/[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 1348 by ph10, Fri Jul 5 10:38:37 2013 UTC revision 1477 by ph10, Wed May 21 17:53:49 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) 1997-2013 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 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 509  static const char error_texts[] = Line 540  static const char error_texts[] =
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"    "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552    ;    ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 648  static const pcre_uint8 ebcdic_chartab[] Line 688  static const pcre_uint8 ebcdic_chartab[]
688  #endif  #endif
689    
690    
691    /* This table is used to check whether auto-possessification is possible
692    between adjacent character-type opcodes. The left-hand (repeated) opcode is
693    used to select the row, and the right-hand opcode is use to select the column.
694    A value of 1 means that auto-possessification is OK. For example, the second
695    value in the first row means that \D+\d can be turned into \D++\d.
696    
697    The Unicode property types (\P and \p) have to be present to fill out the table
698    because of what their opcode values are, but the table values should always be
699    zero because property types are handled separately in the code. The last four
700    columns apply to items that cannot be repeated, so there is no need to have
701    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
702    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
703    
704    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
705    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
706    
707    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
708    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
709      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
710      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
711      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
712      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
713      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
714      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
720      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
724      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
725      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
726    };
727    
728    
729    /* This table is used to check whether auto-possessification is possible
730    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
731    left-hand (repeated) opcode is used to select the row, and the right-hand
732    opcode is used to select the column. The values are as follows:
733    
734      0   Always return FALSE (never auto-possessify)
735      1   Character groups are distinct (possessify if both are OP_PROP)
736      2   Check character categories in the same group (general or particular)
737      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
738    
739      4   Check left general category vs right particular category
740      5   Check right general category vs left particular category
741    
742      6   Left alphanum vs right general category
743      7   Left space vs right general category
744      8   Left word vs right general category
745    
746      9   Right alphanum vs left general category
747     10   Right space vs left general category
748     11   Right word vs left general category
749    
750     12   Left alphanum vs right particular category
751     13   Left space vs right particular category
752     14   Left word vs right particular category
753    
754     15   Right alphanum vs left particular category
755     16   Right space vs left particular category
756     17   Right word vs left particular category
757    */
758    
759    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
760    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
761      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
762      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
763      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
764      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
765      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
766      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
767      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
769      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
770      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
772    };
773    
774    /* This table is used to check whether auto-possessification is possible
775    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
776    specifies a general category and the other specifies a particular category. The
777    row is selected by the general category and the column by the particular
778    category. The value is 1 if the particular category is not part of the general
779    category. */
780    
781    static const pcre_uint8 catposstab[7][30] = {
782    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
783      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
784      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
790    };
791    
792    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
793    a general or particular category. The properties in each row are those
794    that apply to the character set in question. Duplication means that a little
795    unnecessary work is done when checking, but this keeps things much simpler
796    because they can all use the same code. For more details see the comment where
797    this table is used.
798    
799    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
800    "space", but from Perl 5.18 it's included, so both categories are treated the
801    same here. */
802    
803    static const pcre_uint8 posspropstab[3][4] = {
804      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
805      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
806      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807    };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
869  /*************************************************  /*************************************************
870  *            Find an error text                  *  *            Find an error text                  *
# Line 675  return s; Line 892  return s;
892  }  }
893    
894    
895    
896  /*************************************************  /*************************************************
897  *           Expand the workspace                 *  *           Expand the workspace                 *
898  *************************************************/  *************************************************/
# Line 752  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 970  return (*p == CHAR_RIGHT_CURLY_BRACKET);
970  *************************************************/  *************************************************/
971    
972  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
973  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
974  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.
975  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
976  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
977  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981    chptr          points to the data character    chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
# Line 965  else Line 1182  else
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1190      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1191      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1192      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1193      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1194      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 1001  else Line 1222  else
1222          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1223          break;          break;
1224          }          }
1225        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          escape = -s;          escape = -s;
1228          break;          break;
# Line 1009  else Line 1230  else
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1234      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1235      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1243      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1035  else Line 1254  else
1254  #endif  #endif
1255      break;      break;
1256    
1257      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1258      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1259      If not, { is treated as a data character. */  
1260        case CHAR_o:
1261        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262          {
1263          ptr += 2;
1264          c = 0;
1265          overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267            {
1268            register pcre_uint32 cc = *ptr++;
1269            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270    #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272    #endif
1273            c = (c << 3) + cc - CHAR_0 ;
1274    #if defined COMPILE_PCRE8
1275            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE16
1277            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280    #endif
1281            }
1282          if (overflow)
1283            {
1284            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285            *errorcodeptr = ERR34;
1286            }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288            {
1289            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292          }
1293        break;
1294    
1295        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298      case CHAR_x:      case CHAR_x:
1299      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1301        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303          {          {
# Line 1060  else Line 1314  else
1314  #endif  #endif
1315            }            }
1316          }          }
1317        break;        }    /* End JavaScript handling */
       }  
1318    
1319      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        c = 0;      else
1327        overflow = FALSE;        {
1328        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329          {          {
1330          register pcre_uint32 cc = *pt++;          ptr += 2;
1331          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1332            overflow = FALSE;
1333            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1339          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1340  #endif  #endif
1341    
1342  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1346          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1347          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348  #endif  #endif
1349    
1350  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1351          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1353          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1355          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356  #endif  #endif
1357          }            }
1358    
1359        if (overflow)          if (overflow)
1360          {            {
1361          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1363          }            }
1364    
1365        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366          {            {
1367          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368          ptr = pt;            }
         break;  
         }  
1369    
1370        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1371        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1372        }          \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1380        {        else
1381        pcre_uint32 cc;                          /* Some compilers don't like */          {
1382        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1391        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393  #endif  #endif
1394        }            }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1192  if ((options & PCRE_UCP) != 0 && escape Line 1459  if ((options & PCRE_UCP) != 0 && escape
1459  return escape;  return escape;
1460  }  }
1461    
1462    
1463    
1464  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1465  /*************************************************  /*************************************************
1466  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1289  return FALSE; Line 1558  return FALSE;
1558    
1559    
1560    
   
1561  /*************************************************  /*************************************************
1562  *         Read repeat counts                     *  *         Read repeat counts                     *
1563  *************************************************/  *************************************************/
# Line 1315  read_repeat_counts(const pcre_uchar *p, Line 1583  read_repeat_counts(const pcre_uchar *p,
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
1585    
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1587    {    {
1588    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1589    return p;    if (min > 65535)
1590    }      {
1591        *errorcodeptr = ERR5;
1592  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1593  Also, max must not be less than min. */      }
1594      }
1595    
1596  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1597    {    {
1598    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1599      {      {
1600      max = 0;      max = 0;
1601      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1602        {        {
1603        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1604        return p;        if (max > 65535)
1605        }          {
1606            *errorcodeptr = ERR5;
1607            return p;
1608            }
1609          }
1610      if (max < min)      if (max < min)
1611        {        {
1612        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1347  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1615  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1615      }      }
1616    }    }
1617    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1618  *minp = min;  *minp = min;
1619  *maxp = max;  *maxp = max;
1620  return p;  return p;
# Line 1358  return p; Line 1623  return p;
1623    
1624    
1625  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / 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;  
     while (ptr < cd->end_pattern && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
     }  
   
   /* 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;  
 }  
   
   
   
   
 /*************************************************  
1626  *      Find first significant op code            *  *      Find first significant op code            *
1627  *************************************************/  *************************************************/
1628    
# Line 1696  for (;;) Line 1661  for (;;)
1661    
1662      case OP_CALLOUT:      case OP_CALLOUT:
1663      case OP_CREF:      case OP_CREF:
1664      case OP_NCREF:      case OP_DNCREF:
1665      case OP_RREF:      case OP_RREF:
1666      case OP_NRREF:      case OP_DNRREF:
1667      case OP_DEF:      case OP_DEF:
1668      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1669      break;      break;
# Line 1712  for (;;) Line 1677  for (;;)
1677    
1678    
1679    
   
1680  /*************************************************  /*************************************************
1681  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1682  *************************************************/  *************************************************/
# Line 1836  for (;;) Line 1800  for (;;)
1800      case OP_COMMIT:      case OP_COMMIT:
1801      case OP_CREF:      case OP_CREF:
1802      case OP_DEF:      case OP_DEF:
1803        case OP_DNCREF:
1804        case OP_DNRREF:
1805      case OP_DOLL:      case OP_DOLL:
1806      case OP_DOLLM:      case OP_DOLLM:
1807      case OP_EOD:      case OP_EOD:
1808      case OP_EODN:      case OP_EODN:
1809      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1810      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1811      case OP_PRUNE:      case OP_PRUNE:
1812      case OP_REVERSE:      case OP_REVERSE:
# Line 1937  for (;;) Line 1901  for (;;)
1901    
1902      switch (*cc)      switch (*cc)
1903        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1904        case OP_CRSTAR:        case OP_CRSTAR:
1905        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1906          case OP_CRPLUS:
1907          case OP_CRMINPLUS:
1908        case OP_CRQUERY:        case OP_CRQUERY:
1909        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1910          case OP_CRPOSSTAR:
1911          case OP_CRPOSPLUS:
1912          case OP_CRPOSQUERY:
1913        return -1;        return -1;
1914    
1915        case OP_CRRANGE:        case OP_CRRANGE:
1916        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1917          case OP_CRPOSRANGE:
1918        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1919        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1920        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2015  for (;;) Line 1983  for (;;)
1983      case OP_QUERYI:      case OP_QUERYI:
1984      case OP_REF:      case OP_REF:
1985      case OP_REFI:      case OP_REFI:
1986        case OP_DNREF:
1987        case OP_DNREFI:
1988      case OP_SBRA:      case OP_SBRA:
1989      case OP_SBRAPOS:      case OP_SBRAPOS:
1990      case OP_SCBRA:      case OP_SCBRA:
# Line 2051  for (;;) Line 2021  for (;;)
2021    
2022    
2023    
   
2024  /*************************************************  /*************************************************
2025  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2026  *************************************************/  *************************************************/
# Line 2361  Returns:      TRUE if what is matched co Line 2330  Returns:      TRUE if what is matched co
2330  typedef struct recurse_check {  typedef struct recurse_check {
2331    struct recurse_check *prev;    struct recurse_check *prev;
2332    const pcre_uchar *group;    const pcre_uchar *group;
2333  } recurse_check;  } recurse_check;
2334    
2335  static BOOL  static BOOL
2336  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
# Line 2377  for (code = first_significant_code(code Line 2346  for (code = first_significant_code(code
2346    const pcre_uchar *ccode;    const pcre_uchar *ccode;
2347    
2348    c = *code;    c = *code;
2349    
2350    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2351    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2352    
# Line 2405  for (code = first_significant_code(code Line 2374  for (code = first_significant_code(code
2374      NULL. */      NULL. */
2375    
2376      if (cd->start_workspace != NULL)      if (cd->start_workspace != NULL)
2377        {        {
2378        const pcre_uchar *tcode;        const pcre_uchar *tcode;
2379        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2380          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2381        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2382        }        }
2383    
2384      /* If we are scanning a completed pattern, there are no forward references      /* If we are scanning a completed pattern, there are no forward references
2385      and all groups are complete. We need to detect whether this is a recursive      and all groups are complete. We need to detect whether this is a recursive
2386      call, as otherwise there will be an infinite loop. If it is a recursion,      call, as otherwise there will be an infinite loop. If it is a recursion,
2387      just skip over it. Simple recursions are easily detected. For mutual      just skip over it. Simple recursions are easily detected. For mutual
2388      recursions we keep a chain on the stack. */      recursions we keep a chain on the stack. */
2389    
2390      else      else
2391        {        {
2392        recurse_check *r = recurses;        recurse_check *r = recurses;
2393        const pcre_uchar *endgroup = scode;        const pcre_uchar *endgroup = scode;
2394    
2395        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2396        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2397    
2398        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2399          if (r->group == scode) break;          if (r->group == scode) break;
2400        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2436  for (code = first_significant_code(code Line 2405  for (code = first_significant_code(code
2405    
2406      empty_branch = FALSE;      empty_branch = FALSE;
2407      this_recurse.prev = recurses;      this_recurse.prev = recurses;
2408      this_recurse.group = scode;      this_recurse.group = scode;
2409    
2410      do      do
2411        {        {
2412        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
# Line 2538  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2508        case OP_CRQUERY:        case OP_CRQUERY:
2509        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2510          case OP_CRPOSSTAR:
2511          case OP_CRPOSQUERY:
2512        break;        break;
2513    
2514        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2515        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2516        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2517          case OP_CRPOSPLUS:
2518        return FALSE;        return FALSE;
2519    
2520        case OP_CRRANGE:        case OP_CRRANGE:
2521        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2522          case OP_CRPOSRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2524        break;        break;
2525        }        }
# Line 2557  for (code = first_significant_code(code Line 2530  for (code = first_significant_code(code
2530      case OP_ANY:      case OP_ANY:
2531      case OP_ALLANY:      case OP_ALLANY:
2532      case OP_ANYBYTE:      case OP_ANYBYTE:
2533    
2534      case OP_PROP:      case OP_PROP:
2535      case OP_NOTPROP:      case OP_NOTPROP:
2536      case OP_ANYNL:      case OP_ANYNL:
2537    
2538      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2539      case OP_HSPACE:      case OP_HSPACE:
2540      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2541      case OP_VSPACE:      case OP_VSPACE:
2542      case OP_EXTUNI:      case OP_EXTUNI:
2543    
2544      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2545      case OP_DIGIT:      case OP_DIGIT:
2546      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2547      case OP_WHITESPACE:      case OP_WHITESPACE:
2548      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2549      case OP_WORDCHAR:      case OP_WORDCHAR:
2550    
2551      case OP_CHAR:      case OP_CHAR:
2552      case OP_CHARI:      case OP_CHARI:
2553      case OP_NOT:      case OP_NOT:
2554      case OP_NOTI:      case OP_NOTI:
2555    
2556      case OP_PLUS:      case OP_PLUS:
2557      case OP_PLUSI:      case OP_PLUSI:
2558      case OP_MINPLUS:      case OP_MINPLUS:
# Line 2589  for (code = first_significant_code(code Line 2562  for (code = first_significant_code(code
2562      case OP_NOTPLUSI:      case OP_NOTPLUSI:
2563      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2564      case OP_NOTMINPLUSI:      case OP_NOTMINPLUSI:
2565    
2566      case OP_POSPLUS:      case OP_POSPLUS:
2567      case OP_POSPLUSI:      case OP_POSPLUSI:
2568      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2569      case OP_NOTPOSPLUSI:      case OP_NOTPOSPLUSI:
2570    
2571      case OP_EXACT:      case OP_EXACT:
2572      case OP_EXACTI:      case OP_EXACTI:
2573      case OP_NOTEXACT:      case OP_NOTEXACT:
2574      case OP_NOTEXACTI:      case OP_NOTEXACTI:
2575    
2576      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2577      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2578      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2579      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2580    
2581      return FALSE;      return FALSE;
2582    
2583      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2644  for (code = first_significant_code(code Line 2617  for (code = first_significant_code(code
2617  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2618      case OP_STAR:      case OP_STAR:
2619      case OP_STARI:      case OP_STARI:
2620      case OP_NOTSTAR:      case OP_NOTSTAR:
2621      case OP_NOTSTARI:      case OP_NOTSTARI:
2622    
2623      case OP_MINSTAR:      case OP_MINSTAR:
2624      case OP_MINSTARI:      case OP_MINSTARI:
2625      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
2626      case OP_NOTMINSTARI:      case OP_NOTMINSTARI:
2627    
2628      case OP_POSSTAR:      case OP_POSSTAR:
2629      case OP_POSSTARI:      case OP_POSSTARI:
2630      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
2631      case OP_NOTPOSSTARI:      case OP_NOTPOSSTARI:
2632    
2633      case OP_QUERY:      case OP_QUERY:
2634      case OP_QUERYI:      case OP_QUERYI:
2635      case OP_NOTQUERY:      case OP_NOTQUERY:
2636      case OP_NOTQUERYI:      case OP_NOTQUERYI:
2637    
2638      case OP_MINQUERY:      case OP_MINQUERY:
2639      case OP_MINQUERYI:      case OP_MINQUERYI:
2640      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
2641      case OP_NOTMINQUERYI:      case OP_NOTMINQUERYI:
2642    
2643      case OP_POSQUERY:      case OP_POSQUERY:
2644      case OP_POSQUERYI:      case OP_POSQUERYI:
2645      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
2646      case OP_NOTPOSQUERYI:      case OP_NOTPOSQUERYI:
2647    
2648      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2649      break;      break;
2650    
2651      case OP_UPTO:      case OP_UPTO:
2652      case OP_UPTOI:      case OP_UPTOI:
2653      case OP_NOTUPTO:      case OP_NOTUPTO:
2654      case OP_NOTUPTOI:      case OP_NOTUPTOI:
2655    
2656      case OP_MINUPTO:      case OP_MINUPTO:
2657      case OP_MINUPTOI:      case OP_MINUPTOI:
2658      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
2659      case OP_NOTMINUPTOI:      case OP_NOTMINUPTOI:
2660    
2661      case OP_POSUPTO:      case OP_POSUPTO:
2662      case OP_POSUPTOI:      case OP_POSUPTOI:
2663      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
2664      case OP_NOTPOSUPTOI:      case OP_NOTPOSUPTOI:
2665    
2666      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2667      break;      break;
2668  #endif  #endif
# Line 2753  return TRUE; Line 2726  return TRUE;
2726    
2727    
2728  /*************************************************  /*************************************************
2729  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2730  *************************************************/  *************************************************/
2731    
2732  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2733  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2734    
2735  Returns:   TRUE or FALSE  Arguments:  c opcode
2736    Returns:    base opcode for the type
2737  */  */
2738    
2739  static BOOL  static pcre_uchar
2740  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2741  {  {
2742  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2743  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2744  for (++ptr; *ptr != CHAR_NULL; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2745    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2746    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2747      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2748  }  }
2749    
2750    
2751    
2752    #ifdef SUPPORT_UCP
2753  /*************************************************  /*************************************************
2754  *          Check POSIX class name                *  *        Check a character and a property        *
2755  *************************************************/  *************************************************/
2756    
2757  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2758  such as [:alnum:].  is adjacent to a fixed character.
2759    
2760  Arguments:  Arguments:
2761    ptr        points to the first letter    c            the character
2762    len        the length of the name    ptype        the property type
2763      pdata        the data for the type
2764      negated      TRUE if it's a negated property (\P or \p{^)
2765    
2766  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2767  */  */
2768    
2769  static int  static BOOL
2770  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2771      BOOL negated)
2772  {  {
2773  const char *pn = posix_names;  const pcre_uint32 *p;
2774  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2775    
2776  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2777  that is referenced. This means that groups can be replicated for fixed    {
2778  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2779  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2780  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2781  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2782    
2783  This function has been extended with the possibility of forward references for    case PT_GC:
2784  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2785    
2786  Arguments:    case PT_PC:
2787    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2788    
2789  Returns:     nothing    case PT_SC:
2790  */    return (pdata == prop->script) == negated;
2791    
2792  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2793    
2794  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2795    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2796    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2797    
2798    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2799    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2800      at release 8.34. */
2801    
2802    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2803      case PT_PXSPACE:  /* POSIX space */
2804      switch(c)
2805      {      {
2806      offset = (int)GET(hc, 0);      HSPACE_CASES:
2807      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2808        {      return negated;
2809        PUT(hc, 0, offset + adjust);  
2810        break;      default:
2811        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2812      }      }
2813      break;  /* Control never reaches here */
2814    
2815    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2816    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2817              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2818              c == CHAR_UNDERSCORE) == negated;
2819    
2820    if (hc >= cd->hwm)    case PT_CLIST:
2821      p = PRIV(ucd_caseless_sets) + prop->caseset;
2822      for (;;)
2823      {      {
2824      offset = (int)GET(ptr, 1);      if (c < *p) return !negated;
2825      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2826      }      }
2827      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2828    }    }
2829    
2830    return FALSE;
2831  }  }
2832    #endif  /* SUPPORT_UCP */
2833    
2834    
2835    
2836  /*************************************************  /*************************************************
2837  *        Insert an automatic callout point       *  *        Fill the character property list        *
2838  *************************************************/  *************************************************/
2839    
2840  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2841  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2842    
2843  Arguments:  Arguments:
2844    code           current code pointer    code        points to start of expression
2845    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2846    cd             pointers to tables etc    fcc         points to case-flipping table
2847      list        points to output list
2848                  list[0] will be filled with the opcode
2849                  list[1] will be non-zero if this opcode
2850                    can match an empty character string
2851                  list[2..7] depends on the opcode
2852    
2853  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2854                  NULL if *code is not accepted
2855  */  */
2856    
2857  static pcre_uchar *  static const pcre_uchar *
2858  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2859      const pcre_uint8 *fcc, pcre_uint32 *list)
2860  {  {
2861  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2862  *code++ = 255;  pcre_uchar base;
2863  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uchar *end;
2864  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 chr;
 return code + 2 * LINK_SIZE;  
 }  
   
2865    
2866    #ifdef SUPPORT_UCP
2867    pcre_uint32 *clist_dest;
2868    const pcre_uint32 *clist_src;
2869    #else
2870    utf = utf;  /* Suppress "unused parameter" compiler warning */
2871    #endif
2872    
2873  /*************************************************  list[0] = c;
2874  *         Complete a callout item                *  list[1] = FALSE;
2875  *************************************************/  code++;
2876    
2877  /* A callout item contains the length of the next item in the pattern, which  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2878  we can't fill in till after we have reached the relevant point. This is used    {
2879  for both automatic and manual callouts.    base = get_repeat_base(c);
2880      c -= (base - OP_STAR);
2881    
2882  Arguments:    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2883    previous_callout   points to previous callout item      code += IMM2_SIZE;
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
2884    
2885  Returns:             nothing    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 */  
2886    
2887  static void    switch(base)
2888  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      {
2889  {      case OP_STAR:
2890  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));      list[0] = OP_CHAR;
2891  PUT(previous_callout, 2 + LINK_SIZE, length);      break;
 }  
2892    
2893        case OP_STARI:
2894        list[0] = OP_CHARI;
2895        break;
2896    
2897        case OP_NOTSTAR:
2898        list[0] = OP_NOT;
2899        break;
2900    
2901  #ifdef SUPPORT_UCP      case OP_NOTSTARI:
2902  /*************************************************      list[0] = OP_NOTI;
2903  *           Get othercase range                  *      break;
 *************************************************/  
   
 /* This function is passed the start and end of a class range, in UTF-8 mode  
 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.  
   
 Arguments:  
   cptr        points to starting character value; updated  
   d           end value  
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
   
 Yield:        -1 when no more  
                0 when a range is returned  
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
 */  
   
 static int  
 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 unsigned int co;  
   
 /* Find the first character that has an other case. If it has multiple other  
 cases, return its case offset value. */  
2904    
2905  for (c = *cptr; c <= d; c++)      case OP_TYPESTAR:
2906    {      list[0] = *code;
2907    if ((co = UCD_CASESET(c)) != 0)      code++;
2908      {      break;
     *ocptr = c++;   /* Character that has the set */  
     *cptr = c;      /* Rest of input range */  
     return (int)co;  
2909      }      }
2910    if ((othercase = UCD_OTHERCASE(c)) != c) break;    c = list[0];
2911    }    }
2912    
2913  if (c > d) return -1;  /* Reached end of range */  switch(c)
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
2914    {    {
2915    if (UCD_OTHERCASE(c) != next) break;    case OP_NOT_DIGIT:
2916    next++;    case OP_DIGIT:
2917    }    case OP_NOT_WHITESPACE:
2918      case OP_WHITESPACE:
2919  *odptr = next - 1;     /* End of othercase range */    case OP_NOT_WORDCHAR:
2920  *cptr = c;             /* Rest of input range */    case OP_WORDCHAR:
2921  return 0;    case OP_ANY:
2922  }    case OP_ALLANY:
2923      case OP_ANYNL:
2924      case OP_NOT_HSPACE:
2925      case OP_HSPACE:
2926  /*************************************************    case OP_NOT_VSPACE:
2927  *        Check a character and a property        *    case OP_VSPACE:
2928  *************************************************/    case OP_EXTUNI:
2929      case OP_EODN:
2930  /* This function is called by check_auto_possessive() when a property item    case OP_EOD:
2931  is adjacent to a fixed character.    case OP_DOLL:
2932      case OP_DOLLM:
2933      return code;
2934    
2935  Arguments:    case OP_CHAR:
2936    c            the character    case OP_NOT:
2937    ptype        the property type    GETCHARINCTEST(chr, code);
2938    pdata        the data for the type    list[2] = chr;
2939    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2940      return code;
2941    
2942  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2943  */    case OP_NOTI:
2944      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2945      GETCHARINCTEST(chr, code);
2946      list[2] = chr;
2947    
 static BOOL  
 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  
 {  
2948  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2949  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2950        list[3] = fcc[chr];
2951      else
2952        list[3] = UCD_OTHERCASE(chr);
2953    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2954      list[3] = (chr < 256) ? fcc[chr] : chr;
2955    #else
2956      list[3] = fcc[chr];
2957  #endif  #endif
2958    
2959  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
   
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
2960    
2961    case PT_GC:    if (chr == list[3])
2962    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;      list[3] = NOTACHAR;
2963      else
2964        list[4] = NOTACHAR;
2965      return code;
2966    
2967    case PT_PC:  #ifdef SUPPORT_UCP
2968    return (pdata == prop->chartype) == negated;    case OP_PROP:
2969      case OP_NOTPROP:
2970      if (code[0] != PT_CLIST)
2971        {
2972        list[2] = code[0];
2973        list[3] = code[1];
2974        return code + 2;
2975        }
2976    
2977    case PT_SC:    /* Convert only if we have enough space. */
   return (pdata == prop->script) == negated;  
2978    
2979    /* These are specials */    clist_src = PRIV(ucd_caseless_sets) + code[1];
2980      clist_dest = list + 2;
2981      code += 2;
2982    
2983    case PT_ALNUM:    do {
2984    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||       if (clist_dest >= list + 8)
2985            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;         {
2986           /* Early return if there is not enough space. This should never
2987           happen, since all clists are shorter than 5 character now. */
2988           list[2] = code[0];
2989           list[3] = code[1];
2990           return code;
2991           }
2992         *clist_dest++ = *clist_src;
2993         }
2994      while(*clist_src++ != NOTACHAR);
2995    
2996    case PT_SPACE:    /* Perl space */    /* All characters are stored. The terminating NOTACHAR
2997    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    is copied form the clist itself. */
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
2998    
2999    case PT_PXSPACE:  /* POSIX space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3000    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
3001            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #endif
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3002    
3003    case PT_WORD:    case OP_NCLASS:
3004    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    case OP_CLASS:
3005            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3006            c == CHAR_UNDERSCORE) == negated;    case OP_XCLASS:
3007      if (c == OP_XCLASS)
3008        end = code + GET(code, 0) - 1;
3009      else
3010    #endif
3011        end = code + 32 / sizeof(pcre_uchar);
3012    
3013  #ifdef SUPPORT_UCP    switch(*end)
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3014      {      {
3015      if (c < *p) return !negated;      case OP_CRSTAR:
3016      if (c == *p++) return negated;      case OP_CRMINSTAR:
3017        case OP_CRQUERY:
3018        case OP_CRMINQUERY:
3019        case OP_CRPOSSTAR:
3020        case OP_CRPOSQUERY:
3021        list[1] = TRUE;
3022        end++;
3023        break;
3024    
3025        case OP_CRPLUS:
3026        case OP_CRMINPLUS:
3027        case OP_CRPOSPLUS:
3028        end++;
3029        break;
3030    
3031        case OP_CRRANGE:
3032        case OP_CRMINRANGE:
3033        case OP_CRPOSRANGE:
3034        list[1] = (GET2(end, 1) == 0);
3035        end += 1 + 2 * IMM2_SIZE;
3036        break;
3037      }      }
3038    break;  /* Control never reaches here */    list[2] = (pcre_uint32)(end - code);
3039  #endif    return end;
3040    }    }
3041    return NULL;    /* Opcode not accepted */
 return FALSE;  
3042  }  }
 #endif  /* SUPPORT_UCP */  
3043    
3044    
3045    
3046  /*************************************************  /*************************************************
3047  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3048  *************************************************/  *************************************************/
3049    
3050  /* 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
3051  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.  
3052    
3053  Arguments:  Arguments:
3054    previous      pointer to the repeated opcode    code        points to the byte code
3055    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3056    ptr           next character in pattern    cd          static compile data
3057    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3058    
3059  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3060  */  */
3061    
3062  static BOOL  static BOOL
3063  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3064    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3065  {  {
3066  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3067  pcre_uint32 next;  pcre_uint32 list[8];
3068  int escape;  const pcre_uint32 *chr_ptr;
3069  pcre_uchar op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3070    const pcre_uint32 *list_ptr;
3071  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3072    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3073    const pcre_uchar *xclass_flags;
3074    #endif
3075    const pcre_uint8 *class_bitset;
3076    const pcre_uint8 *set1, *set2, *set_end;
3077    pcre_uint32 chr;
3078    BOOL accepted, invert_bits;
3079    
3080    /* Note: the base_list[1] contains whether the current opcode has greedy
3081    (represented by a non-zero value) quantifier. This is a different from
3082    other character type lists, which stores here that the character iterator
3083    matches to an empty string (also represented by a non-zero value). */
3084    
3085  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3086    {    {
3087    for (;;)    /* All operations move the code pointer forward.
3088      Therefore infinite recursions are not possible. */
3089    
3090      c = *code;
3091    
3092      /* Skip over callouts */
3093    
3094      if (c == OP_CALLOUT)
3095      {      {
3096      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3097      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;  
3098      }      }
   }  
3099    
3100  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3101  value is a character, a negative value is an escape value. */      {
3102        do code += GET(code, 1); while (*code == OP_ALT);
3103        c = *code;
3104        }
3105    
3106  if (*ptr == CHAR_BACKSLASH)    switch(c)
3107    {      {
3108    int temperrorcode = 0;      case OP_END:
3109    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,      case OP_KETRPOS:
3110      FALSE);      /* TRUE only in greedy case. The non-greedy case could be replaced by
3111    if (temperrorcode != 0) return FALSE;      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3112    ptr++;    /* Point after the escape sequence */      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;  
3113    
3114  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3115    
3116  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3117    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3118    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3119      {      cannot be converted to a possessive form. */
3120      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3121      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[1] == 0) return FALSE;
3122    
3123        switch(*(code - GET(code, 1)))
3124          {
3125          case OP_ASSERT:
3126          case OP_ASSERT_NOT:
3127          case OP_ASSERTBACK:
3128          case OP_ASSERTBACK_NOT:
3129          case OP_ONCE:
3130          case OP_ONCE_NC:
3131          /* Atomic sub-patterns and assertions can always auto-possessify their
3132          last iterator. */
3133          return TRUE;
3134          }
3135    
3136        code += PRIV(OP_lengths)[c];
3137        continue;
3138    
3139        case OP_ONCE:
3140        case OP_ONCE_NC:
3141        case OP_BRA:
3142        case OP_CBRA:
3143        next_code = code + GET(code, 1);
3144        code += PRIV(OP_lengths)[c];
3145    
3146        while (*next_code == OP_ALT)
3147        {        {
3148        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3149        while (*ptr != CHAR_NULL)        code = next_code + 1 + LINK_SIZE;
3150          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3151        }        }
3152      else break;      continue;
     }  
   }  
3153    
3154  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3155        case OP_BRAMINZERO:
3156    
3157  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3158    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3159      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3160    
3161        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3162    
3163        /* The bracket content will be checked by the
3164        OP_BRA/OP_CBRA case above. */
3165        next_code += 1 + LINK_SIZE;
3166        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3167          return FALSE;
3168    
3169  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3170        continue;
3171        }
3172    
3173  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  
   }  
3174    
3175  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3176  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3177    
3178  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3179    {    characters from that list with another list, or with a property. */
   /* For a caseless UTF match, the next character may have more than one other  
   case, which maps to the special PT_CLIST property. Check this first. */  
3180    
3181  #ifdef SUPPORT_UCP    if (base_list[0] == OP_CHAR)
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3182      {      {
3183      unsigned int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3184      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3185      }      }
3186  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3187      {      {
3188      case OP_CHAR:      chr_ptr = list + 2;
3189      return c != next;      list_ptr = base_list;
3190        }
3191    
3192      /* 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. */  
3193    
3194      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3195      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3196  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3197      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3198        {  #endif
3199        pcre_uint32 othercase;        )
3200        if (next < 128) othercase = cd->fcc[next]; else      {
3201  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3202        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3203  #else  #else
3204        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3205  #endif  #endif
3206        return c != othercase;        {
3207          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3208          list_ptr = list;
3209        }        }
3210      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)  
3211        {        {
3212        pcre_uint32 othercase;        set1 = (pcre_uint8 *)(code - list[2]);
3213        if (next < 128) othercase = cd->fcc[next]; else        list_ptr = base_list;
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
3214        }        }
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3215    
3216      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      invert_bits = FALSE;
3217      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      switch(list_ptr[0])
3218          {
3219      case OP_DIGIT:        case OP_CLASS:
3220      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        case OP_NCLASS:
3221          set2 = (pcre_uint8 *)
3222            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3223          break;
3224    
3225      case OP_NOT_DIGIT:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3226      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;        case OP_XCLASS:
3227          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3228          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3229          if ((*xclass_flags & XCL_MAP) == 0)
3230            {
3231            /* No bits are set for characters < 256. */
3232            if (list[1] == 0) return TRUE;
3233            /* Might be an empty repeat. */
3234            continue;
3235            }
3236          set2 = (pcre_uint8 *)(xclass_flags + 1);
3237          break;
3238    #endif
3239    
3240          case OP_NOT_DIGIT:
3241          invert_bits = TRUE;
3242          /* Fall through */
3243          case OP_DIGIT:
3244          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3245          break;
3246    
3247          case OP_NOT_WHITESPACE:
3248          invert_bits = TRUE;
3249          /* Fall through */
3250          case OP_WHITESPACE:
3251          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3252          break;
3253    
3254          case OP_NOT_WORDCHAR:
3255          invert_bits = TRUE;
3256          /* Fall through */
3257          case OP_WORDCHAR:
3258          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3259          break;
3260    
3261          default:
3262          return FALSE;
3263          }
3264    
3265        /* Because the sets are unaligned, we need
3266        to perform byte comparison here. */
3267        set_end = set1 + 32;
3268        if (invert_bits)
3269          {
3270          do
3271            {
3272            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3273            }
3274          while (set1 < set_end);
3275          }
3276        else
3277          {
3278          do
3279            {
3280            if ((*set1++ & *set2++) != 0) return FALSE;
3281            }
3282          while (set1 < set_end);
3283          }
3284    
3285        if (list[1] == 0) return TRUE;
3286        /* Might be an empty repeat. */
3287        continue;
3288        }
3289    
3290      /* Some property combinations also acceptable. Unicode property opcodes are
3291      processed specially; the rest can be handled with a lookup table. */
3292    
3293      else
3294        {
3295        pcre_uint32 leftop, rightop;
3296    
3297        leftop = base_list[0];
3298        rightop = list[0];
3299    
3300    #ifdef SUPPORT_UCP
3301        accepted = FALSE; /* Always set in non-unicode case. */
3302        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3303          {
3304          if (rightop == OP_EOD)
3305            accepted = TRUE;
3306          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3307            {
3308            int n;
3309            const pcre_uint8 *p;
3310            BOOL same = leftop == rightop;
3311            BOOL lisprop = leftop == OP_PROP;
3312            BOOL risprop = rightop == OP_PROP;
3313            BOOL bothprop = lisprop && risprop;
3314    
3315            /* There's a table that specifies how each combination is to be
3316            processed:
3317              0   Always return FALSE (never auto-possessify)
3318              1   Character groups are distinct (possessify if both are OP_PROP)
3319              2   Check character categories in the same group (general or particular)
3320              3   Return TRUE if the two opcodes are not the same
3321              ... see comments below
3322            */
3323    
3324            n = propposstab[base_list[2]][list[2]];
3325            switch(n)
3326              {
3327              case 0: break;
3328              case 1: accepted = bothprop; break;
3329              case 2: accepted = (base_list[3] == list[3]) != same; break;
3330              case 3: accepted = !same; break;
3331    
3332              case 4:  /* Left general category, right particular category */
3333              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3334              break;
3335    
3336              case 5:  /* Right general category, left particular category */
3337              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3338              break;
3339    
3340              /* This code is logically tricky. Think hard before fiddling with it.
3341              The posspropstab table has four entries per row. Each row relates to
3342              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3343              Only WORD actually needs all four entries, but using repeats for the
3344              others means they can all use the same code below.
3345    
3346              The first two entries in each row are Unicode general categories, and
3347              apply always, because all the characters they include are part of the
3348              PCRE character set. The third and fourth entries are a general and a
3349              particular category, respectively, that include one or more relevant
3350              characters. One or the other is used, depending on whether the check
3351              is for a general or a particular category. However, in both cases the
3352              category contains more characters than the specials that are defined
3353              for the property being tested against. Therefore, it cannot be used
3354              in a NOTPROP case.
3355    
3356              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3357              Underscore is covered by ucp_P or ucp_Po. */
3358    
3359              case 6:  /* Left alphanum vs right general category */
3360              case 7:  /* Left space vs right general category */
3361              case 8:  /* Left word vs right general category */
3362              p = posspropstab[n-6];
3363              accepted = risprop && lisprop ==
3364                (list[3] != p[0] &&
3365                 list[3] != p[1] &&
3366                (list[3] != p[2] || !lisprop));
3367              break;
3368    
3369              case 9:   /* Right alphanum vs left general category */
3370              case 10:  /* Right space vs left general category */
3371              case 11:  /* Right word vs left general category */
3372              p = posspropstab[n-9];
3373              accepted = lisprop && risprop ==
3374                (base_list[3] != p[0] &&
3375                 base_list[3] != p[1] &&
3376                (base_list[3] != p[2] || !risprop));
3377              break;
3378    
3379              case 12:  /* Left alphanum vs right particular category */
3380              case 13:  /* Left space vs right particular category */
3381              case 14:  /* Left word vs right particular category */
3382              p = posspropstab[n-12];
3383              accepted = risprop && lisprop ==
3384                (catposstab[p[0]][list[3]] &&
3385                 catposstab[p[1]][list[3]] &&
3386                (list[3] != p[3] || !lisprop));
3387              break;
3388    
3389              case 15:  /* Right alphanum vs left particular category */
3390              case 16:  /* Right space vs left particular category */
3391              case 17:  /* Right word vs left particular category */
3392              p = posspropstab[n-15];
3393              accepted = lisprop && risprop ==
3394                (catposstab[p[0]][base_list[3]] &&
3395                 catposstab[p[1]][base_list[3]] &&
3396                (base_list[3] != p[3] || !risprop));
3397              break;
3398              }
3399            }
3400          }
3401    
3402        else
3403    #endif  /* SUPPORT_UCP */
3404    
3405        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3406               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3407               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3408    
3409        if (!accepted)
3410          return FALSE;
3411    
3412        if (list[1] == 0) return TRUE;
3413        /* Might be an empty repeat. */
3414        continue;
3415        }
3416    
3417      /* Control reaches here only if one of the items is a small character list.
3418      All characters are checked against the other side. */
3419    
3420      do
3421        {
3422        chr = *chr_ptr;
3423    
3424        switch(list_ptr[0])
3425          {
3426          case OP_CHAR:
3427          ochr_ptr = list_ptr + 2;
3428          do
3429            {
3430            if (chr == *ochr_ptr) return FALSE;
3431            ochr_ptr++;
3432            }
3433          while(*ochr_ptr != NOTACHAR);
3434          break;
3435    
3436          case OP_NOT:
3437          ochr_ptr = list_ptr + 2;
3438          do
3439            {
3440            if (chr == *ochr_ptr)
3441              break;
3442            ochr_ptr++;
3443            }
3444          while(*ochr_ptr != NOTACHAR);
3445          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3446          break;
3447    
3448          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3449          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3450    
3451          case OP_DIGIT:
3452          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3453          break;
3454    
3455          case OP_NOT_DIGIT:
3456          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3457          break;
3458    
3459          case OP_WHITESPACE:
3460          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3461          break;
3462    
3463          case OP_NOT_WHITESPACE:
3464          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3465          break;
3466    
3467          case OP_WORDCHAR:
3468          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3469          break;
3470    
3471          case OP_NOT_WORDCHAR:
3472          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3473          break;
3474    
3475          case OP_HSPACE:
3476          switch(chr)
3477            {
3478            HSPACE_CASES: return FALSE;
3479            default: break;
3480            }
3481          break;
3482    
3483          case OP_NOT_HSPACE:
3484          switch(chr)
3485            {
3486            HSPACE_CASES: break;
3487            default: return FALSE;
3488            }
3489          break;
3490    
3491          case OP_ANYNL:
3492          case OP_VSPACE:
3493          switch(chr)
3494            {
3495            VSPACE_CASES: return FALSE;
3496            default: break;
3497            }
3498          break;
3499    
3500          case OP_NOT_VSPACE:
3501          switch(chr)
3502            {
3503            VSPACE_CASES: break;
3504            default: return FALSE;
3505            }
3506          break;
3507    
3508          case OP_DOLL:
3509          case OP_EODN:
3510          switch (chr)
3511            {
3512            case CHAR_CR:
3513            case CHAR_LF:
3514            case CHAR_VT:
3515            case CHAR_FF:
3516            case CHAR_NEL:
3517    #ifndef EBCDIC
3518            case 0x2028:
3519            case 0x2029:
3520    #endif  /* Not EBCDIC */
3521            return FALSE;
3522            }
3523          break;
3524    
3525          case OP_EOD:    /* Can always possessify before \z */
3526          break;
3527    
3528    #ifdef SUPPORT_UCP
3529          case OP_PROP:
3530          case OP_NOTPROP:
3531          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3532                list_ptr[0] == OP_NOTPROP))
3533            return FALSE;
3534          break;
3535    #endif
3536    
3537          case OP_NCLASS:
3538          if (chr > 255) return FALSE;
3539          /* Fall through */
3540    
3541          case OP_CLASS:
3542          if (chr > 255) break;
3543          class_bitset = (pcre_uint8 *)
3544            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3545          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3546          break;
3547    
3548    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3549          case OP_XCLASS:
3550          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3551              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3552          break;
3553    #endif
3554    
3555          default:
3556          return FALSE;
3557          }
3558    
3559        chr_ptr++;
3560        }
3561      while(*chr_ptr != NOTACHAR);
3562    
3563      /* At least one character must be matched from this opcode. */
3564    
3565      if (list[1] == 0) return TRUE;
3566      }
3567    
3568    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3569    but some compilers complain about an unreachable statement. */
3570    
3571    }
3572    
3573    
3574    
3575    /*************************************************
3576    *    Scan compiled regex for auto-possession     *
3577    *************************************************/
3578    
3579    /* Replaces single character iterations with their possessive alternatives
3580    if appropriate. This function modifies the compiled opcode!
3581    
3582    Arguments:
3583      code        points to start of the byte code
3584      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3585      cd          static compile data
3586    
3587    Returns:      nothing
3588    */
3589    
3590    static void
3591    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3592    {
3593    register pcre_uchar c;
3594    const pcre_uchar *end;
3595    pcre_uchar *repeat_opcode;
3596    pcre_uint32 list[8];
3597    
3598    for (;;)
3599      {
3600      c = *code;
3601    
3602      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3603        {
3604        c -= get_repeat_base(c) - OP_STAR;
3605        end = (c <= OP_MINUPTO) ?
3606          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3607        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3608    
3609        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3610          {
3611          switch(c)
3612            {
3613            case OP_STAR:
3614            *code += OP_POSSTAR - OP_STAR;
3615            break;
3616    
3617            case OP_MINSTAR:
3618            *code += OP_POSSTAR - OP_MINSTAR;
3619            break;
3620    
3621            case OP_PLUS:
3622            *code += OP_POSPLUS - OP_PLUS;
3623            break;
3624    
3625            case OP_MINPLUS:
3626            *code += OP_POSPLUS - OP_MINPLUS;
3627            break;
3628    
3629            case OP_QUERY:
3630            *code += OP_POSQUERY - OP_QUERY;
3631            break;
3632    
3633            case OP_MINQUERY:
3634            *code += OP_POSQUERY - OP_MINQUERY;
3635            break;
3636    
3637            case OP_UPTO:
3638            *code += OP_POSUPTO - OP_UPTO;
3639            break;
3640    
3641            case OP_MINUPTO:
3642            *code += OP_POSUPTO - OP_MINUPTO;
3643            break;
3644            }
3645          }
3646        c = *code;
3647        }
3648      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3649        {
3650    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3651        if (c == OP_XCLASS)
3652          repeat_opcode = code + GET(code, 1);
3653        else
3654    #endif
3655          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3656    
3657        c = *repeat_opcode;
3658        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3659          {
3660          /* end must not be NULL. */
3661          end = get_chr_property_list(code, utf, cd->fcc, list);
3662    
3663          list[1] = (c & 1) == 0;
3664    
3665          if (compare_opcodes(end, utf, cd, list, end))
3666            {
3667            switch (c)
3668              {
3669              case OP_CRSTAR:
3670              case OP_CRMINSTAR:
3671              *repeat_opcode = OP_CRPOSSTAR;
3672              break;
3673    
3674              case OP_CRPLUS:
3675              case OP_CRMINPLUS:
3676              *repeat_opcode = OP_CRPOSPLUS;
3677              break;
3678    
3679              case OP_CRQUERY:
3680              case OP_CRMINQUERY:
3681              *repeat_opcode = OP_CRPOSQUERY;
3682              break;
3683    
3684              case OP_CRRANGE:
3685              case OP_CRMINRANGE:
3686              *repeat_opcode = OP_CRPOSRANGE;
3687              break;
3688              }
3689            }
3690          }
3691        c = *code;
3692        }
3693    
3694      switch(c)
3695        {
3696        case OP_END:
3697        return;
3698    
3699        case OP_TYPESTAR:
3700        case OP_TYPEMINSTAR:
3701        case OP_TYPEPLUS:
3702        case OP_TYPEMINPLUS:
3703        case OP_TYPEQUERY:
3704        case OP_TYPEMINQUERY:
3705        case OP_TYPEPOSSTAR:
3706        case OP_TYPEPOSPLUS:
3707        case OP_TYPEPOSQUERY:
3708        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3709        break;
3710    
3711        case OP_TYPEUPTO:
3712        case OP_TYPEMINUPTO:
3713        case OP_TYPEEXACT:
3714        case OP_TYPEPOSUPTO:
3715        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3716          code += 2;
3717        break;
3718    
3719    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3720        case OP_XCLASS:
3721        code += GET(code, 1);
3722        break;
3723    #endif
3724    
3725        case OP_MARK:
3726        case OP_PRUNE_ARG:
3727        case OP_SKIP_ARG:
3728        case OP_THEN_ARG:
3729        code += code[1];
3730        break;
3731        }
3732    
3733      /* Add in the fixed length from the table */
3734    
3735      code += PRIV(OP_lengths)[c];
3736    
3737      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3738      a multi-byte character. The length in the table is a minimum, so we have to
3739      arrange to skip the extra bytes. */
3740    
3741    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3742      if (utf) switch(c)
3743        {
3744        case OP_CHAR:
3745        case OP_CHARI:
3746        case OP_NOT:
3747        case OP_NOTI:
3748        case OP_STAR:
3749        case OP_MINSTAR:
3750        case OP_PLUS:
3751        case OP_MINPLUS:
3752        case OP_QUERY:
3753        case OP_MINQUERY:
3754        case OP_UPTO:
3755        case OP_MINUPTO:
3756        case OP_EXACT:
3757        case OP_POSSTAR:
3758        case OP_POSPLUS:
3759        case OP_POSQUERY:
3760        case OP_POSUPTO:
3761        case OP_STARI:
3762        case OP_MINSTARI:
3763        case OP_PLUSI:
3764        case OP_MINPLUSI:
3765        case OP_QUERYI:
3766        case OP_MINQUERYI:
3767        case OP_UPTOI:
3768        case OP_MINUPTOI:
3769        case OP_EXACTI:
3770        case OP_POSSTARI:
3771        case OP_POSPLUSI:
3772        case OP_POSQUERYI:
3773        case OP_POSUPTOI:
3774        case OP_NOTSTAR:
3775        case OP_NOTMINSTAR:
3776        case OP_NOTPLUS:
3777        case OP_NOTMINPLUS:
3778        case OP_NOTQUERY:
3779        case OP_NOTMINQUERY:
3780        case OP_NOTUPTO:
3781        case OP_NOTMINUPTO:
3782        case OP_NOTEXACT:
3783        case OP_NOTPOSSTAR:
3784        case OP_NOTPOSPLUS:
3785        case OP_NOTPOSQUERY:
3786        case OP_NOTPOSUPTO:
3787        case OP_NOTSTARI:
3788        case OP_NOTMINSTARI:
3789        case OP_NOTPLUSI:
3790        case OP_NOTMINPLUSI:
3791        case OP_NOTQUERYI:
3792        case OP_NOTMINQUERYI:
3793        case OP_NOTUPTOI:
3794        case OP_NOTMINUPTOI:
3795        case OP_NOTEXACTI:
3796        case OP_NOTPOSSTARI:
3797        case OP_NOTPOSPLUSI:
3798        case OP_NOTPOSQUERYI:
3799        case OP_NOTPOSUPTOI:
3800        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3801        break;
3802        }
3803    #else
3804      (void)(utf);  /* Keep compiler happy by referencing function argument */
3805    #endif
3806      }
3807    }
3808    
3809    
3810    
3811    /*************************************************
3812    *           Check for POSIX class syntax         *
3813    *************************************************/
3814    
3815    /* This function is called when the sequence "[:" or "[." or "[=" is
3816    encountered in a character class. It checks whether this is followed by a
3817    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3818    reach an unescaped ']' without the special preceding character, return FALSE.
3819    
3820    Originally, this function only recognized a sequence of letters between the
3821    terminators, but it seems that Perl recognizes any sequence of characters,
3822    though of course unknown POSIX names are subsequently rejected. Perl gives an
3823    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3824    didn't consider this to be a POSIX class. Likewise for [:1234:].
3825    
3826    The problem in trying to be exactly like Perl is in the handling of escapes. We
3827    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3828    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3829    below handles the special case of \], but does not try to do any other escape
3830    processing. This makes it different from Perl for cases such as [:l\ower:]
3831    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3832    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3833    I think.
3834    
3835    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3836    It seems that the appearance of a nested POSIX class supersedes an apparent
3837    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3838    a digit.
3839    
3840    In Perl, unescaped square brackets may also appear as part of class names. For
3841    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3842    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3843    seem right at all. PCRE does not allow closing square brackets in POSIX class
3844    names.
3845    
3846    Arguments:
3847      ptr      pointer to the initial [
3848      endptr   where to return the end pointer
3849    
3850    Returns:   TRUE or FALSE
3851    */
3852    
3853    static BOOL
3854    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3855    {
3856    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3857    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3858    for (++ptr; *ptr != CHAR_NULL; ptr++)
3859      {
3860      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3861        ptr++;
3862      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3863      else
3864        {
3865        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3866          {
3867          *endptr = ptr;
3868          return TRUE;
3869          }
3870        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3871             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3872              ptr[1] == CHAR_EQUALS_SIGN) &&
3873            check_posix_syntax(ptr, endptr))
3874          return FALSE;
3875        }
3876      }
3877    return FALSE;
3878    }
3879    
3880    
3881    
3882    
3883    /*************************************************
3884    *          Check POSIX class name                *
3885    *************************************************/
3886    
3887    /* This function is called to check the name given in a POSIX-style class entry
3888    such as [:alnum:].
3889    
3890    Arguments:
3891      ptr        points to the first letter
3892      len        the length of the name
3893    
3894    Returns:     a value representing the name, or -1 if unknown
3895    */
3896    
3897    static int
3898    check_posix_name(const pcre_uchar *ptr, int len)
3899    {
3900    const char *pn = posix_names;
3901    register int yield = 0;
3902    while (posix_name_lengths[yield] != 0)
3903      {
3904      if (len == posix_name_lengths[yield] &&
3905        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3906      pn += posix_name_lengths[yield] + 1;
3907      yield++;
3908      }
3909    return -1;
3910    }
3911    
3912    
3913    /*************************************************
3914    *    Adjust OP_RECURSE items in repeated group   *
3915    *************************************************/
3916    
3917    /* OP_RECURSE items contain an offset from the start of the regex to the group
3918    that is referenced. This means that groups can be replicated for fixed
3919    repetition simply by copying (because the recursion is allowed to refer to
3920    earlier groups that are outside the current group). However, when a group is
3921    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3922    inserted before it, after it has been compiled. This means that any OP_RECURSE
3923    items within it that refer to the group itself or any contained groups have to
3924    have their offsets adjusted. That one of the jobs of this function. Before it
3925    is called, the partially compiled regex must be temporarily terminated with
3926    OP_END.
3927    
3928      case OP_WHITESPACE:  This function has been extended with the possibility of forward references for
3929      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  recursions and subroutine calls. It must also check the list of such references
3930    for the group we are dealing with. If it finds that one of the recursions in
3931    the current group is on this list, it adjusts the offset in the list, not the
3932    value in the reference (which is a group number).
3933    
3934      case OP_NOT_WHITESPACE:  Arguments:
3935      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;    group      points to the start of the group
3936      adjust     the amount by which the group is to be moved
3937      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3938      cd         contains pointers to tables etc.
3939      save_hwm   the hwm forward reference pointer at the start of the group
3940    
3941      case OP_WORDCHAR:  Returns:     nothing
3942      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  */
3943    
3944      case OP_NOT_WORDCHAR:  static void
3945      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3946      pcre_uchar *save_hwm)
3947    {
3948    pcre_uchar *ptr = group;
3949    
3950      case OP_HSPACE:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3951      case OP_NOT_HSPACE:    {
3952      switch(next)    int offset;
3953        {    pcre_uchar *hc;
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3954    
3955        default:    /* See if this recursion is on the forward reference list. If so, adjust the
3956        return op_code != OP_NOT_HSPACE;    reference. */
       }  
3957    
3958      case OP_ANYNL:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3959      case OP_VSPACE:      {
3960      case OP_NOT_VSPACE:      offset = (int)GET(hc, 0);
3961      switch(next)      if (cd->start_code + offset == ptr + 1)
3962        {        {
3963        VSPACE_CASES:        PUT(hc, 0, offset + adjust);
3964        return op_code == OP_NOT_VSPACE;        break;
   
       default:  
       return op_code != OP_NOT_VSPACE;  
3965        }        }
   
 #ifdef SUPPORT_UCP  
     case OP_PROP:  
     return check_char_prop(next, previous[0], previous[1], FALSE);  
   
     case OP_NOTPROP:  
     return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
     default:  
     return FALSE;  
3966      }      }
   }  
3967    
3968  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP    /* Otherwise, adjust the recursion offset if it's after the start of this
3969  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are    group. */
 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. */  
3970    
3971  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
3972      {      {
3973      case ESC_d:      offset = (int)GET(ptr, 1);
3974      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3975        }
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3976    
3977      case ESC_S:    ptr += 1 + LINK_SIZE;
3978      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    }
3979    }
3980    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3981    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3982    
3983      case ESC_h:  /*************************************************
3984      case ESC_H:  *        Insert an automatic callout point       *
3985      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
3986    
3987        default:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3988        return escape == ESC_h;  callout points before each pattern item.
       }  
3989    
3990      case ESC_v:  Arguments:
3991      case ESC_V:    code           current code pointer
3992      switch(c)    ptr            current pattern pointer
3993        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return escape != ESC_v;  
3994    
3995        default:  Returns:         new code pointer
3996        return escape == ESC_v;  */
       }  
3997    
3998      /* When PCRE_UCP is set, these values get generated for \d etc. Find  static pcre_uchar *
3999      their substitutions and process them. The result will always be either  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4000      ESC_p or ESC_P. Then fall through to process those values. */  {
4001    *code++ = OP_CALLOUT;
4002    *code++ = 255;
4003    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4004    PUT(code, LINK_SIZE, 0);                       /* Default length */
4005    return code + 2 * LINK_SIZE;
4006    }
4007    
 #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 */  
4008    
     case ESC_p:  
     case ESC_P:  
       {  
       unsigned int ptype = 0, pdata = 0;  
       int errorcodeptr;  
       BOOL negated;  
4009    
4010        ptr--;      /* Make ptr point at the p or P */  /*************************************************
4011        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  *         Complete a callout item                *
4012          return FALSE;  *************************************************/
       ptr++;      /* Point past the final curly ket */  
4013    
4014        /* 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
4015        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
4016        to the original \d etc. At this point, ptr will point to a zero byte. */  for both automatic and manual callouts.
4017    
4018        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  Arguments:
4019          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    previous_callout   points to previous callout item
4020            return FALSE;    ptr                current pattern pointer
4021      cd                 pointers to tables etc
4022    
4023        /* Do the property check. */  Returns:             nothing
4024    */
4025    
4026        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  static void
4027        }  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4028  #endif  {
4029    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4030    PUT(previous_callout, 2 + LINK_SIZE, length);
4031    }
4032    
     default:  
     return FALSE;  
     }  
4033    
   /* 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.) */  
4034    
4035    case OP_DIGIT:  #ifdef SUPPORT_UCP
4036    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  /*************************************************
4037           escape == ESC_h || escape == ESC_v || escape == ESC_R;  *           Get othercase range                  *
4038    *************************************************/
4039    
4040    case OP_NOT_DIGIT:  /* This function is passed the start and end of a class range, in UTF-8 mode
4041    return escape == ESC_d;  with UCP support. It searches up the characters, looking for ranges of
4042    characters in the "other" case. Each call returns the next one, updating the
4043    start address. A character with multiple other cases is returned on its own
4044    with a special return value.
4045    
4046    case OP_WHITESPACE:  Arguments:
4047    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4048      d           end value
4049      ocptr       where to put start of othercase range
4050      odptr       where to put end of othercase range
4051    
4052    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4053    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4054                  >0 the CASESET offset for char with multiple other cases
4055                    in this case, ocptr contains the original
4056    */
4057    
4058    case OP_HSPACE:  static int
4059    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4060           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4061    {
4062    pcre_uint32 c, othercase, next;
4063    unsigned int co;
4064    
4065    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4066    return escape == ESC_h;  cases, return its case offset value. */
4067    
4068    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4069    case OP_ANYNL:    {
4070    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4071    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4072        *ocptr = c++;   /* Character that has the set */
4073        *cptr = c;      /* Rest of input range */
4074        return (int)co;
4075        }
4076      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4077      }
4078    
4079    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4080    
4081    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4082    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
4083           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4084    
4085    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4086    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4087    
4088    default:  for (++c; c <= d; c++)
4089    return FALSE;    {
4090      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4091      next++;
4092    }    }
4093    
4094  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4095    *cptr = c;             /* Rest of input range */
4096    return 0;
4097  }  }
4098    #endif  /* SUPPORT_UCP */
4099    
4100    
4101    
# Line 3531  add_to_class(pcre_uint8 *classbits, pcre Line 4125  add_to_class(pcre_uint8 *classbits, pcre
4125    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4126  {  {
4127  pcre_uint32 c;  pcre_uint32 c;
4128    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4129  int n8 = 0;  int n8 = 0;
4130    
4131  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
# Line 3574  if ((options & PCRE_CASELESS) != 0) Line 4169  if ((options & PCRE_CASELESS) != 0)
4169    
4170    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4171    
4172    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4173      {      {
4174      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4175      n8++;      n8++;
# Line 3599  in all cases. */ Line 4194  in all cases. */
4194    
4195  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4196    
4197  /* 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. */  
4198    
4199  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4200    {    {
4201    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4202      {    SETBIT(classbits, c);
4203      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4204    }    }
4205    
4206  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4207    if (start <= 0xff) start = 0xff + 1;
4208    
4209    if (end >= start)
4210    {    {
4211    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4212  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4213    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4214      {      {
# Line 3654  else Line 4248  else
4248    
4249    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4250    }    }
4251    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4252    
4253  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4254  }  }
# Line 3754  to find out the amount of memory needed, Line 4349  to find out the amount of memory needed,
4349  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4350    
4351  Arguments:  Arguments:
4352    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4353    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4354    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4355    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4356    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
4357    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
4358    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
4359    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
4360    bcptr          points to current branch chain    bcptr             points to current branch chain
4361    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
4362    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
4363    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
4364                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
4365    
4366  Returns:         TRUE on success  Returns:            TRUE on success
4367                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4368  */  */
4369    
4370  static BOOL  static BOOL
# Line 3875  for (;; ptr++) Line 4470  for (;; ptr++)
4470    BOOL reset_bracount;    BOOL reset_bracount;
4471    int class_has_8bitchar;    int class_has_8bitchar;
4472    int class_one_char;    int class_one_char;
4473    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4474      BOOL xclass_has_prop;
4475    #endif
4476    int newoptions;    int newoptions;
4477    int recno;    int recno;
4478    int refsign;    int refsign;
# Line 3992  for (;; ptr++) Line 4590  for (;; ptr++)
4590          }          }
4591        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4592        }        }
4593        /* Control does not reach here. */
4594      }      }
4595    
4596    /* 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
4597    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. */  
4598    
4599    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4600      {      {
4601      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4602        {        {
4603          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4604          if (c != CHAR_NUMBER_SIGN) break;
4605        ptr++;        ptr++;
4606        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4607          {          {
4608          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4609              {                          /* IS_NEWLINE sets cd->nllen. */
4610              ptr += cd->nllen;
4611              break;
4612              }
4613          ptr++;          ptr++;
4614  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4615          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4616  #endif  #endif
4617          }          }
4618        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;  
4619        }        }
4620      }      }
4621    
4622    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4623    
4624      is_quantifier =
4625        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4626        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4627    
4628      /* Fill in length of a previous callout, except when the next thing is a
4629      quantifier or when processing a property substitution string in UCP mode. */
4630    
4631      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4632           after_manual_callout-- <= 0)
4633        {
4634        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4635          complete_callout(previous_callout, ptr, cd);
4636        previous_callout = NULL;
4637        }
4638    
4639      /* Create auto callout, except for quantifiers, or while processing property
4640      strings that are substituted for \w etc in UCP mode. */
4641    
4642    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4643      {      {
4644      previous_callout = code;      previous_callout = code;
4645      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4646      }      }
4647    
4648      /* Process the next pattern item. */
4649    
4650    switch(c)    switch(c)
4651      {      {
4652      /* ===================================================================*/      /* ===================================================================*/
4653      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4654      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4655      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4656      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4121  for (;; ptr++) Line 4728  for (;; ptr++)
4728        }        }
4729      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4730    
4731        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4732        used for "start of word" and "end of word". As these are otherwise illegal
4733        sequences, we don't break anything by recognizing them. They are replaced
4734        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4735        erroneous and are handled by the normal code below. */
4736    
4737      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4738        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4739          {
4740          nestptr = ptr + 7;
4741          ptr = sub_start_of_word - 1;
4742          continue;
4743          }
4744    
4745        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4746          {
4747          nestptr = ptr + 7;
4748          ptr = sub_end_of_word - 1;
4749          continue;
4750          }
4751    
4752        /* Handle a real character class. */
4753    
4754      previous = code;      previous = code;
4755    
4756      /* 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 4178  for (;; ptr++) Line 4807  for (;; ptr++)
4807    
4808      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4809    
4810        /* Extended class (xclass) will be used when characters > 255
4811        might match. */
4812    
4813    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4814        xclass = FALSE;
4815        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4816        class_uchardata_base = class_uchardata;   /* Save the start */
4817    #endif
4818    
4819      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4820      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 <
4821      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
4822      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4823        are present in the class. */
4824    
4825      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4826      class_one_char = 0;      class_one_char = 0;
4827    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4828        xclass_has_prop = FALSE;
4829    #endif
4830    
4831      /* 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
4832      temporary bit of memory, in case the class contains fewer than two      temporary bit of memory, in case the class contains fewer than two
4833      8-bit characters because in that case the compiled code doesn't use the bit      8-bit characters because in that case the compiled code doesn't use the bit
4834      map. */      map. */
4835    
4836      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
   
 #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    
4838      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4839      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
# Line 4224  for (;; ptr++) Line 4860  for (;; ptr++)
4860        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4861          {          {
4862          xclass = TRUE;          xclass = TRUE;
4863          *lengthptr += class_uchardata - class_uchardata_base;          *lengthptr += (int)(class_uchardata - class_uchardata_base);
4864          class_uchardata = class_uchardata_base;          class_uchardata = class_uchardata_base;
4865          }          }
4866  #endif  #endif
# Line 4286  for (;; ptr++) Line 4922  for (;; ptr++)
4922            posix_class = 0;            posix_class = 0;
4923    
4924          /* 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
4925          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4926            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4927            directly. */
4928    
4929  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4930          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4931            {            {
4932              unsigned int ptype = 0;
4933            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4934    
4935              /* The posix_substitutes table specifies which POSIX classes can be
4936              converted to \p or \P items. */
4937    
4938            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4939              {              {
4940              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4941              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4942              continue;              continue;
4943              }              }
4944    
4945              /* There are three other classes that generate special property calls
4946              that are recognized only in an XCLASS. */
4947    
4948              else switch(posix_class)
4949                {
4950                case PC_GRAPH:
4951                ptype = PT_PXGRAPH;
4952                /* Fall through */
4953                case PC_PRINT:
4954                if (ptype == 0) ptype = PT_PXPRINT;
4955                /* Fall through */
4956                case PC_PUNCT:
4957                if (ptype == 0) ptype = PT_PXPUNCT;
4958                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4959                *class_uchardata++ = ptype;
4960                *class_uchardata++ = 0;
4961                xclass_has_prop = TRUE;
4962                ptr = tempptr + 1;
4963                continue;
4964    
4965                /* For all other POSIX classes, no special action is taken in UCP
4966                mode. Fall through to the non_UCP case. */
4967    
4968                default:
4969                break;
4970                }
4971            }            }
4972  #endif  #endif
4973          /* 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
4974          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
4975          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
4976          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
4977            bit map that is being built. */
4978    
4979          posix_class *= 3;          posix_class *= 3;
4980    
# Line 4420  for (;; ptr++) Line 5091  for (;; ptr++)
5091              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5092              continue;              continue;
5093    
5094              /* 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
5095              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
5096              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5097              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5098                we could just adjust the appropriate bit. From PCRE 8.34 we no
5099                longer treat \s and \S specially. */
5100    
5101              case ESC_s:              case ESC_s:
5102              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];  
5103              continue;              continue;
5104    
5105              case ESC_S:              case ESC_S:
5106              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5107              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 */  
5108              continue;              continue;
5109    
5110              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
# Line 4471  for (;; ptr++) Line 5141  for (;; ptr++)
5141                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5142                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5143                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5144                  xclass_has_prop = TRUE;
5145                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5146                continue;                continue;
5147                }                }
# Line 4556  for (;; ptr++) Line 5227  for (;; ptr++)
5227  #endif  #endif
5228          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5229    
5230          /* The second part of a range can be a single-character escape, but          /* The second part of a range can be a single-character escape
5231          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          sequence, but not any of the other escapes. Perl treats a hyphen as a
5232          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5233            warning is given, so PCRE now faults it as it is almost certainly a
5234            mistake on the user's part. */
5235    
5236          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5237            {            {
5238            int descape;            if (d == CHAR_BACKSLASH)
5239            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5240            if (*errorcodeptr != 0) goto FAILED;              int descape;
5241                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5242                if (*errorcodeptr != 0) goto FAILED;
5243    
5244            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5245                special causes an error. */
5246    
5247            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5248                {                {
5249                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5250                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5251                    *errorcodeptr = ERR83;
5252                    goto FAILED;
5253                    }
5254                }                }
5255              }              }
5256    
5257              /* A hyphen followed by a POSIX class is treated in the same way. */
5258    
5259              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5260                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5261                        ptr[1] == CHAR_EQUALS_SIGN) &&
5262                       check_posix_syntax(ptr, &tempptr))
5263                {
5264                *errorcodeptr = ERR83;
5265                goto FAILED;
5266                }
5267            }            }
5268    
5269          /* Check that the two values are in the correct order. Optimize          /* Check that the two values are in the correct order. Optimize
# Line 4748  for (;; ptr++) Line 5436  for (;; ptr++)
5436        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5437        code += LINK_SIZE;        code += LINK_SIZE;
5438        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5439          if (xclass_has_prop) *code |= XCL_HASPROP;
5440    
5441        /* If the map is required, move up the extra data to make room for it;        /* If the map is required, move up the extra data to make room for it;
5442        otherwise just move the code pointer to the end of the extra data. */        otherwise just move the code pointer to the end of the extra data. */
# Line 4757  for (;; ptr++) Line 5446  for (;; ptr++)
5446          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5447          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5448            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5449            if (negate_class && !xclass_has_prop)
5450              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5451          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5452          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5453          }          }
# Line 4839  for (;; ptr++) Line 5530  for (;; ptr++)
5530    
5531      tempcode = previous;      tempcode = previous;
5532    
5533        /* Before checking for a possessive quantifier, we must skip over
5534        whitespace and comments in extended mode because Perl allows white space at
5535        this point. */
5536    
5537        if ((options & PCRE_EXTENDED) != 0)
5538          {
5539          const pcre_uchar *p = ptr + 1;
5540          for (;;)
5541            {
5542            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5543            if (*p != CHAR_NUMBER_SIGN) break;
5544            p++;
5545            while (*p != CHAR_NULL)
5546              {
5547              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5548                {                        /* IS_NEWLINE sets cd->nllen. */
5549                p += cd->nllen;
5550                break;
5551                }
5552              p++;
5553    #ifdef SUPPORT_UTF
5554              if (utf) FORWARDCHAR(p);
5555    #endif
5556              }           /* Loop for comment characters */
5557            }             /* Loop for multiple comments */
5558          ptr = p - 1;    /* Character before the next significant one. */
5559          }
5560    
5561      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5562      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5563      If the next character is '?' this is a minimizing repeat, by default,      If the next character is '?' this is a minimizing repeat, by default,
# Line 4933  for (;; ptr++) Line 5652  for (;; ptr++)
5652            }            }
5653          }          }
5654    
       /* If the repetition is unlimited, it pays to see if the next thing on  
       the line is something that cannot possibly match this character. If so,  
       automatically possessifying this item gains some performance in the case  
       where the match fails. */  
   
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5655        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
5656        }        }
5657    
# Line 4963  for (;; ptr++) Line 5669  for (;; ptr++)
5669        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5670        c = *previous;        c = *previous;
5671    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5672        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5673        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5674          {          {
# Line 5119  for (;; ptr++) Line 5817  for (;; ptr++)
5817      /* If previous was a character class or a back reference, we put the repeat      /* If previous was a character class or a back reference, we put the repeat
5818      stuff after it, but just skip the item if the repeat was {0,0}. */      stuff after it, but just skip the item if the repeat was {0,0}. */
5819    
5820      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5821  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5822               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5823  #endif  #endif
5824               *previous == OP_REF ||               *previous == OP_REF   || *previous == OP_REFI ||
5825               *previous == OP_REFI)               *previous == OP_DNREF || *previous == OP_DNREFI)
5826        {        {
5827        if (repeat_max == 0)        if (repeat_max == 0)
5828          {          {
# Line 5153  for (;; ptr++) Line 5850  for (;; ptr++)
5850      opcodes such as BRA and CBRA, as this is the place where they get converted      opcodes such as BRA and CBRA, as this is the place where they get converted
5851      into the more special varieties such as BRAPOS and SBRA. A test for >=      into the more special varieties such as BRAPOS and SBRA. A test for >=
5852      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5853      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5854      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5855        for Perl compatibility. */
5856    
5857      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5858        {        {
# Line 5172  for (;; ptr++) Line 5870  for (;; ptr++)
5870        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5871        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5872        if the minimum is greater than zero, just ignore the repeat. If the        if the minimum is greater than zero, just ignore the repeat. If the
5873        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5874    
5875        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5876          {          {
# Line 5307  for (;; ptr++) Line 6005  for (;; ptr++)
6005                while (cd->hwm > cd->start_workspace + cd->workspace_size -                while (cd->hwm > cd->start_workspace + cd->workspace_size -
6006                       WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                       WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6007                  {                  {
6008                  int save_offset = save_hwm - cd->start_workspace;                  size_t save_offset = save_hwm - cd->start_workspace;
6009                  int this_offset = this_hwm - cd->start_workspace;                  size_t this_offset = this_hwm - cd->start_workspace;
6010                  *errorcodeptr = expand_workspace(cd);                  *errorcodeptr = expand_workspace(cd);
6011                  if (*errorcodeptr != 0) goto FAILED;                  if (*errorcodeptr != 0) goto FAILED;
6012                  save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;                  save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
# Line 5389  for (;; ptr++) Line 6087  for (;; ptr++)
6087            while (cd->hwm > cd->start_workspace + cd->workspace_size -            while (cd->hwm > cd->start_workspace + cd->workspace_size -
6088                   WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                   WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6089              {              {
6090              int save_offset = save_hwm - cd->start_workspace;              size_t save_offset = save_hwm - cd->start_workspace;
6091              int this_offset = this_hwm - cd->start_workspace;              size_t this_offset = this_hwm - cd->start_workspace;
6092              *errorcodeptr = expand_workspace(cd);              *errorcodeptr = expand_workspace(cd);
6093              if (*errorcodeptr != 0) goto FAILED;              if (*errorcodeptr != 0) goto FAILED;
6094              save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;              save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
# Line 5545  for (;; ptr++) Line 6243  for (;; ptr++)
6243        goto FAILED;        goto FAILED;
6244        }        }
6245    
6246      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6247      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6248      there are special alternative opcodes for this case. For anything else, we      case. For anything else, we wrap the entire repeated item inside OP_ONCE
6249      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6250      notation is just syntactic sugar, taken from Sun's Java package, but the      Sun's Java package, but the special opcodes can optimize it.
     special opcodes can optimize it.  
6251    
6252      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6253      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6254      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6255        tempcode, not at previous, which might be the first part of a string whose
6256      Note that the repeated item starts at tempcode, not at previous, which      (former) last char we repeated. */
     might be the first part of a string whose (former) last char we repeated.  
   
     Possessifying an 'exact' quantifier has no effect, so we can ignore it. But  
     an 'upto' may follow. We skip over an 'exact' item, and then test the  
     length of what remains before proceeding. */  
6257    
6258      if (possessive_quantifier)      if (possessive_quantifier)
6259        {        {
6260        int len;        int len;
6261    
6262        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6263          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6264          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6265          remains is greater than zero, there's a further opcode that can be
6266          handled. If not, do nothing, leaving the EXACT alone. */
6267    
6268          switch(*tempcode)
6269            {
6270            case OP_TYPEEXACT:
6271          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6272            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6273            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6274            break;
6275    
6276        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6277          {  
6278            case OP_CHAR:
6279            case OP_CHARI:
6280            case OP_NOT:
6281            case OP_NOTI:
6282            case OP_EXACT:
6283            case OP_EXACTI:
6284            case OP_NOTEXACT:
6285            case OP_NOTEXACTI:
6286          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6287  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6288          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6289            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6290  #endif  #endif
6291            break;
6292    
6293            /* For the class opcodes, the repeat operator appears at the end;
6294            adjust tempcode to point to it. */
6295    
6296            case OP_CLASS:
6297            case OP_NCLASS:
6298            tempcode += 1 + 32/sizeof(pcre_uchar);
6299            break;
6300    
6301    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6302            case OP_XCLASS:
6303            tempcode += GET(tempcode, 1);
6304            break;
6305    #endif
6306          }          }
6307    
6308          /* If tempcode is equal to code (which points to the end of the repeated
6309          item), it means we have skipped an EXACT item but there is no following
6310          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6311          all other cases, tempcode will be pointing to the repeat opcode, and will
6312          be less than code, so the value of len will be greater than 0. */
6313    
6314        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6315          if (len > 0)
6316            {
6317            unsigned int repcode = *tempcode;
6318    
6319            /* There is a table for possessifying opcodes, all of which are less
6320            than OP_CALLOUT. A zero entry means there is no possessified version.
6321            */
6322    
6323            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6324              *tempcode = opcode_possessify[repcode];
6325    
6326            /* For opcode without a special possessified version, wrap the item in
6327            ONCE brackets. Because we are moving code along, we must ensure that any
6328            pending recursive references are updated. */
6329    
6330            else
6331              {
6332              *code = OP_END;
6333              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6334              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6335              code += 1 + LINK_SIZE;
6336              len += 1 + LINK_SIZE;
6337              tempcode[0] = OP_ONCE;
6338              *code++ = OP_KET;
6339              PUTINC(code, 0, len);
6340              PUT(tempcode, 1, len);
6341              }
6342            }
6343    
6344    #ifdef NEVER
6345        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6346          {          {
6347          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5609  for (;; ptr++) Line 6369  for (;; ptr++)
6369          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6370          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6371    
6372            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6373            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6374            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6375            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6376    
6377          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6378          pending recursive references are updated. */          pending recursive references are updated. */
6379    
# Line 5624  for (;; ptr++) Line 6389  for (;; ptr++)
6389          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6390          break;          break;
6391          }          }
6392    #endif
6393        }        }
6394    
6395      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 5812  for (;; ptr++) Line 6578  for (;; ptr++)
6578          tempptr = ptr;          tempptr = ptr;
6579    
6580          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6581          group), a name (referring to a named group), or 'R', referring to          group's having been set), a name (referring to a named group), or 'R',
6582          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6583            recursion tests.
6584          There are several syntaxes for testing a named group: (?(name)) is used  
6585          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6586            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6587          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6588          be the recursive thing or the name 'R' (and similarly for 'R' followed          There is one unfortunate ambiguity, caused by history. 'R' can be the
6589          by digits), and (b) a number could be a name that consists of digits.          recursive thing or the name 'R' (and similarly for 'R' followed by
6590          In both cases, we look for a name first; if not found, we try the other          digits). We look for a name first; if not found, we try the other case.
         cases.  
6591    
6592          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6593          specified before a condition that is an assertion. First, check for the          specified before a condition that is an assertion. First, check for the
# Line 5846  for (;; ptr++) Line 6611  for (;; ptr++)
6611                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6612            break;            break;
6613    
6614          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6615          below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */          need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6616    
6617          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6618          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6619          refsign = -1;          refsign = -1;     /* => not a number */
6620            namelen = -1;     /* => not a name; must set to avoid warning */
6621            name = NULL;      /* Always set to avoid warning */
6622            recno = 0;        /* Always set to avoid warning */
6623    
6624          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6625    
6626          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6627            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6628            {            {
6629            terminator = -1;            terminator = -1;
6630            ptr += 2;            ptr += 2;
# Line 5863  for (;; ptr++) Line 6632  for (;; ptr++)
6632            }            }
6633    
6634          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
6635          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6636            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6637    
6638          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6639            {            {
6640            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6641            ptr++;            ptr++;
6642            }            }
6643          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6644            {            {
6645            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6646            ptr++;            ptr++;
# Line 5878  for (;; ptr++) Line 6648  for (;; ptr++)
6648          else          else
6649            {            {
6650            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6651            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6652                else if (IS_DIGIT(*ptr)) refsign = 0;
6653            }            }
6654    
6655          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6656    
6657          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6658            {            {
6659            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6660            *errorcodeptr = ERR28;              {
6661            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6662                ptr++;
6663                }
6664            }            }
6665    
6666          /* Read the name, but also get it as a number if it's all digits */          /* Otherwise we expect to read a name; anything else is an error. When
6667            a name is one of a number of duplicates, a different opcode is used and
6668            it needs more memory. Unfortunately we cannot tell whether a name is a
6669            duplicate in the first pass, so we have to allow for more memory. */
6670    
6671          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6672            {            {
6673            if (recno >= 0)            if (IS_DIGIT(*ptr))
6674              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6675            ptr++;              *errorcodeptr = ERR84;
6676                goto FAILED;
6677                }
6678              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6679                {
6680                *errorcodeptr = ERR28;   /* Assertion expected */
6681                goto FAILED;
6682                }
6683              name = ptr++;
6684              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6685                {
6686                ptr++;
6687                }
6688              namelen = (int)(ptr - name);
6689              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6690            }            }
6691          namelen = (int)(ptr - name);  
6692            /* Check the terminator */
6693    
6694          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6695              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6696            {            {
6697            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6698            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6699            goto FAILED;            goto FAILED;
6700            }            }
6701    
# Line 5915  for (;; ptr++) Line 6704  for (;; ptr++)
6704          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6705    
6706          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
6707          reference. If the string started with "+" or "-" we require the rest to          reference. If refsign is not negative, it means we have a number in
6708          be digits, in which case recno will be set. */          recno. */
6709    
6710          if (refsign > 0)          if (refsign >= 0)
6711            {            {
6712            if (recno <= 0)            if (recno <= 0)
6713              {              {
6714              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6715              goto FAILED;              goto FAILED;
6716              }              }
6717            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6718              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6719            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6720              {              {
6721              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 5936  for (;; ptr++) Line 6725  for (;; ptr++)
6725            break;            break;
6726            }            }
6727    
6728          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise look for the name. */
         name. If we find a name, add one to the opcode to change OP_CREF or