992        case OP_SCBRAPOS:
1993        case OP_SCOND:
1994        case OP_SKIPZERO:
1995        case OP_STAR:
1996        case OP_STARI:
1997        case OP_TYPEMINPLUS:
1998        case OP_TYPEMINQUERY:
1999        case OP_TYPEMINSTAR:
2000        case OP_TYPEMINUPTO:
2001        case OP_TYPEPLUS:
2002        case OP_TYPEPOSPLUS:
2003        case OP_TYPEPOSQUERY:
2004        case OP_TYPEPOSSTAR:
2005        case OP_TYPEPOSUPTO:
2006        case OP_TYPEQUERY:
2007        case OP_TYPESTAR:
2008        case OP_TYPEUPTO:
2009        case OP_UPTO:
2010        case OP_UPTOI:
2011      return -1;      return -1;
2012    
2013        /* Catch unrecognized opcodes so that when new ones are added they
2014        are not forgotten, as has happened in the past. */
2015    
2016        default:
2017        return -4;
2018      }      }
2019    }    }
2020  /* Control never gets here */  /* Control never gets here */
# Line 1726  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 1739  length. Line 2034  length.
2034    
2035  Arguments:  Arguments:
2036    code        points to start of expression    code        points to start of expression
2037    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2038    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2039    
2040  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
2041  */  */
2042    
2043  const uschar *  const pcre_uchar *
2044  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045  {  {
2046  for (;;)  for (;;)
2047    {    {
2048    register int c = *code;    register pcre_uchar c = *code;
2049    
2050    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2051    
# Line 1764  for (;;) Line 2059  for (;;)
2059    
2060    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2061      {      {
2062      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2063      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2064      }      }
2065    
2066    /* Handle capturing bracket */    /* Handle capturing bracket */
# Line 1773  for (;;) Line 2068  for (;;)
2068    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2069             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070      {      {
2071      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2072      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2073      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2074      }      }
2075    
2076    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
# Line 1803  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102            code += 2;
2103        break;        break;
2104    
2105        case OP_MARK:        case OP_MARK:
2106        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2108        code += code[1];        case OP_THEN_ARG:
       break;  
   
       case OP_THEN_ARG:  
2109        code += code[1];        code += code[1];
2110        break;        break;
2111        }        }
2112    
2113      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2114    
2115      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2116    
2117    /* 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
2118    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
2119    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2120    
2121  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2122      if (utf8) switch(c)      if (utf) switch(c)
2123        {        {
2124        case OP_CHAR:        case OP_CHAR:
2125        case OP_CHARI:        case OP_CHARI:
# Line 1856  for (;;) Line 2149  for (;;)
2149        case OP_MINQUERYI:        case OP_MINQUERYI:
2150        case OP_POSQUERY:        case OP_POSQUERY:
2151        case OP_POSQUERYI:        case OP_POSQUERYI:
2152        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2153        break;        break;
2154        }        }
2155  #else  #else
2156      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2157  #endif  #endif
2158      }      }
2159    }    }
# Line 1877  instance of OP_RECURSE. Line 2170  instance of OP_RECURSE.
2170    
2171  Arguments:  Arguments:
2172    code        points to start of expression    code        points to start of expression
2173    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2174    
2175  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
2176  */  */
2177    
2178  static const uschar *  static const pcre_uchar *
2179  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2180  {  {
2181  for (;;)  for (;;)
2182    {    {
2183    register int c = *code;    register pcre_uchar c = *code;
2184    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2185    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2186    
# Line 1922  for (;;) Line 2215  for (;;)
2215        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2216        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2217        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2218        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219            code += 2;
2220        break;        break;
2221    
2222        case OP_MARK:        case OP_MARK:
2223        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2224        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2225        case OP_THEN_ARG:        case OP_THEN_ARG:
2226        code += code[1];        code += code[1];
2227        break;        break;
# Line 1938  for (;;) Line 2229  for (;;)
2229    
2230      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2231    
2232      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2233    
2234      /* 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
2235      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
2236      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2237    
2238  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2239      if (utf8) switch(c)      if (utf) switch(c)
2240        {        {
2241        case OP_CHAR:        case OP_CHAR:
2242        case OP_CHARI:        case OP_CHARI:
2243          case OP_NOT:
2244          case OP_NOTI:
2245        case OP_EXACT:        case OP_EXACT:
2246        case OP_EXACTI:        case OP_EXACTI:
2247          case OP_NOTEXACT:
2248          case OP_NOTEXACTI:
2249        case OP_UPTO:        case OP_UPTO:
2250        case OP_UPTOI:        case OP_UPTOI:
2251          case OP_NOTUPTO:
2252          case OP_NOTUPTOI:
2253        case OP_MINUPTO:        case OP_MINUPTO:
2254        case OP_MINUPTOI:        case OP_MINUPTOI:
2255          case OP_NOTMINUPTO:
2256          case OP_NOTMINUPTOI:
2257        case OP_POSUPTO:        case OP_POSUPTO:
2258        case OP_POSUPTOI:        case OP_POSUPTOI:
2259          case OP_NOTPOSUPTO:
2260          case OP_NOTPOSUPTOI:
2261        case OP_STAR:        case OP_STAR:
2262        case OP_STARI:        case OP_STARI:
2263          case OP_NOTSTAR:
2264          case OP_NOTSTARI:
2265        case OP_MINSTAR:        case OP_MINSTAR:
2266        case OP_MINSTARI:        case OP_MINSTARI:
2267          case OP_NOTMINSTAR:
2268          case OP_NOTMINSTARI:
2269        case OP_POSSTAR:        case OP_POSSTAR:
2270        case OP_POSSTARI:        case OP_POSSTARI:
2271          case OP_NOTPOSSTAR:
2272          case OP_NOTPOSSTARI:
2273        case OP_PLUS:        case OP_PLUS:
2274        case OP_PLUSI:        case OP_PLUSI:
2275          case OP_NOTPLUS:
2276          case OP_NOTPLUSI:
2277        case OP_MINPLUS:        case OP_MINPLUS:
2278        case OP_MINPLUSI:        case OP_MINPLUSI:
2279          case OP_NOTMINPLUS:
2280          case OP_NOTMINPLUSI:
2281        case OP_POSPLUS:        case OP_POSPLUS:
2282        case OP_POSPLUSI:        case OP_POSPLUSI:
2283          case OP_NOTPOSPLUS:
2284          case OP_NOTPOSPLUSI:
2285        case OP_QUERY:        case OP_QUERY:
2286        case OP_QUERYI:        case OP_QUERYI:
2287          case OP_NOTQUERY:
2288          case OP_NOTQUERYI:
2289        case OP_MINQUERY:        case OP_MINQUERY:
2290        case OP_MINQUERYI:        case OP_MINQUERYI:
2291          case OP_NOTMINQUERY:
2292          case OP_NOTMINQUERYI:
2293        case OP_POSQUERY:        case OP_POSQUERY:
2294        case OP_POSQUERYI:        case OP_POSQUERYI:
2295        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_NOTPOSQUERY:
2296          case OP_NOTPOSQUERYI:
2297          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2298        break;        break;
2299        }        }
2300  #else  #else
2301      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2302  #endif  #endif
2303      }      }
2304    }    }
# Line 2002  bracket whose current branch will alread Line 2321  bracket whose current branch will alread
2321  Arguments:  Arguments:
2322    code        points to start of search    code        points to start of search
2323    endcode     points to where to stop    endcode     points to where to stop
2324    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2326      recurses    chain of recurse_check to catch mutual recursion
2327    
2328  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2329  */  */
2330    
2331    typedef struct recurse_check {
2332      struct recurse_check *prev;
2333      const pcre_uchar *group;
2334    } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338    compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2339  {  {
2340  register int c;  register pcre_uchar c;
2341  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);  recurse_check this_recurse;
2342    
2343    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346    {    {
2347    const uschar *ccode;    const pcre_uchar *ccode;
2348    
2349    c = *code;    c = *code;
2350    
# Line 2040  for (code = first_significant_code(code Line 2367  for (code = first_significant_code(code
2367    
2368    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2369      {      {
2370      const uschar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371      BOOL empty_branch;      BOOL empty_branch;
2372    
2373      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2374        when called to scan a completed pattern by setting cd->start_workspace to
2375        NULL. */
2376    
2377        if (cd->start_workspace != NULL)
2378          {
2379          const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383          }
2384    
2385        /* If we are scanning a completed pattern, there are no forward references
2386        and all groups are complete. We need to detect whether this is a recursive
2387        call, as otherwise there will be an infinite loop. If it is a recursion,
2388        just skip over it. Simple recursions are easily detected. For mutual
2389        recursions we keep a chain on the stack. */
2390    
2391      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2392        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2393          recurse_check *r = recurses;
2394          const pcre_uchar *endgroup = scode;
2395    
2396      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2400      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2401      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2402          }
2403    
2404        /* Completed reference; scan the referenced group, remembering it on the
2405        stack chain to detect mutual recursions. */
2406    
2407      /* Completed backwards reference */      empty_branch = FALSE;
2408        this_recurse.prev = recurses;
2409        this_recurse.group = scode;
2410    
2411      do      do
2412        {        {
2413        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414          {          {
2415          empty_branch = TRUE;          empty_branch = TRUE;
2416          break;          break;
# Line 2076  for (code = first_significant_code(code Line 2428  for (code = first_significant_code(code
2428    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2429        c == OP_BRAPOSZERO)        c == OP_BRAPOSZERO)
2430      {      {
2431      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2432      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2433      c = *code;      c = *code;
2434      continue;      continue;
# Line 2114  for (code = first_significant_code(code Line 2466  for (code = first_significant_code(code
2466        empty_branch = FALSE;        empty_branch = FALSE;
2467        do        do
2468          {          {
2469          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470            empty_branch = TRUE;            empty_branch = TRUE;
2471          code += GET(code, 1);          code += GET(code, 1);
2472          }          }
# Line 2132  for (code = first_significant_code(code Line 2484  for (code = first_significant_code(code
2484      {      {
2485      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2486      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2487      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2488      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"
2489      here. */      here. */
2490    
2491  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2492      case OP_XCLASS:      case OP_XCLASS:
2493      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2494      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2144  for (code = first_significant_code(code Line 2496  for (code = first_significant_code(code
2496    
2497      case OP_CLASS:      case OP_CLASS:
2498      case OP_NCLASS:      case OP_NCLASS:
2499      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2500    
2501  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2502      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2503  #endif  #endif
2504    
# Line 2156  for (code = first_significant_code(code Line 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        case OP_CRPOSRANGE:
2524        break;        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        }        break;
2526      break;        }
2527        break;
2528    
2529        /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535        case OP_PROP:
2536        case OP_NOTPROP:
2537        case OP_ANYNL:
2538    
2539        case OP_NOT_HSPACE:
2540        case OP_HSPACE:
2541        case OP_NOT_VSPACE:
2542        case OP_VSPACE:
2543        case OP_EXTUNI:
2544    
2545        case OP_NOT_DIGIT:
2546        case OP_DIGIT:
2547        case OP_NOT_WHITESPACE:
2548        case OP_WHITESPACE:
2549        case OP_NOT_WORDCHAR:
2550        case OP_WORDCHAR:
2551    
2552        case OP_CHAR:
2553        case OP_CHARI:
2554        case OP_NOT:
2555        case OP_NOTI:
2556    
2557        case OP_PLUS:
2558        case OP_PLUSI:
2559        case OP_MINPLUS:
2560        case OP_MINPLUSI:
2561    
2562        case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564        case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569        case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574        case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577        case OP_TYPEPLUS:
2578        case OP_TYPEMINPLUS:
2579        case OP_TYPEPOSPLUS:
2580        case OP_TYPEEXACT:
2581    
2582        return FALSE;
2583    
2584        /* These are going to continue, as they may be empty, but we have to
2585        fudge the length for the \p and \P cases. */
2586    
2587        case OP_TYPESTAR:
2588        case OP_TYPEMINSTAR:
2589        case OP_TYPEPOSSTAR:
2590        case OP_TYPEQUERY:
2591        case OP_TYPEMINQUERY:
2592        case OP_TYPEPOSQUERY:
2593        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2594        break;
2595    
2596        /* Same for these */
2597    
2598        case OP_TYPEUPTO:
2599        case OP_TYPEMINUPTO:
2600        case OP_TYPEPOSUPTO:
2601        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602          code += 2;
2603        break;
2604    
2605        /* End of branch */
2606    
2607        case OP_KET:
2608        case OP_KETRMAX:
2609        case OP_KETRMIN:
2610        case OP_KETRPOS:
2611        case OP_ALT:
2612        return TRUE;
2613    
2614        /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615        MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619        case OP_STAR:
2620        case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624        case OP_MINSTAR:
2625        case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629        case OP_POSSTAR:
2630        case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634        case OP_QUERY:
2635        case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639        case OP_MINQUERY:
2640        case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644        case OP_POSQUERY:
2645        case OP_POSQUERYI:
2646        case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650        break;
2651    
2652        case OP_UPTO:
2653        case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657        case OP_MINUPTO:
2658        case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662        case OP_POSUPTO:
2663        case OP_POSUPTOI:
2664        case OP_NOTPOSUPTO:
2665        case OP_NOTPOSUPTOI:
2666    
2667        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668        break;
2669    #endif
2670    
2671        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2672        string. */
2673    
2674        case OP_MARK:
2675        case OP_PRUNE_ARG:
2676        case OP_SKIP_ARG:
2677        case OP_THEN_ARG:
2678        code += code[1];
2679        break;
2680    
2681        /* None of the remaining opcodes are required to match a character. */
2682    
2683        default:
2684        break;
2685        }
2686      }
2687    
2688    return TRUE;
2689    }
2690    
2691    
2692    
2693    /*************************************************
2694    *    Scan compiled regex for non-emptiness       *
2695    *************************************************/
2696    
2697    /* This function is called to check for left recursive calls. We want to check
2698    the current branch of the current pattern to see if it could match the empty
2699    string. If it could, we must look outwards for branches at other levels,
2700    stopping when we pass beyond the bracket which is the subject of the recursion.
2701    This function is called only during the real compile, not during the
2702    pre-compile.
2703    
2704    Arguments:
2705      code        points to start of the recursion
2706      endcode     points to where to stop (current RECURSE item)
2707      bcptr       points to the chain of current (unclosed) branch starts
2708      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709      cd          pointers to tables etc
2710    
2711    Returns:      TRUE if what is matched could be empty
2712    */
2713    
2714    static BOOL
2715    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716      branch_chain *bcptr, BOOL utf, compile_data *cd)
2717    {
2718    while (bcptr != NULL && bcptr->current_branch >= code)
2719      {
2720      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721        return FALSE;
2722      bcptr = bcptr->outer;
2723      }
2724    return TRUE;
2725    }
2726    
2727    
2728    
2729    /*************************************************
2730    *        Base opcode of repeated opcodes         *
2731    *************************************************/
2732    
2733    /* Returns the base opcode for repeated single character type opcodes. If the
2734    opcode is not a repeated character type, it returns with the original value.
2735    
2736    Arguments:  c opcode
2737    Returns:    base opcode for the type
2738    */
2739    
2740    static pcre_uchar
2741    get_repeat_base(pcre_uchar c)
2742    {
2743    return (c > OP_TYPEPOSUPTO)? c :
2744           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747           (c >= OP_STARI)?      OP_STARI :
2748                                 OP_STAR;
2749    }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754    /*************************************************
2755    *        Check a character and a property        *
2756    *************************************************/
2757    
2758    /* This function is called by check_auto_possessive() when a property item
2759    is adjacent to a fixed character.
2760    
2761    Arguments:
2762      c            the character
2763      ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767    Returns:       TRUE if auto-possessifying is OK
2768    */
2769    
2770    static BOOL
2771    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773    {
2774    const pcre_uint32 *p;
2775    const ucd_record *prop = GET_UCD(c);
2776    
2777    switch(ptype)
2778      {
2779      case PT_LAMP:
2780      return (prop->chartype == ucp_Lu ||
2781              prop->chartype == ucp_Ll ||
2782              prop->chartype == ucp_Lt) == negated;
2783    
2784      case PT_GC:
2785      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2786    
2787      case PT_PC:
2788      return (pdata == prop->chartype) == negated;
2789    
2790      case PT_SC:
2791      return (pdata == prop->script) == negated;
2792    
2793      /* These are specials */
2794    
2795      case PT_ALNUM:
2796      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2798    
2799      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800      means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803      case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806        {
2807        HSPACE_CASES:
2808        VSPACE_CASES:
2809        return negated;
2810    
2811        default:
2812        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813        }
2814      break;  /* Control never reaches here */
2815    
2816      case PT_WORD:
2817      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821      case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824        {
2825        if (c < *p) return !negated;
2826        if (c == *p++) return negated;
2827        }
2828      break;  /* Control never reaches here */
2829      }
2830    
2831    return FALSE;
2832    }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837    /*************************************************
2838    *        Fill the character property list        *
2839    *************************************************/
2840    
2841    /* Checks whether the code points to an opcode that can take part in auto-
2842    possessification, and if so, fills a list with its properties.
2843    
2844    Arguments:
2845      code        points to start of expression
2846      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847      fcc         points to case-flipping table
2848      list        points to output list
2849                  list[0] will be filled with the opcode
2850                  list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852                  list[2..7] depends on the opcode
2853    
2854    Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856    */
2857    
2858    static const pcre_uchar *
2859    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861    {
2862    pcre_uchar c = *code;
2863    pcre_uchar base;
2864    const pcre_uchar *end;
2865    pcre_uint32 chr;
2866    
2867    #ifdef SUPPORT_UCP
2868    pcre_uint32 *clist_dest;
2869    const pcre_uint32 *clist_src;
2870    #else
2871    utf = utf;  /* Suppress "unused parameter" compiler warning */
2872    #endif
2873    
2874    list[0] = c;
2875    list[1] = FALSE;
2876    code++;
2877    
2878    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879      {
2880      base = get_repeat_base(c);
2881      c -= (base - OP_STAR);
2882    
2883      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884        code += IMM2_SIZE;
2885    
2886      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2887    
2888      switch(base)
2889        {
2890        case OP_STAR:
2891        list[0] = OP_CHAR;
2892        break;
2893    
2894        case OP_STARI:
2895        list[0] = OP_CHARI;
2896        break;
2897    
2898        case OP_NOTSTAR:
2899        list[0] = OP_NOT;
2900        break;
2901    
2902        case OP_NOTSTARI:
2903        list[0] = OP_NOTI;
2904        break;
2905    
2906        case OP_TYPESTAR:
2907        list[0] = *code;
2908        code++;
2909        break;
2910        }
2911      c = list[0];
2912      }
2913    
2914    switch(c)
2915      {
2916      case OP_NOT_DIGIT:
2917      case OP_DIGIT:
2918      case OP_NOT_WHITESPACE:
2919      case OP_WHITESPACE:
2920      case OP_NOT_WORDCHAR:
2921      case OP_WORDCHAR:
2922      case OP_ANY:
2923      case OP_ALLANY:
2924      case OP_ANYNL:
2925      case OP_NOT_HSPACE:
2926      case OP_HSPACE:
2927      case OP_NOT_VSPACE:
2928      case OP_VSPACE:
2929      case OP_EXTUNI:
2930      case OP_EODN:
2931      case OP_EOD:
2932      case OP_DOLL:
2933      case OP_DOLLM:
2934      return code;
2935    
2936      case OP_CHAR:
2937      case OP_NOT:
2938      GETCHARINCTEST(chr, code);
2939      list[2] = chr;
2940      list[3] = NOTACHAR;
2941      return code;
2942    
2943      case OP_CHARI:
2944      case OP_NOTI:
2945      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946      GETCHARINCTEST(chr, code);
2947      list[2] = chr;
2948    
2949    #ifdef SUPPORT_UCP
2950      if (chr < 128 || (chr < 256 && !utf))
2951        list[3] = fcc[chr];
2952      else
2953        list[3] = UCD_OTHERCASE(chr);
2954    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955      list[3] = (chr < 256) ? fcc[chr] : chr;
2956    #else
2957      list[3] = fcc[chr];
2958    #endif
2959    
2960      /* The othercase might be the same value. */
2961    
2962      if (chr == list[3])
2963        list[3] = NOTACHAR;
2964      else
2965        list[4] = NOTACHAR;
2966      return code;
2967    
2968    #ifdef SUPPORT_UCP
2969      case OP_PROP:
2970      case OP_NOTPROP:
2971      if (code[0] != PT_CLIST)
2972        {
2973        list[2] = code[0];
2974        list[3] = code[1];
2975        return code + 2;
2976        }
2977    
2978      /* Convert only if we have enough space. */
2979    
2980      clist_src = PRIV(ucd_caseless_sets) + code[1];
2981      clist_dest = list + 2;
2982      code += 2;
2983    
2984      do {
2985         if (clist_dest >= list + 8)
2986           {
2987           /* Early return if there is not enough space. This should never
2988           happen, since all clists are shorter than 5 character now. */
2989           list[2] = code[0];
2990           list[3] = code[1];
2991           return code;
2992           }
2993         *clist_dest++ = *clist_src;
2994         }
2995      while(*clist_src++ != NOTACHAR);
2996    
2997      /* All characters are stored. The terminating NOTACHAR
2998      is copied form the clist itself. */
2999    
3000      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001      return code;
3002    #endif
3003    
3004      case OP_NCLASS:
3005      case OP_CLASS:
3006    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007      case OP_XCLASS:
3008      if (c == OP_XCLASS)
3009        end = code + GET(code, 0) - 1;
3010      else
3011    #endif
3012        end = code + 32 / sizeof(pcre_uchar);
3013    
3014      switch(*end)
3015        {
3016        case OP_CRSTAR:
3017        case OP_CRMINSTAR:
3018        case OP_CRQUERY:
3019        case OP_CRMINQUERY:
3020        case OP_CRPOSSTAR:
3021        case OP_CRPOSQUERY:
3022        list[1] = TRUE;
3023        end++;
3024        break;
3025    
3026        case OP_CRPLUS:
3027        case OP_CRMINPLUS:
3028        case OP_CRPOSPLUS:
3029        end++;
3030        break;
3031    
3032        case OP_CRRANGE:
3033        case OP_CRMINRANGE:
3034        case OP_CRPOSRANGE:
3035        list[1] = (GET2(end, 1) == 0);
3036        end += 1 + 2 * IMM2_SIZE;
3037        break;
3038        }
3039      list[2] = end - code;
3040      return end;
3041      }
3042    return NULL;    /* Opcode not accepted */
3043    }
3044    
3045    
3046    
3047    /*************************************************
3048    *    Scan further character sets for match       *
3049    *************************************************/
3050    
3051    /* Checks whether the base and the current opcode have a common character, in
3052    which case the base cannot be possessified.
3053    
3054    Arguments:
3055      code        points to the byte code
3056      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057      cd          static compile data
3058      base_list   the data list of the base opcode
3059    
3060    Returns:      TRUE if the auto-possessification is possible
3061    */
3062    
3063    static BOOL
3064    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066    {
3067    pcre_uchar c;
3068    pcre_uint32 list[8];
3069    const pcre_uint32 *chr_ptr;
3070    const pcre_uint32 *ochr_ptr;
3071    const pcre_uint32 *list_ptr;
3072    const pcre_uchar *next_code;
3073    const pcre_uint8 *class_bitset;
3074    const pcre_uint32 *set1, *set2, *set_end;
3075    pcre_uint32 chr;
3076    BOOL accepted, invert_bits;
3077    
3078    /* Note: the base_list[1] contains whether the current opcode has greedy
3079    (represented by a non-zero value) quantifier. This is a different from
3080    other character type lists, which stores here that the character iterator
3081    matches to an empty string (also represented by a non-zero value). */
3082    
3083    for(;;)
3084      {
3085      /* All operations move the code pointer forward.
3086      Therefore infinite recursions are not possible. */
3087    
3088      c = *code;
3089    
3090      /* Skip over callouts */
3091    
3092      if (c == OP_CALLOUT)
3093        {
3094        code += PRIV(OP_lengths)[c];
3095        continue;
3096        }
3097    
3098      if (c == OP_ALT)
3099        {
3100        do code += GET(code, 1); while (*code == OP_ALT);
3101        c = *code;
3102        }
3103    
3104      switch(c)
3105        {
3106        case OP_END:
3107        case OP_KETRPOS:
3108        /* TRUE only in greedy case. The non-greedy case could be replaced by
3109        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3110        uses more memory, which we cannot get at this stage.) */
3111    
3112        return base_list[1] != 0;
3113    
3114        case OP_KET:
3115        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3116        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3117        cannot be converted to a possessive form. */
3118    
3119        if (base_list[1] == 0) return FALSE;
3120    
3121        switch(*(code - GET(code, 1)))
3122          {
3123          case OP_ASSERT:
3124          case OP_ASSERT_NOT:
3125          case OP_ASSERTBACK:
3126          case OP_ASSERTBACK_NOT:
3127          case OP_ONCE:
3128          case OP_ONCE_NC:
3129          /* Atomic sub-patterns and assertions can always auto-possessify their
3130          last iterator. */
3131          return TRUE;
3132          }
3133    
3134        code += PRIV(OP_lengths)[c];
3135        continue;
3136    
3137        case OP_ONCE:
3138        case OP_ONCE_NC:
3139        case OP_BRA:
3140        case OP_CBRA:
3141        next_code = code + GET(code, 1);
3142        code += PRIV(OP_lengths)[c];
3143    
3144        while (*next_code == OP_ALT)
3145          {
3146          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3147          code = next_code + 1 + LINK_SIZE;
3148          next_code += GET(next_code, 1);
3149          }
3150        continue;
3151    
3152        case OP_BRAZERO:
3153        case OP_BRAMINZERO:
3154    
3155        next_code = code + 1;
3156        if (*next_code != OP_BRA && *next_code != OP_CBRA
3157            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3158    
3159        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3160    
3161        /* The bracket content will be checked by the
3162        OP_BRA/OP_CBRA case above. */
3163        next_code += 1 + LINK_SIZE;
3164        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3165          return FALSE;
3166    
3167        code += PRIV(OP_lengths)[c];
3168        continue;
3169        }
3170    
3171      /* Check for a supported opcode, and load its properties. */
3172    
3173      code = get_chr_property_list(code, utf, cd->fcc, list);
3174      if (code == NULL) return FALSE;    /* Unsupported */
3175    
3176      /* If either opcode is a small character list, set pointers for comparing
3177      characters from that list with another list, or with a property. */
3178    
3179      if (base_list[0] == OP_CHAR)
3180        {
3181        chr_ptr = base_list + 2;
3182        list_ptr = list;
3183        }
3184      else if (list[0] == OP_CHAR)
3185        {
3186        chr_ptr = list + 2;
3187        list_ptr = base_list;
3188        }
3189    
3190      /* Character bitsets can also be compared to certain opcodes. */
3191    
3192      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3193    #ifdef COMPILE_PCRE8
3194          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3195          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3196    #endif
3197          )
3198        {
3199    #ifdef COMPILE_PCRE8
3200        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3201    #else
3202        if (base_list[0] == OP_CLASS)
3203    #endif
3204          {
3205          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3206          list_ptr = list;
3207          }
3208        else
3209          {
3210          set1 = (pcre_uint32 *)(code - list[2]);
3211          list_ptr = base_list;
3212          }
3213    
3214        invert_bits = FALSE;
3215        switch(list_ptr[0])
3216          {
3217          case OP_CLASS:
3218          case OP_NCLASS:
3219          set2 = (pcre_uint32 *)
3220            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3221          break;
3222    
3223          /* OP_XCLASS cannot be supported here, because its bitset
3224          is not necessarily complete. E.g: [a-\0x{200}] is stored
3225          as a character range, and the appropriate bits are not set. */
3226    
3227          case OP_NOT_DIGIT:
3228            invert_bits = TRUE;
3229            /* Fall through */
3230          case OP_DIGIT:
3231            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3232            break;
3233    
3234          case OP_NOT_WHITESPACE:
3235            invert_bits = TRUE;
3236            /* Fall through */
3237          case OP_WHITESPACE:
3238            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3239            break;
3240    
3241          case OP_NOT_WORDCHAR:
3242            invert_bits = TRUE;
3243            /* Fall through */
3244          case OP_WORDCHAR:
3245            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3246            break;
3247    
3248          default:
3249          return FALSE;
3250          }
3251    
3252        /* Compare 4 bytes to improve speed. */
3253        set_end = set1 + (32 / 4);
3254        if (invert_bits)
3255          {
3256          do
3257            {
3258            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3259            }
3260          while (set1 < set_end);
3261          }
3262        else
3263          {
3264          do
3265            {
3266            if ((*set1++ & *set2++) != 0) return FALSE;
3267            }
3268          while (set1 < set_end);
3269          }
3270    
3271        if (list[1] == 0) return TRUE;
3272        /* Might be an empty repeat. */
3273        continue;
3274        }
3275    
3276      /* Some property combinations also acceptable. Unicode property opcodes are
3277      processed specially; the rest can be handled with a lookup table. */
3278    
3279      else
3280        {
3281        pcre_uint32 leftop, rightop;
3282    
3283        leftop = base_list[0];
3284        rightop = list[0];
3285    
3286    #ifdef SUPPORT_UCP
3287        accepted = FALSE; /* Always set in non-unicode case. */
3288        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3289          {
3290          if (rightop == OP_EOD)
3291            accepted = TRUE;
3292          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3293            {
3294            int n;
3295            const pcre_uint8 *p;
3296            BOOL same = leftop == rightop;
3297            BOOL lisprop = leftop == OP_PROP;
3298            BOOL risprop = rightop == OP_PROP;
3299            BOOL bothprop = lisprop && risprop;
3300    
3301            /* There's a table that specifies how each combination is to be
3302            processed:
3303              0   Always return FALSE (never auto-possessify)
3304              1   Character groups are distinct (possessify if both are OP_PROP)
3305              2   Check character categories in the same group (general or particular)
3306              3   Return TRUE if the two opcodes are not the same
3307              ... see comments below
3308            */
3309    
3310            n = propposstab[base_list[2]][list[2]];
3311            switch(n)
3312              {
3313              case 0: break;
3314              case 1: accepted = bothprop; break;
3315              case 2: accepted = (base_list[3] == list[3]) != same; break;
3316              case 3: accepted = !same; break;
3317    
3318              case 4:  /* Left general category, right particular category */
3319              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3320              break;
3321    
3322              case 5:  /* Right general category, left particular category */
3323              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3324              break;
3325    
3326              /* This code is logically tricky. Think hard before fiddling with it.
3327              The posspropstab table has four entries per row. Each row relates to
3328              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3329              Only WORD actually needs all four entries, but using repeats for the
3330              others means they can all use the same code below.
3331    
3332              The first two entries in each row are Unicode general categories, and
3333              apply always, because all the characters they include are part of the
3334              PCRE character set. The third and fourth entries are a general and a
3335              particular category, respectively, that include one or more relevant
3336              characters. One or the other is used, depending on whether the check
3337              is for a general or a particular category. However, in both cases the
3338              category contains more characters than the specials that are defined
3339              for the property being tested against. Therefore, it cannot be used
3340              in a NOTPROP case.
3341    
3342              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3343              Underscore is covered by ucp_P or ucp_Po. */
3344    
3345              case 6:  /* Left alphanum vs right general category */
3346              case 7:  /* Left space vs right general category */
3347              case 8:  /* Left word vs right general category */
3348              p = posspropstab[n-6];
3349              accepted = risprop && lisprop ==
3350                (list[3] != p[0] &&
3351                 list[3] != p[1] &&
3352                (list[3] != p[2] || !lisprop));
3353              break;
3354    
3355              case 9:   /* Right alphanum vs left general category */
3356              case 10:  /* Right space vs left general category */
3357              case 11:  /* Right word vs left general category */
3358              p = posspropstab[n-9];
3359              accepted = lisprop && risprop ==
3360                (base_list[3] != p[0] &&
3361                 base_list[3] != p[1] &&
3362                (base_list[3] != p[2] || !risprop));
3363              break;
3364    
3365              case 12:  /* Left alphanum vs right particular category */
3366              case 13:  /* Left space vs right particular category */
3367              case 14:  /* Left word vs right particular category */
3368              p = posspropstab[n-12];
3369              accepted = risprop && lisprop ==
3370                (catposstab[p[0]][list[3]] &&
3371                 catposstab[p[1]][list[3]] &&
3372                (list[3] != p[3] || !lisprop));
3373              break;
3374    
3375              case 15:  /* Right alphanum vs left particular category */
3376              case 16:  /* Right space vs left particular category */
3377              case 17:  /* Right word vs left particular category */
3378              p = posspropstab[n-15];
3379              accepted = lisprop && risprop ==
3380                (catposstab[p[0]][base_list[3]] &&
3381                 catposstab[p[1]][base_list[3]] &&
3382                (base_list[3] != p[3] || !risprop));
3383              break;
3384              }
3385            }
3386          }
3387    
3388        else
3389    #endif  /* SUPPORT_UCP */
3390    
3391        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3392               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3393               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3394    
3395        if (!accepted)
3396          return FALSE;
3397    
3398        if (list[1] == 0) return TRUE;
3399        /* Might be an empty repeat. */
3400        continue;
3401        }
3402    
3403      /* Control reaches here only if one of the items is a small character list.
3404      All characters are checked against the other side. */
3405    
3406      do
3407        {
3408        chr = *chr_ptr;
3409    
3410        switch(list_ptr[0])
3411          {
3412          case OP_CHAR:
3413          ochr_ptr = list_ptr + 2;
3414          do
3415            {
3416            if (chr == *ochr_ptr) return FALSE;
3417            ochr_ptr++;
3418            }
3419          while(*ochr_ptr != NOTACHAR);
3420          break;
3421    
3422          case OP_NOT:
3423          ochr_ptr = list_ptr + 2;
3424          do
3425            {
3426            if (chr == *ochr_ptr)
3427              break;
3428            ochr_ptr++;
3429            }
3430          while(*ochr_ptr != NOTACHAR);
3431          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3432          break;
3433    
3434          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3435          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3436    
3437          case OP_DIGIT:
3438          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3439          break;
3440    
3441          case OP_NOT_DIGIT:
3442          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3443          break;
3444    
3445          case OP_WHITESPACE:
3446          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3447          break;
3448    
3449          case OP_NOT_WHITESPACE:
3450          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3451          break;
3452    
3453          case OP_WORDCHAR:
3454          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3455          break;
3456    
3457          case OP_NOT_WORDCHAR:
3458          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3459          break;
3460    
3461          case OP_HSPACE:
3462          switch(chr)
3463            {
3464            HSPACE_CASES: return FALSE;
3465            default: break;
3466            }
3467          break;
3468    
3469          case OP_NOT_HSPACE:
3470          switch(chr)
3471            {
3472            HSPACE_CASES: break;
3473            default: return FALSE;
3474            }
3475          break;
3476    
3477          case OP_ANYNL:
3478          case OP_VSPACE:
3479          switch(chr)
3480            {
3481            VSPACE_CASES: return FALSE;
3482            default: break;
3483            }
3484          break;
3485    
3486          case OP_NOT_VSPACE:
3487          switch(chr)
3488            {
3489            VSPACE_CASES: break;
3490            default: return FALSE;
3491            }
3492          break;
3493    
3494          case OP_DOLL:
3495          case OP_EODN:
3496          switch (chr)
3497            {
3498            case CHAR_CR:
3499            case CHAR_LF:
3500            case CHAR_VT:
3501            case CHAR_FF:
3502            case CHAR_NEL:
3503    #ifndef EBCDIC
3504            case 0x2028:
3505            case 0x2029:
3506    #endif  /* Not EBCDIC */
3507            return FALSE;
3508            }
3509          break;
3510    
3511          case OP_EOD:    /* Can always possessify before \z */
3512          break;
3513    
3514    #ifdef SUPPORT_UCP
3515          case OP_PROP:
3516          case OP_NOTPROP:
3517          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3518                list_ptr[0] == OP_NOTPROP))
3519            return FALSE;
3520          break;
3521    #endif
3522    
3523          case OP_NCLASS:
3524          if (chr > 255) return FALSE;
3525          /* Fall through */
3526    
3527          case OP_CLASS:
3528          if (chr > 255) break;
3529          class_bitset = (pcre_uint8 *)
3530            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3531          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3532          break;
3533    
3534    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3535          case OP_XCLASS:
3536          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3537              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3538          break;
3539    #endif
3540    
3541          default:
3542          return FALSE;
3543          }
3544    
3545        chr_ptr++;
3546        }
3547      while(*chr_ptr != NOTACHAR);
3548    
3549      /* At least one character must be matched from this opcode. */
3550    
3551      if (list[1] == 0) return TRUE;
3552      }
3553    
3554    return FALSE;
3555    }
3556    
3557    
3558    
3559    /*************************************************
3560    *    Scan compiled regex for auto-possession     *
3561    *************************************************/
3562    
3563    /* Replaces single character iterations with their possessive alternatives
3564    if appropriate. This function modifies the compiled opcode!
3565    
3566    Arguments:
3567      code        points to start of the byte code
3568      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3569      cd          static compile data
3570    
3571    Returns:      nothing
3572    */
3573    
3574    static void
3575    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3576    {
3577    register pcre_uchar c;
3578    const pcre_uchar *end;
3579    pcre_uchar *repeat_opcode;
3580    pcre_uint32 list[8];
3581    
3582    for (;;)
3583      {
3584      c = *code;
3585    
3586      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3587        {
3588        c -= get_repeat_base(c) - OP_STAR;
3589        end = (c <= OP_MINUPTO) ?
3590          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3591        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3592    
3593        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3594          {
3595          switch(c)
3596            {
3597            case OP_STAR:
3598            *code += OP_POSSTAR - OP_STAR;
3599            break;
3600    
3601            case OP_MINSTAR:
3602            *code += OP_POSSTAR - OP_MINSTAR;
3603            break;
3604    
3605            case OP_PLUS:
3606            *code += OP_POSPLUS - OP_PLUS;
3607            break;
3608    
3609            case OP_MINPLUS:
3610            *code += OP_POSPLUS - OP_MINPLUS;
3611            break;
3612    
3613            case OP_QUERY:
3614            *code += OP_POSQUERY - OP_QUERY;
3615            break;
3616    
3617            case OP_MINQUERY:
3618            *code += OP_POSQUERY - OP_MINQUERY;
3619            break;
3620    
3621            case OP_UPTO:
3622            *code += OP_POSUPTO - OP_UPTO;
3623            break;
3624    
3625            case OP_MINUPTO:
3626            *code += OP_MINUPTO - OP_UPTO;
3627            break;
3628            }
3629          }
3630        c = *code;
3631        }
3632      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3633        {
3634    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3635        if (c == OP_XCLASS)
3636          repeat_opcode = code + GET(code, 1);
3637        else
3638    #endif
3639          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3640    
3641        c = *repeat_opcode;
3642        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3643          {
3644          /* end must not be NULL. */
3645          end = get_chr_property_list(code, utf, cd->fcc, list);
3646    
3647          list[1] = (c & 1) == 0;
3648    
3649      /* Opcodes that must match a character */        if (compare_opcodes(end, utf, cd, list, end))
3650            {
3651            switch (c)
3652              {
3653              case OP_CRSTAR:
3654              case OP_CRMINSTAR:
3655              *repeat_opcode = OP_CRPOSSTAR;
3656              break;
3657    
3658      case OP_PROP:            case OP_CRPLUS:
3659      case OP_NOTPROP:            case OP_CRMINPLUS:
3660      case OP_EXTUNI:            *repeat_opcode = OP_CRPOSPLUS;
3661      case OP_NOT_DIGIT:            break;
     case OP_DIGIT:  
     case OP_NOT_WHITESPACE:  
     case OP_WHITESPACE:  
     case OP_NOT_WORDCHAR:  
     case OP_WORDCHAR:  
     case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
     case OP_CHAR:  
     case OP_CHARI:  
     case OP_NOT:  
     case OP_NOTI:  
     case OP_PLUS:  
     case OP_MINPLUS:  
     case OP_POSPLUS:  
     case OP_EXACT:  
     case OP_NOTPLUS:  
     case OP_NOTMINPLUS:  
     case OP_NOTPOSPLUS:  
     case OP_NOTEXACT:  
     case OP_TYPEPLUS:  
     case OP_TYPEMINPLUS:  
     case OP_TYPEPOSPLUS:  
     case OP_TYPEEXACT:  
     return FALSE;  
3662    
3663      /* These are going to continue, as they may be empty, but we have to            case OP_CRQUERY:
3664      fudge the length for the \p and \P cases. */            case OP_CRMINQUERY:
3665              *repeat_opcode = OP_CRPOSQUERY;
3666              break;
3667    
3668              case OP_CRRANGE:
3669              case OP_CRMINRANGE:
3670              *repeat_opcode = OP_CRPOSRANGE;
3671              break;
3672              }
3673            }