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

Diff of /code/trunk/pcre_compile.c

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

revision 637 by ph10, Sun Jul 24 17:44:12 2011 UTC revision 1408 by ph10, Tue Dec 3 16:27:00 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 *, 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            *
933    *************************************************/
934    
935    /* This function is called when a '{' is encountered in a place where it might
936    start a quantifier. It looks ahead to see if it really is a quantifier or not.
937    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
938    where the ddds are digits.
939    
940    Arguments:
941      p         pointer to the first char after '{'
942    
943    Returns:    TRUE or FALSE
944    */
945    
946    static BOOL
947    is_counted_repeat(const pcre_uchar *p)
948    {
949    if (!IS_DIGIT(*p)) return FALSE;
950    p++;
951    while (IS_DIGIT(*p)) p++;
952    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953    
954    if (*p++ != CHAR_COMMA) return FALSE;
955    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956    
957    if (!IS_DIGIT(*p)) return FALSE;
958    p++;
959    while (IS_DIGIT(*p)) p++;
960    
961    return (*p == CHAR_RIGHT_CURLY_BRACKET);
962    }
963    
964    
965    
966  /*************************************************  /*************************************************
967  *            Handle escapes                      *  *            Handle escapes                      *
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 643  else Line 1037  else
1037    
1038      case CHAR_l:      case CHAR_l:
1039      case CHAR_L:      case CHAR_L:
1040        *errorcodeptr = ERR37;
1041        break;
1042    
1043      case CHAR_u:      case CHAR_u:
1044        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1045          {
1046          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047          Otherwise it is a lowercase u letter. */
1048          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049            && 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;
1054            for (i = 0; i < 4; ++i)
1055              {
1056              register pcre_uint32 cc = *(++ptr);
1057    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1058              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1059              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1060    #else           /* EBCDIC coding */
1061              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1062              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063    #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
1080          *errorcodeptr = ERR37;
1081        break;
1082    
1083      case CHAR_U:      case CHAR_U:
1084      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
1085        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1086      break;      break;
1087    
1088      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
1089        class, \g must be followed by one of a number of specific things:
1090    
1091      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
1092      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 662  else Line 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;
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 675  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 695  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 711  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 719  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 748  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;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1257    
1258      case CHAR_x:      case CHAR_o:
1259      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260        {        {
1261        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1262        c = 0;        c = 0;
1263        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265          {          {
1266          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1267          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268          count++;  #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;
1292    
1293        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296        case CHAR_x:
1297        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298          {
1299          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301            {
1302            c = 0;
1303            for (i = 0; i < 2; ++i)
1304              {
1305              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));
1309  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1310          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 */
1311          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1312  #endif  #endif
1313              }
1314          }          }
1315          }    /* End JavaScript handling */
1316    
1317        if (*pt == CHAR_RIGHT_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        digits. If not, { used to be treated as a data character. However, Perl
1320        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        else
1325          {
1326          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          ptr += 2;
1329          ptr = pt;          c = 0;
1330          break;          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        /* If the sequence of hex digits does not end with '}', then we don't  #ifdef COMPILE_PCRE32
1337        recognize this construct; fall through to the normal \x handling. */            if (c >= 0x10000000l) { overflow = TRUE; break; }
1338        }  #endif
1339    
1340    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343    #else           /* EBCDIC coding */
1344              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));
1346    #endif
1347    
1348      /* Read just a single-byte hex-defined char */  #if defined COMPILE_PCRE8
1349              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350    #elif defined COMPILE_PCRE16
1351              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352    #elif defined COMPILE_PCRE32
1353              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354    #endif
1355              }
1356    
1357      c = 0;          if (overflow)
1358      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)            {
1359        {            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360        int cc;                                  /* Some compilers don't like */            *errorcodeptr = ERR34;
1361        cc = *(++ptr);                           /* ++ in initializers */            }
1362    
1363            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364              {
1365              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366              }
1367    
1368            /* 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            else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377    
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 849  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 886  else Line 1438  else
1438    }    }
1439    
1440  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
1441  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1442    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))
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 917  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 945  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 969  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    
   
 /*************************************************  
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1049  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 1058  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 1073  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 1098  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)
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
1649      {      {
1650      *count += 1;      case OP_ASSERT_NOT:
1651      if (name == NULL && *count == lorn) return *count;      case OP_ASSERTBACK:
1652      ptr++;      case OP_ASSERTBACK_NOT:
1653      }      if (!skipassert) return code;
1654        do code += GET(code, 1); while (*code == OP_ALT);
1655    /* All cases now have (? at the start. Remember when we are in a group      code += PRIV(OP_lengths)[*code];
1656    where the parenthesis numbers are duplicated. */      break;
1657    
1658    else if (ptr[2] == CHAR_VERTICAL_LINE)      case OP_WORD_BOUNDARY:
1659      {      case OP_NOT_WORD_BOUNDARY:
1660      ptr += 3;      if (!skipassert) return code;
1661      dup_parens = TRUE;      /* Fall through */
     }  
1662    
1663    /* Handle comments; all characters are allowed until a ket is reached. */      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[2] == CHAR_NUMBER_SIGN)      default:
1673      {      return code;
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
1674      }      }
1675      }
1676    /* Control never reaches here */
1677    }
1678    
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
1679    
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
1680    
1681    /* Start with (? but not a condition. */  /*************************************************
1682    *        Find the fixed length of a branch       *
1683    *************************************************/
1684    
1685    else  /* Scan a branch and compute the fixed length of subject that will match it,
1686      {  if the length is fixed. This is needed for dealing with backward assertions.
1687      ptr += 2;  In UTF8 mode, the result is in characters rather than bytes. The branch is
1688      if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  temporarily terminated with OP_END when this function is called.
1689    
1690      /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  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      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  However, we cannot do this when the assertion contains subroutine calls,
1693          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  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.
       int term;  
       const uschar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == ptr - thisname &&  
           strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (strncmp((const char *)ptr+2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF8  
       if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf8         TRUE if we are in UTF-8 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,  
   BOOL utf8)  
 {  
 uschar *ptr = (uschar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, it makes sense to skip negative  
 forward and all backward assertions, and also the \b assertion; for others it  
 does not.  
   
 Arguments:  
   code         pointer to the start of the group  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const uschar*  
 first_significant_code(const uschar *code, BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
     {  
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     case OP_WORD_BOUNDARY:  
     case OP_NOT_WORD_BOUNDARY:  
     if (!skipassert) return code;  
     /* Fall through */  
   
     case OP_CALLOUT:  
     case OP_CREF:  
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
     case OP_DEF:  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     default:  
     return code;  
     }  
   }  
 /* Control never reaches here */  
 }  
   
   
   
   
 /*************************************************  
 *        Find the fixed length of a branch       *  
 *************************************************/  
   
 /* Scan a branch and compute the fixed length of subject that will match it,  
 if the length is fixed. This is needed for dealing with backward assertions.  
 In UTF8 mode, the result is in characters rather than bytes. The branch is  
 temporarily terminated with OP_END when this function is called.  
   
 This function is called when a backward assertion is encountered, so that if it  
 fails, the error message can point to the correct place in the pattern.  
 However, we cannot do this when the assertion contains subroutine calls,  
 because they can be forward references. We solve this by remembering this case  
 and doing the check at the end; a flag specifies which mode we are running in.  
1695    
1696  Arguments:  Arguments:
1697    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1698    utf8     TRUE in UTF-8 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1700    cd       the "compile data" structure    cd       the "compile data" structure
1701    
1702  Returns:   the fixed length,  Returns:   the fixed length,
1703               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1704               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1705               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706                 or -4 if an unknown opcode was encountered (internal error)
1707  */  */
1708    
1709  static int  static int
1710  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711  {  {
1712  int length = -1;  int length = -1;
1713    
1714  register int branchlength = 0;  register int branchlength = 0;
1715  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1716    
1717  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1718  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1489  branch, check the length against that of Line 1720  branch, check the length against that of
1720  for (;;)  for (;;)
1721    {    {
1722    int d;    int d;
1723    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1724    register int op = *cc;    register pcre_uchar op = *cc;
1725    
1726    switch (op)    switch (op)
1727      {      {
1728      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729      OP_BRA (normal non-capturing bracket) because the other variants of these      OP_BRA (normal non-capturing bracket) because the other variants of these
1730      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1731      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1732    
1733      case OP_CBRA:      case OP_CBRA:
1734      case OP_BRA:      case OP_BRA:
1735      case OP_ONCE:      case OP_ONCE:
1736        case OP_ONCE_NC:
1737      case OP_COND:      case OP_COND:
1738      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739      if (d < 0) return d;      if (d < 0) return d;
1740      branchlength += d;      branchlength += d;
1741      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1742      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1743      break;      break;
1744    
1745      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1746      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1747      END it's the end of the outer call. All can be handled by the same code.      an ALT. If it is END it's the end of the outer call. All can be handled by
1748      Note that we must not include the OP_KETRxxx opcodes here, because they      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1750    
1751      case OP_ALT:      case OP_ALT:
1752      case OP_KET:      case OP_KET:
1753      case OP_END:      case OP_END:
1754        case OP_ACCEPT:
1755        case OP_ASSERT_ACCEPT:
1756      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1757        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1758      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1532  for (;;) Line 1766  for (;;)
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 1548  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 1581  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 1590  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 1610  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 1622  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 1661  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 1670  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 1683  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 1708  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 1717  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 1747  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:
       code += code[1];  
       break;  
   
2108        case OP_THEN_ARG:        case OP_THEN_ARG:
2109        code += code[1+LINK_SIZE];        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 1800  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 1821  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 1866  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+LINK_SIZE];        code += code[1];
2227        break;        break;
2228        }        }
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 1946  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 1978  for (code = first_significant_code(code Line 2361  for (code = first_significant_code(code
2361    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2362    implies a backward reference subroutine call, we can scan it. If it's a    implies a backward reference subroutine call, we can scan it. If it's a
2363    forward reference subroutine call, we can't. To detect forward reference    forward reference subroutine call, we can't. To detect forward reference
2364    we have to scan up the list that is kept in the workspace. This function is    we have to scan up the list that is kept in the workspace. This function is
2365    called only when doing the real compile, not during the pre-compile that    called only when doing the real compile, not during the pre-compile that
2366    measures the size of the compiled pattern. */    measures the size of the compiled pattern. */
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;
   
     /* Test for forward reference */  
   
     for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)  
       if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;  
2372    
2373      /* Not a forward reference, test for completed backward 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      empty_branch = FALSE;      NULL. */
2376      scode = cd->start_code + GET(code, 1);  
2377      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (cd->start_workspace != NULL)
2378          {
2379      /* Completed backwards reference */        const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381      do          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        else
2392        {        {
2393        if (could_be_empty_branch(scode, endcode, utf8, cd))        recurse_check *r = recurses;
2394          {        const pcre_uchar *endgroup = scode;
2395          empty_branch = TRUE;  
2396          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399          for (r = recurses; r != NULL; r = r->prev)
2400            if (r->group == scode) break;
2401          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        empty_branch = FALSE;
2408        this_recurse.prev = recurses;
2409        this_recurse.group = scode;
2410    
2411        do
2412          {
2413          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414            {
2415            empty_branch = TRUE;
2416          break;          break;
2417          }          }
2418        scode += GET(scode, 1);        scode += GET(scode, 1);
2419        }        }
2420      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2421    
2422      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2423      continue;      continue;
2424      }      }
# Line 2020  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 2041  for (code = first_significant_code(code Line 2449  for (code = first_significant_code(code
2449    
2450    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2451        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2452        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2453          c == OP_COND)
2454      {      {
2455      BOOL empty_branch;      BOOL empty_branch;
2456      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2057  for (code = first_significant_code(code Line 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 2075  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 2087  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 2099  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          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 2115  for (code = first_significant_code(code Line 2528  for (code = first_significant_code(code
2528    
2529      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      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:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARI:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555      case OP_NOTI:      case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558        case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
2560      case OP_POSPLUS:      case OP_MINPLUSI:
2561      case OP_EXACT:  
2562      case OP_NOTPLUS:      case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2162  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2599      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2600      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2601      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602          code += 2;
2603      break;      break;
2604    
2605      /* End of branch */      /* End of branch */
# Line 2175  for (code = first_significant_code(code Line 2612  for (code = first_significant_code(code
2612      return TRUE;      return TRUE;
2613    
2614      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619      case OP_STAR:      case OP_STAR:
2620      case OP_STARI:      case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624      case OP_MINSTAR:      case OP_MINSTAR:
2625      case OP_MINSTARI:      case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629      case OP_POSSTAR:      case OP_POSSTAR:
2630      case OP_POSSTARI:      case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634      case OP_QUERY:      case OP_QUERY:
2635      case OP_QUERYI:      case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639      case OP_MINQUERY:      case OP_MINQUERY:
2640      case OP_MINQUERYI:      case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644      case OP_POSQUERY:      case OP_POSQUERY:
2645      case OP_POSQUERYI:      case OP_POSQUERYI:
2646      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650      break;      break;
2651    
2652      case OP_UPTO:      case OP_UPTO:
2653      case OP_UPTOI:      case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657      case OP_MINUPTO:      case OP_MINUPTO:
2658      case OP_MINUPTOI:      case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662      case OP_POSUPTO:      case OP_POSUPTO:
2663      case OP_POSUPTOI:      case OP_POSUPTOI:
2664      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      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;      break;
2669  #endif  #endif
2670    
# Line 2209  for (code = first_significant_code(code Line 2674  for (code = first_significant_code(code
2674      case OP_MARK:      case OP_MARK:
2675      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2676      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2677      case OP_THEN_ARG:      case OP_THEN_ARG:
2678      code += code[1+LINK_SIZE];      code += code[1];
2679      break;      break;
2680    
2681      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2236  return TRUE; Line 2698  return TRUE;
2698  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2699  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2700  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2701  This function is called only during the real compile, not during the  This function is called only during the real compile, not during the
2702  pre-compile.  pre-compile.
2703    
2704  Arguments:  Arguments:
2705    code        points to start of the recursion    code        points to start of the recursion
2706    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2707    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2708    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709    cd          pointers to tables etc    cd          pointers to tables etc
2710    
2711  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2712  */  */
2713    
2714  static BOOL  static BOOL
2715  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2717  {  {
2718  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2719    {    {
2720    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721      return FALSE;      return FALSE;
2722    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2723    }    }
# Line 2265  return TRUE; Line 2727  return TRUE;
2727    
2728    
2729  /*************************************************  /*************************************************
2730  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2731  *************************************************/  *************************************************/
2732    
2733  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2734  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
2735    
2736  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2737  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2738  though of course unknown POSIX names are subsequently rejected. Perl gives an  */
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
2739    
2740  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2741  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2742  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2743  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2744  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747  I think.         (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:  Arguments:
2762    ptr      pointer to the initial [    c            the character
2763    endptr   where to return the end pointer    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 or FALSE  Returns:       TRUE if auto-possessifying is OK
2768  */  */
2769    
2770  static BOOL  static BOOL
2771  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773  {  {
2774  int terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2775  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2776  for (++ptr; *ptr != 0; ptr++)  
2777    switch(ptype)
2778    {    {
2779    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    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      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;      HSPACE_CASES:
2808      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      VSPACE_CASES:
2809        {      return negated;
2810        *endptr = ptr;  
2811        return TRUE;      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;  return FALSE;
2832  }  }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837  /*************************************************  /*************************************************
2838  *          Check POSIX class name                *  *        Fill the character property list        *
2839  *************************************************/  *************************************************/
2840    
2841  /* This function is called to check the name given in a POSIX-style class entry  /* Checks whether the code points to an opcode that can take part in auto-
2842  such as [:alnum:].  possessification, and if so, fills a list with its properties.
2843    
2844  Arguments:  Arguments:
2845    ptr        points to the first letter    code        points to start of expression
2846    len        the length of the name    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:     a value representing the name, or -1 if unknown  Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856  */  */
2857    
2858  static int  static const pcre_uchar *
2859  check_posix_name(const uschar *ptr, int len)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861  {  {
2862  const char *pn = posix_names;  pcre_uchar c = *code;
2863  register int yield = 0;  const pcre_uchar *end;
2864  while (posix_name_lengths[yield] != 0)  const pcre_uint32 *clist_src;
2865    pcre_uint32 *clist_dest;
2866    pcre_uint32 chr;
2867    pcre_uchar base;
2868    
2869    list[0] = c;
2870    list[1] = FALSE;
2871    code++;
2872    
2873    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2874    {    {
2875    if (len == posix_name_lengths[yield] &&    base = get_repeat_base(c);
2876      strncmp((const char *)ptr, pn, len) == 0) return yield;    c -= (base - OP_STAR);
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
2877    
2878      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2879        code += IMM2_SIZE;
2880    
2881  /*************************************************    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2882    
2883  /* OP_RECURSE items contain an offset from the start of the regex to the group    switch(base)
2884  that is referenced. This means that groups can be replicated for fixed      {
2885  repetition simply by copying (because the recursion is allowed to refer to      case OP_STAR:
2886  earlier groups that are outside the current group). However, when a group is      list[0] = OP_CHAR;
2887  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is      break;
 inserted before it, after it has been compiled. This means that any OP_RECURSE  
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2888    
2889  This function has been extended with the possibility of forward references for      case OP_STARI:
2890  recursions and subroutine calls. It must also check the list of such references      list[0] = OP_CHARI;
2891  for the group we are dealing with. If it finds that one of the recursions in      break;
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2892    
2893  Arguments:      case OP_NOTSTAR:
2894    group      points to the start of the group      list[0] = OP_NOT;
2895    adjust     the amount by which the group is to be moved      break;
   utf8       TRUE in UTF-8 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2896    
2897  Returns:     nothing      case OP_NOTSTARI:
2898  */      list[0] = OP_NOTI;
2899        break;
2900    
2901  static void      case OP_TYPESTAR:
2902  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,      list[0] = *code;
2903    uschar *save_hwm)      code++;
2904  {      break;
2905  uschar *ptr = group;      }
2906      c = list[0];
2907      }
2908    
2909  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  switch(c)
2910    {    {
2911    int offset;    case OP_NOT_DIGIT:
2912    uschar *hc;    case OP_DIGIT:
2913      case OP_NOT_WHITESPACE:
2914      case OP_WHITESPACE:
2915      case OP_NOT_WORDCHAR:
2916      case OP_WORDCHAR:
2917      case OP_ANY:
2918      case OP_ALLANY:
2919      case OP_ANYNL:
2920      case OP_NOT_HSPACE:
2921      case OP_HSPACE:
2922      case OP_NOT_VSPACE:
2923      case OP_VSPACE:
2924      case OP_EXTUNI:
2925      case OP_EODN:
2926      case OP_EOD:
2927      case OP_DOLL:
2928      case OP_DOLLM:
2929      return code;
2930    
2931    /* See if this recursion is on the forward reference list. If so, adjust the    case OP_CHAR:
2932    reference. */    case OP_NOT:
2933      GETCHARINCTEST(chr, code);
2934      list[2] = chr;
2935      list[3] = NOTACHAR;
2936      return code;
2937    
2938    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case OP_CHARI:
2939      case OP_NOTI:
2940      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2941      GETCHARINCTEST(chr, code);
2942      list[2] = chr;
2943    
2944    #ifdef SUPPORT_UCP
2945      if (chr < 128 || (chr < 256 && !utf))
2946        list[3] = fcc[chr];
2947      else
2948        list[3] = UCD_OTHERCASE(chr);
2949    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2950      list[3] = (chr < 256) ? fcc[chr] : chr;
2951    #else
2952      list[3] = fcc[chr];
2953    #endif
2954    
2955      /* The othercase might be the same value. */
2956    
2957      if (chr == list[3])
2958        list[3] = NOTACHAR;
2959      else
2960        list[4] = NOTACHAR;
2961      return code;
2962    
2963    #ifdef SUPPORT_UCP
2964      case OP_PROP:
2965      case OP_NOTPROP:
2966      if (code[0] != PT_CLIST)
2967      {      {
2968      offset = GET(hc, 0);      list[2] = code[0];
2969      if (cd->start_code + offset == ptr + 1)      list[3] = code[1];
2970        {      return code + 2;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
2971      }      }
2972    
2973    /* Otherwise, adjust the recursion offset if it's after the start of this    /* Convert only if we have enough space. */
   group. */  
2974    
2975    if (hc >= cd->hwm)    clist_src = PRIV(ucd_caseless_sets) + code[1];
2976      clist_dest = list + 2;
2977      code += 2;
2978    
2979      do {
2980         if (clist_dest >= list + 8)
2981           {
2982           /* Early return if there is not enough space. This should never
2983           happen, since all clists are shorter than 5 character now. */
2984           list[2] = code[0];
2985           list[3] = code[1];
2986           return code;
2987           }
2988         *clist_dest++ = *clist_src;
2989         }
2990      while(*clist_src++ != NOTACHAR);
2991    
2992      /* All characters are stored. The terminating NOTACHAR
2993      is copied form the clist itself. */
2994    
2995      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2996      return code;
2997    #endif
2998    
2999      case OP_NCLASS:
3000      case OP_CLASS:
3001    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3002      case OP_XCLASS:
3003      if (c == OP_XCLASS)
3004        end = code + GET(code, 0) - 1;
3005      else
3006    #endif
3007        end = code + 32 / sizeof(pcre_uchar);
3008    
3009      switch(*end)
3010      {      {
3011      offset = GET(ptr, 1);      case OP_CRSTAR:
3012      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      case OP_CRMINSTAR:
3013      }      case OP_CRQUERY:
3014        case OP_CRMINQUERY:
3015        case OP_CRPOSSTAR:
3016        case OP_CRPOSQUERY:
3017        list[1] = TRUE;
3018        end++;
3019        break;
3020    
3021    ptr += 1 + LINK_SIZE;      case OP_CRPLUS:
3022        case OP_CRMINPLUS:
3023        case OP_CRPOSPLUS:
3024        end++;
3025        break;
3026    
3027        case OP_CRRANGE:
3028        case OP_CRMINRANGE:
3029        case OP_CRPOSRANGE:
3030        list[1] = (GET2(end, 1) == 0);
3031        end += 1 + 2 * IMM2_SIZE;
3032        break;
3033        }
3034      list[2] = end - code;
3035      return end;
3036    }    }
3037    return NULL;    /* Opcode not accepted */
3038  }  }
3039    
3040    
3041    
3042  /*************************************************  /*************************************************
3043  *        Insert an automatic callout point       *  *    Scan further character sets for match       *
3044  *************************************************/  *************************************************/
3045    
3046  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the base and the current opcode have a common character, in
3047  callout points before each pattern item.  which case the base cannot be possessified.
3048    
3049  Arguments:  Arguments:
3050    code           current code pointer    code        points to the byte code
3051    ptr            current pattern pointer    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3052    cd             pointers to tables etc    cd          static compile data
3053      base_list   the data list of the base opcode
3054    
3055  Returns:         new code pointer  Returns:      TRUE if the auto-possessification is possible
3056  */  */
3057    
3058  static uschar *  static BOOL
3059  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3060      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3061  {  {
3062  *code++ = OP_CALLOUT;  pcre_uchar c;
3063  *code++ = 255;  pcre_uint32 list[8];
3064  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *chr_ptr;
3065  PUT(code, LINK_SIZE, 0);                       /* Default length */  const pcre_uint32 *ochr_ptr;
3066  return code + 2*LINK_SIZE;  const pcre_uint32 *list_ptr;
3067  }  const pcre_uchar *next_code;
3068    const pcre_uint8 *class_bitset;
3069    const pcre_uint32 *set1, *set2, *set_end;
3070    pcre_uint32 chr;
3071    BOOL accepted, invert_bits;
3072    
3073    /* Note: the base_list[1] contains whether the current opcode has greedy
3074    (represented by a non-zero value) quantifier. This is a different from
3075    other character type lists, which stores here that the character iterator
3076    matches to an empty string (also represented by a non-zero value). */
3077    
3078    for(;;)
3079      {
3080      /* All operations move the code pointer forward.
3081      Therefore infinite recursions are not possible. */
3082    
3083      c = *code;
3084    
3085  /*************************************************    /* Skip over callouts */
3086  *         Complete a callout item                *  
3087      if (c == OP_CALLOUT)
3088        {
3089        code += PRIV(OP_lengths)[c];
3090        continue;
3091        }
3092    
3093      if (c == OP_ALT)
3094        {
3095        do code += GET(code, 1); while (*code == OP_ALT);
3096        c = *code;
3097        }
3098    
3099      switch(c)
3100        {
3101        case OP_END:
3102        case OP_KETRPOS:
3103        /* TRUE only in greedy case. The non-greedy case could be replaced by
3104        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3105        uses more memory, which we cannot get at this stage.) */
3106    
3107        return base_list[1] != 0;
3108    
3109        case OP_KET:
3110        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3111        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3112        cannot be converted to a possessive form. */
3113    
3114        if (base_list[1] == 0) return FALSE;
3115    
3116        switch(*(code - GET(code, 1)))
3117          {
3118          case OP_ASSERT:
3119          case OP_ASSERT_NOT:
3120          case OP_ASSERTBACK:
3121          case OP_ASSERTBACK_NOT:
3122          case OP_ONCE:
3123          case OP_ONCE_NC:
3124          /* Atomic sub-patterns and assertions can always auto-possessify their
3125          last iterator. */
3126          return TRUE;
3127          }
3128    
3129        code += PRIV(OP_lengths)[c];
3130        continue;
3131    
3132        case OP_ONCE:
3133        case OP_ONCE_NC:
3134        case OP_BRA:
3135        case OP_CBRA:
3136        next_code = code + GET(code, 1);
3137        code += PRIV(OP_lengths)[c];
3138    
3139        while (*next_code == OP_ALT)
3140          {
3141          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3142          code = next_code + 1 + LINK_SIZE;
3143          next_code += GET(next_code, 1);
3144          }
3145        continue;
3146    
3147        case OP_BRAZERO:
3148        case OP_BRAMINZERO:
3149    
3150        next_code = code + 1;
3151        if (*next_code != OP_BRA && *next_code != OP_CBRA
3152            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3153    
3154        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3155    
3156        /* The bracket content will be checked by the
3157        OP_BRA/OP_CBRA case above. */
3158        next_code += 1 + LINK_SIZE;
3159        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3160          return FALSE;
3161    
3162        code += PRIV(OP_lengths)[c];
3163        continue;
3164        }
3165    
3166      /* Check for a supported opcode, and load its properties. */
3167    
3168      code = get_chr_property_list(code, utf, cd->fcc, list);
3169      if (code == NULL) return FALSE;    /* Unsupported */
3170    
3171      /* If either opcode is a small character list, set pointers for comparing
3172      characters from that list with another list, or with a property. */
3173    
3174      if (base_list[0] == OP_CHAR)
3175        {
3176        chr_ptr = base_list + 2;
3177        list_ptr = list;
3178        }
3179      else if (list[0] == OP_CHAR)
3180        {
3181        chr_ptr = list + 2;
3182        list_ptr = base_list;
3183        }
3184    
3185      /* Character bitsets can also be compared to certain opcodes. */
3186    
3187      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3188    #ifdef COMPILE_PCRE8
3189          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3190          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3191    #endif
3192          )
3193        {
3194    #ifdef COMPILE_PCRE8
3195        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3196    #else
3197        if (base_list[0] == OP_CLASS)
3198    #endif
3199          {
3200          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3201          list_ptr = list;
3202          }
3203        else
3204          {
3205          set1 = (pcre_uint32 *)(code - list[2]);
3206          list_ptr = base_list;
3207          }
3208    
3209        invert_bits = FALSE;
3210        switch(list_ptr[0])
3211          {
3212          case OP_CLASS:
3213          case OP_NCLASS:
3214          set2 = (pcre_uint32 *)
3215            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3216          break;
3217    
3218          /* OP_XCLASS cannot be supported here, because its bitset
3219          is not necessarily complete. E.g: [a-\0x{200}] is stored
3220          as a character range, and the appropriate bits are not set. */
3221    
3222          case OP_NOT_DIGIT:
3223            invert_bits = TRUE;
3224            /* Fall through */
3225          case OP_DIGIT:
3226            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3227            break;
3228    
3229          case OP_NOT_WHITESPACE:
3230            invert_bits = TRUE;
3231            /* Fall through */
3232          case OP_WHITESPACE:
3233            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3234            break;
3235    
3236          case OP_NOT_WORDCHAR:
3237            invert_bits = TRUE;
3238            /* Fall through */
3239          case OP_WORDCHAR:
3240            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3241            break;
3242    
3243          default:
3244          return FALSE;
3245          }
3246    
3247        /* Compare 4 bytes to improve speed. */
3248        set_end = set1 + (32 / 4);
3249        if (invert_bits)
3250          {
3251          do
3252            {
3253            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3254            }
3255          while (set1 < set_end);
3256          }
3257        else
3258          {
3259          do
3260            {
3261            if ((*set1++ & *set2++) != 0) return FALSE;
3262            }
3263          while (set1 < set_end);
3264          }
3265    
3266        if (list[1] == 0) return TRUE;
3267        /* Might be an empty repeat. */
3268        continue;
3269        }
3270    
3271      /* Some property combinations also acceptable. Unicode property opcodes are
3272      processed specially; the rest can be handled with a lookup table. */
3273    
3274      else
3275        {
3276        pcre_uint32 leftop, rightop;
3277    
3278        leftop = base_list[0];
3279        rightop = list[0];
3280    
3281    #ifdef SUPPORT_UCP
3282        accepted = FALSE; /* Always set in non-unicode case. */
3283        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3284          {
3285          if (rightop == OP_EOD)
3286            accepted = TRUE;
3287          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3288            {
3289            int n;
3290            const pcre_uint8 *p;
3291            BOOL same = leftop == rightop;
3292            BOOL lisprop = leftop == OP_PROP;
3293            BOOL risprop = rightop == OP_PROP;
3294            BOOL bothprop = lisprop && risprop;
3295    
3296            /* There's a table that specifies how each combination is to be
3297            processed:
3298              0   Always return FALSE (never auto-possessify)
3299              1   Character groups are distinct (possessify if both are OP_PROP)
3300              2   Check character categories in the same group (general or particular)
3301              3   Return TRUE if the two opcodes are not the same
3302              ... see comments below
3303            */
3304    
3305            n = propposstab[base_list[2]][list[2]];
3306            switch(n)
3307              {
3308              case 0: break;
3309              case 1: accepted = bothprop; break;
3310              case 2: accepted = (base_list[3] == list[3]) != same; break;
3311              case 3: accepted = !same; break;
3312    
3313              case 4:  /* Left general category, right particular category */
3314              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3315              break;
3316    
3317              case 5:  /* Right general category, left particular category */
3318              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3319              break;
3320    
3321              /* This code is logically tricky. Think hard before fiddling with it.
3322              The posspropstab table has four entries per row. Each row relates to
3323              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3324              Only WORD actually needs all four entries, but using repeats for the
3325              others means they can all use the same code below.
3326    
3327              The first two entries in each row are Unicode general categories, and
3328              apply always, because all the characters they include are part of the
3329              PCRE character set. The third and fourth entries are a general and a
3330              particular category, respectively, that include one or more relevant
3331              characters. One or the other is used, depending on whether the check
3332              is for a general or a particular category. However, in both cases the
3333              category contains more characters than the specials that are defined
3334              for the property being tested against. Therefore, it cannot be used
3335              in a NOTPROP case.
3336    
3337              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3338              Underscore is covered by ucp_P or ucp_Po. */
3339    
3340              case 6:  /* Left alphanum vs right general category */
3341              case 7:  /* Left space vs right general category */
3342              case 8:  /* Left word vs right general category */
3343              p = posspropstab[n-6];
3344              accepted = risprop && lisprop ==
3345                (list[3] != p[0] &&
3346                 list[3] != p[1] &&
3347                (list[3] != p[2] || !lisprop));
3348              break;
3349    
3350              case 9:   /* Right alphanum vs left general category */
3351              case 10:  /* Right space vs left general category */
3352              case 11:  /* Right word vs left general category */
3353              p = posspropstab[n-9];
3354              accepted = lisprop && risprop ==
3355                (base_list[3] != p[0] &&
3356                 base_list[3] != p[1] &&
3357                (base_list[3] != p[2] || !risprop));
3358              break;
3359    
3360              case 12:  /* Left alphanum vs right particular category */
3361              case 13:  /* Left space vs right particular category */
3362              case 14:  /* Left word vs right particular category */
3363              p = posspropstab[n-12];
3364              accepted = risprop && lisprop ==
3365                (catposstab[p[0]][list[3]] &&
3366                 catposstab[p[1]][list[3]] &&
3367                (list[3] != p[3] || !lisprop));
3368              break;
3369    
3370              case 15:  /* Right alphanum vs left particular category */
3371              case 16:  /* Right space vs left particular category */
3372              case 17:  /* Right word vs left particular category */
3373              p = posspropstab[n-15];
3374              accepted = lisprop && risprop ==
3375                (catposstab[p[0]][base_list[3]] &&
3376                 catposstab[p[1]][base_list[3]] &&
3377                (base_list[3] != p[3] || !risprop));
3378              break;
3379              }
3380            }
3381          }
3382    
3383        else
3384    #endif  /* SUPPORT_UCP */
3385    
3386        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3387               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3388               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3389    
3390        if (!accepted)
3391          return FALSE;
3392    
3393        if (list[1] == 0) return TRUE;
3394        /* Might be an empty repeat. */
3395        continue;
3396        }
3397    
3398      /* Control reaches here only if one of the items is a small character list.
3399      All characters are checked against the other side. */
3400    
3401      do
3402        {
3403        chr = *chr_ptr;
3404    
3405        switch(list_ptr[0])
3406          {
3407          case OP_CHAR:
3408          ochr_ptr = list_ptr + 2;
3409          do
3410            {
3411            if (chr == *ochr_ptr) return FALSE;
3412            ochr_ptr++;
3413            }
3414          while(*ochr_ptr != NOTACHAR);
3415          break;
3416    
3417          case OP_NOT:
3418          ochr_ptr = list_ptr + 2;
3419          do
3420            {
3421            if (chr == *ochr_ptr)
3422              break;
3423            ochr_ptr++;
3424            }
3425          while(*ochr_ptr != NOTACHAR);
3426          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3427          break;
3428    
3429          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3430          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3431    
3432          case OP_DIGIT:
3433          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3434          break;
3435    
3436          case OP_NOT_DIGIT:
3437          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3438          break;
3439    
3440          case OP_WHITESPACE:
3441          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3442          break;
3443    
3444          case OP_NOT_WHITESPACE:
3445          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3446          break;
3447    
3448          case OP_WORDCHAR:
3449          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3450          break;
3451    
3452          case OP_NOT_WORDCHAR:
3453          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3454          break;
3455    
3456          case OP_HSPACE:
3457          switch(chr)
3458            {
3459            HSPACE_CASES: return FALSE;
3460            default: break;
3461            }
3462          break;
3463    
3464          case OP_NOT_HSPACE:
3465          switch(chr)
3466            {
3467            HSPACE_CASES: break;
3468            default: return FALSE;
3469            }
3470          break;
3471    
3472          case OP_ANYNL:
3473          case OP_VSPACE:
3474          switch(chr)
3475            {
3476            VSPACE_CASES: return FALSE;
3477            default: break;
3478            }
3479          break;
3480    
3481          case OP_NOT_VSPACE:
3482          switch(chr)
3483            {
3484            VSPACE_CASES: break;
3485            default: return FALSE;
3486            }
3487          break;
3488    
3489          case OP_DOLL:
3490          case OP_EODN:
3491          switch (chr)
3492            {
3493            case CHAR_CR:
3494            case CHAR_LF:
3495            case CHAR_VT:
3496            case CHAR_FF:
3497            case CHAR_NEL:
3498    #ifndef EBCDIC
3499            case 0x2028:
3500            case 0x2029:
3501    #endif  /* Not EBCDIC */
3502            return FALSE;
3503            }
3504          break;
3505    
3506          case OP_EOD:    /* Can always possessify before \z */
3507          break;
3508    
3509    #ifdef SUPPORT_UCP
3510          case OP_PROP:
3511          case OP_NOTPROP:
3512          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3513                list_ptr[0] == OP_NOTPROP))
3514            return FALSE;
3515          break;
3516    #endif
3517    
3518          case OP_NCLASS:
3519          if (chr > 255) return FALSE;
3520          /* Fall through */
3521    
3522          case OP_CLASS:
3523          if (chr > 255) break;
3524          class_bitset = (pcre_uint8 *)
3525            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3526          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3527          break;
3528    
3529    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3530          case OP_XCLASS:
3531          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3532              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3533          break;
3534    #endif
3535    
3536          default:
3537          return FALSE;
3538          }
3539    
3540        chr_ptr++;
3541        }
3542      while(*chr_ptr != NOTACHAR);
3543    
3544      /* At least one character must be matched from this opcode. */
3545    
3546      if (list[1] == 0) return TRUE;
3547      }
3548    
3549    return FALSE;
3550    }
3551    
3552    
3553    
3554    /*************************************************
3555    *    Scan compiled regex for auto-possession     *
3556    *************************************************/
3557    
3558    /* Replaces single character iterations with their possessive alternatives
3559    if appropriate. This function modifies the compiled opcode!
3560    
3561    Arguments:
3562      code        points to start of the byte code
3563      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3564      cd          static compile data
3565    
3566    Returns:      nothing
3567    */
3568    
3569    static void
3570    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3571    {
3572    register pcre_uchar c;
3573    const pcre_uchar *end;
3574    pcre_uchar *repeat_opcode;
3575    pcre_uint32 list[8];
3576    
3577    for (;;)
3578      {
3579      c = *code;
3580    
3581      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3582        {
3583        c -= get_repeat_base(c) - OP_STAR;
3584        end = (c <= OP_MINUPTO) ?
3585          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3586        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3587    
3588        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3589          {
3590          switch(c)
3591            {
3592            case OP_STAR:
3593            *code += OP_POSSTAR - OP_STAR;
3594            break;
3595    
3596            case OP_MINSTAR:
3597            *code += OP_POSSTAR - OP_MINSTAR;
3598            break;
3599    
3600            case OP_PLUS:
3601            *code += OP_POSPLUS - OP_PLUS;
3602            break;
3603    
3604            case OP_MINPLUS:
3605            *code += OP_POSPLUS - OP_MINPLUS;
3606            break;
3607    
3608            case OP_QUERY:
3609            *code += OP_POSQUERY - OP_QUERY;
3610            break;
3611    
3612            case OP_MINQUERY:
3613            *code += OP_POSQUERY - OP_MINQUERY;
3614            break;
3615    
3616            case OP_UPTO:
3617            *code += OP_POSUPTO - OP_UPTO;
3618            break;
3619    
3620            case OP_MINUPTO:
3621            *code += OP_MINUPTO - OP_UPTO;
3622            break;
3623            }
3624          }
3625        c = *code;
3626        }
3627      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3628        {
3629    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3630        if (c == OP_XCLASS)
3631          repeat_opcode = code + GET(code, 1);
3632        else
3633    #endif
3634          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3635    
3636        c = *repeat_opcode;
3637        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3638          {
3639          /* end must not be NULL. */
3640          end = get_chr_property_list(code, utf, cd->fcc, list);
3641    
3642          list[1] = (c & 1) == 0;
3643    
3644          if (compare_opcodes(end, utf, cd, list, end))
3645            {
3646            switch (c)
3647              {
3648              case OP_CRSTAR:
3649              case OP_CRMINSTAR:
3650              *repeat_opcode = OP_CRPOSSTAR;
3651              break;
3652    
3653              case OP_CRPLUS:
3654              case OP_CRMINPLUS:
3655              *repeat_opcode = OP_CRPOSPLUS;
3656              break;
3657    
3658              case OP_CRQUERY:
3659              case OP_CRMINQUERY:
3660              *repeat_opcode = OP_CRPOSQUERY;
3661              break;
3662    
3663              case OP_CRRANGE:
3664              case OP_CRMINRANGE:
3665              *repeat_opcode = OP_CRPOSRANGE;
3666              break;
3667              }
3668            }
3669          }
3670        c = *code;
3671        }
3672    
3673      switch(c)
3674        {
3675        case OP_END:
3676        return;
3677    
3678        case OP_TYPESTAR:
3679        case OP_TYPEMINSTAR:
3680        case OP_TYPEPLUS:
3681        case OP_TYPEMINPLUS:
3682        case OP_TYPEQUERY:
3683        case OP_TYPEMINQUERY:
3684        case OP_TYPEPOSSTAR:
3685        case OP_TYPEPOSPLUS:
3686        case OP_TYPEPOSQUERY:
3687        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3688        break;
3689    
3690        case OP_TYPEUPTO:
3691        case OP_TYPEMINUPTO:
3692        case OP_TYPEEXACT:
3693        case OP_TYPEPOSUPTO:
3694        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3695          code += 2;
3696        break;
3697    
3698    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3699        case OP_XCLASS:
3700        code += GET(code, 1);
3701        break;
3702    #endif
3703    
3704        case OP_MARK:
3705        case OP_PRUNE_ARG:
3706        case OP_SKIP_ARG:
3707        case OP_THEN_ARG:
3708        code += code[1];
3709        break;
3710        }
3711    
3712      /* Add in the fixed length from the table */
3713    
3714      code += PRIV(OP_lengths)[c];
3715    
3716      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3717      a multi-byte character. The length in the table is a minimum, so we have to
3718      arrange to skip the extra bytes. */
3719    
3720    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3721      if (utf) switch(c)
3722        {
3723        case OP_CHAR:
3724        case OP_CHARI:
3725        case OP_NOT:
3726        case OP_NOTI:
3727        case OP_STAR:
3728        case OP_MINSTAR:
3729        case OP_PLUS:
3730        case OP_MINPLUS:
3731        case OP_QUERY:
3732        case OP_MINQUERY:
3733        case OP_UPTO:
3734        case OP_MINUPTO:
3735        case OP_EXACT:
3736        case OP_POSSTAR:
3737        case OP_POSPLUS:
3738        case OP_POSQUERY:
3739        case OP_POSUPTO:
3740        case OP_STARI:
3741        case OP_MINSTARI:
3742        case OP_PLUSI:
3743        case OP_MINPLUSI:
3744        case OP_QUERYI:
3745        case OP_MINQUERYI:
3746        case OP_UPTOI:
3747        case OP_MINUPTOI:
3748        case OP_EXACTI:
3749        case OP_POSSTARI:
3750        case OP_POSPLUSI:
3751        case OP_POSQUERYI:
3752        case OP_POSUPTOI:
3753        case OP_NOTSTAR:
3754        case OP_NOTMINSTAR:
3755        case OP_NOTPLUS:
3756        case OP_NOTMINPLUS:
3757        case OP_NOTQUERY:
3758        case OP_NOTMINQUERY:
3759        case OP_NOTUPTO:
3760        case OP_NOTMINUPTO:
3761        case OP_NOTEXACT:
3762        case OP_NOTPOSSTAR:
3763        case OP_NOTPOSPLUS:
3764        case OP_NOTPOSQUERY:
3765        case OP_NOTPOSUPTO:
3766        case OP_NOTSTARI:
3767        case OP_NOTMINSTARI:
3768        case OP_NOTPLUSI:
3769        case OP_NOTMINPLUSI:
3770        case OP_NOTQUERYI:
3771        case OP_NOTMINQUERYI:
3772        case OP_NOTUPTOI:
3773        case OP_NOTMINUPTOI:
3774        case OP_NOTEXACTI:
3775        case OP_NOTPOSSTARI:
3776        case OP_NOTPOSPLUSI:
3777        case OP_NOTPOSQUERYI:
3778        case OP_NOTPOSUPTOI:
3779        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3780        break;
3781        }
3782    #else
3783      (void)(utf);  /* Keep compiler happy by referencing function argument */
3784    #endif
3785      }
3786    }
3787    
3788    
3789    
3790    /*************************************************
3791    *           Check for POSIX class syntax         *
3792  *************************************************/  *************************************************/
3793    
3794  /* A callout item contains the length of the next item in the pattern, which  /* This function is called when the sequence "[:" or "[." or "[=" is
3795  we can't fill in till after we have reached the relevant point. This is used  encountered in a character class. It checks whether this is followed by a
3796  for both automatic and manual callouts.  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3797    reach an unescaped ']' without the special preceding character, return FALSE.
3798    
3799    Originally, this function only recognized a sequence of letters between the
3800    terminators, but it seems that Perl recognizes any sequence of characters,
3801    though of course unknown POSIX names are subsequently rejected. Perl gives an
3802    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3803    didn't consider this to be a POSIX class. Likewise for [:1234:].
3804    
3805    The problem in trying to be exactly like Perl is in the handling of escapes. We
3806    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3807    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3808    below handles the special case of \], but does not try to do any other escape
3809    processing. This makes it different from Perl for cases such as [:l\ower:]
3810    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3811    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3812    I think.
3813