/[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 964 by ph10, Fri May 4 13:03:39 2012 UTC revision 1481 by ph10, Tue May 27 18:24:42 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is 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. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 68  COMPILE_PCREx macro will already be appr Line 68  COMPILE_PCREx macro will already be appr
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 77  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 103  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #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 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    
# Line 110  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #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. */  /* Repeated character flags. */
139    
# Line 238  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 253  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 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 280  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 pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 307  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 360  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 375  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^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(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 440  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF 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 472  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 490  static const char error_texts[] = Line 536  static const char error_texts[] =
536    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
538    /* 75 */    /* 75 */
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553    ;    ;
554    
555  /* 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 630  static const pcre_uint8 ebcdic_chartab[] Line 689  static const pcre_uint8 ebcdic_chartab[]
689  #endif  #endif
690    
691    
692  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730  static BOOL  /* This table is used to check whether auto-possessification is possible
731    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732      int *, int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* 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 */
784      { 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 */
785      { 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 */
786      { 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 */
787      { 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 */
788      { 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 */
789      { 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 */
790      { 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 */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
# Line 657  find_error_text(int n) Line 886  find_error_text(int n)
886  const char *s = error_texts;  const char *s = error_texts;
887  for (; n > 0; n--)  for (; n > 0; n--)
888    {    {
889    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
890    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
891    }    }
892  return s;  return s;
893  }  }
894    
895    
896    
897  /*************************************************  /*************************************************
898  *           Expand the workspace                 *  *           Expand the workspace                 *
899  *************************************************/  *************************************************/
# Line 741  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 971  return (*p == CHAR_RIGHT_CURLY_BRACKET);
971  *************************************************/  *************************************************/
972    
973  /* 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
974  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
975  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.
976  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
977  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
978  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982      chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
986    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
987    
988  Returns:         zero or positive => a data character  Returns:         zero => a data character
989                   negative => a special escape sequence                   positive => a special escape sequence
990                     negative => a back reference
991                   on error, errorcodeptr is set                   on error, errorcodeptr is set
992  */  */
993    
994  static int  static int
995  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
997  {  {
998  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
1000  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
1001  pcre_int32 c;  pcre_uint32 c;
1002    int escape = 0;
1003  int i;  int i;
1004    
1005  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 775  ptr--;                            /* Set Line 1007  ptr--;                            /* Set
1007    
1008  /* 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. */
1009    
1010  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011    
1012  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013  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.
# Line 784  Otherwise further processing may be requ Line 1016  Otherwise further processing may be requ
1016  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017  /* Not alphanumeric */  /* Not alphanumeric */
1018  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1019  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1020      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021    
1022  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1023  /* Not alphanumeric */  /* Not alphanumeric */
1024  else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1025  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1026  #endif  #endif
1027    
1028  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 797  else if ((i = escapes[c - 0x48]) != 0) Line 1030  else if ((i = escapes[c - 0x48]) != 0)
1030  else  else
1031    {    {
1032    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1033    BOOL braced, negated;    BOOL braced, negated, overflow;
1034      int s;
1035    
1036    switch (c)    switch (c)
1037      {      {
# Line 822  else Line 1056  else
1056          c = 0;          c = 0;
1057          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1058            {            {
1059            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1060  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1061            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1062            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 831  else Line 1065  else
1065            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066  #endif  #endif
1067            }            }
1068    
1069    #if defined COMPILE_PCRE8
1070            if (c > (utf ? 0x10ffffU : 0xffU))
1071    #elif defined COMPILE_PCRE16
1072            if (c > (utf ? 0x10ffffU : 0xffffU))
1073    #elif defined COMPILE_PCRE32
1074            if (utf && c > 0x10ffffU)
1075    #endif
1076              {
1077              *errorcodeptr = ERR76;
1078              }
1079            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080          }          }
1081        }        }
1082      else      else
# Line 857  else Line 1103  else
1103      (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
1104      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1105      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1107    
1108      case CHAR_g:      case CHAR_g:
1109      if (isclass) break;      if (isclass) break;
1110      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111        {        {
1112        c = -ESC_g;        escape = ESC_g;
1113        break;        break;
1114        }        }
1115    
# Line 872  else Line 1118  else
1118      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119        {        {
1120        const pcre_uchar *p;        const pcre_uchar *p;
1121        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++)
1122          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124          {          {
1125          c = -ESC_k;          escape = ESC_k;
1126          break;          break;
1127          }          }
1128        braced = TRUE;        braced = TRUE;
# Line 892  else Line 1138  else
1138      else negated = FALSE;      else negated = FALSE;
1139    
1140      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1141      c = 0;      s = 0;
1142        overflow = FALSE;
1143      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1144        {        {
1145        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146          {          {
1147          c = -1;          overflow = TRUE;
1148          break;          break;
1149          }          }
1150        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151        }        }
1152      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1153        {        {
1154        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1155          ptr++;          ptr++;
# Line 916  else Line 1163  else
1163        break;        break;
1164        }        }
1165    
1166      if (c == 0)      if (s == 0)
1167        {        {
1168        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1169        break;        break;
# Line 924  else Line 1171  else
1171    
1172      if (negated)      if (negated)
1173        {        {
1174        if (c > bracount)        if (s > bracount)
1175          {          {
1176          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1177          break;          break;
1178          }          }
1179        c = bracount - (c - 1);        s = bracount - (s - 1);
1180        }        }
1181    
1182      c = -(ESC_REF + c);      escape = -s;
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      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
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      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
1191      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
1192      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
1193      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
1194      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
1195      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
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      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:
1202      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 954  else Line 1205  else
1205        {        {
1206        oldptr = ptr;        oldptr = ptr;
1207        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1208        c -= CHAR_0;        s = (int)(c -CHAR_0);
1209          overflow = FALSE;
1210        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1211          {          {
1212          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213            {            {
1214            c = -1;            overflow = TRUE;
1215            break;            break;
1216            }            }
1217          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218          }          }
1219        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1220          {          {
1221          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1222            ptr++;            ptr++;
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          c = -(ESC_REF + c);          escape = -s;
1229          break;          break;
1230          }          }
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* 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
1235      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
1236      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
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \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
1244      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
# Line 1005  else Line 1255  else
1255  #endif  #endif
1256      break;      break;
1257    
1258      /* \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
1259      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1260      If not, { is treated as a data character. */  
1261        case CHAR_o:
1262        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264          {
1265          ptr += 2;
1266          c = 0;
1267          overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269            {
1270            register pcre_uint32 cc = *ptr++;
1271            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272    #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274    #endif
1275            c = (c << 3) + cc - CHAR_0 ;
1276    #if defined COMPILE_PCRE8
1277            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE16
1279            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282    #endif
1283            }
1284          if (overflow)
1285            {
1286            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287            *errorcodeptr = ERR34;
1288            }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290            {
1291            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294          }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      case CHAR_x:      case CHAR_x:
1301      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1303        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305          {          {
1306          c = 0;          c = 0;
1307          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1308            {            {
1309            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1310  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1311            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1312            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 1030  else Line 1316  else
1316  #endif  #endif
1317            }            }
1318          }          }
1319        break;        }    /* End JavaScript handling */
       }  
1320    
1321      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        c = 0;      else
1329        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1330          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331          {          {
1332          register int cc = *pt++;          ptr += 2;
1333          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345    #ifdef COMPILE_PCRE32
1346              if (c >= 0x10000000l) { overflow = TRUE; break; }
1347    #endif
1348    
1349  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1353          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 */
1354          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355  #endif  #endif
1356    
1357  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1358          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359  #else  #elif defined COMPILE_PCRE16
1360  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1362              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363  #endif  #endif
1364  #endif            }
         }  
1365    
1366        if (c < 0)          if (overflow)
1367          {            {
1368          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1370          }            }
1371    
1372        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373          {            {
1374          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375          ptr = pt;            }
         break;  
         }  
1376    
1377        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1378        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1379        }          \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1387        {        else
1388        int cc;                                  /* Some compilers don't like */          {
1389        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1398        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400  #endif  #endif
1401        }            }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* 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 1101  else Line 1410  else
1410    
1411      case CHAR_c:      case CHAR_c:
1412      c = *(++ptr);      c = *(++ptr);
1413      if (c == 0)      if (c == CHAR_NULL)
1414        {        {
1415        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1416        break;        break;
# Line 1141  else Line 1450  else
1450  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1451  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1452    
1453  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1455    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1456    
1457  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1458    
1459  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1461    
1462  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1463    
1464  *ptrptr = ptr;  *ptrptr = ptr;
1465  return c;  *chptr = c;
1466    return escape;
1467  }  }
1468    
1469    
# Line 1171  escape sequence. Line 1481  escape sequence.
1481  Argument:  Argument:
1482    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1483    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
1484    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
1485      pdataptr       points to an unsigned int that is set to the detailed property value
1486    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1487    
1488  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
1489  */  */
1490    
1491  static int  static BOOL
1492  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493      unsigned int *pdataptr, int *errorcodeptr)
1494  {  {
1495  int c, i, bot, top;  pcre_uchar c;
1496    int i, bot, top;
1497  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1498  pcre_uchar name[32];  pcre_uchar name[32];
1499    
1500  c = *(++ptr);  c = *(++ptr);
1501  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1502    
1503  *negptr = FALSE;  *negptr = FALSE;
1504    
# Line 1202  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1515  if (c == CHAR_LEFT_CURLY_BRACKET)
1515    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516      {      {
1517      c = *(++ptr);      c = *(++ptr);
1518      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1519      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520      name[i] = c;      name[i] = c;
1521      }      }
# Line 1227  top = PRIV(utt_size); Line 1540  top = PRIV(utt_size);
1540    
1541  while (bot < top)  while (bot < top)
1542    {    {
1543      int r;
1544    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1545    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1546    if (c == 0)    if (r == 0)
1547      {      {
1548      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1549      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1550        return TRUE;
1551      }      }
1552    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1553    }    }
1554    
1555  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1556  *ptrptr = ptr;  *ptrptr = ptr;
1557  return -1;  return FALSE;
1558    
1559  ERROR_RETURN:  ERROR_RETURN:
1560  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1561  *ptrptr = ptr;  *ptrptr = ptr;
1562  return -1;  return FALSE;
1563  }  }
1564  #endif  #endif
1565    
1566    
1567    
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 1276  read_repeat_counts(const pcre_uchar *p, Line 1590  read_repeat_counts(const pcre_uchar *p,
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597    }      {
1598        *errorcodeptr = ERR5;
1599  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1600  Also, max must not be less than min. */      }
1601      }
1602    
1603  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604    {    {
1605    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612        }          {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616          }
1617      if (max < min)      if (max < min)
1618        {        {
1619        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1308  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 1319  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It 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. 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.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   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  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 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)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* 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). */  
   
   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++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       int term;  
       const pcre_uchar *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_UC_UC(name, 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_UC_C8(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_UTF  
       if (utf) FORWARDCHAR(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, utf, 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  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)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, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1633  *      Find first significant op code            *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
# Line 1652  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671      case OP_NCREF:      case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673      case OP_NRREF:      case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
# Line 1668  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
# Line 1686  and doing the check at the end; a flag s Line 1701  and doing the check at the end; a flag s
1701    
1702  Arguments:  Arguments:
1703    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1704    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1706    cd       the "compile data" structure    cd       the "compile data" structure
1707    
# Line 1712  for (;;) Line 1727  for (;;)
1727    {    {
1728    int d;    int d;
1729    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1730    register int op = *cc;    register pcre_uchar op = *cc;
1731    
1732    switch (op)    switch (op)
1733      {      {
# Line 1792  for (;;) Line 1807  for (;;)
1807      case OP_COMMIT:      case OP_COMMIT:
1808      case OP_CREF:      case OP_CREF:
1809      case OP_DEF:      case OP_DEF:
1810        case OP_DNCREF:
1811        case OP_DNRREF:
1812      case OP_DOLL:      case OP_DOLL:
1813      case OP_DOLLM:      case OP_DOLLM:
1814      case OP_EOD:      case OP_EOD:
1815      case OP_EODN:      case OP_EODN:
1816      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1817      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1818      case OP_PRUNE:      case OP_PRUNE:
1819      case OP_REVERSE:      case OP_REVERSE:
# Line 1832  for (;;) Line 1847  for (;;)
1847      case OP_EXACTI:      case OP_EXACTI:
1848      case OP_NOTEXACT:      case OP_NOTEXACT:
1849      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1850      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1851      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1852  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1853      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1841  for (;;) Line 1856  for (;;)
1856    
1857      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1858      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1859      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1860          cc += 2;
1861      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1862      break;      break;
1863    
# Line 1876  for (;;) Line 1892  for (;;)
1892    
1893      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1894    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1895      case OP_CLASS:      case OP_CLASS:
1896      case OP_NCLASS:      case OP_NCLASS:
1897    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1898        case OP_XCLASS:
1899        /* The original code caused an unsigned overflow in 64 bit systems,
1900        so now we use a conditional statement. */
1901        if (op == OP_XCLASS)
1902          cc += GET(cc, 1);
1903        else
1904          cc += PRIV(OP_lengths)[OP_CLASS];
1905    #else
1906      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1907    #endif
1908    
1909      switch (*cc)      switch (*cc)
1910        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1911        case OP_CRSTAR:        case OP_CRSTAR:
1912        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRQUERY:        case OP_CRQUERY:
1916        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1917          case OP_CRPOSSTAR:
1918          case OP_CRPOSPLUS:
1919          case OP_CRPOSQUERY:
1920        return -1;        return -1;
1921    
1922        case OP_CRRANGE:        case OP_CRRANGE:
1923        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1924          case OP_CRPOSRANGE:
1925        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1927        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1928        break;        break;
1929    
# Line 1966  for (;;) Line 1990  for (;;)
1990      case OP_QUERYI:      case OP_QUERYI:
1991      case OP_REF:      case OP_REF:
1992      case OP_REFI:      case OP_REFI:
1993        case OP_DNREF:
1994        case OP_DNREFI:
1995      case OP_SBRA:      case OP_SBRA:
1996      case OP_SBRAPOS:      case OP_SBRAPOS:
1997      case OP_SCBRA:      case OP_SCBRA:
# Line 2002  for (;;) Line 2028  for (;;)
2028    
2029    
2030    
   
2031  /*************************************************  /*************************************************
2032  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2033  *************************************************/  *************************************************/
# Line 2015  length. Line 2040  length.
2040    
2041  Arguments:  Arguments:
2042    code        points to start of expression    code        points to start of expression
2043    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2044    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2045    
2046  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
# Line 2026  PRIV(find_bracket)(const pcre_uchar *cod Line 2051  PRIV(find_bracket)(const pcre_uchar *cod
2051  {  {
2052  for (;;)  for (;;)
2053    {    {
2054    register int c = *code;    register pcre_uchar c = *code;
2055    
2056    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2057    
# Line 2049  for (;;) Line 2074  for (;;)
2074    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2075             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2076      {      {
2077      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2078      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2079      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2080      }      }
# Line 2079  for (;;) Line 2104  for (;;)
2104        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2105        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2106        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2107        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2108          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2109        break;        break;
2110    
2111        case OP_MARK:        case OP_MARK:
2112        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2113        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2114        case OP_THEN_ARG:        case OP_THEN_ARG:
2115        code += code[1];        code += code[1];
2116        break;        break;
# Line 2102  for (;;) Line 2124  for (;;)
2124    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
2125    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2126    
2127  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2128      if (utf) switch(c)      if (utf) switch(c)
2129        {        {
2130        case OP_CHAR:        case OP_CHAR:
# Line 2154  instance of OP_RECURSE. Line 2176  instance of OP_RECURSE.
2176    
2177  Arguments:  Arguments:
2178    code        points to start of expression    code        points to start of expression
2179    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2180    
2181  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
2182  */  */
# Line 2164  find_recurse(const pcre_uchar *code, BOO Line 2186  find_recurse(const pcre_uchar *code, BOO
2186  {  {
2187  for (;;)  for (;;)
2188    {    {
2189    register int c = *code;    register pcre_uchar c = *code;
2190    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2191    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2192    
# Line 2199  for (;;) Line 2221  for (;;)
2221        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2222        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2223        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2224        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2225          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2226        break;        break;
2227    
2228        case OP_MARK:        case OP_MARK:
2229        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2230        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2231        case OP_THEN_ARG:        case OP_THEN_ARG:
2232        code += code[1];        code += code[1];
2233        break;        break;
# Line 2222  for (;;) Line 2241  for (;;)
2241      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
2242      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2243    
2244  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2245      if (utf) switch(c)      if (utf) switch(c)
2246        {        {
2247        case OP_CHAR:        case OP_CHAR:
# Line 2308  bracket whose current branch will alread Line 2327  bracket whose current branch will alread
2327  Arguments:  Arguments:
2328    code        points to start of search    code        points to start of search
2329    endcode     points to where to stop    endcode     points to where to stop
2330    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2331    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2332      recurses    chain of recurse_check to catch mutual recursion
2333    
2334  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2335  */  */
2336    
2337    typedef struct recurse_check {
2338      struct recurse_check *prev;
2339      const pcre_uchar *group;
2340    } recurse_check;
2341    
2342  static BOOL  static BOOL
2343  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2344    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2345  {  {
2346  register int c;  register pcre_uchar c;
2347    recurse_check this_recurse;
2348    
2349  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2350       code < endcode;       code < endcode;
2351       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2346  for (code = first_significant_code(code Line 2373  for (code = first_significant_code(code
2373    
2374    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2375      {      {
2376      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2377      BOOL empty_branch;      BOOL empty_branch;
2378    
2379      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2380        when called to scan a completed pattern by setting cd->start_workspace to
2381        NULL. */
2382    
2383        if (cd->start_workspace != NULL)
2384          {
2385          const pcre_uchar *tcode;
2386          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2387            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2388          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2389          }
2390    
2391        /* If we are scanning a completed pattern, there are no forward references
2392        and all groups are complete. We need to detect whether this is a recursive
2393        call, as otherwise there will be an infinite loop. If it is a recursion,
2394        just skip over it. Simple recursions are easily detected. For mutual
2395        recursions we keep a chain on the stack. */
2396    
2397        else
2398          {
2399          recurse_check *r = recurses;
2400          const pcre_uchar *endgroup = scode;
2401    
2402      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2403        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2404    
2405      /* Not a forward reference, test for completed backward reference */        for (r = recurses; r != NULL; r = r->prev)
2406            if (r->group == scode) break;
2407          if (r != NULL) continue;   /* Mutual recursion */
2408          }
2409    
2410      empty_branch = FALSE;      /* Completed reference; scan the referenced group, remembering it on the
2411      scode = cd->start_code + GET(code, 1);      stack chain to detect mutual recursions. */
     if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */  
2412    
2413      /* Completed backwards reference */      empty_branch = FALSE;
2414        this_recurse.prev = recurses;
2415        this_recurse.group = scode;
2416    
2417      do      do
2418        {        {
2419        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2420          {          {
2421          empty_branch = TRUE;          empty_branch = TRUE;
2422          break;          break;
# Line 2420  for (code = first_significant_code(code Line 2472  for (code = first_significant_code(code
2472        empty_branch = FALSE;        empty_branch = FALSE;
2473        do        do
2474          {          {
2475          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2476            empty_branch = TRUE;            empty_branch = TRUE;
2477          code += GET(code, 1);          code += GET(code, 1);
2478          }          }
# Line 2462  for (code = first_significant_code(code Line 2514  for (code = first_significant_code(code
2514        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2515        case OP_CRQUERY:        case OP_CRQUERY:
2516        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2517          case OP_CRPOSSTAR:
2518          case OP_CRPOSQUERY:
2519        break;        break;
2520    
2521        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2522        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2523        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2524          case OP_CRPOSPLUS:
2525        return FALSE;        return FALSE;
2526    
2527        case OP_CRRANGE:        case OP_CRRANGE:
2528        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2529          case OP_CRPOSRANGE:
2530        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2531        break;        break;
2532        }        }
# Line 2478  for (code = first_significant_code(code Line 2534  for (code = first_significant_code(code
2534    
2535      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2536    
2537        case OP_ANY:
2538        case OP_ALLANY:
2539        case OP_ANYBYTE:
2540    
2541      case OP_PROP:      case OP_PROP:
2542      case OP_NOTPROP:      case OP_NOTPROP:
2543        case OP_ANYNL:
2544    
2545        case OP_NOT_HSPACE:
2546        case OP_HSPACE:
2547        case OP_NOT_VSPACE:
2548        case OP_VSPACE:
2549      case OP_EXTUNI:      case OP_EXTUNI:
2550    
2551      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2552      case OP_DIGIT:      case OP_DIGIT:
2553      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2554      case OP_WHITESPACE:      case OP_WHITESPACE:
2555      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2556      case OP_WORDCHAR:      case OP_WORDCHAR:
2557      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2558      case OP_CHAR:      case OP_CHAR:
2559      case OP_CHARI:      case OP_CHARI:
2560      case OP_NOT:      case OP_NOT:
2561      case OP_NOTI:      case OP_NOTI:
2562    
2563      case OP_PLUS:      case OP_PLUS:
2564        case OP_PLUSI:
2565      case OP_MINPLUS:      case OP_MINPLUS:
2566      case OP_POSPLUS:      case OP_MINPLUSI:
2567      case OP_EXACT:  
2568      case OP_NOTPLUS:      case OP_NOTPLUS:
2569        case OP_NOTPLUSI:
2570      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2571        case OP_NOTMINPLUSI:
2572    
2573        case OP_POSPLUS:
2574        case OP_POSPLUSI:
2575      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2576        case OP_NOTPOSPLUSI:
2577    
2578        case OP_EXACT:
2579        case OP_EXACTI:
2580      case OP_NOTEXACT:      case OP_NOTEXACT:
2581        case OP_NOTEXACTI:
2582    
2583      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2584      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2585      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2586      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2587    
2588      return FALSE;      return FALSE;
2589    
2590      /* 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 2525  for (code = first_significant_code(code Line 2604  for (code = first_significant_code(code
2604      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2605      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2606      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2607      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2608        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2609      break;      break;
2610    
2611      /* End of branch */      /* End of branch */
# Line 2539  for (code = first_significant_code(code Line 2618  for (code = first_significant_code(code
2618      return TRUE;      return TRUE;
2619    
2620      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2621      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2622        followed by a multibyte character. */
2623    
2624  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2625      case OP_STAR:      case OP_STAR:
2626      case OP_STARI:      case OP_STARI:
2627        case OP_NOTSTAR:
2628        case OP_NOTSTARI:
2629    
2630      case OP_MINSTAR:      case OP_MINSTAR:
2631      case OP_MINSTARI:      case OP_MINSTARI:
2632        case OP_NOTMINSTAR:
2633        case OP_NOTMINSTARI:
2634    
2635      case OP_POSSTAR:      case OP_POSSTAR:
2636      case OP_POSSTARI:      case OP_POSSTARI:
2637        case OP_NOTPOSSTAR:
2638        case OP_NOTPOSSTARI:
2639    
2640      case OP_QUERY:      case OP_QUERY:
2641      case OP_QUERYI:      case OP_QUERYI:
2642        case OP_NOTQUERY:
2643        case OP_NOTQUERYI:
2644    
2645      case OP_MINQUERY:      case OP_MINQUERY:
2646      case OP_MINQUERYI:      case OP_MINQUERYI:
2647        case OP_NOTMINQUERY:
2648        case OP_NOTMINQUERYI:
2649    
2650      case OP_POSQUERY:      case OP_POSQUERY:
2651      case OP_POSQUERYI:      case OP_POSQUERYI:
2652        case OP_NOTPOSQUERY:
2653        case OP_NOTPOSQUERYI:
2654    
2655      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2656      break;      break;
2657    
2658      case OP_UPTO:      case OP_UPTO:
2659      case OP_UPTOI:      case OP_UPTOI:
2660        case OP_NOTUPTO:
2661        case OP_NOTUPTOI:
2662    
2663      case OP_MINUPTO:      case OP_MINUPTO:
2664      case OP_MINUPTOI:      case OP_MINUPTOI:
2665        case OP_NOTMINUPTO:
2666        case OP_NOTMINUPTOI:
2667    
2668      case OP_POSUPTO:      case OP_POSUPTO:
2669      case OP_POSUPTOI:      case OP_POSUPTOI:
2670        case OP_NOTPOSUPTO:
2671        case OP_NOTPOSUPTOI:
2672    
2673      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2674      break;      break;
2675  #endif  #endif
# Line 2573  for (code = first_significant_code(code Line 2680  for (code = first_significant_code(code
2680      case OP_MARK:      case OP_MARK:
2681      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2682      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2683      case OP_THEN_ARG:      case OP_THEN_ARG:
2684      code += code[1];      code += code[1];
2685      break;      break;
# Line 2607  Arguments: Line 2711  Arguments:
2711    code        points to start of the recursion    code        points to start of the recursion
2712    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2713    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2714    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2715    cd          pointers to tables etc    cd          pointers to tables etc
2716    
2717  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2619  could_be_empty(const pcre_uchar *code, c Line 2723  could_be_empty(const pcre_uchar *code, c
2723  {  {
2724  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2725    {    {
2726    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2727      return FALSE;      return FALSE;
2728    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2729    }    }
# Line 2629  return TRUE; Line 2733  return TRUE;
2733    
2734    
2735  /*************************************************  /*************************************************
2736  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2737  *************************************************/  *************************************************/
2738    
2739  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2740  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2741    
2742  Returns:   TRUE or FALSE  Arguments:  c opcode
2743    Returns:    base opcode for the type
2744  */  */
2745    
2746  static BOOL  static pcre_uchar
2747  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2748  {  {
2749  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2750  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2751  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2752    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2753    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2754      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2755  }  }
2756    
2757    
2758    
2759    #ifdef SUPPORT_UCP
2760  /*************************************************  /*************************************************
2761  *          Check POSIX class name                *  *        Check a character and a property        *
2762  *************************************************/  *************************************************/
2763    
2764  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2765  such as [:alnum:].  is adjacent to a fixed character.
2766    
2767  Arguments:  Arguments:
2768    ptr        points to the first letter    c            the character
2769    len        the length of the name    ptype        the property type
2770      pdata        the data for the type
2771      negated      TRUE if it's a negated property (\P or \p{^)
2772    
2773  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2774  */  */
2775    
2776  static int  static BOOL
2777  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2778      BOOL negated)
2779  {  {
2780  const char *pn = posix_names;  const pcre_uint32 *p;
2781  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2782    
2783  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2784  that is referenced. This means that groups can be replicated for fixed    {
2785  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2786  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2787  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2788  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2789    
2790  This function has been extended with the possibility of forward references for    case PT_GC:
2791  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2792    
2793  Arguments:    case PT_PC:
2794    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2795    
2796  Returns:     nothing    case PT_SC:
2797  */    return (pdata == prop->script) == negated;
2798    
2799  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2800    
2801  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2802    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2803    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2804    
2805    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2806    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2807      at release 8.34. */
2808    
2809    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2810      case PT_PXSPACE:  /* POSIX space */
2811      switch(c)
2812      {      {
2813      offset = GET(hc, 0);      HSPACE_CASES:
2814      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2815        {      return negated;
2816        PUT(hc, 0, offset + adjust);  
2817        break;      default:
2818        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2819      }      }
2820      break;  /* Control never reaches here */
2821    
2822    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2823    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2824              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2825              c == CHAR_UNDERSCORE) == negated;
2826    
2827    if (hc >= cd->hwm)    case PT_CLIST:
2828      p = PRIV(ucd_caseless_sets) + prop->caseset;
2829      for (;;)
2830      {      {
2831      offset = GET(ptr, 1);      if (c < *p) return !negated;
2832      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2833      }      }
2834      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2835    }    }
2836    
2837    return FALSE;
2838  }  }
2839    #endif  /* SUPPORT_UCP */
2840    
2841    
2842    
2843  /*************************************************  /*************************************************
2844  *        Insert an automatic callout point       *  *        Fill the character property list        *
2845    *************************************************/
2846    
2847    /* Checks whether the code points to an opcode that can take part in auto-
2848    possessification, and if so, fills a list with its properties.
2849    
2850    Arguments:
2851      code        points to start of expression
2852      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2853      fcc         points to case-flipping table
2854      list        points to output list
2855                  list[0] will be filled with the opcode
2856                  list[1] will be non-zero if this opcode
2857                    can match an empty character string
2858                  list[2..7] depends on the opcode
2859    
2860    Returns:      points to the start of the next opcode if *code is accepted
2861                  NULL if *code is not accepted
2862    */
2863    
2864    static const pcre_uchar *
2865    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2866      const pcre_uint8 *fcc, pcre_uint32 *list)
2867    {
2868    pcre_uchar c = *code;
2869    pcre_uchar base;
2870    const pcre_uchar *end;
2871    pcre_uint32 chr;
2872    
2873    #ifdef SUPPORT_UCP
2874    pcre_uint32 *clist_dest;
2875    const pcre_uint32 *clist_src;
2876    #else
2877    utf = utf;  /* Suppress "unused parameter" compiler warning */
2878    #endif
2879    
2880    list[0] = c;
2881    list[1] = FALSE;
2882    code++;
2883    
2884    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2885      {
2886      base = get_repeat_base(c);
2887      c -= (base - OP_STAR);
2888    
2889      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2890        code += IMM2_SIZE;
2891    
2892      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2893    
2894      switch(base)
2895        {
2896        case OP_STAR:
2897        list[0] = OP_CHAR;
2898        break;
2899    
2900        case OP_STARI:
2901        list[0] = OP_CHARI;
2902        break;
2903    
2904        case OP_NOTSTAR:
2905        list[0] = OP_NOT;
2906        break;
2907    
2908        case OP_NOTSTARI:
2909        list[0] = OP_NOTI;
2910        break;
2911    
2912        case OP_TYPESTAR:
2913        list[0] = *code;
2914        code++;
2915        break;
2916        }
2917      c = list[0];
2918      }
2919    
2920    switch(c)
2921      {
2922      case OP_NOT_DIGIT:
2923      case OP_DIGIT:
2924      case OP_NOT_WHITESPACE:
2925      case OP_WHITESPACE:
2926      case OP_NOT_WORDCHAR:
2927      case OP_WORDCHAR:
2928      case OP_ANY:
2929      case OP_ALLANY:
2930      case OP_ANYNL:
2931      case OP_NOT_HSPACE:
2932      case OP_HSPACE:
2933      case OP_NOT_VSPACE:
2934      case OP_VSPACE:
2935      case OP_EXTUNI:
2936      case OP_EODN:
2937      case OP_EOD:
2938      case OP_DOLL:
2939      case OP_DOLLM:
2940      return code;
2941    
2942      case OP_CHAR:
2943      case OP_NOT:
2944      GETCHARINCTEST(chr, code);
2945      list[2] = chr;
2946      list[3] = NOTACHAR;
2947      return code;
2948    
2949      case OP_CHARI:
2950      case OP_NOTI:
2951      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2952      GETCHARINCTEST(chr, code);
2953      list[2] = chr;
2954    
2955    #ifdef SUPPORT_UCP
2956      if (chr < 128 || (chr < 256 && !utf))
2957        list[3] = fcc[chr];
2958      else
2959        list[3] = UCD_OTHERCASE(chr);
2960    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2961      list[3] = (chr < 256) ? fcc[chr] : chr;
2962    #else
2963      list[3] = fcc[chr];
2964    #endif
2965    
2966      /* The othercase might be the same value. */
2967    
2968      if (chr == list[3])
2969        list[3] = NOTACHAR;
2970      else
2971        list[4] = NOTACHAR;
2972      return code;
2973    
2974    #ifdef SUPPORT_UCP
2975      case OP_PROP:
2976      case OP_NOTPROP:
2977      if (code[0] != PT_CLIST)
2978        {
2979        list[2] = code[0];
2980        list[3] = code[1];
2981        return code + 2;
2982        }
2983    
2984      /* Convert only if we have enough space. */
2985    
2986      clist_src = PRIV(ucd_caseless_sets) + code[1];
2987      clist_dest = list + 2;
2988      code += 2;
2989    
2990      do {
2991         if (clist_dest >= list + 8)
2992           {
2993           /* Early return if there is not enough space. This should never
2994           happen, since all clists are shorter than 5 character now. */
2995           list[2] = code[0];
2996           list[3] = code[1];
2997           return code;
2998           }
2999         *clist_dest++ = *clist_src;
3000         }
3001      while(*clist_src++ != NOTACHAR);
3002    
3003      /* All characters are stored. The terminating NOTACHAR
3004      is copied form the clist itself. */
3005    
3006      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3007      return code;
3008    #endif
3009    
3010      case OP_NCLASS:
3011      case OP_CLASS:
3012    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3013      case OP_XCLASS:
3014      if (c == OP_XCLASS)
3015        end = code + GET(code, 0) - 1;
3016      else
3017    #endif
3018        end = code + 32 / sizeof(pcre_uchar);
3019    
3020      switch(*end)
3021        {
3022        case OP_CRSTAR:
3023        case OP_CRMINSTAR:
3024        case OP_CRQUERY:
3025        case OP_CRMINQUERY:
3026        case OP_CRPOSSTAR:
3027        case OP_CRPOSQUERY:
3028        list[1] = TRUE;
3029        end++;
3030        break;
3031    
3032        case OP_CRPLUS:
3033        case OP_CRMINPLUS:
3034        case OP_CRPOSPLUS:
3035        end++;
3036        break;
3037    
3038        case OP_CRRANGE:
3039        case OP_CRMINRANGE:
3040        case OP_CRPOSRANGE:
3041        list[1] = (GET2(end, 1) == 0);
3042        end += 1 + 2 * IMM2_SIZE;
3043        break;
3044        }
3045      list[2] = (pcre_uint32)(end - code);
3046      return end;
3047      }
3048    return NULL;    /* Opcode not accepted */
3049    }
3050    
3051    
3052    
3053    /*************************************************
3054    *    Scan further character sets for match       *
3055    *************************************************/
3056    
3057    /* Checks whether the base and the current opcode have a common character, in
3058    which case the base cannot be possessified.
3059    
3060    Arguments:
3061      code        points to the byte code
3062      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3063      cd          static compile data
3064      base_list   the data list of the base opcode
3065    
3066    Returns:      TRUE if the auto-possessification is possible
3067    */
3068    
3069    static BOOL
3070    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3071      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3072    {
3073    pcre_uchar c;
3074    pcre_uint32 list[8];
3075    const pcre_uint32 *chr_ptr;
3076    const pcre_uint32 *ochr_ptr;
3077    const pcre_uint32 *list_ptr;
3078    const pcre_uchar *next_code;
3079    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3080    const pcre_uchar *xclass_flags;
3081    #endif
3082    const pcre_uint8 *class_bitset;
3083    const pcre_uint8 *set1, *set2, *set_end;
3084    pcre_uint32 chr;
3085    BOOL accepted, invert_bits;
3086    BOOL entered_a_group = FALSE;
3087    
3088    /* Note: the base_list[1] contains whether the current opcode has greedy
3089    (represented by a non-zero value) quantifier. This is a different from
3090    other character type lists, which stores here that the character iterator
3091    matches to an empty string (also represented by a non-zero value). */
3092    
3093    for(;;)
3094      {
3095      /* All operations move the code pointer forward.
3096      Therefore infinite recursions are not possible. */
3097    
3098      c = *code;
3099    
3100      /* Skip over callouts */
3101    
3102      if (c == OP_CALLOUT)
3103        {
3104        code += PRIV(OP_lengths)[c];
3105        continue;
3106        }
3107    
3108      if (c == OP_ALT)
3109        {
3110        do code += GET(code, 1); while (*code == OP_ALT);
3111        c = *code;
3112        }
3113    
3114      switch(c)
3115        {
3116        case OP_END:
3117        case OP_KETRPOS:
3118        /* TRUE only in greedy case. The non-greedy case could be replaced by
3119        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3120        uses more memory, which we cannot get at this stage.) */
3121    
3122        return base_list[1] != 0;
3123    
3124        case OP_KET:
3125        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3126        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3127        cannot be converted to a possessive form. */
3128    
3129        if (base_list[1] == 0) return FALSE;
3130    
3131        switch(*(code - GET(code, 1)))
3132          {
3133          case OP_ASSERT:
3134          case OP_ASSERT_NOT:
3135          case OP_ASSERTBACK:
3136          case OP_ASSERTBACK_NOT:
3137          case OP_ONCE:
3138          case OP_ONCE_NC:
3139          /* Atomic sub-patterns and assertions can always auto-possessify their
3140          last iterator. However, if the group was entered as a result of checking
3141          a previous iterator, this is not possible. */
3142    
3143          return !entered_a_group;
3144          }
3145    
3146        code += PRIV(OP_lengths)[c];
3147        continue;
3148    
3149        case OP_ONCE:
3150        case OP_ONCE_NC:
3151        case OP_BRA:
3152        case OP_CBRA:
3153        next_code = code + GET(code, 1);
3154        code += PRIV(OP_lengths)[c];
3155    
3156        while (*next_code == OP_ALT)
3157          {
3158          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3159          code = next_code + 1 + LINK_SIZE;
3160          next_code += GET(next_code, 1);
3161          }
3162    
3163        entered_a_group = TRUE;
3164        continue;
3165    
3166        case OP_BRAZERO:
3167        case OP_BRAMINZERO:
3168    
3169        next_code = code + 1;
3170        if (*next_code != OP_BRA && *next_code != OP_CBRA
3171            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3172    
3173        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3174    
3175        /* The bracket content will be checked by the
3176        OP_BRA/OP_CBRA case above. */
3177        next_code += 1 + LINK_SIZE;
3178        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3179          return FALSE;
3180    
3181        code += PRIV(OP_lengths)[c];
3182        continue;
3183    
3184        default:
3185        break;
3186        }
3187    
3188      /* Check for a supported opcode, and load its properties. */
3189    
3190      code = get_chr_property_list(code, utf, cd->fcc, list);
3191      if (code == NULL) return FALSE;    /* Unsupported */
3192    
3193      /* If either opcode is a small character list, set pointers for comparing
3194      characters from that list with another list, or with a property. */
3195    
3196      if (base_list[0] == OP_CHAR)
3197        {
3198        chr_ptr = base_list + 2;
3199        list_ptr = list;
3200        }
3201      else if (list[0] == OP_CHAR)
3202        {
3203        chr_ptr = list + 2;
3204        list_ptr = base_list;
3205        }
3206    
3207      /* Character bitsets can also be compared to certain opcodes. */
3208    
3209      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3210    #ifdef COMPILE_PCRE8
3211          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3212          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3213    #endif
3214          )
3215        {
3216    #ifdef COMPILE_PCRE8
3217        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3218    #else
3219        if (base_list[0] == OP_CLASS)
3220    #endif
3221          {
3222          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3223          list_ptr = list;
3224          }
3225        else
3226          {
3227          set1 = (pcre_uint8 *)(code - list[2]);
3228          list_ptr = base_list;
3229          }
3230    
3231        invert_bits = FALSE;
3232        switch(list_ptr[0])
3233          {
3234          case OP_CLASS:
3235          case OP_NCLASS:
3236          set2 = (pcre_uint8 *)
3237            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3238          break;
3239    
3240    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3241          case OP_XCLASS:
3242          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3243          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3244          if ((*xclass_flags & XCL_MAP) == 0)
3245            {
3246            /* No bits are set for characters < 256. */
3247            if (list[1] == 0) return TRUE;
3248            /* Might be an empty repeat. */
3249            continue;
3250            }
3251          set2 = (pcre_uint8 *)(xclass_flags + 1);
3252          break;
3253    #endif
3254    
3255          case OP_NOT_DIGIT:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_DIGIT:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3260          break;
3261    
3262          case OP_NOT_WHITESPACE:
3263          invert_bits = TRUE;
3264          /* Fall through */
3265          case OP_WHITESPACE:
3266          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3267          break;
3268    
3269          case OP_NOT_WORDCHAR:
3270          invert_bits = TRUE;
3271          /* Fall through */
3272          case OP_WORDCHAR:
3273          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3274          break;
3275    
3276          default:
3277          return FALSE;
3278          }
3279    
3280        /* Because the sets are unaligned, we need
3281        to perform byte comparison here. */
3282        set_end = set1 + 32;
3283        if (invert_bits)
3284          {
3285          do
3286            {
3287            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3288            }
3289          while (set1 < set_end);
3290          }
3291        else
3292          {
3293          do
3294            {
3295            if ((*set1++ & *set2++) != 0) return FALSE;
3296            }
3297          while (set1 < set_end);
3298          }
3299    
3300        if (list[1] == 0) return TRUE;
3301        /* Might be an empty repeat. */
3302        continue;
3303        }
3304    
3305      /* Some property combinations also acceptable. Unicode property opcodes are
3306      processed specially; the rest can be handled with a lookup table. */
3307    
3308      else
3309        {
3310        pcre_uint32 leftop, rightop;
3311    
3312        leftop = base_list[0];
3313        rightop = list[0];
3314    
3315    #ifdef SUPPORT_UCP
3316        accepted = FALSE; /* Always set in non-unicode case. */
3317        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3318          {
3319          if (rightop == OP_EOD)
3320            accepted = TRUE;
3321          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3322            {
3323            int n;
3324            const pcre_uint8 *p;
3325            BOOL same = leftop == rightop;
3326            BOOL lisprop = leftop == OP_PROP;
3327            BOOL risprop = rightop == OP_PROP;
3328            BOOL bothprop = lisprop && risprop;
3329    
3330            /* There's a table that specifies how each combination is to be
3331            processed:
3332              0   Always return FALSE (never auto-possessify)
3333              1   Character groups are distinct (possessify if both are OP_PROP)
3334              2   Check character categories in the same group (general or particular)
3335              3   Return TRUE if the two opcodes are not the same
3336              ... see comments below
3337            */
3338    
3339            n = propposstab[base_list[2]][list[2]];
3340            switch(n)
3341              {
3342              case 0: break;
3343              case 1: accepted = bothprop; break;
3344              case 2: accepted = (base_list[3] == list[3]) != same; break;
3345              case 3: accepted = !same; break;
3346    
3347              case 4:  /* Left general category, right particular category */
3348              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3349              break;
3350    
3351              case 5:  /* Right general category, left particular category */
3352              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3353              break;
3354    
3355              /* This code is logically tricky. Think hard before fiddling with it.
3356              The posspropstab table has four entries per row. Each row relates to
3357              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3358              Only WORD actually needs all four entries, but using repeats for the
3359              others means they can all use the same code below.
3360    
3361              The first two entries in each row are Unicode general categories, and
3362              apply always, because all the characters they include are part of the
3363              PCRE character set. The third and fourth entries are a general and a
3364              particular category, respectively, that include one or more relevant
3365              characters. One or the other is used, depending on whether the check
3366              is for a general or a particular category. However, in both cases the
3367              category contains more characters than the specials that are defined
3368              for the property being tested against. Therefore, it cannot be used
3369              in a NOTPROP case.
3370    
3371              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3372              Underscore is covered by ucp_P or ucp_Po. */
3373    
3374              case 6:  /* Left alphanum vs right general category */
3375              case 7:  /* Left space vs right general category */
3376              case 8:  /* Left word vs right general category */
3377              p = posspropstab[n-6];
3378              accepted = risprop && lisprop ==
3379                (list[3] != p[0] &&
3380                 list[3] != p[1] &&
3381                (list[3] != p[2] || !lisprop));
3382              break;
3383    
3384              case 9:   /* Right alphanum vs left general category */
3385              case 10:  /* Right space vs left general category */
3386              case 11:  /* Right word vs left general category */
3387              p = posspropstab[n-9];
3388              accepted = lisprop && risprop ==
3389                (base_list[3] != p[0] &&
3390                 base_list[3] != p[1] &&
3391                (base_list[3] != p[2] || !risprop));
3392              break;
3393    
3394              case 12:  /* Left alphanum vs right particular category */
3395              case 13:  /* Left space vs right particular category */
3396              case 14:  /* Left word vs right particular category */
3397              p = posspropstab[n-12];
3398              accepted = risprop && lisprop ==
3399                (catposstab[p[0]][list[3]] &&
3400                 catposstab[p[1]][list[3]] &&
3401                (list[3] != p[3] || !lisprop));
3402              break;
3403    
3404              case 15:  /* Right alphanum vs left particular category */
3405              case 16:  /* Right space vs left particular category */
3406              case 17:  /* Right word vs left particular category */
3407              p = posspropstab[n-15];
3408              accepted = lisprop && risprop ==
3409                (catposstab[p[0]][base_list[3]] &&
3410                 catposstab[p[1]][base_list[3]] &&
3411                (base_list[3] != p[3] || !risprop));
3412              break;
3413              }
3414            }
3415          }
3416    
3417        else
3418    #endif  /* SUPPORT_UCP */
3419    
3420        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3421               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3422               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3423    
3424        if (!accepted) return FALSE;
3425    
3426        if (list[1] == 0) return TRUE;
3427        /* Might be an empty repeat. */
3428        continue;
3429        }
3430    
3431      /* Control reaches here only if one of the items is a small character list.
3432      All characters are checked against the other side. */
3433    
3434      do
3435        {
3436        chr = *chr_ptr;
3437    
3438        switch(list_ptr[0])
3439          {
3440          case OP_CHAR:
3441          ochr_ptr = list_ptr + 2;
3442          do
3443            {
3444            if (chr == *ochr_ptr) return FALSE;
3445            ochr_ptr++;
3446            }
3447          while(*ochr_ptr != NOTACHAR);
3448          break;
3449    
3450          case OP_NOT:
3451          ochr_ptr = list_ptr + 2;
3452          do
3453            {
3454            if (chr == *ochr_ptr)
3455              break;
3456            ochr_ptr++;
3457            }
3458          while(*ochr_ptr != NOTACHAR);
3459          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3460          break;
3461    
3462          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3463          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3464    
3465          case OP_DIGIT:
3466          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3467          break;
3468    
3469          case OP_NOT_DIGIT:
3470          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3471          break;
3472    
3473          case OP_WHITESPACE:
3474          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3475          break;
3476    
3477          case OP_NOT_WHITESPACE:
3478          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3479          break;
3480    
3481          case OP_WORDCHAR:
3482          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3483          break;
3484    
3485          case OP_NOT_WORDCHAR:
3486          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3487          break;
3488    
3489          case OP_HSPACE:
3490          switch(chr)
3491            {
3492            HSPACE_CASES: return FALSE;
3493            default: break;
3494            }
3495          break;
3496    
3497          case OP_NOT_HSPACE:
3498          switch(chr)
3499            {
3500            HSPACE_CASES: break;
3501            default: return FALSE;
3502            }
3503          break;
3504    
3505          case OP_ANYNL:
3506          case OP_VSPACE:
3507          switch(chr)
3508            {
3509            VSPACE_CASES: return FALSE;
3510            default: break;
3511            }
3512          break;
3513    
3514          case OP_NOT_VSPACE:
3515          switch(chr)
3516            {
3517            VSPACE_CASES: break;
3518            default: return FALSE;
3519            }
3520          break;
3521    
3522          case OP_DOLL:
3523          case OP_EODN:
3524          switch (chr)
3525            {
3526            case CHAR_CR:
3527            case CHAR_LF:
3528            case CHAR_VT:
3529            case CHAR_FF:
3530            case CHAR_NEL:
3531    #ifndef EBCDIC
3532            case 0x2028:
3533            case 0x2029:
3534    #endif  /* Not EBCDIC */
3535            return FALSE;
3536            }
3537          break;
3538    
3539          case OP_EOD:    /* Can always possessify before \z */
3540          break;
3541    
3542    #ifdef SUPPORT_UCP
3543          case OP_PROP:
3544          case OP_NOTPROP:
3545          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3546                list_ptr[0] == OP_NOTPROP))
3547            return FALSE;
3548          break;
3549    #endif
3550    
3551          case OP_NCLASS:
3552          if (chr > 255) return FALSE;
3553          /* Fall through */
3554    
3555          case OP_CLASS:
3556          if (chr > 255) break;
3557          class_bitset = (pcre_uint8 *)
3558            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3559          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3560          break;
3561    
3562    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3563          case OP_XCLASS:
3564          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3565              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3566          break;
3567    #endif
3568    
3569          default:
3570          return FALSE;
3571          }
3572    
3573        chr_ptr++;
3574        }
3575      while(*chr_ptr != NOTACHAR);
3576    
3577      /* At least one character must be matched from this opcode. */
3578    
3579      if (list[1] == 0) return TRUE;
3580      }
3581    
3582    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3583    but some compilers complain about an unreachable statement. */
3584    
3585    }
3586    
3587    
3588    
3589    /*************************************************
3590    *    Scan compiled regex for auto-possession     *
3591    *************************************************/
3592    
3593    /* Replaces single character iterations with their possessive alternatives
3594    if appropriate. This function modifies the compiled opcode!
3595    
3596    Arguments:
3597      code        points to start of the byte code
3598      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3599      cd          static compile data
3600    
3601    Returns:      nothing
3602    */
3603    
3604    static void
3605    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3606    {
3607    register pcre_uchar c;
3608    const pcre_uchar *end;
3609    pcre_uchar *repeat_opcode;
3610    pcre_uint32 list[8];
3611    
3612    for (;;)
3613      {
3614      c = *code;
3615    
3616      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3617        {
3618        c -= get_repeat_base(c) - OP_STAR;
3619        end = (c <= OP_MINUPTO) ?
3620          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3621        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3622    
3623        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3624          {
3625          switch(c)
3626            {
3627            case OP_STAR:
3628            *code += OP_POSSTAR - OP_STAR;
3629            break;
3630    
3631            case OP_MINSTAR:
3632            *code += OP_POSSTAR - OP_MINSTAR;
3633            break;
3634    
3635            case OP_PLUS:
3636            *code += OP_POSPLUS - OP_PLUS;
3637            break;
3638    
3639            case OP_MINPLUS:
3640            *code += OP_POSPLUS - OP_MINPLUS;
3641            break;
3642    
3643            case OP_QUERY:
3644            *code += OP_POSQUERY - OP_QUERY;
3645            break;
3646    
3647            case OP_MINQUERY:
3648            *code += OP_POSQUERY - OP_MINQUERY;
3649            break;
3650    
3651            case OP_UPTO:
3652            *code += OP_POSUPTO - OP_UPTO;
3653            break;
3654    
3655            case OP_MINUPTO:
3656            *code += OP_POSUPTO - OP_MINUPTO;
3657            break;
3658            }
3659          }
3660        c = *code;
3661        }
3662      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3663        {
3664    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3665        if (c == OP_XCLASS)
3666          repeat_opcode = code + GET(code, 1);
3667        else
3668    #endif
3669          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3670    
3671        c = *repeat_opcode;
3672        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3673          {
3674          /* end must not be NULL. */
3675          end = get_chr_property_list(code, utf, cd->fcc, list);
3676    
3677          list[1] = (c & 1) == 0;
3678    
3679          if (compare_opcodes(end, utf, cd, list, end))
3680            {
3681            switch (c)
3682              {
3683              case OP_CRSTAR:
3684              case OP_CRMINSTAR:
3685              *repeat_opcode = OP_CRPOSSTAR;
3686              break;
3687    
3688              case OP_CRPLUS:
3689              case OP_CRMINPLUS:
3690              *repeat_opcode = OP_CRPOSPLUS;
3691              break;
3692    
3693              case OP_CRQUERY:
3694              case OP_CRMINQUERY:
3695              *repeat_opcode = OP_CRPOSQUERY;
3696              break;
3697    
3698              case OP_CRRANGE:
3699              case OP_CRMINRANGE:
3700              *repeat_opcode = OP_CRPOSRANGE;
3701              break;
3702              }
3703            }
3704          }
3705        c = *code;
3706        }
3707    
3708      switch(c)
3709        {
3710        case OP_END:
3711        return;
3712    
3713        case OP_TYPESTAR:
3714        case OP_TYPEMINSTAR:
3715        case OP_TYPEPLUS:
3716        case OP_TYPEMINPLUS:
3717        case OP_TYPEQUERY:
3718        case OP_TYPEMINQUERY:
3719        case OP_TYPEPOSSTAR:
3720        case OP_TYPEPOSPLUS:
3721        case OP_TYPEPOSQUERY:
3722        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3723        break;
3724    
3725        case OP_TYPEUPTO:
3726        case OP_TYPEMINUPTO:
3727        case OP_TYPEEXACT:
3728        case OP_TYPEPOSUPTO:
3729        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3730          code += 2;
3731        break;
3732    
3733    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3734        case OP_XCLASS:
3735        code += GET(code, 1);
3736        break;
3737    #endif
3738    
3739        case OP_MARK:
3740        case OP_PRUNE_ARG:
3741        case OP_SKIP_ARG:
3742        case OP_THEN_ARG:
3743        code += code[1];
3744        break;
3745        }
3746    
3747      /* Add in the fixed length from the table */
3748    
3749      code += PRIV(OP_lengths)[c];
3750    
3751      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3752      a multi-byte character. The length in the table is a minimum, so we have to
3753      arrange to skip the extra bytes. */
3754    
3755    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3756      if (utf) switch(c)
3757        {
3758        case OP_CHAR:
3759        case OP_CHARI:
3760        case OP_NOT:
3761        case OP_NOTI:
3762        case OP_STAR:
3763        case OP_MINSTAR:
3764        case OP_PLUS:
3765        case OP_MINPLUS:
3766        case OP_QUERY:
3767        case OP_MINQUERY:
3768        case OP_UPTO:
3769        case OP_MINUPTO:
3770        case OP_EXACT:
3771        case OP_POSSTAR:
3772        case OP_POSPLUS:
3773        case OP_POSQUERY:
3774        case OP_POSUPTO:
3775        case OP_STARI:
3776        case OP_MINSTARI:
3777        case OP_PLUSI:
3778        case OP_MINPLUSI:
3779        case OP_QUERYI:
3780        case OP_MINQUERYI:
3781        case OP_UPTOI:
3782        case OP_MINUPTOI:
3783        case OP_EXACTI:
3784        case OP_POSSTARI:
3785        case OP_POSPLUSI:
3786        case OP_POSQUERYI:
3787        case OP_POSUPTOI:
3788        case OP_NOTSTAR:
3789        case OP_NOTMINSTAR:
3790        case OP_NOTPLUS:
3791        case OP_NOTMINPLUS:
3792        case OP_NOTQUERY:
3793        case OP_NOTMINQUERY:
3794        case OP_NOTUPTO:
3795        case OP_NOTMINUPTO:
3796        case OP_NOTEXACT:
3797        case OP_NOTPOSSTAR:
3798        case OP_NOTPOSPLUS:
3799        case OP_NOTPOSQUERY:
3800        case OP_NOTPOSUPTO:
3801        case OP_NOTSTARI:
3802        case OP_NOTMINSTARI:
3803        case OP_NOTPLUSI:
3804        case OP_NOTMINPLUSI:
3805        case OP_NOTQUERYI:
3806        case OP_NOTMINQUERYI:
3807        case OP_NOTUPTOI:
3808        case OP_NOTMINUPTOI:
3809        case OP_NOTEXACTI:
3810        case OP_NOTPOSSTARI:
3811        case OP_NOTPOSPLUSI:
3812        case OP_NOTPOSQUERYI:
3813        case OP_NOTPOSUPTOI:
3814        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3815        break;
3816        }
3817    #else
3818      (void)(utf);  /* Keep compiler happy by referencing function argument */
3819    #endif
3820      }
3821    }
3822    
3823    
3824    
3825    /*************************************************
3826    *           Check for POSIX class syntax         *
3827    *************************************************/
3828    
3829    /* This function is called when the sequence "[:" or "[." or "[=" is
3830    encountered in a character class. It checks whether this is followed by a
3831    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3832    reach an unescaped ']' without the special preceding character, return FALSE.
3833    
3834    Originally, this function only recognized a sequence of letters between the
3835    terminators, but it seems that Perl recognizes any sequence of characters,
3836    though of course unknown POSIX names are subsequently rejected. Perl gives an
3837    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3838    didn't consider this to be a POSIX class. Likewise for [:1234:].
3839    
3840    The problem in trying to be exactly like Perl is in the handling of escapes. We
3841    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3842    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3843    below handles the special case of \], but does not try to do any other escape
3844    processing. This makes it different from Perl for cases such as [:l\ower:]
3845    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3846    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3847    I think.
3848    
3849    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3850    It seems that the appearance of a nested POSIX class supersedes an apparent
3851    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3852    a digit.
3853    
3854    In Perl, unescaped square brackets may also appear as part of class names. For
3855    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3856    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3857    seem right at all. PCRE does not allow closing square brackets in POSIX class
3858    names.
3859    
3860    Arguments:
3861      ptr      pointer to the initial [
3862      endptr   where to return the end pointer
3863    
3864    Returns:   TRUE or FALSE
3865    */
3866    
3867    static BOOL
3868    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3869    {
3870    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3871    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3872    for (++ptr; *ptr != CHAR_NULL; ptr++)
3873      {
3874      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3875        ptr++;
3876      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3877      else
3878        {
3879        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3880          {
3881          *endptr = ptr;
3882          return TRUE;
3883          }
3884        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3885             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3886              ptr[1] == CHAR_EQUALS_SIGN) &&
3887            check_posix_syntax(ptr, endptr))
3888          return FALSE;
3889        }
3890      }
3891    return FALSE;
3892    }
3893    
3894    
3895    
3896    
3897    /*************************************************
3898    *          Check POSIX class name                *
3899    *************************************************/
3900    
3901    /* This function is called to check the name given in a POSIX-style class entry
3902    such as [:alnum:].
3903    
3904    Arguments:
3905      ptr        points to the first letter
3906      len        the length of the name
3907    
3908    Returns:     a value representing the name, or -1 if unknown
3909    */
3910    
3911    static int
3912    check_posix_name(const pcre_uchar *ptr, int len)
3913    {
3914    const char *pn = posix_names;
3915    register int yield = 0;
3916    while (posix_name_lengths[yield] != 0)
3917      {
3918      if (len == posix_name_lengths[yield] &&
3919        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3920      pn += posix_name_lengths[yield] + 1;
3921      yield++;
3922      }
3923    return -1;
3924    }
3925    
3926    
3927    /*************************************************
3928    *    Adjust OP_RECURSE items in repeated group   *
3929    *************************************************/
3930    
3931    /* OP_RECURSE items contain an offset from the start of the regex to the group
3932    that is referenced. This means that groups can be replicated for fixed
3933    repetition simply by copying (because the recursion is allowed to refer to
3934    earlier groups that are outside the current group). However, when a group is
3935    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3936    inserted before it, after it has been compiled. This means that any OP_RECURSE
3937    items within it that refer to the group itself or any contained groups have to
3938    have their offsets adjusted. That one of the jobs of this function. Before it
3939    is called, the partially compiled regex must be temporarily terminated with
3940    OP_END.
3941    
3942    This function has been extended with the possibility of forward references for
3943    recursions and subroutine calls. It must also check the list of such references
3944    for the group we are dealing with. If it finds that one of the recursions in
3945    the current group is on this list, it adjusts the offset in the list, not the
3946    value in the reference (which is a group number).
3947    
3948    Arguments:
3949      group      points to the start of the group
3950      adjust     the amount by which the group is to be moved
3951      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3952      cd         contains pointers to tables etc.
3953      save_hwm   the hwm forward reference pointer at the start of the group
3954    
3955    Returns:     nothing
3956    */
3957    
3958    static void
3959    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3960      pcre_uchar *save_hwm)
3961    {
3962    pcre_uchar *ptr = group;
3963    
3964    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3965      {
3966      int offset;
3967      pcre_uchar *hc;
3968    
3969      /* See if this recursion is on the forward reference list. If so, adjust the
3970      reference. */
3971    
3972      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3973        {
3974        offset = (int)GET(hc, 0);
3975        if (cd->start_code + offset == ptr + 1)
3976          {
3977          PUT(hc, 0, offset + adjust);
3978          break;
3979          }
3980        }
3981    
3982      /* Otherwise, adjust the recursion offset if it's after the start of this
3983      group. */
3984    
3985      if (hc >= cd->hwm)
3986        {
3987        offset = (int)GET(ptr, 1);
3988        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3989        }
3990    
3991      ptr += 1 + LINK_SIZE;
3992      }
3993    }
3994    
3995    
3996    
3997    /*************************************************
3998    *        Insert an automatic callout point       *
3999  *************************************************/  *************************************************/
4000    
4001  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
# Line 2858  PUT(previous_callout, 2 + LINK_SIZE, len Line 4052  PUT(previous_callout, 2 + LINK_SIZE, len
4052  *************************************************/  *************************************************/
4053    
4054  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
4055  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
4056  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
4057  start address.  start address. A character with multiple other cases is returned on its own
4058    with a special return value.
4059    
4060  Arguments:  Arguments:
4061    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2868  Arguments: Line 4063  Arguments:
4063    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4064    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4065    
4066  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4067                   0 when a range is returned
4068                  >0 the CASESET offset for char with multiple other cases
4069                    in this case, ocptr contains the original
4070  */  */
4071    
4072  static BOOL  static int
4073  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4074    unsigned int *odptr)    pcre_uint32 *odptr)
4075  {  {
4076  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4077    unsigned int co;
4078    
4079    /* Find the first character that has an other case. If it has multiple other
4080    cases, return its case offset value. */
4081    
4082  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4083    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4084      if ((co = UCD_CASESET(c)) != 0)
4085        {
4086        *ocptr = c++;   /* Character that has the set */
4087        *cptr = c;      /* Rest of input range */
4088        return (int)co;
4089        }
4090      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4091      }
4092    
4093    if (c > d) return -1;  /* Reached end of range */
4094    
4095  if (c > d) return FALSE;  /* Found a character that has a single other case. Search for the end of the
4096    range, which is either the end of the input range, or a character that has zero
4097    or more than one other cases. */
4098    
4099  *ocptr = othercase;  *ocptr = othercase;
4100  next = othercase + 1;  next = othercase + 1;
4101    
4102  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4103    {    {
4104    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4105    next++;    next++;
4106    }    }
4107    
4108  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4109  *cptr = c;  *cptr = c;             /* Rest of input range */
4110    return 0;
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *        Check a character and a property        *  
 *************************************************/  
   
 /* This function is called by check_auto_possessive() when a property item  
 is adjacent to a fixed character.  
   
 Arguments:  
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
   
 Returns:       TRUE if auto-possessifying is OK  
 */  
   
 static BOOL  
 check_char_prop(int c, int ptype, int pdata, BOOL negated)  
 {  
 const ucd_record *prop = GET_UCD(c);  
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
   
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
   
   case PT_SC:  
   return (pdata == prop->script) == negated;  
   
   /* These are specials */  
   
   case PT_ALNUM:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
   
   case PT_SPACE:    /* Perl space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_PXSPACE:  /* POSIX space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_WORD:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
4111  }  }
4112  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4113    
4114    
4115    
4116  /*************************************************  /*************************************************
4117  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
4118  *************************************************/  *************************************************/
4119    
4120  /* This function is called for unlimited repeats of certain items, to see  /* This function packages up the logic of adding a character or range of
4121  whether the next thing could possibly match the repeated item. If not, it makes  characters to a class. The character values in the arguments will be within the
4122  sense to automatically possessify the repeated item.  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4123    mutually recursive with the function immediately below.
4124    
4125  Arguments:  Arguments:
4126    previous      pointer to the repeated opcode    classbits     the bit map for characters < 256
4127    utf           TRUE in UTF-8 / UTF-16 mode    uchardptr     points to the pointer for extra data
4128    ptr           next character in pattern    options       the options word
   options       options bits  
4129    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4130      start         start of range character
4131      end           end of range character
4132    
4133  Returns:        TRUE if possessifying is wanted  Returns:        the number of < 256 characters added
4134                    the pointer to extra data is updated
4135  */  */
4136    
4137  static BOOL  static int
4138  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4139    const pcre_uchar *ptr, int options, compile_data *cd)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4140  {  {
4141  pcre_int32 c, next;  pcre_uint32 c;
4142  int op_code = *previous++;  pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4143    int n8 = 0;
4144    
4145    /* If caseless matching is required, scan the range and process alternate
4146    cases. In Unicode, there are 8-bit characters that have alternate cases that
4147    are greater than 255 and vice-versa. Sometimes we can just extend the original
4148    range. */
4149    
4150  /* Skip whitespace and comments in extended mode */  if ((options & PCRE_CASELESS) != 0)
   
 if ((options & PCRE_EXTENDED) != 0)  
4151    {    {
4152    for (;;)  #ifdef SUPPORT_UCP
4153      if ((options & PCRE_UTF8) != 0)
4154      {      {
4155      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      int rc;
4156      if (*ptr == CHAR_NUMBER_SIGN)      pcre_uint32 oc, od;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next item is one that we can handle, get its value. A non-negative  
 value is a character, a negative value is an escape value. */  
   
 if (*ptr == CHAR_BACKSLASH)  
   {  
   int temperrorcode = 0;  
   next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
4157    
4158  /* Skip whitespace and comments in extended mode */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4159        c = start;
4160    
4161  if ((options & PCRE_EXTENDED) != 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
4162        {        {
4163        ptr++;        /* Handle a single character that has more than one other case. */
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
4164    
4165  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4166    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)          PRIV(ucd_caseless_sets) + rc, oc);
     return FALSE;  
4167    
4168  /* Now compare the next item with the previous opcode. First, handle cases when        /* Do nothing if the other case range is within the original range. */
 the next item is a character. */  
4169    
4170  if (next >= 0) switch(op_code)        else if (oc >= start && od <= end) continue;
   {  
   case OP_CHAR:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
4171    
4172    /* For CHARI (caseless character) we must check the other case. If we have        /* Extend the original range if there is overlap, noting that if oc < c, we
4173    Unicode property support, we can use it to test the other case of        can't have od > end because a subrange is always shorter than the basic
4174    high-valued characters. */        range. Otherwise, use a recursive call to add the additional range. */
4175    
4176    case OP_CHARI:        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4177  #ifdef SUPPORT_UTF        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4178    GETCHARTEST(c, previous);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4179  #else        }
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
4180      }      }
4181    else    else
4182  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UCP */
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
4183    
4184    case OP_NOT:    /* Not UTF-mode, or no UCP */
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c == next;  
4185    
4186    case OP_NOTI:    for (c = start; c <= classbits_end; c++)
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
4187      {      {
4188      unsigned int othercase;      SETBIT(classbits, cd->fcc[c]);
4189      if (next < 128) othercase = cd->fcc[next]; else      n8++;
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c == othercase;  
4190      }      }
4191    else    }
 #endif  /* SUPPORT_UTF */  
   return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
   
   /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  
   When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  
   
   case OP_DIGIT:  
   return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
   
   case OP_NOT_DIGIT:  
   return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
   
   case OP_WHITESPACE:  
   return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  
   
   case OP_NOT_WHITESPACE:  
   return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  
   
   case OP_WORDCHAR:  
   return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  
4192    
4193    case OP_NOT_WORDCHAR:  /* Now handle the original range. Adjust the final value according to the bit
4194    return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  length - this means that the same lists of (e.g.) horizontal spaces can be used
4195    in all cases. */
4196    
4197    case OP_HSPACE:  #if defined COMPILE_PCRE8
4198    case OP_NOT_HSPACE:  #ifdef SUPPORT_UTF
4199    switch(next)    if ((options & PCRE_UTF8) == 0)
4200      {  #endif
4201      case 0x09:    if (end > 0xff) end = 0xff;
     case 0x20:  
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001:  
     case 0x2002:  
     case 0x2003:  
     case 0x2004:  
     case 0x2005:  
     case 0x2006:  
     case 0x2007:  
     case 0x2008:  
     case 0x2009:  
     case 0x200A:  
     case 0x202f:  
     case 0x205f:  
     case 0x3000:  
     return op_code == OP_NOT_HSPACE;  
     default:  
     return op_code != OP_NOT_HSPACE;  
     }  
4202    
4203    case OP_ANYNL:  #elif defined COMPILE_PCRE16
4204    case OP_VSPACE:  #ifdef SUPPORT_UTF
4205    case OP_NOT_VSPACE:    if ((options & PCRE_UTF16) == 0)
4206    switch(next)  #endif
4207      {    if (end > 0xffff) end = 0xffff;
     case 0x0a:  
     case 0x0b:  
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
     }  
4208    
4209  #ifdef SUPPORT_UCP  #endif /* COMPILE_PCRE[8|16] */
   case OP_PROP:  
   return check_char_prop(next, previous[0], previous[1], FALSE);  
4210    
4211    case OP_NOTPROP:  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
   return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
4212    
4213    default:  for (c = start; c <= classbits_end; c++)
4214    return FALSE;    {
4215      /* Regardless of start, c will always be <= 255. */
4216      SETBIT(classbits, c);
4217      n8++;
4218    }    }
4219    
4220    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4221    if (start <= 0xff) start = 0xff + 1;
4222    
4223  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  if (end >= start)
 is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
   
 switch(op_code)  
4224    {    {
4225    case OP_CHAR:    pcre_uchar *uchardata = *uchardptr;
   case OP_CHARI:  
4226  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4227    GETCHARTEST(c, previous);    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
 #else  
   c = *previous;  
 #endif  
   switch(-next)  
4228      {      {
4229      case ESC_d:      if (start < end)
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
   
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
   
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
   
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
   
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
   
     case ESC_h:  
     case ESC_H:  
     switch(c)  
4230        {        {
4231        case 0x09:        *uchardata++ = XCL_RANGE;
4232        case 0x20:        uchardata += PRIV(ord2utf)(start, uchardata);
4233        case 0xa0:        uchardata += PRIV(ord2utf)(end, uchardata);
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
       default:  
       return -next == ESC_h;  
4234        }        }
4235        else if (start == end)
     case ESC_v:  
     case ESC_V:  
     switch(c)  
4236        {        {
4237        case 0x0a:        *uchardata++ = XCL_SINGLE;
4238        case 0x0b:        uchardata += PRIV(ord2utf)(start, uchardata);
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
       default:  
       return -next == ESC_v;  
       }  
   
     /* When PCRE_UCP is set, these values get generated for \d etc. Find  
     their substitutions and process them. The result will always be either  
     -ESC_p or -ESC_P. Then fall through to process those values. */  
   
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
4239        }        }
4240      /* Fall through */      }
4241      else
4242    #endif  /* SUPPORT_UTF */
4243    
4244      case ESC_p:    /* Without UTF support, character values are constrained by the bit length,
4245      case ESC_P:    and can only be > 256 for 16-bit and 32-bit libraries. */
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
4246    
4247        ptr--;      /* Make ptr point at the p or P */  #ifdef COMPILE_PCRE8
4248        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);      {}
4249        if (ptype < 0) return FALSE;  #else
4250        ptr++;      /* Point past the final curly ket */    if (start < end)
4251        {
4252        *uchardata++ = XCL_RANGE;
4253        *uchardata++ = start;
4254        *uchardata++ = end;
4255        }
4256      else if (start == end)
4257        {
4258        *uchardata++ = XCL_SINGLE;
4259        *uchardata++ = start;
4260        }
4261    #endif
4262    
4263        /* If the property item is optional, we have to give up. (When generated    *uchardptr = uchardata;   /* Updata extra data pointer */
4264        from \d etc by PCRE_UCP, this test will have been applied much earlier,    }
4265        to the original \d etc. At this point, ptr will point to a zero byte. */  #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4266    
4267        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  return n8;    /* Number of 8-bit characters */
4268          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  }
           return FALSE;  
4269    
       /* Do the property check. */  
4270    
       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  
       }  
 #endif  
4271    
     default:  
     return FALSE;  
     }  
4272    
4273    /* In principle, support for Unicode properties should be integrated here as  /*************************************************
4274    well. It means re-organizing the above code so as to get hold of the property  *        Add a list of characters to a class     *
4275    values before switching on the op-code. However, I wonder how many patterns  *************************************************/
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
4276    
4277    case OP_DIGIT:  /* This function is used for adding a list of case-equivalent characters to a
4278    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  class, and also for adding a list of horizontal or vertical whitespace. If the
4279           next == -ESC_h || next == -ESC_v || next == -ESC_R;  list is in order (which it should be), ranges of characters are detected and
4280    handled appropriately. This function is mutually recursive with the function
4281    above.
4282    
4283    case OP_NOT_DIGIT:  Arguments:
4284    return next == -ESC_d;    classbits     the bit map for characters < 256
4285      uchardptr     points to the pointer for extra data
4286      options       the options word
4287      cd            contains pointers to tables etc.
4288      p             points to row of 32-bit values, terminated by NOTACHAR
4289      except        character to omit; this is used when adding lists of
4290                      case-equivalent characters to avoid including the one we
4291                      already know about
4292    
4293    case OP_WHITESPACE:  Returns:        the number of < 256 characters added
4294    return next == -ESC_S || next == -ESC_d || next == -ESC_w;                  the pointer to extra data is updated
4295    */
4296    
4297    case OP_NOT_WHITESPACE:  static int
4298    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;  add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4299      compile_data *cd, const pcre_uint32 *p, unsigned int except)
4300    {
4301    int n8 = 0;
4302    while (p[0] < NOTACHAR)
4303      {
4304      int n = 0;
4305      if (p[0] != except)
4306        {
4307        while(p[n+1] == p[0] + n + 1) n++;
4308        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4309        }
4310      p += n + 1;
4311      }
4312    return n8;
4313    }
4314    
   case OP_HSPACE:  
   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||  
          next == -ESC_w || next == -ESC_v || next == -ESC_R;  
4315    
   case OP_NOT_HSPACE:  
   return next == -ESC_h;  
4316    
4317    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /*************************************************
4318    case OP_ANYNL:  *    Add characters not in a list to a class     *
4319    case OP_VSPACE:  *************************************************/
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
4320    
4321    case OP_NOT_VSPACE:  /* This function is used for adding the complement of a list of horizontal or
4322    return next == -ESC_v || next == -ESC_R;  vertical whitespace to a class. The list must be in order.
4323    
4324    case OP_WORDCHAR:  Arguments:
4325    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    classbits     the bit map for characters < 256
4326           next == -ESC_v || next == -ESC_R;    uchardptr     points to the pointer for extra data
4327      options       the options word
4328      cd            contains pointers to tables etc.
4329      p             points to row of 32-bit values, terminated by NOTACHAR
4330    
4331    case OP_NOT_WORDCHAR:  Returns:        the number of < 256 characters added
4332    return next == -ESC_w || next == -ESC_d;                  the pointer to extra data is updated
4333    */
4334    
4335    default:  static int
4336    return FALSE;  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4337      int options, compile_data *cd, const pcre_uint32 *p)
4338    {
4339    BOOL utf = (options & PCRE_UTF8) != 0;
4340    int n8 = 0;
4341    if (p[0] > 0)
4342      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4343    while (p[0] < NOTACHAR)
4344      {
4345      while (p[1] == p[0] + 1) p++;
4346      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4347        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4348      p++;
4349    }    }
4350    return n8;
 /* Control does not reach here */  
4351  }  }
4352    
4353    
# Line 3398  to find out the amount of memory needed, Line 4363  to find out the amount of memory needed,
4363  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4364    
4365  Arguments:  Arguments:
4366    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4367    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4368    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4369    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4370    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4371    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4372    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4373    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4374    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4375    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4376                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4377      lengthptr         NULL during the real compile phase
4378                        points to length accumulator during pre-compile phase
4379    
4380  Returns:         TRUE on success  Returns:            TRUE on success
4381                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4382  */  */
4383    
4384  static BOOL  static BOOL
4385  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4386    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4387    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4388      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4389      branch_chain *bcptr, int cond_depth,