/[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 903 by ph10, Sat Jan 21 16:37:17 2012 UTC revision 1480 by ph10, Tue May 27 18:02:51 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 489  static const char error_texts[] = Line 535  static const char error_texts[] =
535    "too many forward references\0"    "too many forward references\0"
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 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552    ;    ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 628  static const pcre_uint8 ebcdic_chartab[] Line 688  static const pcre_uint8 ebcdic_chartab[]
688  #endif  #endif
689    
690    
691  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
692    between adjacent character-type opcodes. The left-hand (repeated) opcode is
693    used to select the row, and the right-hand opcode is use to select the column.
694    A value of 1 means that auto-possessification is OK. For example, the second
695    value in the first row means that \D+\d can be turned into \D++\d.
696    
697    The Unicode property types (\P and \p) have to be present to fill out the table
698    because of what their opcode values are, but the table values should always be
699    zero because property types are handled separately in the code. The last four
700    columns apply to items that cannot be repeated, so there is no need to have
701    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
702    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
703    
704    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
705    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
706    
707    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
708    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
709      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
710      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
711      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
712      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
713      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
714      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
720      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
724      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
725      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
726    };
727    
728    
729  static BOOL  /* This table is used to check whether auto-possessification is possible
730    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
731      int *, int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
732    opcode is used to select the column. The values are as follows:
733    
734      0   Always return FALSE (never auto-possessify)
735      1   Character groups are distinct (possessify if both are OP_PROP)
736      2   Check character categories in the same group (general or particular)
737      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
738    
739      4   Check left general category vs right particular category
740      5   Check right general category vs left particular category
741    
742      6   Left alphanum vs right general category
743      7   Left space vs right general category
744      8   Left word vs right general category
745    
746      9   Right alphanum vs left general category
747     10   Right space vs left general category
748     11   Right word vs left general category
749    
750     12   Left alphanum vs right particular category
751     13   Left space vs right particular category
752     14   Left word vs right particular category
753    
754     15   Right alphanum vs left particular category
755     16   Right space vs left particular category
756     17   Right word vs left particular category
757    */
758    
759    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
760    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
761      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
762      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
763      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
764      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
765      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
766      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
767      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
769      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
770      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
772    };
773    
774    /* This table is used to check whether auto-possessification is possible
775    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
776    specifies a general category and the other specifies a particular category. The
777    row is selected by the general category and the column by the particular
778    category. The value is 1 if the particular category is not part of the general
779    category. */
780    
781    static const pcre_uint8 catposstab[7][30] = {
782    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
783      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
784      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
790    };
791    
792    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
793    a general or particular category. The properties in each row are those
794    that apply to the character set in question. Duplication means that a little
795    unnecessary work is done when checking, but this keeps things much simpler
796    because they can all use the same code. For more details see the comment where
797    this table is used.
798    
799    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
800    "space", but from Perl 5.18 it's included, so both categories are treated the
801    same here. */
802    
803    static const pcre_uint8 posspropstab[3][4] = {
804      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
805      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
806      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807    };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
# Line 655  find_error_text(int n) Line 885  find_error_text(int n)
885  const char *s = error_texts;  const char *s = error_texts;
886  for (; n > 0; n--)  for (; n > 0; n--)
887    {    {
888    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
889    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
890    }    }
891  return s;  return s;
892  }  }
893    
894    
895    
896  /*************************************************  /*************************************************
897  *           Expand the workspace                 *  *           Expand the workspace                 *
898  *************************************************/  *************************************************/
# Line 739  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 970  return (*p == CHAR_RIGHT_CURLY_BRACKET);
970  *************************************************/  *************************************************/
971    
972  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
973  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
974  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
975  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
976  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
977  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981      chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
985    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
986    
987  Returns:         zero or positive => a data character  Returns:         zero => a data character
988                   negative => a special escape sequence                   positive => a special escape sequence
989                     negative => a back reference
990                   on error, errorcodeptr is set                   on error, errorcodeptr is set
991  */  */
992    
993  static int  static int
994  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
995    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
996  {  {
997  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
998  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
999  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
1000  pcre_int32 c;  pcre_uint32 c;
1001    int escape = 0;
1002  int i;  int i;
1003    
1004  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 773  ptr--;                            /* Set Line 1006  ptr--;                            /* Set
1006    
1007  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1008    
1009  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1010    
1011  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1012  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 782  Otherwise further processing may be requ Line 1015  Otherwise further processing may be requ
1015  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1016  /* Not alphanumeric */  /* Not alphanumeric */
1017  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1018  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1019      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1020    
1021  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1022  /* Not alphanumeric */  /* Not alphanumeric */
1023  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)) {}
1024  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; }
1025  #endif  #endif
1026    
1027  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 795  else if ((i = escapes[c - 0x48]) != 0) Line 1029  else if ((i = escapes[c - 0x48]) != 0)
1029  else  else
1030    {    {
1031    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1032    BOOL braced, negated;    BOOL braced, negated, overflow;
1033      int s;
1034    
1035    switch (c)    switch (c)
1036      {      {
# Line 820  else Line 1055  else
1055          c = 0;          c = 0;
1056          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1057            {            {
1058            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1059  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1060            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1061            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 829  else Line 1064  else
1064            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1065  #endif  #endif
1066            }            }
1067    
1068    #if defined COMPILE_PCRE8
1069            if (c > (utf ? 0x10ffffU : 0xffU))
1070    #elif defined COMPILE_PCRE16
1071            if (c > (utf ? 0x10ffffU : 0xffffU))
1072    #elif defined COMPILE_PCRE32
1073            if (utf && c > 0x10ffffU)
1074    #endif
1075              {
1076              *errorcodeptr = ERR76;
1077              }
1078            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1079          }          }
1080        }        }
1081      else      else
# Line 855  else Line 1102  else
1102      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1103      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1104      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1105      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1106    
1107      case CHAR_g:      case CHAR_g:
1108      if (isclass) break;      if (isclass) break;
1109      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1110        {        {
1111        c = -ESC_g;        escape = ESC_g;
1112        break;        break;
1113        }        }
1114    
# Line 870  else Line 1117  else
1117      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1118        {        {
1119        const pcre_uchar *p;        const pcre_uchar *p;
1120        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1121          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1122        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1123          {          {
1124          c = -ESC_k;          escape = ESC_k;
1125          break;          break;
1126          }          }
1127        braced = TRUE;        braced = TRUE;
# Line 890  else Line 1137  else
1137      else negated = FALSE;      else negated = FALSE;
1138    
1139      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1140      c = 0;      s = 0;
1141        overflow = FALSE;
1142      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1143        {        {
1144        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1145          {          {
1146          c = -1;          overflow = TRUE;
1147          break;          break;
1148          }          }
1149        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1150        }        }
1151      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1152        {        {
1153        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1154          ptr++;          ptr++;
# Line 914  else Line 1162  else
1162        break;        break;
1163        }        }
1164    
1165      if (c == 0)      if (s == 0)
1166        {        {
1167        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1168        break;        break;
# Line 922  else Line 1170  else
1170    
1171      if (negated)      if (negated)
1172        {        {
1173        if (c > bracount)        if (s > bracount)
1174          {          {
1175          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1176          break;          break;
1177          }          }
1178        c = bracount - (c - 1);        s = bracount - (s - 1);
1179        }        }
1180    
1181      c = -(ESC_REF + c);      escape = -s;
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1190      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1191      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1192      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1193      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1194      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 952  else Line 1204  else
1204        {        {
1205        oldptr = ptr;        oldptr = ptr;
1206        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1207        c -= CHAR_0;        s = (int)(c -CHAR_0);
1208          overflow = FALSE;
1209        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1210          {          {
1211          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1212            {            {
1213            c = -1;            overflow = TRUE;
1214            break;            break;
1215            }            }
1216          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1217          }          }
1218        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1219          {          {
1220          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1221            ptr++;            ptr++;
1222          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1223          break;          break;
1224          }          }
1225        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          c = -(ESC_REF + c);          escape = -s;
1228          break;          break;
1229          }          }
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1234      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1235      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1243      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1003  else Line 1254  else
1254  #endif  #endif
1255      break;      break;
1256    
1257      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1258      than 0xff in 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}. */
1259      If not, { is treated as a data character. */  
1260        case CHAR_o:
1261        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262          {
1263          ptr += 2;
1264          c = 0;
1265          overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267            {
1268            register pcre_uint32 cc = *ptr++;
1269            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270    #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272    #endif
1273            c = (c << 3) + cc - CHAR_0 ;
1274    #if defined COMPILE_PCRE8
1275            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE16
1277            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280    #endif
1281            }
1282          if (overflow)
1283            {
1284            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285            *errorcodeptr = ERR34;
1286            }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288            {
1289            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292          }
1293        break;
1294    
1295        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298      case CHAR_x:      case CHAR_x:
1299      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1301        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303          {          {
1304          c = 0;          c = 0;
1305          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1306            {            {
1307            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1308  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1309            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1310            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 1028  else Line 1314  else
1314  #endif  #endif
1315            }            }
1316          }          }
1317        break;        }    /* End JavaScript handling */
       }  
1318    
1319      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        c = 0;      else
1327        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1328          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329          {          {
1330          register int cc = *pt++;          ptr += 2;
1331          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1332            overflow = FALSE;
1333            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338    #ifdef COMPILE_PCRE32
1339              if (c >= 0x10000000l) { overflow = TRUE; break; }
1340    #endif
1341    
1342  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1346          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 */
1347          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348  #endif  #endif
1349    
1350  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1351          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352  #else  #elif defined COMPILE_PCRE16
1353  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1355              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356  #endif  #endif
1357  #endif            }
         }  
1358    
1359        if (c < 0)          if (overflow)
1360          {            {
1361          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1363          }            }
1364    
1365        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366          {            {
1367          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368          ptr = pt;            }
         break;  
         }  
1369    
1370        /* 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.
1371        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1372        }          \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1380        {        else
1381        int cc;                                  /* Some compilers don't like */          {
1382        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1391        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393  #endif  #endif
1394        }            }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1099  else Line 1403  else
1403    
1404      case CHAR_c:      case CHAR_c:
1405      c = *(++ptr);      c = *(++ptr);
1406      if (c == 0)      if (c == CHAR_NULL)
1407        {        {
1408        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1409        break;        break;
# Line 1139  else Line 1443  else
1443  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1444  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1445    
1446  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1447       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1448    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1449    
1450  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1451    
1452  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1453    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1454    
1455  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1456    
1457  *ptrptr = ptr;  *ptrptr = ptr;
1458  return c;  *chptr = c;
1459    return escape;
1460  }  }
1461    
1462    
# Line 1169  escape sequence. Line 1474  escape sequence.
1474  Argument:  Argument:
1475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1476    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1477    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1478      pdataptr       points to an unsigned int that is set to the detailed property value
1479    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1480    
1481  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1482  */  */
1483    
1484  static int  static BOOL
1485  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1486      unsigned int *pdataptr, int *errorcodeptr)
1487  {  {
1488  int c, i, bot, top;  pcre_uchar c;
1489    int i, bot, top;
1490  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1491  pcre_uchar name[32];  pcre_uchar name[32];
1492    
1493  c = *(++ptr);  c = *(++ptr);
1494  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1495    
1496  *negptr = FALSE;  *negptr = FALSE;
1497    
# Line 1200  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1508  if (c == CHAR_LEFT_CURLY_BRACKET)
1508    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1509      {      {
1510      c = *(++ptr);      c = *(++ptr);
1511      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1512      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1513      name[i] = c;      name[i] = c;
1514      }      }
# Line 1225  top = PRIV(utt_size); Line 1533  top = PRIV(utt_size);
1533    
1534  while (bot < top)  while (bot < top)
1535    {    {
1536      int r;
1537    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1538    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1539    if (c == 0)    if (r == 0)
1540      {      {
1541      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1542      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1543        return TRUE;
1544      }      }
1545    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1546    }    }
1547    
1548  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1549  *ptrptr = ptr;  *ptrptr = ptr;
1550  return -1;  return FALSE;
1551    
1552  ERROR_RETURN:  ERROR_RETURN:
1553  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1554  *ptrptr = ptr;  *ptrptr = ptr;
1555  return -1;  return FALSE;
1556  }  }
1557  #endif  #endif
1558    
1559    
1560    
   
1561  /*************************************************  /*************************************************
1562  *         Read repeat counts                     *  *         Read repeat counts                     *
1563  *************************************************/  *************************************************/
# Line 1274  read_repeat_counts(const pcre_uchar *p, Line 1583  read_repeat_counts(const pcre_uchar *p,
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
1585    
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  
 if (min < 0 || min > 65535)  
1587    {    {
1588    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1589    return p;    if (min > 65535)
1590    }      {
1591        *errorcodeptr = ERR5;
1592  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1593  Also, max must not be less than min. */      }
1594      }
1595    
1596  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1597    {    {
1598    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1599      {      {
1600      max = 0;      max = 0;
1601      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1602        {        {
1603        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1604        return p;        if (max > 65535)
1605        }          {
1606            *errorcodeptr = ERR5;
1607            return p;
1608            }
1609          }
1610      if (max < min)      if (max < min)
1611        {        {
1612        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1306  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1615  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1615      }      }
1616    }    }
1617    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1618  *minp = min;  *minp = min;
1619  *maxp = max;  *maxp = max;
1620  return p;  return p;
# Line 1317  return p; Line 1623  return p;
1623    
1624    
1625  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1626  *      Find first significant op code            *  *      Find first significant op code            *
1627  *************************************************/  *************************************************/
1628    
# Line 1650  for (;;) Line 1661  for (;;)
1661    
1662      case OP_CALLOUT:      case OP_CALLOUT:
1663      case OP_CREF:      case OP_CREF:
1664      case OP_NCREF:      case OP_DNCREF:
1665      case OP_RREF:      case OP_RREF:
1666      case OP_NRREF:      case OP_DNRREF:
1667      case OP_DEF:      case OP_DEF:
1668      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1669      break;      break;
# Line 1666  for (;;) Line 1677  for (;;)
1677    
1678    
1679    
   
1680  /*************************************************  /*************************************************
1681  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1682  *************************************************/  *************************************************/
# Line 1684  and doing the check at the end; a flag s Line 1694  and doing the check at the end; a flag s
1694    
1695  Arguments:  Arguments:
1696    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1697    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1698    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1699    cd       the "compile data" structure    cd       the "compile data" structure
1700    
# Line 1710  for (;;) Line 1720  for (;;)
1720    {    {
1721    int d;    int d;
1722    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1723    register int op = *cc;    register pcre_uchar op = *cc;
1724    
1725    switch (op)    switch (op)
1726      {      {
# Line 1790  for (;;) Line 1800  for (;;)
1800      case OP_COMMIT:      case OP_COMMIT:
1801      case OP_CREF:      case OP_CREF:
1802      case OP_DEF:      case OP_DEF:
1803        case OP_DNCREF:
1804        case OP_DNRREF:
1805      case OP_DOLL:      case OP_DOLL:
1806      case OP_DOLLM:      case OP_DOLLM:
1807      case OP_EOD:      case OP_EOD:
1808      case OP_EODN:      case OP_EODN:
1809      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1810      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1811      case OP_PRUNE:      case OP_PRUNE:
1812      case OP_REVERSE:      case OP_REVERSE:
# Line 1830  for (;;) Line 1840  for (;;)
1840      case OP_EXACTI:      case OP_EXACTI:
1841      case OP_NOTEXACT:      case OP_NOTEXACT:
1842      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1843      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1844      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1845  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1846      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1839  for (;;) Line 1849  for (;;)
1849    
1850      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1851      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1852      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)
1853          cc += 2;
1854      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1855      break;      break;
1856    
# Line 1874  for (;;) Line 1885  for (;;)
1885    
1886      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1887    
1888  #if defined SUPPORT_UTF || defined COMPILE_PCRE16      case OP_CLASS:
1889        case OP_NCLASS:
1890    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1891      case OP_XCLASS:      case OP_XCLASS:
1892      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];      /* The original code caused an unsigned overflow in 64 bit systems,
1893      /* Fall through */      so now we use a conditional statement. */
1894        if (op == OP_XCLASS)
1895          cc += GET(cc, 1);
1896        else
1897          cc += PRIV(OP_lengths)[OP_CLASS];
1898    #else
1899        cc += PRIV(OP_lengths)[OP_CLASS];
1900  #endif  #endif
1901    
     case OP_CLASS:  
     case OP_NCLASS:  
     cc += PRIV(OP_lengths)[OP_CLASS];  
   
1902      switch (*cc)      switch (*cc)
1903        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1904        case OP_CRSTAR:        case OP_CRSTAR:
1905        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1906          case OP_CRPLUS:
1907          case OP_CRMINPLUS:
1908        case OP_CRQUERY:        case OP_CRQUERY:
1909        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1910          case OP_CRPOSSTAR:
1911          case OP_CRPOSPLUS:
1912          case OP_CRPOSQUERY:
1913        return -1;        return -1;
1914    
1915        case OP_CRRANGE:        case OP_CRRANGE:
1916        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1917          case OP_CRPOSRANGE:
1918        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1919        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1920        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1921        break;        break;
1922    
# Line 1964  for (;;) Line 1983  for (;;)
1983      case OP_QUERYI:      case OP_QUERYI:
1984      case OP_REF:      case OP_REF:
1985      case OP_REFI:      case OP_REFI:
1986        case OP_DNREF:
1987        case OP_DNREFI:
1988      case OP_SBRA:      case OP_SBRA:
1989      case OP_SBRAPOS:      case OP_SBRAPOS:
1990      case OP_SCBRA:      case OP_SCBRA:
# Line 2000  for (;;) Line 2021  for (;;)
2021    
2022    
2023    
   
2024  /*************************************************  /*************************************************
2025  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2026  *************************************************/  *************************************************/
# Line 2013  length. Line 2033  length.
2033    
2034  Arguments:  Arguments:
2035    code        points to start of expression    code        points to start of expression
2036    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2037    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2038    
2039  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
# Line 2024  PRIV(find_bracket)(const pcre_uchar *cod Line 2044  PRIV(find_bracket)(const pcre_uchar *cod
2044  {  {
2045  for (;;)  for (;;)
2046    {    {
2047    register int c = *code;    register pcre_uchar c = *code;
2048    
2049    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2050    
# Line 2047  for (;;) Line 2067  for (;;)
2067    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2068             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2069      {      {
2070      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2071      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2072      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2073      }      }
# Line 2077  for (;;) Line 2097  for (;;)
2097        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2098        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2099        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2100        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2101          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2102        break;        break;
2103    
2104        case OP_MARK:        case OP_MARK:
2105        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2106        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2107        case OP_THEN_ARG:        case OP_THEN_ARG:
2108        code += code[1];        code += code[1];
2109        break;        break;
# Line 2100  for (;;) Line 2117  for (;;)
2117    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2118    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2119    
2120  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2121      if (utf) switch(c)      if (utf) switch(c)
2122        {        {
2123        case OP_CHAR:        case OP_CHAR:
# Line 2152  instance of OP_RECURSE. Line 2169  instance of OP_RECURSE.
2169    
2170  Arguments:  Arguments:
2171    code        points to start of expression    code        points to start of expression
2172    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2173    
2174  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2175  */  */
# Line 2162  find_recurse(const pcre_uchar *code, BOO Line 2179  find_recurse(const pcre_uchar *code, BOO
2179  {  {
2180  for (;;)  for (;;)
2181    {    {
2182    register int c = *code;    register pcre_uchar c = *code;
2183    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2184    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2185    
# Line 2197  for (;;) Line 2214  for (;;)
2214        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2215        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2216        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2217        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2218          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2219        break;        break;
2220    
2221        case OP_MARK:        case OP_MARK:
2222        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2223        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2224        case OP_THEN_ARG:        case OP_THEN_ARG:
2225        code += code[1];        code += code[1];
2226        break;        break;
# Line 2220  for (;;) Line 2234  for (;;)
2234      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2235      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2236    
2237  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2238      if (utf) switch(c)      if (utf) switch(c)
2239        {        {
2240        case OP_CHAR:        case OP_CHAR:
2241        case OP_CHARI:        case OP_CHARI:
2242          case OP_NOT:
2243          case OP_NOTI:
2244        case OP_EXACT:        case OP_EXACT:
2245        case OP_EXACTI:        case OP_EXACTI:
2246          case OP_NOTEXACT:
2247          case OP_NOTEXACTI:
2248        case OP_UPTO:        case OP_UPTO:
2249        case OP_UPTOI:        case OP_UPTOI:
2250          case OP_NOTUPTO:
2251          case OP_NOTUPTOI:
2252        case OP_MINUPTO:        case OP_MINUPTO:
2253        case OP_MINUPTOI:        case OP_MINUPTOI:
2254          case OP_NOTMINUPTO:
2255          case OP_NOTMINUPTOI:
2256        case OP_POSUPTO:        case OP_POSUPTO:
2257        case OP_POSUPTOI:        case OP_POSUPTOI:
2258          case OP_NOTPOSUPTO:
2259          case OP_NOTPOSUPTOI:
2260        case OP_STAR:        case OP_STAR:
2261        case OP_STARI:        case OP_STARI:
2262          case OP_NOTSTAR:
2263          case OP_NOTSTARI:
2264        case OP_MINSTAR:        case OP_MINSTAR:
2265        case OP_MINSTARI:        case OP_MINSTARI:
2266          case OP_NOTMINSTAR:
2267          case OP_NOTMINSTARI:
2268        case OP_POSSTAR:        case OP_POSSTAR:
2269        case OP_POSSTARI:        case OP_POSSTARI:
2270          case OP_NOTPOSSTAR:
2271          case OP_NOTPOSSTARI:
2272        case OP_PLUS:        case OP_PLUS:
2273        case OP_PLUSI:        case OP_PLUSI:
2274          case OP_NOTPLUS:
2275          case OP_NOTPLUSI:
2276        case OP_MINPLUS:        case OP_MINPLUS:
2277        case OP_MINPLUSI:        case OP_MINPLUSI:
2278          case OP_NOTMINPLUS:
2279          case OP_NOTMINPLUSI:
2280        case OP_POSPLUS:        case OP_POSPLUS:
2281        case OP_POSPLUSI:        case OP_POSPLUSI:
2282          case OP_NOTPOSPLUS:
2283          case OP_NOTPOSPLUSI:
2284        case OP_QUERY:        case OP_QUERY:
2285        case OP_QUERYI:        case OP_QUERYI:
2286          case OP_NOTQUERY:
2287          case OP_NOTQUERYI:
2288        case OP_MINQUERY:        case OP_MINQUERY:
2289        case OP_MINQUERYI:        case OP_MINQUERYI:
2290          case OP_NOTMINQUERY:
2291          case OP_NOTMINQUERYI:
2292        case OP_POSQUERY:        case OP_POSQUERY:
2293        case OP_POSQUERYI:        case OP_POSQUERYI:
2294          case OP_NOTPOSQUERY:
2295          case OP_NOTPOSQUERYI:
2296        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2297        break;        break;
2298        }        }
# Line 2278  bracket whose current branch will alread Line 2320  bracket whose current branch will alread
2320  Arguments:  Arguments:
2321    code        points to start of search    code        points to start of search
2322    endcode     points to where to stop    endcode     points to where to stop
2323    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2324    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2325      recurses    chain of recurse_check to catch mutual recursion
2326    
2327  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2328  */  */
2329    
2330    typedef struct recurse_check {
2331      struct recurse_check *prev;
2332      const pcre_uchar *group;
2333    } recurse_check;
2334    
2335  static BOOL  static BOOL
2336  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2337    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2338  {  {
2339  register int c;  register pcre_uchar c;
2340    recurse_check this_recurse;
2341    
2342  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2343       code < endcode;       code < endcode;
2344       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2316  for (code = first_significant_code(code Line 2366  for (code = first_significant_code(code
2366    
2367    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2368      {      {
2369      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2370      BOOL empty_branch;      BOOL empty_branch;
2371    
2372      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2373        when called to scan a completed pattern by setting cd->start_workspace to
2374        NULL. */
2375    
2376        if (cd->start_workspace != NULL)
2377          {
2378          const pcre_uchar *tcode;
2379          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2380            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2381          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2382          }
2383    
2384        /* If we are scanning a completed pattern, there are no forward references
2385        and all groups are complete. We need to detect whether this is a recursive
2386        call, as otherwise there will be an infinite loop. If it is a recursion,
2387        just skip over it. Simple recursions are easily detected. For mutual
2388        recursions we keep a chain on the stack. */
2389    
2390        else
2391          {
2392          recurse_check *r = recurses;
2393          const pcre_uchar *endgroup = scode;
2394    
2395      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2396        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2397    
2398      /* Not a forward reference, test for completed backward reference */        for (r = recurses; r != NULL; r = r->prev)
2399            if (r->group == scode) break;
2400          if (r != NULL) continue;   /* Mutual recursion */
2401          }
2402    
2403      empty_branch = FALSE;      /* Completed reference; scan the referenced group, remembering it on the
2404      scode = cd->start_code + GET(code, 1);      stack chain to detect mutual recursions. */
     if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */  
2405    
2406      /* Completed backwards reference */      empty_branch = FALSE;
2407        this_recurse.prev = recurses;
2408        this_recurse.group = scode;
2409    
2410      do      do
2411        {        {
2412        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2413          {          {
2414          empty_branch = TRUE;          empty_branch = TRUE;
2415          break;          break;
# Line 2390  for (code = first_significant_code(code Line 2465  for (code = first_significant_code(code
2465        empty_branch = FALSE;        empty_branch = FALSE;
2466        do        do
2467          {          {
2468          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2469            empty_branch = TRUE;            empty_branch = TRUE;
2470          code += GET(code, 1);          code += GET(code, 1);
2471          }          }
# Line 2432  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2508        case OP_CRQUERY:        case OP_CRQUERY:
2509        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2510          case OP_CRPOSSTAR:
2511          case OP_CRPOSQUERY:
2512        break;        break;
2513    
2514        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2515        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2516        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2517          case OP_CRPOSPLUS:
2518        return FALSE;        return FALSE;
2519    
2520        case OP_CRRANGE:        case OP_CRRANGE:
2521        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2522          case OP_CRPOSRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2524        break;        break;
2525        }        }
# Line 2448  for (code = first_significant_code(code Line 2527  for (code = first_significant_code(code
2527    
2528      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2529    
2530        case OP_ANY:
2531        case OP_ALLANY:
2532        case OP_ANYBYTE:
2533    
2534      case OP_PROP:      case OP_PROP:
2535      case OP_NOTPROP:      case OP_NOTPROP:
2536        case OP_ANYNL:
2537    
2538        case OP_NOT_HSPACE:
2539        case OP_HSPACE:
2540        case OP_NOT_VSPACE:
2541        case OP_VSPACE:
2542      case OP_EXTUNI:      case OP_EXTUNI:
2543    
2544      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2545      case OP_DIGIT:      case OP_DIGIT:
2546      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2547      case OP_WHITESPACE:      case OP_WHITESPACE:
2548      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2549      case OP_WORDCHAR:      case OP_WORDCHAR:
2550      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2551      case OP_CHAR:      case OP_CHAR:
2552      case OP_CHARI:      case OP_CHARI:
2553      case OP_NOT:      case OP_NOT:
2554      case OP_NOTI:      case OP_NOTI:
2555    
2556      case OP_PLUS:      case OP_PLUS:
2557        case OP_PLUSI:
2558      case OP_MINPLUS:      case OP_MINPLUS:
2559      case OP_POSPLUS:      case OP_MINPLUSI:
2560      case OP_EXACT:  
2561      case OP_NOTPLUS:      case OP_NOTPLUS:
2562        case OP_NOTPLUSI:
2563      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2564        case OP_NOTMINPLUSI:
2565    
2566        case OP_POSPLUS:
2567        case OP_POSPLUSI:
2568      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2569        case OP_NOTPOSPLUSI:
2570    
2571        case OP_EXACT:
2572        case OP_EXACTI:
2573      case OP_NOTEXACT:      case OP_NOTEXACT:
2574        case OP_NOTEXACTI:
2575    
2576      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2577      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2578      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2579      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2580    
2581      return FALSE;      return FALSE;
2582    
2583      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2495  for (code = first_significant_code(code Line 2597  for (code = first_significant_code(code
2597      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2598      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2599      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2600      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2601        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2602      break;      break;
2603    
2604      /* End of branch */      /* End of branch */
# Line 2509  for (code = first_significant_code(code Line 2611  for (code = first_significant_code(code
2611      return TRUE;      return TRUE;
2612    
2613      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2614      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2615        followed by a multibyte character. */
2616    
2617  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2618      case OP_STAR:      case OP_STAR:
2619      case OP_STARI:      case OP_STARI:
2620        case OP_NOTSTAR:
2621        case OP_NOTSTARI:
2622    
2623      case OP_MINSTAR:      case OP_MINSTAR:
2624      case OP_MINSTARI:      case OP_MINSTARI:
2625        case OP_NOTMINSTAR:
2626        case OP_NOTMINSTARI:
2627    
2628      case OP_POSSTAR:      case OP_POSSTAR:
2629      case OP_POSSTARI:      case OP_POSSTARI:
2630        case OP_NOTPOSSTAR:
2631        case OP_NOTPOSSTARI:
2632    
2633      case OP_QUERY:      case OP_QUERY:
2634      case OP_QUERYI:      case OP_QUERYI:
2635        case OP_NOTQUERY:
2636        case OP_NOTQUERYI:
2637    
2638      case OP_MINQUERY:      case OP_MINQUERY:
2639      case OP_MINQUERYI:      case OP_MINQUERYI:
2640        case OP_NOTMINQUERY:
2641        case OP_NOTMINQUERYI:
2642    
2643      case OP_POSQUERY:      case OP_POSQUERY:
2644      case OP_POSQUERYI:      case OP_POSQUERYI:
2645        case OP_NOTPOSQUERY:
2646        case OP_NOTPOSQUERYI:
2647    
2648      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2649      break;      break;
2650    
2651      case OP_UPTO:      case OP_UPTO:
2652      case OP_UPTOI:      case OP_UPTOI:
2653        case OP_NOTUPTO:
2654        case OP_NOTUPTOI:
2655    
2656      case OP_MINUPTO:      case OP_MINUPTO:
2657      case OP_MINUPTOI:      case OP_MINUPTOI:
2658        case OP_NOTMINUPTO:
2659        case OP_NOTMINUPTOI:
2660    
2661      case OP_POSUPTO:      case OP_POSUPTO:
2662      case OP_POSUPTOI:      case OP_POSUPTOI:
2663        case OP_NOTPOSUPTO:
2664        case OP_NOTPOSUPTOI:
2665    
2666      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]);
2667      break;      break;
2668  #endif  #endif
# Line 2543  for (code = first_significant_code(code Line 2673  for (code = first_significant_code(code
2673      case OP_MARK:      case OP_MARK:
2674      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2675      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2676      case OP_THEN_ARG:      case OP_THEN_ARG:
2677      code += code[1];      code += code[1];
2678      break;      break;
# Line 2577  Arguments: Line 2704  Arguments:
2704    code        points to start of the recursion    code        points to start of the recursion
2705    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2706    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2707    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2708    cd          pointers to tables etc    cd          pointers to tables etc
2709    
2710  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2589  could_be_empty(const pcre_uchar *code, c Line 2716  could_be_empty(const pcre_uchar *code, c
2716  {  {
2717  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2718    {    {
2719    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2720      return FALSE;      return FALSE;
2721    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2722    }    }
# Line 2599  return TRUE; Line 2726  return TRUE;
2726    
2727    
2728  /*************************************************  /*************************************************
2729  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2730  *************************************************/  *************************************************/
2731    
2732  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2733  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 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  
2734    
2735  Returns:   TRUE or FALSE  Arguments:  c opcode
2736    Returns:    base opcode for the type
2737  */  */
2738    
2739  static BOOL  static pcre_uchar
2740  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2741  {  {
2742  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2743  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2744  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2745    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2746    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2747      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;  
2748  }  }
2749    
2750    
2751    
2752    #ifdef SUPPORT_UCP
2753  /*************************************************  /*************************************************
2754  *          Check POSIX class name                *  *        Check a character and a property        *
2755  *************************************************/  *************************************************/
2756    
2757  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2758  such as [:alnum:].  is adjacent to a fixed character.
2759    
2760  Arguments:  Arguments:
2761    ptr        points to the first letter    c            the character
2762    len        the length of the name    ptype        the property type
2763      pdata        the data for the type
2764      negated      TRUE if it's a negated property (\P or \p{^)
2765    
2766  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2767  */  */
2768    
2769  static int  static BOOL
2770  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2771      BOOL negated)
2772  {  {
2773  const char *pn = posix_names;  const pcre_uint32 *p;
2774  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2775    
2776  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2777  that is referenced. This means that groups can be replicated for fixed    {
2778  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2779  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2780  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2781  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2782    
2783  This function has been extended with the possibility of forward references for    case PT_GC:
2784  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2785    
2786  Arguments:    case PT_PC:
2787    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   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  
2788    
2789  Returns:     nothing    case PT_SC:
2790  */    return (pdata == prop->script) == negated;
2791    
2792  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;  
2793    
2794  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2795    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2796    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2797    
2798    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2799    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2800      at release 8.34. */
2801    
2802    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2803      case PT_PXSPACE:  /* POSIX space */
2804      switch(c)
2805      {      {
2806      offset = GET(hc, 0);      HSPACE_CASES:
2807      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2808        {      return negated;
2809        PUT(hc, 0, offset + adjust);  
2810        break;      default:
2811        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2812      }      }
2813      break;  /* Control never reaches here */
2814    
2815    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2816    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2817              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2818              c == CHAR_UNDERSCORE) == negated;
2819    
2820    if (hc >= cd->hwm)    case PT_CLIST:
2821      p = PRIV(ucd_caseless_sets) + prop->caseset;
2822      for (;;)
2823      {      {
2824      offset = GET(ptr, 1);      if (c < *p) return !negated;
2825      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2826      }      }
2827      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2828    }    }
2829    
2830    return FALSE;
2831  }  }
2832    #endif  /* SUPPORT_UCP */
2833    
2834    
2835    
2836  /*************************************************  /*************************************************
2837  *        Insert an automatic callout point       *  *        Fill the character property list        *
2838  *************************************************/  *************************************************/
2839    
2840  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2841  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2842    
2843    Arguments:
2844      code        points to start of expression
2845      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2846      fcc         points to case-flipping table
2847      list        points to output list
2848                  list[0] will be filled with the opcode
2849                  list[1] will be non-zero if this opcode
2850                    can match an empty character string
2851                  list[2..7] depends on the opcode
2852    
2853    Returns:      points to the start of the next opcode if *code is accepted
2854                  NULL if *code is not accepted
2855    */
2856    
2857    static const pcre_uchar *
2858    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2859      const pcre_uint8 *fcc, pcre_uint32 *list)
2860    {
2861    pcre_uchar c = *code;
2862    pcre_uchar base;
2863    const pcre_uchar *end;
2864    pcre_uint32 chr;
2865    
2866    #ifdef SUPPORT_UCP
2867    pcre_uint32 *clist_dest;
2868    const pcre_uint32 *clist_src;
2869    #else
2870    utf = utf;  /* Suppress "unused parameter" compiler warning */
2871    #endif
2872    
2873    list[0] = c;
2874    list[1] = FALSE;
2875    code++;
2876    
2877    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2878      {
2879      base = get_repeat_base(c);
2880      c -= (base - OP_STAR);
2881    
2882      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2883        code += IMM2_SIZE;
2884    
2885      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2886    
2887      switch(base)
2888        {
2889        case OP_STAR:
2890        list[0] = OP_CHAR;
2891        break;
2892    
2893        case OP_STARI:
2894        list[0] = OP_CHARI;
2895        break;
2896    
2897        case OP_NOTSTAR:
2898        list[0] = OP_NOT;
2899        break;
2900    
2901        case OP_NOTSTARI:
2902        list[0] = OP_NOTI;
2903        break;
2904    
2905        case OP_TYPESTAR:
2906        list[0] = *code;
2907        code++;
2908        break;
2909        }
2910      c = list[0];
2911      }
2912    
2913    switch(c)
2914      {
2915      case OP_NOT_DIGIT:
2916      case OP_DIGIT:
2917      case OP_NOT_WHITESPACE:
2918      case OP_WHITESPACE:
2919      case OP_NOT_WORDCHAR:
2920      case OP_WORDCHAR:
2921      case OP_ANY:
2922      case OP_ALLANY:
2923      case OP_ANYNL:
2924      case OP_NOT_HSPACE:
2925      case OP_HSPACE:
2926      case OP_NOT_VSPACE:
2927      case OP_VSPACE:
2928      case OP_EXTUNI:
2929      case OP_EODN:
2930      case OP_EOD:
2931      case OP_DOLL:
2932      case OP_DOLLM:
2933      return code;
2934    
2935      case OP_CHAR:
2936      case OP_NOT:
2937      GETCHARINCTEST(chr, code);
2938      list[2] = chr;
2939      list[3] = NOTACHAR;
2940      return code;
2941    
2942      case OP_CHARI:
2943      case OP_NOTI:
2944      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2945      GETCHARINCTEST(chr, code);
2946      list[2] = chr;
2947    
2948    #ifdef SUPPORT_UCP
2949      if (chr < 128 || (chr < 256 && !utf))
2950        list[3] = fcc[chr];
2951      else
2952        list[3] = UCD_OTHERCASE(chr);
2953    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2954      list[3] = (chr < 256) ? fcc[chr] : chr;
2955    #else
2956      list[3] = fcc[chr];
2957    #endif
2958    
2959      /* The othercase might be the same value. */
2960    
2961      if (chr == list[3])
2962        list[3] = NOTACHAR;
2963      else
2964        list[4] = NOTACHAR;
2965      return code;
2966    
2967    #ifdef SUPPORT_UCP
2968      case OP_PROP:
2969      case OP_NOTPROP:
2970      if (code[0] != PT_CLIST)
2971        {
2972        list[2] = code[0];
2973        list[3] = code[1];
2974        return code + 2;
2975        }
2976    
2977      /* Convert only if we have enough space. */
2978    
2979      clist_src = PRIV(ucd_caseless_sets) + code[1];
2980      clist_dest = list + 2;
2981      code += 2;
2982    
2983      do {
2984         if (clist_dest >= list + 8)
2985           {
2986           /* Early return if there is not enough space. This should never
2987           happen, since all clists are shorter than 5 character now. */
2988           list[2] = code[0];
2989           list[3] = code[1];
2990           return code;
2991           }
2992         *clist_dest++ = *clist_src;
2993         }
2994      while(*clist_src++ != NOTACHAR);
2995    
2996      /* All characters are stored. The terminating NOTACHAR
2997      is copied form the clist itself. */
2998    
2999      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3000      return code;
3001    #endif
3002    
3003      case OP_NCLASS:
3004      case OP_CLASS:
3005    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3006      case OP_XCLASS:
3007      if (c == OP_XCLASS)
3008        end = code + GET(code, 0) - 1;
3009      else
3010    #endif
3011        end = code + 32 / sizeof(pcre_uchar);
3012    
3013      switch(*end)
3014        {
3015        case OP_CRSTAR:
3016        case OP_CRMINSTAR:
3017        case OP_CRQUERY:
3018        case OP_CRMINQUERY:
3019        case OP_CRPOSSTAR:
3020        case OP_CRPOSQUERY:
3021        list[1] = TRUE;
3022        end++;
3023        break;
3024    
3025        case OP_CRPLUS:
3026        case OP_CRMINPLUS:
3027        case OP_CRPOSPLUS:
3028        end++;
3029        break;
3030    
3031        case OP_CRRANGE:
3032        case OP_CRMINRANGE:
3033        case OP_CRPOSRANGE:
3034        list[1] = (GET2(end, 1) == 0);
3035        end += 1 + 2 * IMM2_SIZE;
3036        break;
3037        }
3038      list[2] = (pcre_uint32)(end - code);
3039      return end;
3040      }
3041    return NULL;    /* Opcode not accepted */
3042    }
3043    
3044    
3045    
3046    /*************************************************
3047    *    Scan further character sets for match       *
3048    *************************************************/
3049    
3050    /* Checks whether the base and the current opcode have a common character, in
3051    which case the base cannot be possessified.
3052    
3053    Arguments:
3054      code        points to the byte code
3055      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3056      cd          static compile data
3057      base_list   the data list of the base opcode
3058    
3059    Returns:      TRUE if the auto-possessification is possible
3060    */
3061    
3062    static BOOL
3063    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3064      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3065    {
3066    pcre_uchar c;
3067    pcre_uint32 list[8];
3068    const pcre_uint32 *chr_ptr;
3069    const pcre_uint32 *ochr_ptr;
3070    const pcre_uint32 *list_ptr;
3071    const pcre_uchar *next_code;
3072    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3073    const pcre_uchar *xclass_flags;
3074    #endif
3075    const pcre_uint8 *class_bitset;
3076    const pcre_uint8 *set1, *set2, *set_end;
3077    pcre_uint32 chr;
3078    BOOL accepted, invert_bits;
3079    BOOL entered_a_group = FALSE;
3080    
3081    /* Note: the base_list[1] contains whether the current opcode has greedy
3082    (represented by a non-zero value) quantifier. This is a different from
3083    other character type lists, which stores here that the character iterator
3084    matches to an empty string (also represented by a non-zero value). */
3085    
3086    for(;;)
3087      {
3088      /* All operations move the code pointer forward.
3089      Therefore infinite recursions are not possible. */
3090    
3091      c = *code;
3092    
3093      /* Skip over callouts */
3094    
3095      if (c == OP_CALLOUT)
3096        {
3097        code += PRIV(OP_lengths)[c];
3098        continue;
3099        }
3100    
3101      if (c == OP_ALT)
3102        {
3103        do code += GET(code, 1); while (*code == OP_ALT);
3104        c = *code;
3105        }
3106    
3107      switch(c)
3108        {
3109        case OP_END:
3110        case OP_KETRPOS:
3111        /* TRUE only in greedy case. The non-greedy case could be replaced by
3112        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3113        uses more memory, which we cannot get at this stage.) */
3114    
3115        return base_list[1] != 0;
3116    
3117        case OP_KET:
3118        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3119        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3120        cannot be converted to a possessive form. */
3121    
3122        if (base_list[1] == 0) return FALSE;
3123    
3124        switch(*(code - GET(code, 1)))
3125          {
3126          case OP_ASSERT:
3127          case OP_ASSERT_NOT:
3128          case OP_ASSERTBACK:
3129          case OP_ASSERTBACK_NOT:
3130          case OP_ONCE:
3131          case OP_ONCE_NC:
3132          /* Atomic sub-patterns and assertions can always auto-possessify their
3133          last iterator. However, if the group was entered as a result of checking
3134          a previous iterator, this is not possible. */
3135    
3136          return !entered_a_group;
3137          }
3138    
3139        code += PRIV(OP_lengths)[c];
3140        continue;
3141    
3142        case OP_ONCE:
3143        case OP_ONCE_NC:
3144        case OP_BRA:
3145        case OP_CBRA:
3146        next_code = code + GET(code, 1);
3147        code += PRIV(OP_lengths)[c];
3148    
3149        while (*next_code == OP_ALT)
3150          {
3151          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3152          code = next_code + 1 + LINK_SIZE;
3153          next_code += GET(next_code, 1);
3154          }
3155    
3156        entered_a_group = TRUE;
3157        continue;
3158    
3159        case OP_BRAZERO:
3160        case OP_BRAMINZERO:
3161    
3162        next_code = code + 1;
3163        if (*next_code != OP_BRA && *next_code != OP_CBRA
3164            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3165    
3166        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3167    
3168        /* The bracket content will be checked by the
3169        OP_BRA/OP_CBRA case above. */
3170        next_code += 1 + LINK_SIZE;
3171        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3172          return FALSE;
3173    
3174        code += PRIV(OP_lengths)[c];
3175        continue;
3176    
3177        default:
3178        break;
3179        }
3180    
3181      /* Check for a supported opcode, and load its properties. */
3182    
3183      code = get_chr_property_list(code, utf, cd->fcc, list);
3184      if (code == NULL) return FALSE;    /* Unsupported */
3185    
3186      /* If either opcode is a small character list, set pointers for comparing
3187      characters from that list with another list, or with a property. */
3188    
3189      if (base_list[0] == OP_CHAR)
3190        {
3191        chr_ptr = base_list + 2;
3192        list_ptr = list;
3193        }
3194      else if (list[0] == OP_CHAR)
3195        {
3196        chr_ptr = list + 2;
3197        list_ptr = base_list;
3198        }
3199    
3200      /* Character bitsets can also be compared to certain opcodes. */
3201    
3202      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3203    #ifdef COMPILE_PCRE8
3204          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3205          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3206    #endif
3207          )
3208        {
3209    #ifdef COMPILE_PCRE8
3210        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3211    #else
3212        if (base_list[0] == OP_CLASS)
3213    #endif
3214          {
3215          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3216          list_ptr = list;
3217          }
3218        else
3219          {
3220          set1 = (pcre_uint8 *)(code - list[2]);
3221          list_ptr = base_list;
3222          }
3223    
3224        invert_bits = FALSE;
3225        switch(list_ptr[0])
3226          {
3227          case OP_CLASS:
3228          case OP_NCLASS:
3229          set2 = (pcre_uint8 *)
3230            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3231          break;
3232    
3233    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3234          case OP_XCLASS:
3235          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3236          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3237          if ((*xclass_flags & XCL_MAP) == 0)
3238            {
3239            /* No bits are set for characters < 256. */
3240            if (list[1] == 0) return TRUE;
3241            /* Might be an empty repeat. */
3242            continue;
3243            }
3244          set2 = (pcre_uint8 *)(xclass_flags + 1);
3245          break;
3246    #endif
3247    
3248          case OP_NOT_DIGIT:
3249          invert_bits = TRUE;
3250          /* Fall through */
3251          case OP_DIGIT:
3252          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3253          break;
3254    
3255          case OP_NOT_WHITESPACE:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_WHITESPACE:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3260          break;
3261    
3262          case OP_NOT_WORDCHAR:
3263          invert_bits = TRUE;
3264          /* Fall through */
3265          case OP_WORDCHAR:
3266          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3267          break;
3268    
3269          default:
3270          return FALSE;
3271          }
3272    
3273        /* Because the sets are unaligned, we need
3274        to perform byte comparison here. */
3275        set_end = set1 + 32;
3276        if (invert_bits)
3277          {
3278          do
3279            {
3280            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3281            }
3282          while (set1 < set_end);
3283          }
3284        else
3285          {
3286          do
3287            {
3288            if ((*set1++ & *set2++) != 0) return FALSE;
3289            }
3290          while (set1 < set_end);
3291          }
3292    
3293        if (list[1] == 0) return TRUE;
3294        /* Might be an empty repeat. */
3295        continue;
3296        }
3297    
3298      /* Some property combinations also acceptable. Unicode property opcodes are
3299      processed specially; the rest can be handled with a lookup table. */
3300    
3301      else
3302        {
3303        pcre_uint32 leftop, rightop;
3304    
3305        leftop = base_list[0];
3306        rightop = list[0];
3307    
3308    #ifdef SUPPORT_UCP
3309        accepted = FALSE; /* Always set in non-unicode case. */
3310        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3311          {
3312          if (rightop == OP_EOD)
3313            accepted = TRUE;
3314          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3315            {
3316            int n;
3317            const pcre_uint8 *p;
3318            BOOL same = leftop == rightop;
3319            BOOL lisprop = leftop == OP_PROP;
3320            BOOL risprop = rightop == OP_PROP;
3321            BOOL bothprop = lisprop && risprop;
3322    
3323            /* There's a table that specifies how each combination is to be
3324            processed:
3325              0   Always return FALSE (never auto-possessify)
3326              1   Character groups are distinct (possessify if both are OP_PROP)
3327              2   Check character categories in the same group (general or particular)
3328              3   Return TRUE if the two opcodes are not the same
3329              ... see comments below
3330            */
3331    
3332            n = propposstab[base_list[2]][list[2]];
3333            switch(n)
3334              {
3335              case 0: break;
3336              case 1: accepted = bothprop; break;
3337              case 2: accepted = (base_list[3] == list[3]) != same; break;
3338              case 3: accepted = !same; break;
3339    
3340              case 4:  /* Left general category, right particular category */
3341              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3342              break;
3343    
3344              case 5:  /* Right general category, left particular category */
3345              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3346              break;
3347    
3348              /* This code is logically tricky. Think hard before fiddling with it.
3349              The posspropstab table has four entries per row. Each row relates to
3350              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3351              Only WORD actually needs all four entries, but using repeats for the
3352              others means they can all use the same code below.
3353    
3354              The first two entries in each row are Unicode general categories, and
3355              apply always, because all the characters they include are part of the
3356              PCRE character set. The third and fourth entries are a general and a
3357              particular category, respectively, that include one or more relevant
3358              characters. One or the other is used, depending on whether the check
3359              is for a general or a particular category. However, in both cases the
3360              category contains more characters than the specials that are defined
3361              for the property being tested against. Therefore, it cannot be used
3362              in a NOTPROP case.
3363    
3364              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3365              Underscore is covered by ucp_P or ucp_Po. */
3366    
3367              case 6:  /* Left alphanum vs right general category */
3368              case 7:  /* Left space vs right general category */
3369              case 8:  /* Left word vs right general category */
3370              p = posspropstab[n-6];
3371              accepted = risprop && lisprop ==
3372                (list[3] != p[0] &&
3373                 list[3] != p[1] &&
3374                (list[3] != p[2] || !lisprop));
3375              break;
3376    
3377              case 9:   /* Right alphanum vs left general category */
3378              case 10:  /* Right space vs left general category */
3379              case 11:  /* Right word vs left general category */
3380              p = posspropstab[n-9];
3381              accepted = lisprop && risprop ==
3382                (base_list[3] != p[0] &&
3383                 base_list[3] != p[1] &&
3384                (base_list[3] != p[2] || !risprop));
3385              break;
3386    
3387              case 12:  /* Left alphanum vs right particular category */
3388              case 13:  /* Left space vs right particular category */
3389              case 14:  /* Left word vs right particular category */
3390              p = posspropstab[n-12];
3391              accepted = risprop && lisprop ==
3392                (catposstab[p[0]][list[3]] &&
3393                 catposstab[p[1]][list[3]] &&
3394                (list[3] != p[3] || !lisprop));
3395              break;
3396    
3397              case 15:  /* Right alphanum vs left particular category */
3398              case 16:  /* Right space vs left particular category */
3399              case 17:  /* Right word vs left particular category */
3400              p = posspropstab[n-15];
3401              accepted = lisprop && risprop ==
3402                (catposstab[p[0]][base_list[3]] &&
3403                 catposstab[p[1]][base_list[3]] &&
3404                (base_list[3] != p[3] || !risprop));
3405              break;
3406              }
3407            }
3408          }
3409    
3410        else
3411    #endif  /* SUPPORT_UCP */
3412    
3413        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3414               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3415               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3416    
3417        if (!accepted) return FALSE;
3418    
3419        if (list[1] == 0) return TRUE;
3420        /* Might be an empty repeat. */
3421        continue;
3422        }
3423    
3424      /* Control reaches here only if one of the items is a small character list.
3425      All characters are checked against the other side. */
3426    
3427      do
3428        {
3429        chr = *chr_ptr;
3430    
3431        switch(list_ptr[0])
3432          {
3433          case OP_CHAR:
3434          ochr_ptr = list_ptr + 2;
3435          do
3436            {
3437            if (chr == *ochr_ptr) return FALSE;
3438            ochr_ptr++;
3439            }
3440          while(*ochr_ptr != NOTACHAR);
3441          break;
3442    
3443          case OP_NOT:
3444          ochr_ptr = list_ptr + 2;
3445          do
3446            {
3447            if (chr == *ochr_ptr)
3448              break;
3449            ochr_ptr++;
3450            }
3451          while(*ochr_ptr != NOTACHAR);
3452          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3453          break;
3454    
3455          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3456          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3457    
3458          case OP_DIGIT:
3459          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3460          break;
3461    
3462          case OP_NOT_DIGIT:
3463          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3464          break;
3465    
3466          case OP_WHITESPACE:
3467          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3468          break;
3469    
3470          case OP_NOT_WHITESPACE:
3471          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3472          break;
3473    
3474          case OP_WORDCHAR:
3475          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3476          break;
3477    
3478          case OP_NOT_WORDCHAR:
3479          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3480          break;
3481    
3482          case OP_HSPACE:
3483          switch(chr)
3484            {
3485            HSPACE_CASES: return FALSE;
3486            default: break;
3487            }
3488          break;
3489    
3490          case OP_NOT_HSPACE:
3491          switch(chr)
3492            {
3493            HSPACE_CASES: break;
3494            default: return FALSE;
3495            }
3496          break;
3497    
3498          case OP_ANYNL:
3499          case OP_VSPACE:
3500          switch(chr)
3501            {
3502            VSPACE_CASES: return FALSE;
3503            default: break;
3504            }
3505          break;
3506    
3507          case OP_NOT_VSPACE:
3508          switch(chr)
3509            {
3510            VSPACE_CASES: break;
3511            default: return FALSE;
3512            }
3513          break;
3514    
3515          case OP_DOLL:
3516          case OP_EODN:
3517          switch (chr)
3518            {
3519            case CHAR_CR:
3520            case CHAR_LF:
3521            case CHAR_VT:
3522            case CHAR_FF:
3523            case CHAR_NEL:
3524    #ifndef EBCDIC
3525            case 0x2028:
3526            case 0x2029:
3527    #endif  /* Not EBCDIC */
3528            return FALSE;
3529            }
3530          break;
3531    
3532          case OP_EOD:    /* Can always possessify before \z */
3533          break;
3534    
3535    #ifdef SUPPORT_UCP
3536          case OP_PROP:
3537          case OP_NOTPROP:
3538          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3539                list_ptr[0] == OP_NOTPROP))
3540            return FALSE;
3541          break;
3542    #endif
3543    
3544          case OP_NCLASS:
3545          if (chr > 255) return FALSE;
3546          /* Fall through */
3547    
3548          case OP_CLASS:
3549          if (chr > 255) break;
3550          class_bitset = (pcre_uint8 *)
3551            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3552          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3553          break;
3554    
3555    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3556          case OP_XCLASS:
3557          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3558              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3559          break;
3560    #endif
3561    
3562          default:
3563          return FALSE;
3564          }
3565    
3566        chr_ptr++;
3567        }
3568      while(*chr_ptr != NOTACHAR);
3569    
3570      /* At least one character must be matched from this opcode. */
3571    
3572      if (list[1] == 0) return TRUE;
3573      }
3574    
3575    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3576    but some compilers complain about an unreachable statement. */
3577    
3578    }
3579    
3580    
3581    
3582    /*************************************************
3583    *    Scan compiled regex for auto-possession     *
3584    *************************************************/
3585    
3586    /* Replaces single character iterations with their possessive alternatives
3587    if appropriate. This function modifies the compiled opcode!
3588    
3589    Arguments:
3590      code        points to start of the byte code
3591      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3592      cd          static compile data
3593    
3594    Returns:      nothing
3595    */
3596    
3597    static void
3598    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3599    {
3600    register pcre_uchar c;
3601    const pcre_uchar *end;
3602    pcre_uchar *repeat_opcode;
3603    pcre_uint32 list[8];
3604    
3605    for (;;)
3606      {
3607      c = *code;
3608    
3609      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3610        {
3611        c -= get_repeat_base(c) - OP_STAR;
3612        end = (c <= OP_MINUPTO) ?
3613          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3614        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3615    
3616        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3617          {
3618          switch(c)
3619            {
3620            case OP_STAR:
3621            *code += OP_POSSTAR - OP_STAR;
3622            break;
3623    
3624            case OP_MINSTAR:
3625            *code += OP_POSSTAR - OP_MINSTAR;
3626            break;
3627    
3628            case OP_PLUS:
3629            *code += OP_POSPLUS - OP_PLUS;
3630            break;
3631    
3632            case OP_MINPLUS:
3633            *code += OP_POSPLUS - OP_MINPLUS;
3634            break;
3635    
3636            case OP_QUERY:
3637            *code += OP_POSQUERY - OP_QUERY;
3638            break;
3639    
3640            case OP_MINQUERY:
3641            *code += OP_POSQUERY - OP_MINQUERY;
3642            break;
3643    
3644            case OP_UPTO:
3645            *code += OP_POSUPTO - OP_UPTO;
3646            break;
3647    
3648            case OP_MINUPTO:
3649            *code += OP_POSUPTO - OP_MINUPTO;
3650            break;
3651            }
3652          }
3653        c = *code;
3654        }
3655      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3656        {
3657    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3658        if (c == OP_XCLASS)
3659          repeat_opcode = code + GET(code, 1);
3660        else
3661    #endif
3662          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3663    
3664        c = *repeat_opcode;
3665        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3666          {
3667          /* end must not be NULL. */
3668          end = get_chr_property_list(code, utf, cd->fcc, list);
3669    
3670          list[1] = (c & 1) == 0;
3671    
3672          if (compare_opcodes(end, utf, cd, list, end))
3673            {
3674            switch (c)
3675              {
3676              case OP_CRSTAR:
3677              case OP_CRMINSTAR:
3678              *repeat_opcode = OP_CRPOSSTAR;
3679              break;
3680    
3681              case OP_CRPLUS:
3682              case OP_CRMINPLUS:
3683              *repeat_opcode = OP_CRPOSPLUS;
3684              break;
3685    
3686              case OP_CRQUERY:
3687              case OP_CRMINQUERY:
3688              *repeat_opcode = OP_CRPOSQUERY;
3689              break;
3690    
3691              case OP_CRRANGE:
3692              case OP_CRMINRANGE:
3693              *repeat_opcode = OP_CRPOSRANGE;
3694              break;
3695              }
3696            }
3697          }
3698        c = *code;
3699        }
3700    
3701      switch(c)
3702        {
3703        case OP_END:
3704        return;
3705    
3706        case OP_TYPESTAR:
3707        case OP_TYPEMINSTAR:
3708        case OP_TYPEPLUS:
3709        case OP_TYPEMINPLUS:
3710        case OP_TYPEQUERY:
3711        case OP_TYPEMINQUERY:
3712        case OP_TYPEPOSSTAR:
3713        case OP_TYPEPOSPLUS:
3714        case OP_TYPEPOSQUERY:
3715        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3716        break;
3717    
3718        case OP_TYPEUPTO:
3719        case OP_TYPEMINUPTO:
3720        case OP_TYPEEXACT:
3721        case OP_TYPEPOSUPTO:
3722        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3723          code += 2;
3724        break;
3725    
3726    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3727        case OP_XCLASS:
3728        code += GET(code, 1);
3729        break;
3730    #endif
3731    
3732        case OP_MARK:
3733        case OP_PRUNE_ARG:
3734        case OP_SKIP_ARG:
3735        case OP_THEN_ARG:
3736        code += code[1];
3737        break;
3738        }
3739    
3740      /* Add in the fixed length from the table */
3741    
3742      code += PRIV(OP_lengths)[c];
3743    
3744      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3745      a multi-byte character. The length in the table is a minimum, so we have to
3746      arrange to skip the extra bytes. */
3747    
3748    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3749      if (utf) switch(c)
3750        {
3751        case OP_CHAR:
3752        case OP_CHARI:
3753        case OP_NOT:
3754        case OP_NOTI:
3755        case OP_STAR:
3756        case OP_MINSTAR:
3757        case OP_PLUS:
3758        case OP_MINPLUS:
3759        case OP_QUERY:
3760        case OP_MINQUERY:
3761        case OP_UPTO:
3762        case OP_MINUPTO:
3763        case OP_EXACT:
3764        case OP_POSSTAR:
3765        case OP_POSPLUS:
3766        case OP_POSQUERY:
3767        case OP_POSUPTO:
3768        case OP_STARI:
3769        case OP_MINSTARI:
3770        case OP_PLUSI:
3771        case OP_MINPLUSI:
3772        case OP_QUERYI:
3773        case OP_MINQUERYI:
3774        case OP_UPTOI:
3775        case OP_MINUPTOI:
3776        case OP_EXACTI:
3777        case OP_POSSTARI:
3778        case OP_POSPLUSI:
3779        case OP_POSQUERYI:
3780        case OP_POSUPTOI:
3781        case OP_NOTSTAR:
3782        case OP_NOTMINSTAR:
3783        case OP_NOTPLUS:
3784        case OP_NOTMINPLUS:
3785        case OP_NOTQUERY:
3786        case OP_NOTMINQUERY:
3787        case OP_NOTUPTO:
3788        case OP_NOTMINUPTO:
3789        case OP_NOTEXACT:
3790        case OP_NOTPOSSTAR:
3791        case OP_NOTPOSPLUS:
3792        case OP_NOTPOSQUERY:
3793        case OP_NOTPOSUPTO:
3794        case OP_NOTSTARI:
3795        case OP_NOTMINSTARI:
3796        case OP_NOTPLUSI:
3797        case OP_NOTMINPLUSI:
3798        case OP_NOTQUERYI:
3799        case OP_NOTMINQUERYI:
3800        case OP_NOTUPTOI:
3801        case OP_NOTMINUPTOI:
3802        case OP_NOTEXACTI:
3803        case OP_NOTPOSSTARI:
3804        case OP_NOTPOSPLUSI:
3805        case OP_NOTPOSQUERYI:
3806        case OP_NOTPOSUPTOI:
3807        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3808        break;
3809        }
3810    #else
3811      (void)(utf);  /* Keep compiler happy by referencing function argument */
3812    #endif
3813      }
3814    }
3815    
3816    
3817    
3818    /*************************************************
3819    *           Check for POSIX class syntax         *
3820    *************************************************/
3821    
3822    /* This function is called when the sequence "[:" or "[." or "[=" is
3823    encountered in a character class. It checks whether this is followed by a
3824    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3825    reach an unescaped ']' without the special preceding character, return FALSE.
3826    
3827    Originally, this function only recognized a sequence of letters between the
3828    terminators, but it seems that Perl recognizes any sequence of characters,
3829    though of course unknown POSIX names are subsequently rejected. Perl gives an
3830    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3831    didn't consider this to be a POSIX class. Likewise for [:1234:].
3832    
3833    The problem in trying to be exactly like Perl is in the handling of escapes. We
3834    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3835    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3836    below handles the special case of \], but does not try to do any other escape
3837    processing. This makes it different from Perl for cases such as [:l\ower:]
3838    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3839    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3840    I think.
3841    
3842    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3843    It seems that the appearance of a nested POSIX class supersedes an apparent
3844    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3845    a digit.
3846    
3847    In Perl, unescaped square brackets may also appear as part of class names. For
3848    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3849    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3850    seem right at all. PCRE does not allow closing square brackets in POSIX class
3851    names.
3852    
3853  Arguments:  Arguments:
3854    code           current code pointer    ptr      pointer to the initial [
3855    ptr            current pattern pointer    endptr   where to return the end pointer
   cd             pointers to tables etc  
3856    
3857  Returns:         new code pointer  Returns:   TRUE or FALSE
3858  */  */
3859    
3860  static pcre_uchar *  static BOOL
3861  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3862  {  {
3863  *code++ = OP_CALLOUT;  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3864  *code++ = 255;  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3865  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  for (++ptr; *ptr != CHAR_NULL; ptr++)
3866  PUT(code, LINK_SIZE, 0);                       /* Default length */    {
3867  return code + 2 * LINK_SIZE;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3868        ptr++;
3869      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3870      else
3871        {
3872        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3873          {
3874          *endptr = ptr;
3875          return TRUE;
3876          }
3877        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3878             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3879              ptr[1] == CHAR_EQUALS_SIGN) &&
3880            check_posix_syntax(ptr, endptr))
3881          return FALSE;
3882        }
3883      }
3884    return FALSE;
3885  }  }
3886    
3887    
3888    
3889    
3890  /*************************************************  /*************************************************
3891  *         Complete a callout item                *  *          Check POSIX class name                *
3892  *************************************************/  *************************************************/
3893    
3894  /* A callout item contains the length of the next item in the pattern, which  /* This function is called to check the name given in a POSIX-style class entry
3895  we can't fill in till after we have reached the relevant point. This is used  such as [:alnum:].
 for both automatic and manual callouts.  
3896    
3897  Arguments:  Arguments:
3898    previous_callout   points to previous callout item    ptr        points to the first letter
3899    ptr                current pattern pointer    len        the length of the name
   cd                 pointers to tables etc  
3900    
3901  Returns:             nothing  Returns:     a value representing the name, or -1 if unknown
3902  */  */
3903    
3904  static void  static int
3905  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  check_posix_name(const pcre_uchar *ptr, int len)
3906  {  {
3907  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  const char *pn = posix_names;
3908  PUT(previous_callout, 2 + LINK_SIZE, length);  register int yield = 0;
3909    while (posix_name_lengths[yield] != 0)
3910      {
3911      if (len == posix_name_lengths[yield] &&
3912        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3913      pn += posix_name_lengths[yield] + 1;
3914      yield++;
3915      }
3916    return -1;
3917  }  }
3918    
3919    
   
 #ifdef SUPPORT_UCP  
3920  /*************************************************  /*************************************************
3921  *           Get othercase range                  *  *    Adjust OP_RECURSE items in repeated group   *
3922  *************************************************/  *************************************************/
3923    
3924  /* This function is passed the start and end of a class range, in UTF-8 mode  /* OP_RECURSE items contain an offset from the start of the regex to the group
3925  with UCP support. It searches up the characters, looking for internal ranges of  that is referenced. This means that groups can be replicated for fixed
3926  characters in the "other" case. Each call returns the next one, updating the  repetition simply by copying (because the recursion is allowed to refer to
3927  start address.  earlier groups that are outside the current group). However, when a group is
3928    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3929    inserted before it, after it has been compiled. This means that any OP_RECURSE
3930    items within it that refer to the group itself or any contained groups have to
3931    have their offsets adjusted. That one of the jobs of this function. Before it
3932    is called, the partially compiled regex must be temporarily terminated with
3933    OP_END.
3934    
3935    This function has been extended with the possibility of forward references for
3936    recursions and subroutine calls. It must also check the list of such references
3937    for the group we are dealing with. If it finds that one of the recursions in
3938    the current group is on this list, it adjusts the offset in the list, not the
3939    value in the reference (which is a group number).
3940    
3941  Arguments:  Arguments:
3942    cptr        points to starting character value; updated    group      points to the start of the group
3943    d           end value    adjust     the amount by which the group is to be moved
3944    ocptr       where to put start of othercase range    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3945    odptr       where to put end of othercase range    cd         contains pointers to tables etc.
3946      save_hwm   the hwm forward reference pointer at the start of the group
3947    
3948  Yield:        TRUE when range returned; FALSE when no more  Returns:     nothing
3949  */  */
3950    
3951  static BOOL  static void
3952  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3953    unsigned int *odptr)    pcre_uchar *save_hwm)
3954  {  {
3955  unsigned int c, othercase, next;  pcre_uchar *ptr = group;
3956    
3957  for (c = *cptr; c <= d; c++)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3958    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
3959      int offset;
3960      pcre_uchar *hc;
3961    
3962  if (c > d) return FALSE;    /* See if this recursion is on the forward reference list. If so, adjust the
3963      reference. */
3964    
3965  *ocptr = othercase;    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3966  next = othercase + 1;      {
3967        offset = (int)GET(hc, 0);
3968        if (cd->start_code + offset == ptr + 1)
3969          {
3970          PUT(hc, 0, offset + adjust);
3971          break;
3972          }
3973        }
3974    
3975  for (++c; c <= d; c++)    /* Otherwise, adjust the recursion offset if it's after the start of this
3976    {    group. */
   if (UCD_OTHERCASE(c) != next) break;  
   next++;  
   }  
3977    
3978  *odptr = next - 1;    if (hc >= cd->hwm)
3979  *cptr = c;      {
3980        offset = (int)GET(ptr, 1);
3981        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3982        }
3983    
3984  return TRUE;    ptr += 1 + LINK_SIZE;
3985      }
3986  }  }
3987    
3988    
3989    
3990  /*************************************************  /*************************************************
3991  *        Check a character and a property        *  *        Insert an automatic callout point       *
3992  *************************************************/  *************************************************/
3993    
3994  /* This function is called by check_auto_possessive() when a property item  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3995  is adjacent to a fixed character.  callout points before each pattern item.
3996    
3997  Arguments:  Arguments:
3998    c            the character    code           current code pointer
3999    ptype        the property type    ptr            current pattern pointer
4000    pdata        the data for the type    cd             pointers to tables etc
   negated      TRUE if it's a negated property (\P or \p{^)  
4001    
4002  Returns:       TRUE if auto-possessifying is OK  Returns:         new code pointer
4003  */  */
4004    
4005  static BOOL  static pcre_uchar *
4006  check_char_prop(int c, int ptype, int pdata, BOOL negated)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4007  {  {
4008  const ucd_record *prop = GET_UCD(c);  *code++ = OP_CALLOUT;
4009  switch(ptype)  *code++ = 255;
4010    {  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4011    case PT_LAMP:  PUT(code, LINK_SIZE, 0);                       /* Default length */
4012    return (prop->chartype == ucp_Lu ||  return code + 2 * LINK_SIZE;
4013            prop->chartype == ucp_Ll ||  }
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
4014    
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
4015    
   case PT_SC:  
   return (pdata == prop->script) == negated;  
4016    
4017    /* These are specials */  /*************************************************
4018    *         Complete a callout item                *
4019    *************************************************/
4020    
4021    case PT_ALNUM:  /* A callout item contains the length of the next item in the pattern, which
4022    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  we can't fill in till after we have reached the relevant point. This is used
4023            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  for both automatic and manual callouts.
4024    
4025    case PT_SPACE:    /* Perl space */  Arguments:
4026    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    previous_callout   points to previous callout item
4027            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)    ptr                current pattern pointer
4028            == negated;    cd                 pointers to tables etc
4029    
4030    case PT_PXSPACE:  /* POSIX space */  Returns:             nothing
4031    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  */
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
4032    
4033    case PT_WORD:  static void
4034    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4035            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  {
4036            c == CHAR_UNDERSCORE) == negated;  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4037    }  PUT(previous_callout, 2 + LINK_SIZE, length);
 return FALSE;  
4038  }  }
 #endif  /* SUPPORT_UCP */  
4039    
4040    
4041    
4042    #ifdef SUPPORT_UCP
4043  /*************************************************  /*************************************************
4044  *     Check if auto-possessifying is possible    *  *           Get othercase range                  *
4045  *************************************************/  *************************************************/
4046    
4047  /* This function is called for unlimited repeats of certain items, to see  /* This function is passed the start and end of a class range, in UTF-8 mode
4048  whether the next thing could possibly match the repeated item. If not, it makes  with UCP support. It searches up the characters, looking for ranges of
4049  sense to automatically possessify the repeated item.  characters in the "other" case. Each call returns the next one, updating the
4050    start address. A character with multiple other cases is returned on its own
4051    with a special return value.
4052    
4053  Arguments:  Arguments:
4054    previous      pointer to the repeated opcode    cptr        points to starting character value; updated
4055    utf           TRUE in UTF-8 / UTF-16 mode    d           end value
4056    ptr           next character in pattern    ocptr       where to put start of othercase range
4057    options       options bits    odptr       where to put end of othercase range
   cd            contains pointers to tables etc.  
4058    
4059  Returns:        TRUE if possessifying is wanted  Yield:        -1 when no more
4060                   0 when a range is returned
4061                  >0 the CASESET offset for char with multiple other cases
4062                    in this case, ocptr contains the original
4063  */  */
4064    
4065  static BOOL  static int
4066  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4067    const pcre_uchar *ptr, int options, compile_data *cd)    pcre_uint32 *odptr)
4068  {  {
4069  pcre_int32 c, next;  pcre_uint32 c, othercase, next;
4070  int op_code = *previous++;  unsigned int co;
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
4071    
4072  /* If the next item is one that we can handle, get its value. A non-negative  /* Find the first character that has an other case. If it has multiple other
4073  value is a character, a negative value is an escape value. */  cases, return its case offset value. */
4074    
4075  if (*ptr == CHAR_BACKSLASH)  for (c = *cptr; c <= d; c++)
   {  
   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)  
4076    {    {
4077  #ifdef SUPPORT_UTF    if ((co = UCD_CASESET(c)) != 0)
4078    if (utf) { GETCHARINC(next, ptr); } else      {
4079  #endif      *ocptr = c++;   /* Character that has the set */
4080    next = *ptr++;      *cptr = c;      /* Rest of input range */
4081        return (int)co;
4082        }
4083      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4084    }    }
 else return FALSE;  
4085    
4086  /* Skip whitespace and comments in extended mode */  if (c > d) return -1;  /* Reached end of range */
4087    
4088  if ((options & PCRE_EXTENDED) != 0)  /* Found a character that has a single other case. Search for the end of the
4089    range, which is either the end of the input range, or a character that has zero
4090    or more than one other cases. */
4091    
4092    *ocptr = othercase;
4093    next = othercase + 1;
4094    
4095    for (++c; c <= d; c++)
4096    {    {
4097    for (;;)    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4098      {    next++;
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
4099    }    }
4100    
4101  /* If the next thing is itself optional, we have to give up. */  *odptr = next - 1;     /* End of othercase range */
4102    *cptr = c;             /* Rest of input range */
4103    return 0;
4104    }
4105    #endif  /* SUPPORT_UCP */
4106    
 if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
4107    
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
4108    
4109  if (next >= 0) switch(op_code)  /*************************************************
4110    {  *        Add a character or range to a class     *
4111    case OP_CHAR:  *************************************************/
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
4112    
4113    /* For CHARI (caseless character) we must check the other case. If we have  /* This function packages up the logic of adding a character or range of
4114    Unicode property support, we can use it to test the other case of  characters to a class. The character values in the arguments will be within the
4115    high-valued characters. */  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4116    mutually recursive with the function immediately below.
4117    
4118    case OP_CHARI:  Arguments:
4119  #ifdef SUPPORT_UTF    classbits     the bit map for characters < 256
4120    GETCHARTEST(c, previous);    uchardptr     points to the pointer for extra data
4121  #else    options       the options word
4122    c = *previous;    cd            contains pointers to tables etc.
4123  #endif    start         start of range character
4124    if (c == next) return FALSE;    end           end of range character
 #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;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */  
4125    
4126    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These  Returns:        the number of < 256 characters added
4127    opcodes are not used for multi-byte characters, because they are coded using                  the pointer to extra data is updated
4128    an XCLASS instead. */  */
4129    
4130    case OP_NOT:  static int
4131    return (c = *previous) == next;  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4132      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4133    {
4134    pcre_uint32 c;
4135    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4136    int n8 = 0;
4137    
4138    /* If caseless matching is required, scan the range and process alternate
4139    cases. In Unicode, there are 8-bit characters that have alternate cases that
4140    are greater than 255 and vice-versa. Sometimes we can just extend the original
4141    range. */
4142    
4143    case OP_NOTI:  if ((options & PCRE_CASELESS) != 0)
4144    if ((c = *previous) == next) return TRUE;    {
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
4145  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4146      othercase = UCD_OTHERCASE(next);    if ((options & PCRE_UTF8) != 0)
4147  #else      {
4148      othercase = NOTACHAR;      int rc;
4149  #endif      pcre_uint32 oc, od;
     return (unsigned int)c == othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c == TABLE_GET(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. */  
4150    
4151    case OP_DIGIT:      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4152    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      c = start;
4153    
4154    case OP_NOT_DIGIT:      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4155    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;        {
4156          /* Handle a single character that has more than one other case. */
4157    
4158    case OP_WHITESPACE:        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4159    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;          PRIV(ucd_caseless_sets) + rc, oc);
4160    
4161    case OP_NOT_WHITESPACE:        /* Do nothing if the other case range is within the original range. */
   return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;  
4162    
4163    case OP_WORDCHAR:        else if (oc >= start && od <= end) continue;
   return next > 127 || (cd->ctypes[next] & ctype_word) == 0;  
4164    
4165    case OP_NOT_WORDCHAR:        /* Extend the original range if there is overlap, noting that if oc < c, we
4166    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;        can't have od > end because a subrange is always shorter than the basic
4167          range. Otherwise, use a recursive call to add the additional range. */
4168    
4169    case OP_HSPACE:        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4170    case OP_NOT_HSPACE:        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4171    switch(next)        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4172      {        }
     case 0x09:  
     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;  
4173      }      }
4174      else
4175    #endif  /* SUPPORT_UCP */
4176    
4177    case OP_ANYNL:    /* Not UTF-mode, or no UCP */
4178    case OP_VSPACE:  
4179    case OP_NOT_VSPACE:    for (c = start; c <= classbits_end; c++)
   switch(next)  
4180      {      {
4181      case 0x0a:      SETBIT(classbits, cd->fcc[c]);
4182      case 0x0b:      n8++;
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
4183      }      }
   
 #ifdef SUPPORT_UCP  
   case OP_PROP:  
   return check_char_prop(next, previous[0], previous[1], FALSE);  
   
   case OP_NOTPROP:  
   return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
   default:  
   return FALSE;  
4184    }    }
4185    
4186    /* Now handle the original range. Adjust the final value according to the bit
4187    length - this means that the same lists of (e.g.) horizontal spaces can be used
4188    in all cases. */
4189    
4190  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  #if defined COMPILE_PCRE8
 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)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
4191  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4192    GETCHARTEST(c, previous);    if ((options & PCRE_UTF8) == 0)
 #else  
   c = *previous;  
4193  #endif  #endif
4194    switch(-next)    if (end > 0xff) end = 0xff;
     {  
     case ESC_d:  
     return c > 127 || (cd->ctypes[c] &