/[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 1219 by ph10, Sun Nov 11 18:04:37 2012 UTC revision 1438 by ph10, Fri Jan 10 16:13:10 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 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89      compile_data *, int *);      compile_data *, int *);
90    
91    
# Line 115  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 253  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 268  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 295  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 322  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 375  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 390  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 455  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 487  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 508  static const char error_texts[] = Line 539  static const char error_texts[] =
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550    ;    ;
551    
552  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 647  static const pcre_uint8 ebcdic_chartab[] Line 686  static const pcre_uint8 ebcdic_chartab[]
686  #endif  #endif
687    
688    
689    /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726    
727    /* This table is used to check whether auto-possessification is possible
728    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729    left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
782      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
783      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
784      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
867  /*************************************************  /*************************************************
868  *            Find an error text                  *  *            Find an error text                  *
# Line 674  return s; Line 890  return s;
890  }  }
891    
892    
893    
894  /*************************************************  /*************************************************
895  *           Expand the workspace                 *  *           Expand the workspace                 *
896  *************************************************/  *************************************************/
# Line 751  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 968  return (*p == CHAR_RIGHT_CURLY_BRACKET);
968  *************************************************/  *************************************************/
969    
970  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
971  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
972  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
973  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
974  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979    chptr          points to the data character    chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
# Line 773  Returns:         zero => a data characte Line 989  Returns:         zero => a data characte
989  */  */
990    
991  static int  static int
992  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
994  {  {
995  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 797  Otherwise further processing may be requ Line 1013  Otherwise further processing may be requ
1013  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1014  /* Not alphanumeric */  /* Not alphanumeric */
1015  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1016  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
1017      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018    
1019  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1020  /* Not alphanumeric */  /* Not alphanumeric */
# Line 847  else Line 1064  else
1064            }            }
1065    
1066  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1067          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1068  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1069          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1070  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1071          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1072  #endif  #endif
1073            {            {
1074            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 963  else Line 1180  else
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1188      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1189      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1190      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1191      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195        Inside a character class, \ followed by a digit is always either a literal
1196        8 or 9 or an octal number. */
1197    
1198      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 999  else Line 1220  else
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          escape = -s;          escape = -s;
1226          break;          break;
# Line 1007  else Line 1228  else
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1232      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1241      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1033  else Line 1252  else
1252  #endif  #endif
1253      break;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in 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}. */
1257      If not, { is treated as a data character. */  
1258        case CHAR_o:
1259        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260          {
1261          ptr += 2;
1262          c = 0;
1263          overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265            {
1266            register pcre_uint32 cc = *ptr++;
1267            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268    #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270    #endif
1271            c = (c << 3) + cc - CHAR_0 ;
1272    #if defined COMPILE_PCRE8
1273            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274    #elif defined COMPILE_PCRE16
1275            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278    #endif
1279            }
1280          if (overflow)
1281            {
1282            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283            *errorcodeptr = ERR34;
1284            }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286            {
1287            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290          }
1291        break;
1292    
1293        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296      case CHAR_x:      case CHAR_x:
1297      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1299        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301          {          {
# Line 1058  else Line 1312  else
1312  #endif  #endif
1313            }            }
1314          }          }
1315        break;        }    /* End JavaScript handling */
       }  
1316    
1317      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1320        seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        c = 0;      else
1325        overflow = FALSE;        {
1326        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          register pcre_uint32 cc = *pt++;          ptr += 2;
1329          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1330            overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1337          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1338  #endif  #endif
1339    
1340  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1344          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1345          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346  #endif  #endif
1347    
1348  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1349          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1351          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1353          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354  #endif  #endif
1355          }            }
1356    
1357        if (overflow)          if (overflow)
1358          {            {
1359          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1361          }            }
1362    
1363        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364          {            {
1365          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366          ptr = pt;            }
         break;  
         }  
1367    
1368        /* 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.
1369        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1370        }          \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1378        {        else
1379        pcre_uint32 cc;                          /* Some compilers don't like */          {
1380        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1389        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391  #endif  #endif
1392        }            }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1190  if ((options & PCRE_UCP) != 0 && escape Line 1457  if ((options & PCRE_UCP) != 0 && escape
1457  return escape;  return escape;
1458  }  }
1459    
1460    
1461    
1462  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1463  /*************************************************  /*************************************************
1464  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1211  Returns:         TRUE if the type value Line 1480  Returns:         TRUE if the type value
1480  */  */
1481    
1482  static BOOL  static BOOL
1483  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484    unsigned int *pdataptr, int *errorcodeptr)    unsigned int *pdataptr, int *errorcodeptr)
1485  {  {
1486  pcre_uchar c;  pcre_uchar c;
# Line 1287  return FALSE; Line 1556  return FALSE;
1556    
1557    
1558    
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1356  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
 *  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 / UTF-32 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 != CHAR_NULL; 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 != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != CHAR_NULL) 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)  
       {  
       pcre_uchar 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 == (int)(ptr - thisname) &&  
           STRNCMP_UC_UC(name, thisname, (unsigned int)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) == CHAR_NULL) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == CHAR_NULL) 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 == CHAR_NULL) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == CHAR_NULL) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != CHAR_NULL)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == CHAR_NULL) 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 == CHAR_NULL) 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 / UTF-32 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++ == CHAR_NULL) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1627  *      Find first significant op code            *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
# Line 1690  for (;;) Line 1662  for (;;)
1662    
1663      case OP_CALLOUT:      case OP_CALLOUT:
1664      case OP_CREF:      case OP_CREF:
1665      case OP_NCREF:      case OP_DNCREF:
1666      case OP_RREF:      case OP_RREF:
1667      case OP_NRREF:      case OP_DNRREF:
1668      case OP_DEF:      case OP_DEF:
1669      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1670      break;      break;
# Line 1706  for (;;) Line 1678  for (;;)
1678    
1679    
1680    
   
1681  /*************************************************  /*************************************************
1682  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1683  *************************************************/  *************************************************/
# Line 1830  for (;;) Line 1801  for (;;)
1801      case OP_COMMIT:      case OP_COMMIT:
1802      case OP_CREF:      case OP_CREF:
1803      case OP_DEF:      case OP_DEF:
1804        case OP_DNCREF:
1805        case OP_DNRREF:
1806      case OP_DOLL:      case OP_DOLL:
1807      case OP_DOLLM:      case OP_DOLLM:
1808      case OP_EOD:      case OP_EOD:
1809      case OP_EODN:      case OP_EODN:
1810      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812      case OP_PRUNE:      case OP_PRUNE:
1813      case OP_REVERSE:      case OP_REVERSE:
# Line 1879  for (;;) Line 1850  for (;;)
1850    
1851      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1852      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1853      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854        cc += 2;        cc += 2;
1855      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1856      break;      break;
# Line 1931  for (;;) Line 1902  for (;;)
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918          case OP_CRPOSRANGE:
1919        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1921        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2009  for (;;) Line 1984  for (;;)
1984      case OP_QUERYI:      case OP_QUERYI:
1985      case OP_REF:      case OP_REF:
1986      case OP_REFI:      case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989      case OP_SBRA:      case OP_SBRA:
1990      case OP_SBRAPOS:      case OP_SBRAPOS:
1991      case OP_SCBRA:      case OP_SCBRA:
# Line 2045  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 2122  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102          code += 2;          code += 2;
2103        break;        break;
2104    
2105        case OP_MARK:        case OP_MARK:
2106        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2108        case OP_THEN_ARG:        case OP_THEN_ARG:
2109        code += code[1];        code += code[1];
2110        break;        break;
# Line 2242  for (;;) Line 2215  for (;;)
2215        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2216        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2217        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2218        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219          code += 2;          code += 2;
2220        break;        break;
2221    
2222        case OP_MARK:        case OP_MARK:
2223        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2224        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2225        case OP_THEN_ARG:        case OP_THEN_ARG:
2226        code += code[1];        code += code[1];
2227        break;        break;
# Line 2353  Arguments: Line 2323  Arguments:
2323    endcode     points to where to stop    endcode     points to where to stop
2324    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2326      recurses    chain of recurse_check to catch mutual recursion
2327    
2328  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2329  */  */
2330    
2331    typedef struct recurse_check {
2332      struct recurse_check *prev;
2333      const pcre_uchar *group;
2334    } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2339  {  {
2340  register pcre_uchar c;  register pcre_uchar c;
2341    recurse_check this_recurse;
2342    
2343  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2389  for (code = first_significant_code(code Line 2367  for (code = first_significant_code(code
2367    
2368    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2369      {      {
2370      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371      BOOL empty_branch;      BOOL empty_branch;
2372    
2373      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2374        when called to scan a completed pattern by setting cd->start_workspace to
2375        NULL. */
2376    
2377        if (cd->start_workspace != NULL)
2378          {
2379          const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383          }
2384    
2385        /* If we are scanning a completed pattern, there are no forward references
2386        and all groups are complete. We need to detect whether this is a recursive
2387        call, as otherwise there will be an infinite loop. If it is a recursion,
2388        just skip over it. Simple recursions are easily detected. For mutual
2389        recursions we keep a chain on the stack. */
2390    
2391        else
2392          {
2393          recurse_check *r = recurses;
2394          const pcre_uchar *endgroup = scode;
2395    
2396      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397        if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399      /* Not a forward reference, test for completed backward reference */        for (r = recurses; r != NULL; r = r->prev)
2400            if (r->group == scode) break;
2401          if (r != NULL) continue;   /* Mutual recursion */
2402          }
2403    
2404      empty_branch = FALSE;      /* Completed reference; scan the referenced group, remembering it on the
2405      scode = cd->start_code + GET(code, 1);      stack chain to detect mutual recursions. */
     if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */  
2406    
2407      /* Completed backwards reference */      empty_branch = FALSE;
2408        this_recurse.prev = recurses;
2409        this_recurse.group = scode;
2410    
2411      do      do
2412        {        {
2413        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414          {          {
2415          empty_branch = TRUE;          empty_branch = TRUE;
2416          break;          break;
# Line 2463  for (code = first_significant_code(code Line 2466  for (code = first_significant_code(code
2466        empty_branch = FALSE;        empty_branch = FALSE;
2467        do        do
2468          {          {
2469          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470            empty_branch = TRUE;            empty_branch = TRUE;
2471          code += GET(code, 1);          code += GET(code, 1);
2472          }          }
# Line 2505  for (code = first_significant_code(code Line 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 2521  for (code = first_significant_code(code Line 2528  for (code = first_significant_code(code
2528    
2529      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      case OP_NOTPROP:
2537        case OP_ANYNL:
2538    
2539        case OP_NOT_HSPACE:
2540        case OP_HSPACE:
2541        case OP_NOT_VSPACE:
2542        case OP_VSPACE:
2543      case OP_EXTUNI:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARI:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555      case OP_NOTI:      case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558        case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
2560      case OP_POSPLUS:      case OP_MINPLUSI:
2561      case OP_EXACT:  
2562      case OP_NOTPLUS:      case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2568  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2599      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2600      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2601      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602        code += 2;        code += 2;
2603      break;      break;
2604    
# Line 2582  for (code = first_significant_code(code Line 2612  for (code = first_significant_code(code
2612      return TRUE;      return TRUE;
2613    
2614      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619      case OP_STAR:      case OP_STAR:
2620      case OP_STARI:      case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624      case OP_MINSTAR:      case OP_MINSTAR:
2625      case OP_MINSTARI:      case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629      case OP_POSSTAR:      case OP_POSSTAR:
2630      case OP_POSSTARI:      case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634      case OP_QUERY:      case OP_QUERY:
2635      case OP_QUERYI:      case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639      case OP_MINQUERY:      case OP_MINQUERY:
2640      case OP_MINQUERYI:      case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644      case OP_POSQUERY:      case OP_POSQUERY:
2645      case OP_POSQUERYI:      case OP_POSQUERYI:
2646        case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650      break;      break;
2651    
2652      case OP_UPTO:      case OP_UPTO:
2653      case OP_UPTOI:      case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657      case OP_MINUPTO:      case OP_MINUPTO:
2658      case OP_MINUPTOI:      case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662      case OP_POSUPTO:      case OP_POSUPTO:
2663      case OP_POSUPTOI:      case OP_POSUPTOI:
2664        case OP_NOTPOSUPTO:
2665        case OP_NOTPOSUPTOI:
2666    
2667      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]);
2668      break;      break;
2669  #endif  #endif
# Line 2616  for (code = first_significant_code(code Line 2674  for (code = first_significant_code(code
2674      case OP_MARK:      case OP_MARK:
2675      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2676      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2677      case OP_THEN_ARG:      case OP_THEN_ARG:
2678      code += code[1];      code += code[1];
2679      break;      break;
# Line 2662  could_be_empty(const pcre_uchar *code, c Line 2717  could_be_empty(const pcre_uchar *code, c
2717  {  {
2718  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2719    {    {
2720    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721      return FALSE;      return FALSE;
2722    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2723    }    }
# Line 2672  return TRUE; Line 2727  return TRUE;
2727    
2728    
2729  /*************************************************  /*************************************************
2730  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2731  *************************************************/  *************************************************/
2732    
2733  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2734  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 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  
2735    
2736  Returns:   TRUE or FALSE  Arguments:  c opcode
2737    Returns:    base opcode for the type
2738  */  */
2739    
2740  static BOOL  static pcre_uchar
2741  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2742  {  {
2743  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2744  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745  for (++ptr; *ptr != CHAR_NULL; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2748      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;  
2749  }  }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754  /*************************************************  /*************************************************
2755  *          Check POSIX class name                *  *        Check a character and a property        *
2756  *************************************************/  *************************************************/
2757    
2758  /* 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
2759  such as [:alnum:].  is adjacent to a fixed character.
2760    
2761  Arguments:  Arguments:
2762    ptr        points to the first letter    c            the character
2763    len        the length of the name    ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2768  */  */
2769    
2770  static int  static BOOL
2771  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773  {  {
2774  const char *pn = posix_names;  const pcre_uint32 *p;
2775  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, (unsigned int)len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2776    
2777  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2778  that is referenced. This means that groups can be replicated for fixed    {
2779  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2780  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2781  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2782  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.  
2783    
2784  This function has been extended with the possibility of forward references for    case PT_GC:
2785  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).  
2786    
2787  Arguments:    case PT_PC:
2788    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 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2789    
2790  Returns:     nothing    case PT_SC:
2791  */    return (pdata == prop->script) == negated;
2792    
2793  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;  
2794    
2795  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2796    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2798    
2799    /* 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
2800    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806      {      {
2807      offset = (int)GET(hc, 0);      HSPACE_CASES:
2808      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2809        {      return negated;
2810        PUT(hc, 0, offset + adjust);  
2811        break;      default:
2812        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813      }      }
2814      break;  /* Control never reaches here */
2815    
2816    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2817    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821    if (hc >= cd->hwm)    case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824      {      {
2825      offset = (int)GET(ptr, 1);      if (c < *p) return !negated;
2826      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2827      }      }
2828      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2829    }    }
2830    
2831    return FALSE;
2832  }  }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837  /*************************************************  /*************************************************
2838  *        Insert an automatic callout point       *  *        Fill the character property list        *
2839  *************************************************/  *************************************************/
2840    
2841  /* 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-
2842  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2843    
2844  Arguments:  Arguments:
2845    code           current code pointer    code        points to start of expression
2846    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847    cd             pointers to tables etc    fcc         points to case-flipping table
2848      list        points to output list
2849                  list[0] will be filled with the opcode
2850                  list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852                  list[2..7] depends on the opcode
2853    
2854  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856  */  */
2857    
2858  static pcre_uchar *  static const pcre_uchar *
2859  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861  {  {
2862  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2863  *code++ = 255;  pcre_uchar base;
2864  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uchar *end;
2865  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 chr;
 return code + 2 * LINK_SIZE;  
 }  
2866    
2867    #ifdef SUPPORT_UCP
2868    pcre_uint32 *clist_dest;
2869    const pcre_uint32 *clist_src;
2870    #else
2871    utf = utf;  /* Suppress "unused parameter" compiler warning */
2872    #endif
2873    
2874    list[0] = c;
2875    list[1] = FALSE;
2876    code++;
2877    
2878  /*************************************************  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879  *         Complete a callout item                *    {
2880  *************************************************/    base = get_repeat_base(c);
2881      c -= (base - OP_STAR);
2882    
2883  /* A callout item contains the length of the next item in the pattern, which    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884  we can't fill in till after we have reached the relevant point. This is used      code += IMM2_SIZE;
 for both automatic and manual callouts.  
   
 Arguments:  
   previous_callout   points to previous callout item  
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
   
 Returns:             nothing  
 */  
   
 static void  
 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  
 {  
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
   
   
   
 #ifdef SUPPORT_UCP  
 /*************************************************  
 *           Get othercase range                  *  
 *************************************************/  
2885    
2886  /* This function is passed the start and end of a class range, in UTF-8 mode    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 with UCP support. It searches up the characters, looking for ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
2887    
2888  Arguments:    switch(base)
2889    cptr        points to starting character value; updated      {
2890    d           end value      case OP_STAR:
2891    ocptr       where to put start of othercase range      list[0] = OP_CHAR;
2892    odptr       where to put end of othercase range      break;
2893    
2894  Yield:        -1 when no more      case OP_STARI:
2895                 0 when a range is returned      list[0] = OP_CHARI;
2896                >0 the CASESET offset for char with multiple other cases      break;
                 in this case, ocptr contains the original  
 */  
2897    
2898  static int      case OP_NOTSTAR:
2899  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      list[0] = OP_NOT;
2900    pcre_uint32 *odptr)      break;
 {  
 pcre_uint32 c, othercase, next;  
 unsigned int co;  
2901    
2902  /* Find the first character that has an other case. If it has multiple other      case OP_NOTSTARI:
2903  cases, return its case offset value. */      list[0] = OP_NOTI;
2904        break;
2905    
2906  for (c = *cptr; c <= d; c++)      case OP_TYPESTAR:
2907    {      list[0] = *code;
2908    if ((co = UCD_CASESET(c)) != 0)      code++;
2909      {      break;
2910      *ocptr = c++;   /* Character that has the set */      }
2911      *cptr = c;      /* Rest of input range */    c = list[0];
     return (int)co;  
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
2912    }    }
2913    
2914  if (c > d) return -1;  /* Reached end of range */  switch(c)
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
2915    {    {
2916    if (UCD_OTHERCASE(c) != next) break;    case OP_NOT_DIGIT:
2917    next++;    case OP_DIGIT:
2918    }    case OP_NOT_WHITESPACE:
2919      case OP_WHITESPACE:
2920  *odptr = next - 1;     /* End of othercase range */    case OP_NOT_WORDCHAR:
2921  *cptr = c;             /* Rest of input range */    case OP_WORDCHAR:
2922  return 0;    case OP_ANY:
2923  }    case OP_ALLANY:
2924      case OP_ANYNL:
2925      case OP_NOT_HSPACE:
2926      case OP_HSPACE:
2927  /*************************************************    case OP_NOT_VSPACE:
2928  *        Check a character and a property        *    case OP_VSPACE:
2929  *************************************************/    case OP_EXTUNI:
2930      case OP_EODN:
2931  /* This function is called by check_auto_possessive() when a property item    case OP_EOD:
2932  is adjacent to a fixed character.    case OP_DOLL:
2933      case OP_DOLLM:
2934      return code;
2935    
2936  Arguments:    case OP_CHAR:
2937    c            the character    case OP_NOT:
2938    ptype        the property type    GETCHARINCTEST(chr, code);
2939    pdata        the data for the type    list[2] = chr;
2940    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2941      return code;
2942    
2943  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2944  */    case OP_NOTI:
2945      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946      GETCHARINCTEST(chr, code);
2947      list[2] = chr;
2948    
 static BOOL  
 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  
 {  
2949  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2950  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2951        list[3] = fcc[chr];
2952      else
2953        list[3] = UCD_OTHERCASE(chr);
2954    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955      list[3] = (chr < 256) ? fcc[chr] : chr;
2956    #else
2957      list[3] = fcc[chr];
2958  #endif  #endif
2959    
2960  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
2961    
2962  switch(ptype)    if (chr == list[3])
2963    {      list[3] = NOTACHAR;
2964    case PT_LAMP:    else
2965    return (prop->chartype == ucp_Lu ||      list[4] = NOTACHAR;
2966            prop->chartype == ucp_Ll ||    return code;
           prop->chartype == ucp_Lt) == negated;  
2967    
2968    case PT_GC:  #ifdef SUPPORT_UCP
2969    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;    case OP_PROP:
2970      case OP_NOTPROP:
2971      if (code[0] != PT_CLIST)
2972        {
2973        list[2] = code[0];
2974        list[3] = code[1];
2975        return code + 2;
2976        }
2977    
2978    case PT_PC:    /* Convert only if we have enough space. */
   return (pdata == prop->chartype) == negated;  
2979    
2980    case PT_SC:    clist_src = PRIV(ucd_caseless_sets) + code[1];
2981    return (pdata == prop->script) == negated;    clist_dest = list + 2;
2982      code += 2;
2983    
2984    /* These are specials */    do {
2985         if (clist_dest >= list + 8)
2986           {
2987           /* Early return if there is not enough space. This should never
2988           happen, since all clists are shorter than 5 character now. */
2989           list[2] = code[0];
2990           list[3] = code[1];
2991           return code;
2992           }
2993         *clist_dest++ = *clist_src;
2994         }
2995      while(*clist_src++ != NOTACHAR);
2996    
2997    case PT_ALNUM:    /* All characters are stored. The terminating NOTACHAR
2998    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    is copied form the clist itself. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
2999    
3000    case PT_SPACE:    /* Perl space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
3002            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  #endif
           == negated;  
3003    
3004    case PT_PXSPACE:  /* POSIX space */    case OP_NCLASS:
3005    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    case OP_CLASS:
3006            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007            c == CHAR_FF || c == CHAR_CR)    case OP_XCLASS:
3008            == negated;    if (c == OP_XCLASS)
3009        end = code + GET(code, 0) - 1;
3010      else
3011    #endif
3012        end = code + 32 / sizeof(pcre_uchar);
3013    
3014    case PT_WORD:    switch(*end)
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   
 #ifdef SUPPORT_UCP  
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3015      {      {
3016      if (c < *p) return !negated;      case OP_CRSTAR:
3017      if (c == *p++) return negated;      case OP_CRMINSTAR:
3018        case OP_CRQUERY:
3019        case OP_CRMINQUERY:
3020        case OP_CRPOSSTAR:
3021        case OP_CRPOSQUERY:
3022        list[1] = TRUE;
3023        end++;
3024        break;
3025    
3026        case OP_CRPLUS:
3027        case OP_CRMINPLUS:
3028        case OP_CRPOSPLUS:
3029        end++;
3030        break;
3031    
3032        case OP_CRRANGE:
3033        case OP_CRMINRANGE:
3034        case OP_CRPOSRANGE:
3035        list[1] = (GET2(end, 1) == 0);
3036        end += 1 + 2 * IMM2_SIZE;
3037        break;
3038      }      }
3039    break;  /* Control never reaches here */    list[2] = end - code;
3040  #endif    return end;
3041    }    }
3042    return NULL;    /* Opcode not accepted */
 return FALSE;  
3043  }  }
 #endif  /* SUPPORT_UCP */  
3044    
3045    
3046    
3047  /*************************************************  /*************************************************
3048  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3049  *************************************************/  *************************************************/
3050    
3051  /* This function is called for unlimited repeats of certain items, to see  /* Checks whether the base and the current opcode have a common character, in
3052  whether the next thing could possibly match the repeated item. If not, it makes  which case the base cannot be possessified.
 sense to automatically possessify the repeated item.  
3053    
3054  Arguments:  Arguments:
3055    previous      pointer to the repeated opcode    code        points to the byte code
3056    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057    ptr           next character in pattern    cd          static compile data
3058    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3059    
3060  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3061  */  */
3062    
3063  static BOOL  static BOOL
3064  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066  {  {
3067  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3068  pcre_uint32 next;  pcre_uint32 list[8];
3069  int escape;  const pcre_uint32 *chr_ptr;
3070  pcre_uchar op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3071    const pcre_uint32 *list_ptr;
3072  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3073    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3074    const pcre_uchar *xclass_flags;
3075    #endif
3076    const pcre_uint8 *class_bitset;
3077    const pcre_uint8 *set1, *set2, *set_end;
3078    pcre_uint32 chr;
3079    BOOL accepted, invert_bits;
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  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3087    {    {
3088    for (;;)    /* 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      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3098      if (*ptr == CHAR_NUMBER_SIGN)      continue;
       {  
       ptr++;  
       while (*ptr != CHAR_NULL)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3099      }      }
   }  
3100    
3101  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3102  value is a character, a negative value is an escape value. */      {
3103        do code += GET(code, 1); while (*code == OP_ALT);
3104        c = *code;
3105        }
3106    
3107  if (*ptr == CHAR_BACKSLASH)    switch(c)
3108    {      {
3109    int temperrorcode = 0;      case OP_END:
3110    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);      case OP_KETRPOS:
3111    if (temperrorcode != 0) return FALSE;      /* TRUE only in greedy case. The non-greedy case could be replaced by
3112    ptr++;    /* Point after the escape sequence */      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.) */
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3114    
3115  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3116    
3117  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3118    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3119    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3120      {      cannot be converted to a possessive form. */
3121      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3122      if (*ptr == CHAR_NUMBER_SIGN)      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. */
3134          return TRUE;
3135          }
3136    
3137        code += PRIV(OP_lengths)[c];
3138        continue;
3139    
3140        case OP_ONCE:
3141        case OP_ONCE_NC:
3142        case OP_BRA:
3143        case OP_CBRA:
3144        next_code = code + GET(code, 1);
3145        code += PRIV(OP_lengths)[c];
3146    
3147        while (*next_code == OP_ALT)
3148        {        {
3149        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3150        while (*ptr != CHAR_NULL)        code = next_code + 1 + LINK_SIZE;
3151          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3152        }        }
3153      else break;      continue;
     }  
   }  
3154    
3155  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3156        case OP_BRAMINZERO:
3157    
3158  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3159    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3160      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3161    
3162        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3163    
3164        /* The bracket content will be checked by the
3165        OP_BRA/OP_CBRA case above. */
3166        next_code += 1 + LINK_SIZE;
3167        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3168          return FALSE;
3169    
3170  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3171        continue;
3172        }
3173    
3174  if (op_code == OP_CHAR || op_code == OP_CHARI ||    /* Check for a supported opcode, and load its properties. */
     op_code == OP_NOT || op_code == OP_NOTI)  
   {  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3175    
3176  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3177  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3178    
3179  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3180    {    characters from that list with another list, or with a property. */
3181    /* For a caseless UTF match, the next character may have more than one other  
3182    case, which maps to the special PT_CLIST property. Check this first. */    if (base_list[0] == OP_CHAR)
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3183      {      {
3184      unsigned int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3185      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3186      }      }
3187  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3188      {      {
3189      case OP_CHAR:      chr_ptr = list + 2;
3190      return c != next;      list_ptr = base_list;
3191        }
3192    
3193      /* For CHARI (caseless character) we must check the other case. If we have    /* Character bitsets can also be compared to certain opcodes. */
     Unicode property support, we can use it to test the other case of  
     high-valued characters. We know that next can have only one other case,  
     because multi-other-case characters are dealt with above. */  
3194    
3195      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3196      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3197  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3198      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3199        {  #endif
3200        pcre_uint32 othercase;        )
3201        if (next < 128) othercase = cd->fcc[next]; else      {
3202  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3203        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3204  #else  #else
3205        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3206  #endif  #endif
3207        return c != othercase;        {
3208          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3209          list_ptr = list;
3210        }        }
3211      else      else
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
   
     case OP_NOT:  
     return c == next;  
   
     case OP_NOTI:  
     if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
3212        {        {
3213        pcre_uint32 othercase;        set1 = (pcre_uint8 *)(code - list[2]);
3214        if (next < 128) othercase = cd->fcc[next]; else        list_ptr = base_list;
3215  #ifdef SUPPORT_UCP        }
3216        othercase = UCD_OTHERCASE(next);  
3217  #else      invert_bits = FALSE;
3218        othercase = NOTACHAR;      switch(list_ptr[0])
3219          {
3220          case OP_CLASS:
3221          case OP_NCLASS:
3222          set2 = (pcre_uint8 *)
3223            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3224          break;
3225    
3226    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3227          case OP_XCLASS:
3228          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3229          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3230          if ((*xclass_flags & XCL_MAP) == 0)
3231            {
3232            /* No bits are set for characters < 256. */
3233            if (list[1] == 0) return TRUE;
3234            /* Might be an empty repeat. */
3235            continue;
3236            }
3237          set2 = (pcre_uint8 *)(xclass_flags + 1);
3238          break;
3239  #endif  #endif
3240        return c == othercase;  
3241          case OP_NOT_DIGIT:
3242          invert_bits = TRUE;
3243          /* Fall through */
3244          case OP_DIGIT:
3245          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3246          break;
3247    
3248          case OP_NOT_WHITESPACE:
3249          invert_bits = TRUE;
3250          /* Fall through */
3251          case OP_WHITESPACE:
3252          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3253          break;
3254    
3255          case OP_NOT_WORDCHAR:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_WORDCHAR:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3260          break;
3261    
3262          default:
3263          return FALSE;
3264          }
3265    
3266        /* Because the sets are unaligned, we need
3267        to perform byte comparison here. */
3268        set_end = set1 + 32;
3269        if (invert_bits)
3270          {
3271          do
3272            {
3273            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3274            }
3275          while (set1 < set_end);
3276        }        }
3277      else      else
3278  #endif  /* SUPPORT_UTF */        {
3279      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */        do
3280            {
3281            if ((*set1++ & *set2++) != 0) return FALSE;
3282            }
3283          while (set1 < set_end);
3284          }
3285    
3286      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      if (list[1] == 0) return TRUE;
3287      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      /* Might be an empty repeat. */
3288        continue;
3289        }
3290    
3291      case OP_DIGIT:    /* Some property combinations also acceptable. Unicode property opcodes are
3292      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;    processed specially; the rest can be handled with a lookup table. */
3293    
3294      case OP_NOT_DIGIT:    else
3295      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;      {
3296        pcre_uint32 leftop, rightop;
3297    
3298      case OP_WHITESPACE:      leftop = base_list[0];
3299      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;      rightop = list[0];
3300    
3301      case OP_NOT_WHITESPACE:  #ifdef SUPPORT_UCP
3302      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;      accepted = FALSE; /* Always set in non-unicode case. */
3303        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3304          {
3305          if (rightop == OP_EOD)
3306            accepted = TRUE;
3307          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3308            {
3309            int n;
3310            const pcre_uint8 *p;
3311            BOOL same = leftop == rightop;
3312            BOOL lisprop = leftop == OP_PROP;
3313            BOOL risprop = rightop == OP_PROP;
3314            BOOL bothprop = lisprop && risprop;
3315    
3316            /* There's a table that specifies how each combination is to be
3317            processed:
3318              0   Always return FALSE (never auto-possessify)
3319              1   Character groups are distinct (possessify if both are OP_PROP)
3320              2   Check character categories in the same group (general or particular)
3321              3   Return TRUE if the two opcodes are not the same
3322              ... see comments below
3323            */
3324    
3325            n = propposstab[base_list[2]][list[2]];
3326            switch(n)
3327              {
3328              case 0: break;
3329              case 1: accepted = bothprop; break;
3330              case 2: accepted = (base_list[3] == list[3]) != same; break;
3331              case 3: accepted = !same; break;
3332    
3333      case OP_WORDCHAR:            case 4:  /* Left general category, right particular category */
3334      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3335              break;
3336    
3337      case OP_NOT_WORDCHAR:            case 5:  /* Right general category, left particular category */
3338      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3339              break;
3340    
3341      case OP_HSPACE:            /* This code is logically tricky. Think hard before fiddling with it.
3342      case OP_NOT_HSPACE:            The posspropstab table has four entries per row. Each row relates to
3343      switch(next)            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3344              Only WORD actually needs all four entries, but using repeats for the
3345              others means they can all use the same code below.
3346    
3347              The first two entries in each row are Unicode general categories, and
3348              apply always, because all the characters they include are part of the
3349              PCRE character set. The third and fourth entries are a general and a
3350              particular category, respectively, that include one or more relevant
3351              characters. One or the other is used, depending on whether the check
3352              is for a general or a particular category. However, in both cases the
3353              category contains more characters than the specials that are defined
3354              for the property being tested against. Therefore, it cannot be used
3355              in a NOTPROP case.
3356    
3357              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3358              Underscore is covered by ucp_P or ucp_Po. */
3359    
3360              case 6:  /* Left alphanum vs right general category */
3361              case 7:  /* Left space vs right general category */
3362              case 8:  /* Left word vs right general category */
3363              p = posspropstab[n-6];
3364              accepted = risprop && lisprop ==
3365                (list[3] != p[0] &&
3366                 list[3] != p[1] &&
3367                (list[3] != p[2] || !lisprop));
3368              break;
3369    
3370              case 9:   /* Right alphanum vs left general category */
3371              case 10:  /* Right space vs left general category */
3372              case 11:  /* Right word vs left general category */
3373              p = posspropstab[n-9];
3374              accepted = lisprop && risprop ==
3375                (base_list[3] != p[0] &&
3376                 base_list[3] != p[1] &&
3377                (base_list[3] != p[2] || !risprop));
3378              break;
3379    
3380              case 12:  /* Left alphanum vs right particular category */
3381              case 13:  /* Left space vs right particular category */
3382              case 14:  /* Left word vs right particular category */
3383              p = posspropstab[n-12];
3384              accepted = risprop && lisprop ==
3385                (catposstab[p[0]][list[3]] &&
3386                 catposstab[p[1]][list[3]] &&
3387                (list[3] != p[3] || !lisprop));
3388              break;
3389    
3390              case 15:  /* Right alphanum vs left particular category */
3391              case 16:  /* Right space vs left particular category */
3392              case 17:  /* Right word vs left particular category */
3393              p = posspropstab[n-15];
3394              accepted = lisprop && risprop ==
3395                (catposstab[p[0]][base_list[3]] &&
3396                 catposstab[p[1]][base_list[3]] &&
3397                (base_list[3] != p[3] || !risprop));
3398              break;
3399              }
3400            }
3401          }
3402    
3403        else
3404    #endif  /* SUPPORT_UCP */
3405    
3406        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3407               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3408               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3409    
3410        if (!accepted)
3411          return FALSE;
3412    
3413        if (list[1] == 0) return TRUE;
3414        /* Might be an empty repeat. */
3415        continue;
3416        }
3417    
3418      /* Control reaches here only if one of the items is a small character list.
3419      All characters are checked against the other side. */
3420    
3421      do
3422        {
3423        chr = *chr_ptr;
3424    
3425        switch(list_ptr[0])
3426        {        {
3427        HSPACE_CASES:        case OP_CHAR:
3428        return op_code == OP_NOT_HSPACE;        ochr_ptr = list_ptr + 2;
3429          do
3430            {
3431            if (chr == *ochr_ptr) return FALSE;
3432            ochr_ptr++;
3433            }
3434          while(*ochr_ptr != NOTACHAR);
3435          break;
3436    
3437          case OP_NOT:
3438          ochr_ptr = list_ptr + 2;
3439          do
3440            {
3441            if (chr == *ochr_ptr)
3442              break;
3443            ochr_ptr++;
3444            }
3445          while(*ochr_ptr != NOTACHAR);
3446          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3447          break;
3448    
3449          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3450          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3451    
3452          case OP_DIGIT:
3453          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3454          break;
3455    
3456          case OP_NOT_DIGIT:
3457          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3458          break;
3459    
3460          case OP_WHITESPACE:
3461          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3462          break;
3463    
3464          case OP_NOT_WHITESPACE:
3465          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3466          break;
3467    
3468          case OP_WORDCHAR:
3469          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3470          break;
3471    
3472          case OP_NOT_WORDCHAR:
3473          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3474          break;
3475    
3476          case OP_HSPACE:
3477          switch(chr)
3478            {
3479            HSPACE_CASES: return FALSE;
3480            default: break;
3481            }
3482          break;
3483    
3484          case OP_NOT_HSPACE:
3485          switch(chr)
3486            {
3487            HSPACE_CASES: break;
3488            default: return FALSE;
3489            }
3490          break;
3491    
3492          case OP_ANYNL:
3493          case OP_VSPACE:
3494          switch(chr)
3495            {
3496            VSPACE_CASES: return FALSE;
3497            default: break;
3498            }
3499          break;
3500    
3501          case OP_NOT_VSPACE:
3502          switch(chr)
3503            {
3504            VSPACE_CASES: break;
3505            default: return FALSE;
3506            }
3507          break;
3508    
3509          case OP_DOLL:
3510          case OP_EODN:
3511          switch (chr)
3512            {
3513            case CHAR_CR:
3514            case CHAR_LF:
3515            case CHAR_VT:
3516            case CHAR_FF:
3517            case CHAR_NEL:
3518    #ifndef EBCDIC
3519            case 0x2028:
3520            case 0x2029:
3521    #endif  /* Not EBCDIC */
3522            return FALSE;
3523            }
3524          break;
3525    
3526          case OP_EOD:    /* Can always possessify before \z */
3527          break;
3528    
3529    #ifdef SUPPORT_UCP
3530          case OP_PROP:
3531          case OP_NOTPROP:
3532          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3533                list_ptr[0] == OP_NOTPROP))
3534            return FALSE;
3535          break;
3536    #endif
3537    
3538          case OP_NCLASS:
3539          if (chr > 255) return FALSE;
3540          /* Fall through */
3541    
3542          case OP_CLASS:
3543          if (chr > 255) break;
3544          class_bitset = (pcre_uint8 *)
3545            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3546          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3547          break;
3548    
3549    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3550          case OP_XCLASS:
3551          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3552              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3553          break;
3554    #endif
3555    
3556        default:        default:
3557        return op_code != OP_NOT_HSPACE;        return FALSE;
3558        }        }
3559    
3560      case OP_ANYNL:      chr_ptr++;
3561      case OP_VSPACE:      }
3562      case OP_NOT_VSPACE:    while(*chr_ptr != NOTACHAR);
3563      switch(next)  
3564      /* At least one character must be matched from this opcode. */
3565    
3566      if (list[1] == 0) return TRUE;
3567      }
3568    
3569    return FALSE;
3570    }
3571    
3572    
3573    
3574    /*************************************************
3575    *    Scan compiled regex for auto-possession     *
3576    *************************************************/
3577    
3578    /* Replaces single character iterations with their possessive alternatives
3579    if appropriate. This function modifies the compiled opcode!
3580    
3581    Arguments:
3582      code        points to start of the byte code
3583      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3584      cd          static compile data
3585    
3586    Returns:      nothing
3587    */
3588    
3589    static void
3590    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3591    {
3592    register pcre_uchar c;
3593    const pcre_uchar *end;
3594    pcre_uchar *repeat_opcode;
3595    pcre_uint32 list[8];
3596    
3597    for (;;)
3598      {
3599      c = *code;
3600    
3601      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3602        {
3603        c -= get_repeat_base(c) - OP_STAR;
3604        end = (c <= OP_MINUPTO) ?
3605          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3606        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3607    
3608        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3609        {        {
3610        VSPACE_CASES:        switch(c)
3611        return op_code == OP_NOT_VSPACE;          {
3612            case OP_STAR:
3613            *code += OP_POSSTAR - OP_STAR;
3614            break;
3615    
3616        default:          case OP_MINSTAR:
3617        return op_code != OP_NOT_VSPACE;          *code += OP_POSSTAR - OP_MINSTAR;
3618            break;
3619    
3620            case OP_PLUS:
3621            *code += OP_POSPLUS - OP_PLUS;
3622            break;
3623    
3624            case OP_MINPLUS:
3625            *code += OP_POSPLUS - OP_MINPLUS;
3626            break;
3627    
3628            case OP_QUERY:
3629            *code += OP_POSQUERY - OP_QUERY;
3630            break;
3631    
3632            case OP_MINQUERY:
3633            *code += OP_POSQUERY - OP_MINQUERY;
3634            break;
3635    
3636            case OP_UPTO:
3637            *code += OP_POSUPTO - OP_UPTO;
3638            break;
3639    
3640            case OP_MINUPTO:
3641            *code += OP_MINUPTO - OP_UPTO;
3642            break;
3643            }
3644        }        }
3645        c = *code;
3646        }
3647      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3648        {
3649    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3650        if (c == OP_XCLASS)
3651          repeat_opcode = code + GET(code, 1);
3652        else
3653    #endif
3654          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3655    
3656  #ifdef SUPPORT_UCP      c = *repeat_opcode;
3657      case OP_PROP:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3658      return check_char_prop(next, previous[0], previous[1], FALSE);        {
3659          /* end must not be NULL. */
3660          end = get_chr_property_list(code, utf, cd->fcc, list);
3661    
3662      case OP_NOTPROP:        list[1] = (c & 1) == 0;
3663      return check_char_prop(next, previous[0], previous[1], TRUE);  
3664          if (compare_opcodes(end, utf, cd, list, end))
3665            {
3666            switch (c)
3667              {
3668              case OP_CRSTAR:
3669              case OP_CRMINSTAR:
3670              *repeat_opcode = OP_CRPOSSTAR;
3671              break;
3672    
3673              case OP_CRPLUS:
3674              case OP_CRMINPLUS:
3675              *repeat_opcode = OP_CRPOSPLUS;
3676              break;
3677    
3678              case OP_CRQUERY:
3679              case OP_CRMINQUERY:
3680              *repeat_opcode = OP_CRPOSQUERY;
3681              break;
3682    
3683              case OP_CRRANGE:
3684              case OP_CRMINRANGE:
3685              *repeat_opcode = OP_CRPOSRANGE;
3686              break;
3687              }
3688            }
3689          }
3690        c = *code;
3691        }
3692    
3693      switch(c)
3694        {
3695        case OP_END:
3696        return;
3697    
3698        case OP_TYPESTAR:
3699        case OP_TYPEMINSTAR:
3700        case OP_TYPEPLUS:
3701        case OP_TYPEMINPLUS:
3702        case OP_TYPEQUERY:
3703        case OP_TYPEMINQUERY:
3704        case OP_TYPEPOSSTAR:
3705        case OP_TYPEPOSPLUS:
3706        case OP_TYPEPOSQUERY:
3707        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3708        break;
3709    
3710        case OP_TYPEUPTO:
3711        case OP_TYPEMINUPTO:
3712        case OP_TYPEEXACT:
3713        case OP_TYPEPOSUPTO:
3714        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3715          code += 2;
3716        break;
3717    
3718    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3719        case OP_XCLASS:
3720        code += GET(code, 1);
3721        break;
3722  #endif  #endif
3723    
3724      default:      case OP_MARK:
3725      return FALSE;      case OP_PRUNE_ARG:
3726        case OP_SKIP_ARG:
3727        case OP_THEN_ARG:
3728        code += code[1];
3729        break;
3730      }      }
3731    
3732      /* Add in the fixed length from the table */
3733    
3734      code += PRIV(OP_lengths)[c];
3735    
3736      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3737      a multi-byte character. The length in the table is a minimum, so we have to
3738      arrange to skip the extra bytes. */
3739    
3740    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3741      if (utf) switch(c)
3742        {
3743        case OP_CHAR:
3744        case OP_CHARI:
3745        case OP_NOT:
3746        case OP_NOTI:
3747        case OP_STAR:
3748        case OP_MINSTAR:
3749        case OP_PLUS:
3750        case OP_MINPLUS:
3751        case OP_QUERY:
3752        case OP_MINQUERY:
3753        case OP_UPTO:
3754        case OP_MINUPTO:
3755        case OP_EXACT:
3756        case OP_POSSTAR:
3757        case OP_POSPLUS:
3758        case OP_POSQUERY:
3759        case OP_POSUPTO:
3760        case OP_STARI:
3761        case OP_MINSTARI:
3762        case OP_PLUSI:
3763        case OP_MINPLUSI:
3764        case OP_QUERYI:
3765        case OP_MINQUERYI:
3766        case OP_UPTOI:
3767        case OP_MINUPTOI:
3768        case OP_EXACTI:
3769        case OP_POSSTARI:
3770        case OP_POSPLUSI:
3771        case OP_POSQUERYI:
3772        case OP_POSUPTOI:
3773        case OP_NOTSTAR:
3774        case OP_NOTMINSTAR:
3775        case OP_NOTPLUS:
3776        case OP_NOTMINPLUS:
3777        case OP_NOTQUERY:
3778        case OP_NOTMINQUERY:
3779        case OP_NOTUPTO:
3780        case OP_NOTMINUPTO:
3781        case OP_NOTEXACT:
3782        case OP_NOTPOSSTAR:
3783        case OP_NOTPOSPLUS:
3784        case OP_NOTPOSQUERY:
3785        case OP_NOTPOSUPTO:
3786        case OP_NOTSTARI:
3787        case OP_NOTMINSTARI:
3788        case OP_NOTPLUSI:
3789        case OP_NOTMINPLUSI:
3790        case OP_NOTQUERYI:
3791        case OP_NOTMINQUERYI:
3792        case OP_NOTUPTOI:
3793        case OP_NOTMINUPTOI:
3794        case OP_NOTEXACTI:
3795        case OP_NOTPOSSTARI:
3796        case OP_NOTPOSPLUSI:
3797        case OP_NOTPOSQUERYI:
3798        case OP_NOTPOSUPTOI:
3799        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3800        break;
3801        }
3802    #else
3803      (void)(utf);  /* Keep compiler happy by referencing function argument */
3804    #endif
3805    }    }
3806    }
3807    
 /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  
 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. */  
3808    
3809  switch(op_code)  
3810    /*************************************************
3811    *           Check for POSIX class syntax         *
3812    *************************************************/
3813    
3814    /* This function is called when the sequence "[:" or "[." or "[=" is
3815    encountered in a character class. It checks whether this is followed by a
3816    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3817    reach an unescaped ']' without the special preceding character, return FALSE.
3818    
3819    Originally, this function only recognized a sequence of letters between the
3820    terminators, but it seems that Perl recognizes any sequence of characters,
3821    though of course unknown POSIX names are subsequently rejected. Perl gives an
3822    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3823    didn't consider this to be a POSIX class. Likewise for [:1234:].
3824    
3825    The problem in trying to be exactly like Perl is in the handling of escapes. We
3826    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3827    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3828    below handles the special case of \], but does not try to do any other escape
3829    processing. This makes it different from Perl for cases such as [:l\ower:]
3830    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3831    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3832    I think.
3833    
3834    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3835    It seems that the appearance of a nested POSIX class supersedes an apparent
3836    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3837    a digit.
3838    
3839    In Perl, unescaped square brackets may also appear as part of class names. For
3840    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3841    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3842    seem right at all. PCRE does not allow closing square brackets in POSIX class
3843    names.
3844    
3845    Arguments:
3846      ptr      pointer to the initial [
3847      endptr   where to return the end pointer
3848    
3849    Returns:   TRUE or FALSE
3850    */
3851    
3852    static BOOL
3853    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3854    {
3855    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3856    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3857    for (++ptr; *ptr != CHAR_NULL; ptr++)
3858    {    {
3859    case OP_CHAR:    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3860    case OP_CHARI:      ptr++;
3861    switch(escape)    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3862      else
3863      {      {
3864      case ESC_d:      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3865      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;        {
3866          *endptr = ptr;
3867          return TRUE;
3868          }
3869        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3870             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3871              ptr[1] == CHAR_EQUALS_SIGN) &&
3872            check_posix_syntax(ptr, endptr))
3873          return FALSE;
3874        }
3875      }
3876    return FALSE;
3877    }
3878    
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
3879    
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3880    
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
3881    
3882      case ESC_w:  /*************************************************
3883      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  *          Check POSIX class name                *
3884    *************************************************/
3885    
3886      case ESC_W:  /* This function is called to check the name given in a POSIX-style class entry
3887      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  such as [:alnum:].
3888    
3889      case ESC_h:  Arguments:
3890      case ESC_H:    ptr        points to the first letter
3891      switch(c)    len        the length of the name
3892    
3893    Returns:     a value representing the name, or -1 if unknown
3894    */
3895    
3896    static int
3897    check_posix_name(const pcre_uchar *ptr, int len)
3898    {
3899    const char *pn = posix_names;
3900    register int yield = 0;
3901    while (posix_name_lengths[yield] != 0)
3902      {
3903      if (len == posix_name_lengths[yield] &&
3904        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3905      pn += posix_name_lengths[yield] + 1;
3906      yield++;
3907      }
3908    return -1;
3909    }
3910    
3911    
3912    /*************************************************
3913    *    Adjust OP_RECURSE items in repeated group   *
3914    *************************************************/
3915    
3916    /* OP_RECURSE items contain an offset from the start of the regex to the group
3917    that is referenced. This means that groups can be replicated for fixed
3918    repetition simply by copying (because the recursion is allowed to refer to
3919    earlier groups that are outside the current group). However, when a group is
3920    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3921    inserted before it, after it has been compiled. This means that any OP_RECURSE
3922    items within it that refer to the group itself or any contained groups have to
3923    have their offsets adjusted. That one of the jobs of this function. Before it
3924    is called, the partially compiled regex must be temporarily terminated with
3925    OP_END.
3926    
3927    This function has been extended with the possibility of forward references for
3928    recursions and subroutine calls. It must also check the list of such references
3929    for the group we are dealing with. If it finds that one of the recursions in
3930    the current group is on this list, it adjusts the offset in the list, not the
3931    value in the reference (which is a group number).
3932    
3933    Arguments:
3934      group      points to the start of the group
3935      adjust     the amount by which the group is to be moved
3936      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3937      cd         contains pointers to tables etc.
3938      save_hwm   the hwm forward reference pointer at the start of the group
3939    
3940    Returns:     nothing
3941    */
3942    
3943    static void
3944    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3945      pcre_uchar *save_hwm)
3946    {
3947    pcre_uchar *ptr = group;
3948    
3949    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3950      {
3951      int offset;
3952      pcre_uchar *hc;
3953    
3954      /* See if this recursion is on the forward reference list. If so, adjust the
3955      reference. */
3956    
3957      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3958        {
3959        offset = (int)GET(hc, 0);
3960        if (cd->start_code + offset == ptr + 1)
3961        {        {
3962        HSPACE_CASES:        PUT(hc, 0, offset + adjust);
3963        return escape != ESC_h;        break;
   
       default:  
       return escape == ESC_h;  
3964        }        }
3965        }
3966    
3967      case ESC_v:    /* Otherwise, adjust the recursion offset if it's after the start of this
3968      case ESC_V:    group. */
3969      switch(c)  
3970        {    if (hc >= cd->hwm)
3971        VSPACE_CASES:      {
3972        return escape != ESC_v;      offset = (int)GET(ptr, 1);
3973        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3974        }
3975    
3976      ptr += 1 + LINK_SIZE;
3977      }
3978    }
3979    
3980    
3981    
3982    /*************************************************
3983    *        Insert an automatic callout point       *
3984    *************************************************/
3985    
3986    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3987    callout points before each pattern item.
3988    
3989    Arguments:
3990      code           current code pointer
3991      ptr            current pattern pointer
3992      cd             pointers to tables etc
3993    
3994        default:  Returns:         new code pointer
3995        return escape == ESC_v;  */
       }  
3996    
3997      /* When PCRE_UCP is set, these values get generated for \d etc. Find  static pcre_uchar *
3998      their substitutions and process them. The result will always be either  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3999      ESC_p or ESC_P. Then fall through to process those values. */  {
4000    *code++ = OP_CALLOUT;
4001    *code++ = 255;
4002    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4003    PUT(code, LINK_SIZE, 0);                       /* Default length */
4004    return code + 2 * LINK_SIZE;
4005    }
4006    
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[escape - ESC_DU];  
       escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
4007    
     case ESC_p:  
     case ESC_P:  
       {  
       unsigned int ptype = 0, pdata = 0;  
       int errorcodeptr;  
       BOOL negated;  
4008    
4009        ptr--;      /* Make ptr point at the p or P */  /*************************************************
4010        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  *         Complete a callout item                *
4011          return FALSE;  *************************************************/
       ptr++;      /* Point past the final curly ket */  
4012    
4013        /* If the property item is optional, we have to give up. (When generated  /* A callout item contains the length of the next item in the pattern, which
4014        from \d etc by PCRE_UCP, this test will have been applied much earlier,  we can't fill in till after we have reached the relevant point. This is used
4015        to the original \d etc. At this point, ptr will point to a zero byte. */  for both automatic and manual callouts.
4016    
4017        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  Arguments:
4018          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    previous_callout   points to previous callout item
4019            return FALSE;    ptr                current pattern pointer
4020      cd                 pointers to tables etc
4021    
4022        /* Do the property check. */  Returns:             nothing
4023    */
4024    
4025        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  static void
4026        }  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4027  #endif  {
4028    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4029    PUT(previous_callout, 2 + LINK_SIZE, length);
4030    }
4031    
     default:  
     return FALSE;  
     }  
4032    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
4033    
4034    case OP_DIGIT:  #ifdef SUPPORT_UCP
4035    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  /*************************************************
4036           escape == ESC_h || escape == ESC_v || escape == ESC_R;  *           Get othercase range                  *
4037    *************************************************/
4038    
4039    case OP_NOT_DIGIT:  /* This function is passed the start and end of a class range, in UTF-8 mode
4040    return escape == ESC_d;  with UCP support. It searches up the characters, looking for ranges of
4041    characters in the "other" case. Each call returns the next one, updating the
4042    start address. A character with multiple other cases is returned on its own
4043    with a special return value.
4044    
4045    case OP_WHITESPACE:  Arguments:
4046    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4047      d           end value
4048      ocptr       where to put start of othercase range
4049      odptr       where to put end of othercase range
4050    
4051    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4052    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4053                  >0 the CASESET offset for char with multiple other cases
4054                    in this case, ocptr contains the original
4055    */
4056    
4057    case OP_HSPACE:  static int
4058    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4059           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4060    {
4061    pcre_uint32 c, othercase, next;
4062    unsigned int co;
4063    
4064    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4065    return escape == ESC_h;  cases, return its case offset value. */
4066    
4067    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4068    case OP_ANYNL:    {
4069    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4070    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4071        *ocptr = c++;   /* Character that has the set */
4072        *cptr = c;      /* Rest of input range */
4073        return (int)co;
4074        }
4075      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4076      }
4077    
4078    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4079    
4080    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4081    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  range, which is either the end of the input range, or a character that has zero
4082           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4083    
4084    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4085    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4086    
4087    default:  for (++c; c <= d; c++)
4088    return FALSE;    {
4089      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4090      next++;
4091    }    }
4092    
4093  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4094    *cptr = c;             /* Rest of input range */
4095    return 0;
4096  }  }
4097    #endif  /* SUPPORT_UCP */
4098    
4099    
4100    
# Line 3428  switch(op_code) Line 4103  switch(op_code)
4103  *************************************************/  *************************************************/
4104    
4105  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
4106  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
4107  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4108  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4109    
4110  Arguments:  Arguments:
4111    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4112    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4113    options       the options word    options       the options word
4114    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4115    start         start of range character    start         start of range character
4116    end           end of range character    end           end of range character
4117    
4118  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4119                  the pointer to extra data is updated                  the pointer to extra data is updated
4120  */  */
# Line 3449  add_to_class(pcre_uint8 *classbits, pcre Line 4124  add_to_class(pcre_uint8 *classbits, pcre
4124    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4125  {  {
4126  pcre_uint32 c;  pcre_uint32 c;
4127    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4128  int n8 = 0;  int n8 = 0;
4129    
4130  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4131  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
4132  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
4133  range. */  range. */
4134    
4135  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4136    {    {
4137  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4138    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4139      {      {
4140      int rc;      int rc;
4141      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4142    
4143      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4144      c = start;      c = start;
4145    
4146      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4147        {        {
4148        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4149    
4150        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4151          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4152    
4153        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4154    
4155        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4156    
4157        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
4158        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
4159        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4160    
4161        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4162        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4163        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3491  if ((options & PCRE_CASELESS) != 0) Line 4167  if ((options & PCRE_CASELESS) != 0)
4167  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4168    
4169    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4170    
4171    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4172      {      {
4173      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4174      n8++;      n8++;
4175      }      }
4176    }    }
4177    
4178  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
4179  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
4180  in all cases. */  in all cases. */
# Line 3517  in all cases. */ Line 4193  in all cases. */
4193    
4194  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4195    
4196  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
 data. */  
4197    
4198  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4199    {    {
4200    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4201      {    SETBIT(classbits, c);
4202      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4203    }    }
4204    
4205  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4206    {  if (start <= 0xff) start = 0xff + 1;
4207    
4208    if (end >= start)
4209      {
4210    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4211  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4212    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4213      {      {
4214      if (start < end)      if (start < end)
4215        {        {
4216        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4217        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4218        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4219        }        }
4220      else if (start == end)      else if (start == end)
4221        {        {
4222        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4223        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4224        }        }
4225      }      }
4226    else    else
4227  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4228    
4229    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4230    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4231    
4232  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4233      {}      {}
4234  #else  #else
4235    if (start < end)    if (start < end)
4236      {      {
4237      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3567  else Line 4242  else
4242      {      {
4243      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4244      *uchardata++ = start;      *uchardata++ = start;
4245      }      }
4246  #endif  #endif
4247    
4248    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4249    }    }
4250    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4251    
4252  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4253  }  }
4254    
4255    
4256    
4257    
4258  /*************************************************  /*************************************************
4259  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4260  *************************************************/  *************************************************/
4261    
4262  /* This function is used for adding a list of case-equivalent characters to a  /* This function is used for adding a list of case-equivalent characters to a
4263  class, and also for adding a list of horizontal or vertical whitespace. If the  class, and also for adding a list of horizontal or vertical whitespace. If the
4264  list is in order (which it should be), ranges of characters are detected and  list is in order (which it should be), ranges of characters are detected and
4265  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3593  Arguments: Line 4269  Arguments:
4269    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4270    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4271    options       the options word    options       the options word
4272    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4273    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4274    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4275                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4276                    already know about                    already know about
4277    
4278  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4279                  the pointer to extra data is updated                  the pointer to extra data is updated
4280  */  */
# Line 3612  while (p[0] < NOTACHAR) Line 4288  while (p[0] < NOTACHAR)
4288    {    {
4289    int n = 0;    int n = 0;
4290    if (p[0] != except)    if (p[0] != except)
4291      {      {
4292      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4293      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4294      }      }
4295    p += n + 1;    p += n + 1;
4296    }    }
4297  return n8;  return n8;
4298  }  }
4299    
4300    
4301    
# Line 3634  Arguments: Line 4310  Arguments:
4310    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4311    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4312    options       the options word    options       the options word
4313    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4314    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4315    
4316  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4317                  the pointer to extra data is updated                  the pointer to extra data is updated
4318  */  */
4319    
4320  static int  static int
4321  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4322    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4323  {  {
4324  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3654  while (p[0] < NOTACHAR) Line 4330  while (p[0] < NOTACHAR)
4330    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4331    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4332      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4333    p++;    p++;
4334    }    }
4335  return n8;  return n8;
4336  }  }
4337    
4338    
4339    
# Line 3672  to find out the amount of memory needed, Line 4348  to find out the amount of memory needed,
4348  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4349    
4350  Arguments:  Arguments:
4351    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4352    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4353    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4354    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4355    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
4356    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
4357    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
4358    reqcharflagsptr place to put the last required character flags, or a negative number    reqcharflagsptr   place to put the last required character flags, or a negative number
4359    bcptr          points to current branch chain    bcptr             points to current branch chain
4360    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
4361    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
4362    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
4363                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
4364    
4365  Returns:         TRUE on success  Returns:            TRUE on success
4366                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4367  */  */
4368    
4369  static BOOL  static BOOL
# Line 3741  BOOL utf = FALSE; Line 4417  BOOL utf = FALSE;
4417  #endif  #endif
4418    
4419  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4420  class_uchardata always so that it can be passed to add_to_class() always,  class_uchardata always so that it can be passed to add_to_class() always,
4421  though it will not be used in non-UTF 8-bit cases. This avoids having to supply  though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4422  alternative calls for the different cases. */  alternative calls for the different cases. */
4423    
4424  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3793  for (;; ptr++) Line 4469  for (;; ptr++)
4469    BOOL reset_bracount;    BOOL reset_bracount;
4470    int class_has_8bitchar;    int class_has_8bitchar;
4471    int class_one_char;    int class_one_char;
4472    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4473      BOOL xclass_has_prop;
4474    #endif
4475    int newoptions;    int newoptions;
4476    int recno;    int recno;
4477    int refsign;    int refsign;
# Line 3910  for (;; ptr++) Line 4589  for (;; ptr++)
4589          }          }
4590        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4591        }        }
4592        /* Control does not reach here. */
4593      }      }
4594    
4595    /* Fill in length of a previous callout, except when the next thing is    /* In extended mode, skip white space and comments. We need a loop in order
4596    a quantifier. */    to check for more white space and more comments after a comment. */
   
   is_quantifier =  
     c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||  
     (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));  
   
   if (!is_quantifier && previous_callout != NULL &&  
        after_manual_callout-- <= 0)  
     {  
     if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */  
       complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
   
   /* In extended mode, skip white space and comments. */  
4597    
4598    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4599      {      {
4600      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4601        {        {
4602          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4603          if (c != CHAR_NUMBER_SIGN) break;
4604        ptr++;        ptr++;
4605        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4606          {          {
4607          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4608              {                          /* IS_NEWLINE sets cd->nllen. */
4609              ptr += cd->nllen;
4610              break;
4611              }
4612          ptr++;          ptr++;
4613  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4614          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4615  #endif  #endif
4616          }          }
4617        if (*ptr != CHAR_NULL) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4618        }        }
4619      }      }
4620    
4621    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4622    
4623      is_quantifier =
4624        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4625        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4626    
4627      /* Fill in length of a previous callout, except when the next thing is a
4628      quantifier or when processing a property substitution string in UCP mode. */
4629    
4630      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4631           after_manual_callout-- <= 0)
4632        {
4633        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4634          complete_callout(previous_callout, ptr, cd);
4635        previous_callout = NULL;
4636        }
4637    
4638      /* Create auto callout, except for quantifiers, or while processing property
4639      strings that are substituted for \w etc in UCP mode. */
4640    
4641    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4642      {      {
4643      previous_callout = code;      previous_callout = code;
4644      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4645      }      }
4646    
4647      /* Process the next pattern item. */
4648    
4649    switch(c)    switch(c)
4650      {      {
4651      /* ===================================================================*/      /* ===================================================================*/
4652      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4653      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4654      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4655      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4039  for (;; ptr++) Line 4727  for (;; ptr++)
4727        }        }
4728      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4729    
4730        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4731        used for "start of word" and "end of word". As these are otherwise illegal
4732        sequences, we don't break anything by recognizing them. They are replaced
4733        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4734        erroneous and are handled by the normal code below. */
4735    
4736      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4737        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4738          {
4739          nestptr = ptr + 7;
4740          ptr = sub_start_of_word - 1;
4741          continue;
4742          }
4743    
4744        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4745          {
4746          nestptr = ptr + 7;
4747          ptr = sub_end_of_word - 1;
4748          continue;
4749          }
4750    
4751        /* Handle a real character class. */
4752    
4753      previous = code;      previous = code;
4754    
4755      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
# Line 4096  for (;; ptr++) Line 4806  for (;; ptr++)
4806    
4807      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4808    
4809        /* Extended class (xclass) will be used when characters > 255
4810        might match. */
4811    
4812    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4813        xclass = FALSE;
4814        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4815        class_uchardata_base = class_uchardata;   /* Save the start */
4816    #endif
4817    
4818      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4819      class_has_8bitchar will be non-zero if the class contains at least one <      class_has_8bitchar will be non-zero if the class contains at least one <
4820      256 character; class_one_char will be 1 if the class contains just one      256 character; class_one_char will be 1 if the class contains just one
4821      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4822        are present in the class. */
4823    
4824      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4825      class_one_char = 0;      class_one_char = 0;
4826    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4827        xclass_has_prop = FALSE;
4828    #endif
4829    
4830      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
4831      temporary bit of memory, in case the class contains fewer than two      temporary bit of memory, in case the class contains fewer than two
# Line 4111  for (;; ptr++) Line 4834  for (;; ptr++)
4834    
4835      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4836    
 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  
     xclass = FALSE;  
     class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */  
     class_uchardata_base = class_uchardata;   /* Save the start */  
 #endif  
   
4837      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4838      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4839      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4136  for (;; ptr++) Line 4853  for (;; ptr++)
4853        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4854        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
4855        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4856        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4857        however. */        however. */
4858    
4859        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4200  for (;; ptr++) Line 4917  for (;; ptr++)
4917          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4918          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4919    
4920          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4921            posix_class = 0;            posix_class = 0;
4922    
4923          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
4924          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4925            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4926            directly. */
4927    
4928  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4929          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4930            {            {
4931              unsigned int ptype = 0;
4932            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4933    
4934              /* The posix_substitutes table specifies which POSIX classes can be
4935              converted to \p or \P items. */
4936    
4937            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4938              {              {
4939              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4940              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4941              continue;              continue;
4942              }              }
4943    
4944              /* There are three other classes that generate special property calls
4945              that are recognized only in an XCLASS. */
4946    
4947              else switch(posix_class)
4948                {
4949                case PC_GRAPH:
4950                ptype = PT_PXGRAPH;
4951                /* Fall through */
4952                case PC_PRINT:
4953                if (ptype == 0) ptype = PT_PXPRINT;
4954                /* Fall through */
4955                case PC_PUNCT:
4956                if (ptype == 0) ptype = PT_PXPUNCT;
4957                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4958                *class_uchardata++ = ptype;
4959                *class_uchardata++ = 0;
4960                xclass_has_prop = TRUE;
4961                ptr = tempptr + 1;
4962                continue;
4963    
4964                /* For all other POSIX classes, no special action is taken in UCP
4965                mode. Fall through to the non_UCP case. */
4966    
4967                default:
4968                break;
4969                }
4970            }            }
4971  #endif  #endif
4972          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, or when UCP makes no difference, we build the
4973          chunk of local store because we may be adding and subtracting from it,          bit map for the POSIX class in a chunk of local store because we may be
4974          and we don't want to subtract bits that may be in the main map already.          adding and subtracting from it, and we don't want to subtract bits that
4975          At the end we or the result into the bit map that is being built. */          may be in the main map already. At the end we or the result into the
4976            bit map that is being built. */
4977    
4978          posix_class *= 3;          posix_class *= 3;
4979    
# Line 4277  for (;; ptr++) Line 5029  for (;; ptr++)
5029    
5030        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
5031          {          {
5032          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5033              TRUE);
5034          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
5035            if (escape == 0) c = ec;
         if (escape == 0)  
           c = ec;  
5036          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5037          else if (escape == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
5038            {            {
5039            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
5040            goto FAILED;            goto FAILED;
# Line 4340  for (;; ptr++) Line 5090  for (;; ptr++)
5090              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5091              continue;              continue;
5092    
5093              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5094              if it was previously set by something earlier in the character              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5095              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5096              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5097                we could just adjust the appropriate bit. From PCRE 8.34 we no
5098                longer treat \s and \S specially. */
5099    
5100              case ESC_s:              case ESC_s:
5101              classbits[0] |= cbits[cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
             classbits[1] |= cbits[cbit_space+1] & ~0x08;  
             for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  
5102              continue;              continue;
5103    
5104              case ESC_S:              case ESC_S:
5105              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5106              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
5107              continue;              continue;
5108    
5109              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
5110    
5111              case ESC_h:              case ESC_h:
5112              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5113                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
5114              continue;              continue;
5115    
5116              case ESC_H:              case ESC_H:
5117              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
5118                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
5119              continue;              continue;
5120    
5121              case ESC_v:              case ESC_v:
5122              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5123                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
5124              continue;              continue;
5125    
5126              case ESC_V:              case ESC_V:
5127              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
5128                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
5129              continue;              continue;
5130    
5131  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4391  for (;; ptr++) Line 5140  for (;; ptr++)
5140                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5141                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5142                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5143                  xclass_has_prop = TRUE;
5144                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5145                continue;                continue;
5146                }                }
# Line 4414  for (;; ptr++) Line 5164  for (;; ptr++)
5164    
5165          /* Fall through if the escape just defined a single character (c >= 0).          /* Fall through if the escape just defined a single character (c >= 0).
5166          This may be greater than 256. */          This may be greater than 256. */
5167    
5168          escape = 0;