/[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 1446 by ph10, Mon Jan 13 17:00:49 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    ;    ;
551    
552  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 648  static const pcre_uint8 ebcdic_chartab[] Line 686  static const pcre_uint8 ebcdic_chartab[]
686  #endif  #endif
687    
688    
689    /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726    
727    /* This table is used to check whether auto-possessification is possible
728    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729    left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
782      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
783      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
784      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
867  /*************************************************  /*************************************************
868  *            Find an error text                  *  *            Find an error text                  *
# Line 675  return s; Line 890  return s;
890  }  }
891    
892    
893    
894  /*************************************************  /*************************************************
895  *           Expand the workspace                 *  *           Expand the workspace                 *
896  *************************************************/  *************************************************/
# Line 752  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 968  return (*p == CHAR_RIGHT_CURLY_BRACKET);
968  *************************************************/  *************************************************/
969    
970  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
971  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
972  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
973  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
974  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979    chptr          points to the data character    chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
# Line 965  else Line 1180  else
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1188      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1189      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1190      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1191      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195        Inside a character class, \ followed by a digit is always either a literal
1196        8 or 9 or an octal number. */
1197    
1198      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 1001  else Line 1220  else
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          escape = -s;          escape = -s;
1226          break;          break;
# Line 1009  else Line 1228  else
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1232      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1241      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1035  else Line 1252  else
1252  #endif  #endif
1253      break;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1257      If not, { is treated as a data character. */  
1258        case CHAR_o:
1259        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260          {
1261          ptr += 2;
1262          c = 0;
1263          overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265            {
1266            register pcre_uint32 cc = *ptr++;
1267            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268    #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270    #endif
1271            c = (c << 3) + cc - CHAR_0 ;
1272    #if defined COMPILE_PCRE8
1273            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274    #elif defined COMPILE_PCRE16
1275            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278    #endif
1279            }
1280          if (overflow)
1281            {
1282            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283            *errorcodeptr = ERR34;
1284            }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286            {
1287            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290          }
1291        break;
1292    
1293        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296      case CHAR_x:      case CHAR_x:
1297      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1299        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301          {          {
# Line 1060  else Line 1312  else
1312  #endif  #endif
1313            }            }
1314          }          }
1315        break;        }    /* End JavaScript handling */
       }  
1316    
1317      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1320        seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        c = 0;      else
1325        overflow = FALSE;        {
1326        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          register pcre_uint32 cc = *pt++;          ptr += 2;
1329          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1330            overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1337          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1338  #endif  #endif
1339    
1340  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1344          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1345          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346  #endif  #endif
1347    
1348  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1349          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1351          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1353          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354  #endif  #endif
1355          }            }
1356    
1357        if (overflow)          if (overflow)
1358          {            {
1359          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1361          }            }
1362    
1363        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364          {            {
1365          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366          ptr = pt;            }
         break;  
         }  
1367    
1368        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1369        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1370        }          \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1378        {        else
1379        pcre_uint32 cc;                          /* Some compilers don't like */          {
1380        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1389        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391  #endif  #endif
1392        }            }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1192  if ((options & PCRE_UCP) != 0 && escape Line 1457  if ((options & PCRE_UCP) != 0 && escape
1457  return escape;  return escape;
1458  }  }
1459    
1460    
1461    
1462  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1463  /*************************************************  /*************************************************
1464  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1289  return FALSE; Line 1556  return FALSE;
1556    
1557    
1558    
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1358  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK)  
     {  
     ptr += 2;  
     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;  
 }  
   
   
   
   
 /*************************************************  
1627  *      Find first significant op code            *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
# Line 1696  for (;;) Line 1662  for (;;)
1662    
1663      case OP_CALLOUT:      case OP_CALLOUT:
1664      case OP_CREF:      case OP_CREF:
1665      case OP_NCREF:      case OP_DNCREF:
1666      case OP_RREF:      case OP_RREF:
1667      case OP_NRREF:      case OP_DNRREF:
1668      case OP_DEF:      case OP_DEF:
1669      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1670      break;      break;
# Line 1712  for (;;) Line 1678  for (;;)
1678    
1679    
1680    
   
1681  /*************************************************  /*************************************************
1682  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1683  *************************************************/  *************************************************/
# Line 1836  for (;;) Line 1801  for (;;)
1801      case OP_COMMIT:      case OP_COMMIT:
1802      case OP_CREF:      case OP_CREF:
1803      case OP_DEF:      case OP_DEF:
1804        case OP_DNCREF:
1805        case OP_DNRREF:
1806      case OP_DOLL:      case OP_DOLL:
1807      case OP_DOLLM:      case OP_DOLLM:
1808      case OP_EOD:      case OP_EOD:
1809      case OP_EODN:      case OP_EODN:
1810      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812      case OP_PRUNE:      case OP_PRUNE:
1813      case OP_REVERSE:      case OP_REVERSE:
# Line 1937  for (;;) Line 1902  for (;;)
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918          case OP_CRPOSRANGE:
1919        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1921        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2015  for (;;) Line 1984  for (;;)
1984      case OP_QUERYI:      case OP_QUERYI:
1985      case OP_REF:      case OP_REF:
1986      case OP_REFI:      case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989      case OP_SBRA:      case OP_SBRA:
1990      case OP_SBRAPOS:      case OP_SBRAPOS:
1991      case OP_SCBRA:      case OP_SCBRA:
# Line 2051  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 2361  Returns:      TRUE if what is matched co Line 2331  Returns:      TRUE if what is matched co
2331  typedef struct recurse_check {  typedef struct recurse_check {
2332    struct recurse_check *prev;    struct recurse_check *prev;
2333    const pcre_uchar *group;    const pcre_uchar *group;
2334  } recurse_check;  } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
# Line 2377  for (code = first_significant_code(code Line 2347  for (code = first_significant_code(code
2347    const pcre_uchar *ccode;    const pcre_uchar *ccode;
2348    
2349    c = *code;    c = *code;
2350    
2351    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2352    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2353    
# Line 2405  for (code = first_significant_code(code Line 2375  for (code = first_significant_code(code
2375      NULL. */      NULL. */
2376    
2377      if (cd->start_workspace != NULL)      if (cd->start_workspace != NULL)
2378        {        {
2379        const pcre_uchar *tcode;        const pcre_uchar *tcode;
2380        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383        }        }
2384    
2385      /* If we are scanning a completed pattern, there are no forward references      /* If we are scanning a completed pattern, there are no forward references
2386      and all groups are complete. We need to detect whether this is a recursive      and all groups are complete. We need to detect whether this is a recursive
2387      call, as otherwise there will be an infinite loop. If it is a recursion,      call, as otherwise there will be an infinite loop. If it is a recursion,
2388      just skip over it. Simple recursions are easily detected. For mutual      just skip over it. Simple recursions are easily detected. For mutual
2389      recursions we keep a chain on the stack. */      recursions we keep a chain on the stack. */
2390    
2391      else      else
2392        {        {
2393        recurse_check *r = recurses;        recurse_check *r = recurses;
2394        const pcre_uchar *endgroup = scode;        const pcre_uchar *endgroup = scode;
2395    
2396        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2400          if (r->group == scode) break;          if (r->group == scode) break;
2401        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2436  for (code = first_significant_code(code Line 2406  for (code = first_significant_code(code
2406    
2407      empty_branch = FALSE;      empty_branch = FALSE;
2408      this_recurse.prev = recurses;      this_recurse.prev = recurses;
2409      this_recurse.group = scode;      this_recurse.group = scode;
2410    
2411      do      do
2412        {        {
2413        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 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 2557  for (code = first_significant_code(code Line 2531  for (code = first_significant_code(code
2531      case OP_ANY:      case OP_ANY:
2532      case OP_ALLANY:      case OP_ALLANY:
2533      case OP_ANYBYTE:      case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      case OP_NOTPROP:
2537      case OP_ANYNL:      case OP_ANYNL:
2538    
2539      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2540      case OP_HSPACE:      case OP_HSPACE:
2541      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2542      case OP_VSPACE:      case OP_VSPACE:
2543      case OP_EXTUNI:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551    
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARI:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555      case OP_NOTI:      case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558      case OP_PLUSI:      case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
# Line 2589  for (code = first_significant_code(code Line 2563  for (code = first_significant_code(code
2563      case OP_NOTPLUSI:      case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565      case OP_NOTMINPLUSI:      case OP_NOTMINPLUSI:
2566    
2567      case OP_POSPLUS:      case OP_POSPLUS:
2568      case OP_POSPLUSI:      case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570      case OP_NOTPOSPLUSI:      case OP_NOTPOSPLUSI:
2571    
2572      case OP_EXACT:      case OP_EXACT:
2573      case OP_EXACTI:      case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575      case OP_NOTEXACTI:      case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2644  for (code = first_significant_code(code Line 2618  for (code = first_significant_code(code
2618  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619      case OP_STAR:      case OP_STAR:
2620      case OP_STARI:      case OP_STARI:
2621      case OP_NOTSTAR:      case OP_NOTSTAR:
2622      case OP_NOTSTARI:      case OP_NOTSTARI:
2623    
2624      case OP_MINSTAR:      case OP_MINSTAR:
2625      case OP_MINSTARI:      case OP_MINSTARI:
2626      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
2627      case OP_NOTMINSTARI:      case OP_NOTMINSTARI:
2628    
2629      case OP_POSSTAR:      case OP_POSSTAR:
2630      case OP_POSSTARI:      case OP_POSSTARI:
2631      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
2632      case OP_NOTPOSSTARI:      case OP_NOTPOSSTARI:
2633    
2634      case OP_QUERY:      case OP_QUERY:
2635      case OP_QUERYI:      case OP_QUERYI:
2636      case OP_NOTQUERY:      case OP_NOTQUERY:
2637      case OP_NOTQUERYI:      case OP_NOTQUERYI:
2638    
2639      case OP_MINQUERY:      case OP_MINQUERY:
2640      case OP_MINQUERYI:      case OP_MINQUERYI:
2641      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
2642      case OP_NOTMINQUERYI:      case OP_NOTMINQUERYI:
2643    
2644      case OP_POSQUERY:      case OP_POSQUERY:
2645      case OP_POSQUERYI:      case OP_POSQUERYI:
2646      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
2647      case OP_NOTPOSQUERYI:      case OP_NOTPOSQUERYI:
2648    
2649      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650      break;      break;
2651    
2652      case OP_UPTO:      case OP_UPTO:
2653      case OP_UPTOI:      case OP_UPTOI:
2654      case OP_NOTUPTO:      case OP_NOTUPTO:
2655      case OP_NOTUPTOI:      case OP_NOTUPTOI:
2656    
2657      case OP_MINUPTO:      case OP_MINUPTO:
2658      case OP_MINUPTOI:      case OP_MINUPTOI:
2659      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
2660      case OP_NOTMINUPTOI:      case OP_NOTMINUPTOI:
2661    
2662      case OP_POSUPTO:      case OP_POSUPTO:
2663      case OP_POSUPTOI:      case OP_POSUPTOI:
2664      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
2665      case OP_NOTPOSUPTOI:      case OP_NOTPOSUPTOI:
2666    
2667      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668      break;      break;
2669  #endif  #endif
# Line 2753  return TRUE; Line 2727  return TRUE;
2727    
2728    
2729  /*************************************************  /*************************************************
2730  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2731  *************************************************/  *************************************************/
2732    
2733  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2734  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2735    
2736  Returns:   TRUE or FALSE  Arguments:  c opcode
2737    Returns:    base opcode for the type
2738  */  */
2739    
2740  static BOOL  static pcre_uchar
2741  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2742  {  {
2743  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2744  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745  for (++ptr; *ptr != CHAR_NULL; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2748      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2749  }  }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754  /*************************************************  /*************************************************
2755  *          Check POSIX class name                *  *        Check a character and a property        *
2756  *************************************************/  *************************************************/
2757    
2758  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2759  such as [:alnum:].  is adjacent to a fixed character.
2760    
2761  Arguments:  Arguments:
2762    ptr        points to the first letter    c            the character
2763    len        the length of the name    ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2768  */  */
2769    
2770  static int  static BOOL
2771  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773  {  {
2774  const char *pn = posix_names;  const pcre_uint32 *p;
2775  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2776    
2777  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2778  that is referenced. This means that groups can be replicated for fixed    {
2779  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2780  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2781  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2782  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2783    
2784  This function has been extended with the possibility of forward references for    case PT_GC:
2785  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2786    
2787  Arguments:    case PT_PC:
2788    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2789    
2790  Returns:     nothing    case PT_SC:
2791  */    return (pdata == prop->script) == negated;
2792    
2793  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2794    
2795  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2796    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2798    
2799    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806      {      {
2807      offset = (int)GET(hc, 0);      HSPACE_CASES:
2808      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2809        {      return negated;
2810        PUT(hc, 0, offset + adjust);  
2811        break;      default:
2812        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813      }      }
2814      break;  /* Control never reaches here */
2815    
2816    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2817    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821    if (hc >= cd->hwm)    case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824      {      {
2825      offset = (int)GET(ptr, 1);      if (c < *p) return !negated;
2826      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2827      }      }
2828      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2829    }    }
2830    
2831    return FALSE;
2832  }  }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837  /*************************************************  /*************************************************
2838  *        Insert an automatic callout point       *  *        Fill the character property list        *
2839  *************************************************/  *************************************************/
2840    
2841  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2842  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2843    
2844  Arguments:  Arguments:
2845    code           current code pointer    code        points to start of expression
2846    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847    cd             pointers to tables etc    fcc         points to case-flipping table
2848      list        points to output list
2849                  list[0] will be filled with the opcode
2850                  list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852                  list[2..7] depends on the opcode
2853    
2854  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856  */  */
2857    
2858  static pcre_uchar *  static const pcre_uchar *
2859  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861  {  {
2862  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2863  *code++ = 255;  pcre_uchar base;
2864  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uchar *end;
2865  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 chr;
 return code + 2 * LINK_SIZE;  
 }  
   
2866    
2867    #ifdef SUPPORT_UCP
2868    pcre_uint32 *clist_dest;
2869    const pcre_uint32 *clist_src;
2870    #else
2871    utf = utf;  /* Suppress "unused parameter" compiler warning */
2872    #endif
2873    
2874  /*************************************************  list[0] = c;
2875  *         Complete a callout item                *  list[1] = FALSE;
2876  *************************************************/  code++;
2877    
2878  /* A callout item contains the length of the next item in the pattern, which  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879  we can't fill in till after we have reached the relevant point. This is used    {
2880  for both automatic and manual callouts.    base = get_repeat_base(c);
2881      c -= (base - OP_STAR);
2882    
2883  Arguments:    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884    previous_callout   points to previous callout item      code += IMM2_SIZE;
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
2885    
2886  Returns:             nothing    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 */  
2887    
2888  static void    switch(base)
2889  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      {
2890  {      case OP_STAR:
2891  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));      list[0] = OP_CHAR;
2892  PUT(previous_callout, 2 + LINK_SIZE, length);      break;
 }  
2893    
2894        case OP_STARI:
2895        list[0] = OP_CHARI;
2896        break;
2897    
2898        case OP_NOTSTAR:
2899        list[0] = OP_NOT;
2900        break;
2901    
2902  #ifdef SUPPORT_UCP      case OP_NOTSTARI:
2903  /*************************************************      list[0] = OP_NOTI;
2904  *           Get othercase range                  *      break;
 *************************************************/  
2905    
2906  /* This function is passed the start and end of a class range, in UTF-8 mode      case OP_TYPESTAR:
2907  with UCP support. It searches up the characters, looking for ranges of      list[0] = *code;
2908  characters in the "other" case. Each call returns the next one, updating the      code++;
2909  start address. A character with multiple other cases is returned on its own      break;
2910  with a special return value.      }
2911      c = list[0];
2912      }
2913    
2914  Arguments:  switch(c)
2915    cptr        points to starting character value; updated    {
2916    d           end value    case OP_NOT_DIGIT:
2917    ocptr       where to put start of othercase range    case OP_DIGIT:
2918    odptr       where to put end of othercase range    case OP_NOT_WHITESPACE:
2919      case OP_WHITESPACE:
2920      case OP_NOT_WORDCHAR:
2921      case OP_WORDCHAR:
2922      case OP_ANY:
2923      case OP_ALLANY:
2924      case OP_ANYNL:
2925      case OP_NOT_HSPACE:
2926      case OP_HSPACE:
2927      case OP_NOT_VSPACE:
2928      case OP_VSPACE:
2929      case OP_EXTUNI:
2930      case OP_EODN:
2931      case OP_EOD:
2932      case OP_DOLL:
2933      case OP_DOLLM:
2934      return code;
2935    
2936  Yield:        -1 when no more    case OP_CHAR:
2937                 0 when a range is returned    case OP_NOT:
2938                >0 the CASESET offset for char with multiple other cases    GETCHARINCTEST(chr, code);
2939                  in this case, ocptr contains the original    list[2] = chr;
2940  */    list[3] = NOTACHAR;
2941      return code;
2942    
2943  static int    case OP_CHARI:
2944  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,    case OP_NOTI:
2945    pcre_uint32 *odptr)    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946  {    GETCHARINCTEST(chr, code);
2947  pcre_uint32 c, othercase, next;    list[2] = chr;
 unsigned int co;  
2948    
2949  /* Find the first character that has an other case. If it has multiple other  #ifdef SUPPORT_UCP
2950  cases, return its case offset value. */    if (chr < 128 || (chr < 256 && !utf))
2951        list[3] = fcc[chr];
2952      else
2953        list[3] = UCD_OTHERCASE(chr);
2954    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955      list[3] = (chr < 256) ? fcc[chr] : chr;
2956    #else
2957      list[3] = fcc[chr];
2958    #endif
2959    
2960  for (c = *cptr; c <= d; c++)    /* The othercase might be the same value. */
2961    {  
2962    if ((co = UCD_CASESET(c)) != 0)    if (chr == list[3])
2963        list[3] = NOTACHAR;
2964      else
2965        list[4] = NOTACHAR;
2966      return code;
2967    
2968    #ifdef SUPPORT_UCP
2969      case OP_PROP:
2970      case OP_NOTPROP:
2971      if (code[0] != PT_CLIST)
2972      {      {
2973      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2974      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2975      return (int)co;      return code + 2;
2976      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2977    
2978  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2979    
2980  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2981  next = othercase + 1;    clist_dest = list + 2;
2982      code += 2;
2983    
2984  for (++c; c <= d; c++)    do {
2985    {       if (clist_dest >= list + 8)
2986    if (UCD_OTHERCASE(c) != next) break;         {
2987    next++;         /* Early return if there is not enough space. This should never
2988    }         happen, since all clists are shorter than 5 character now. */
2989           list[2] = code[0];
2990           list[3] = code[1];
2991           return code;
2992           }
2993         *clist_dest++ = *clist_src;
2994         }
2995      while(*clist_src++ != NOTACHAR);
2996    
2997  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2998  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2999  return 0;  
3000      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001      return code;
3002    #endif
3003    
3004      case OP_NCLASS:
3005      case OP_CLASS:
3006    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007      case OP_XCLASS:
3008      if (c == OP_XCLASS)
3009        end = code + GET(code, 0) - 1;
3010      else
3011    #endif
3012        end = code + 32 / sizeof(pcre_uchar);
3013    
3014      switch(*end)
3015        {
3016        case OP_CRSTAR:
3017        case OP_CRMINSTAR:
3018        case OP_CRQUERY:
3019        case OP_CRMINQUERY:
3020        case OP_CRPOSSTAR:
3021        case OP_CRPOSQUERY:
3022        list[1] = TRUE;
3023        end++;
3024        break;
3025    
3026        case OP_CRPLUS:
3027        case OP_CRMINPLUS:
3028        case OP_CRPOSPLUS:
3029        end++;
3030        break;
3031    
3032        case OP_CRRANGE:
3033        case OP_CRMINRANGE:
3034        case OP_CRPOSRANGE:
3035        list[1] = (GET2(end, 1) == 0);
3036        end += 1 + 2 * IMM2_SIZE;
3037        break;
3038        }
3039      list[2] = end - code;
3040      return end;
3041      }
3042    return NULL;    /* Opcode not accepted */
3043  }  }
3044    
3045    
3046    
3047  /*************************************************  /*************************************************
3048  *        Check a character and a property        *  *    Scan further character sets for match       *
3049  *************************************************/  *************************************************/
3050    
3051  /* This function is called by check_auto_possessive() when a property item  /* Checks whether the base and the current opcode have a common character, in
3052  is adjacent to a fixed character.  which case the base cannot be possessified.
3053    
3054  Arguments:  Arguments:
3055    c            the character    code        points to the byte code
3056    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057    pdata        the data for the type    cd          static compile data
3058    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3059    
3060  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3061  */  */
3062    
3063  static BOOL  static BOOL
3064  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066  {  {
3067  #ifdef SUPPORT_UCP  pcre_uchar c;
3068  const pcre_uint32 *p;  pcre_uint32 list[8];
3069    const pcre_uint32 *chr_ptr;
3070    const pcre_uint32 *ochr_ptr;
3071    const pcre_uint32 *list_ptr;
3072    const pcre_uchar *next_code;
3073    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3074    const pcre_uchar *xclass_flags;
3075  #endif  #endif
3076    const pcre_uint8 *class_bitset;
3077    const pcre_uint8 *set1, *set2, *set_end;
3078    pcre_uint32 chr;
3079    BOOL accepted, invert_bits;
3080    
3081    /* Note: the base_list[1] contains whether the current opcode has greedy
3082    (represented by a non-zero value) quantifier. This is a different from
3083    other character type lists, which stores here that the character iterator
3084    matches to an empty string (also represented by a non-zero value). */
3085    
3086  const ucd_record *prop = GET_UCD(c);  for(;;)
   
 switch(ptype)  
3087    {    {
3088    case PT_LAMP:    /* All operations move the code pointer forward.
3089    return (prop->chartype == ucp_Lu ||    Therefore infinite recursions are not possible. */
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3090    
3091    case PT_GC:    c = *code;
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3092    
3093    case PT_PC:    /* Skip over callouts */
   return (pdata == prop->chartype) == negated;  
3094    
3095    case PT_SC:    if (c == OP_CALLOUT)
3096    return (pdata == prop->script) == negated;      {
3097        code += PRIV(OP_lengths)[c];
3098        continue;
3099        }
3100    
3101    /* These are specials */    if (c == OP_ALT)
3102        {
3103        do code += GET(code, 1); while (*code == OP_ALT);
3104        c = *code;
3105        }
3106    
3107    case PT_ALNUM:    switch(c)
3108    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      {
3109            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;      case OP_END:
3110        case OP_KETRPOS:
3111        /* TRUE only in greedy case. The non-greedy case could be replaced by
3112        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3113        uses more memory, which we cannot get at this stage.) */
3114    
3115    case PT_SPACE:    /* Perl space */      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3116    
3117    case PT_PXSPACE:  /* POSIX space */      case OP_KET:
3118    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3119            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3120            c == CHAR_FF || c == CHAR_CR)      cannot be converted to a possessive form. */
3121            == negated;  
3122        if (base_list[1] == 0) return FALSE;
3123    
3124        switch(*(code - GET(code, 1)))
3125          {
3126          case OP_ASSERT:
3127          case OP_ASSERT_NOT:
3128          case OP_ASSERTBACK:
3129          case OP_ASSERTBACK_NOT:
3130          case OP_ONCE:
3131          case OP_ONCE_NC:
3132          /* Atomic sub-patterns and assertions can always auto-possessify their
3133          last iterator. */
3134          return TRUE;
3135          }
3136    
3137    case PT_WORD:      code += PRIV(OP_lengths)[c];
3138    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      continue;
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
3139    
3140  #ifdef SUPPORT_UCP      case OP_ONCE:
3141    case PT_CLIST:      case OP_ONCE_NC:
3142    p = PRIV(ucd_caseless_sets) + prop->caseset;      case OP_BRA:
3143    for (;;)      case OP_CBRA:
3144      {      next_code = code + GET(code, 1);
3145      if (c < *p) return !negated;      code += PRIV(OP_lengths)[c];
     if (c == *p++) return negated;  
     }  
   break;  /* Control never reaches here */  
 #endif  
   }  
3146    
3147  return FALSE;      while (*next_code == OP_ALT)
3148  }        {
3149  #endif  /* SUPPORT_UCP */        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3150          code = next_code + 1 + LINK_SIZE;
3151          next_code += GET(next_code, 1);
3152          }
3153        continue;
3154    
3155        case OP_BRAZERO:
3156        case OP_BRAMINZERO:
3157    
3158        next_code = code + 1;
3159        if (*next_code != OP_BRA && *next_code != OP_CBRA
3160            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3161    
3162        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3163    
3164        /* The bracket content will be checked by the
3165        OP_BRA/OP_CBRA case above. */
3166        next_code += 1 + LINK_SIZE;
3167        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3168          return FALSE;
3169    
3170  /*************************************************      code += PRIV(OP_lengths)[c];
3171  *     Check if auto-possessifying is possible    *      continue;
3172  *************************************************/      }
3173    
3174  /* This function is called for unlimited repeats of certain items, to see    /* Check for a supported opcode, and load its properties. */
 whether the next thing could possibly match the repeated item. If not, it makes  
 sense to automatically possessify the repeated item.  
3175    
3176  Arguments:    code = get_chr_property_list(code, utf, cd->fcc, list);
3177    previous      pointer to the repeated opcode    if (code == NULL) return FALSE;    /* Unsupported */
   utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3178    
3179  Returns:        TRUE if possessifying is wanted    /* If either opcode is a small character list, set pointers for comparing
3180  */    characters from that list with another list, or with a property. */
3181    
3182  static BOOL    if (base_list[0] == OP_CHAR)
3183  check_auto_possessive(const pcre_uchar *previous, BOOL utf,      {
3184    const pcre_uchar *ptr, int options, compile_data *cd)      chr_ptr = base_list + 2;
3185  {      list_ptr = list;
3186  pcre_uint32 c = NOTACHAR;      }
3187  pcre_uint32 next;    else if (list[0] == OP_CHAR)
3188  int escape;      {
3189  pcre_uchar op_code = *previous++;      chr_ptr = list + 2;
3190        list_ptr = base_list;
3191        }
3192    
3193  /* Skip whitespace and comments in extended mode */    /* Character bitsets can also be compared to certain opcodes. */
3194    
3195  if ((options & PCRE_EXTENDED) != 0)    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3196    {  #ifdef COMPILE_PCRE8
3197    for (;;)        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3198          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3199    #endif
3200          )
3201      {      {
3202      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  #ifdef COMPILE_PCRE8
3203      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3204        {  #else
3205        ptr++;      if (base_list[0] == OP_CLASS)
       while (*ptr != CHAR_NULL)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
3206  #endif  #endif
3207          }        {
3208          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3209          list_ptr = list;
3210          }
3211        else
3212          {
3213          set1 = (pcre_uint8 *)(code - list[2]);
3214          list_ptr = base_list;
3215        }        }
     else break;  
     }  
   }  
3216    
3217  /* If the next item is one that we can handle, get its value. A non-negative      invert_bits = FALSE;
3218  value is a character, a negative value is an escape value. */      switch(list_ptr[0])
3219          {
3220          case OP_CLASS:
3221          case OP_NCLASS:
3222          set2 = (pcre_uint8 *)
3223            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3224          break;
3225    
3226  if (*ptr == CHAR_BACKSLASH)  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3227    {        case OP_XCLASS:
3228    int temperrorcode = 0;        xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3229    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,        if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3230      FALSE);        if ((*xclass_flags & XCL_MAP) == 0)
3231    if (temperrorcode != 0) return FALSE;          {
3232    ptr++;    /* Point after the escape sequence */          /* No bits are set for characters < 256. */
3233    }          if (list[1] == 0) return TRUE;
3234  else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)          /* Might be an empty repeat. */
3235    {          continue;
3236    escape = 0;          }
3237  #ifdef SUPPORT_UTF        set2 = (pcre_uint8 *)(xclass_flags + 1);
3238    if (utf) { GETCHARINC(next, ptr); } else        break;
3239  #endif  #endif
   next = *ptr++;  
   }  
 else return FALSE;  
3240    
3241  /* Skip whitespace and comments in extended mode */        case OP_NOT_DIGIT:
3242          invert_bits = TRUE;
3243          /* Fall through */
3244          case OP_DIGIT:
3245          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3246          break;
3247    
3248  if ((options & PCRE_EXTENDED) != 0)        case OP_NOT_WHITESPACE:
3249    {        invert_bits = TRUE;
3250    for (;;)        /* Fall through */
3251      {        case OP_WHITESPACE:
3252      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;        set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3253      if (*ptr == CHAR_NUMBER_SIGN)        break;
3254    
3255          case OP_NOT_WORDCHAR:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_WORDCHAR:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3260          break;
3261    
3262          default:
3263          return FALSE;
3264          }
3265    
3266        /* Because the sets are unaligned, we need
3267        to perform byte comparison here. */
3268        set_end = set1 + 32;
3269        if (invert_bits)
3270        {        {
3271        ptr++;        do
       while (*ptr != CHAR_NULL)  
3272          {          {
3273          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3274          ptr++;          }
3275  #ifdef SUPPORT_UTF        while (set1 < set_end);
3276          if (utf) FORWARDCHAR(ptr);        }
3277  #endif      else
3278          {
3279          do
3280            {
3281            if ((*set1++ & *set2++) != 0) return FALSE;
3282          }          }
3283          while (set1 < set_end);
3284        }        }
3285      else break;  
3286        if (list[1] == 0) return TRUE;
3287        /* Might be an empty repeat. */
3288        continue;
3289      }      }
   }  
3290    
3291  /* If the next thing is itself optional, we have to give up. */    /* Some property combinations also acceptable. Unicode property opcodes are
3292      processed specially; the rest can be handled with a lookup table. */
3293    
3294  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||    else
3295    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      {
3296      return FALSE;      pcre_uint32 leftop, rightop;
3297    
3298  /* If the previous item is a character, get its value. */      leftop = base_list[0];
3299        rightop = list[0];
3300    
3301  if (op_code == OP_CHAR || op_code == OP_CHARI ||  #ifdef SUPPORT_UCP
3302      op_code == OP_NOT || op_code == OP_NOTI)      accepted = FALSE; /* Always set in non-unicode case. */
3303    {      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3304  #ifdef SUPPORT_UTF        {
3305    GETCHARTEST(c, previous);        if (rightop == OP_EOD)
3306  #else          accepted = TRUE;
3307    c = *previous;        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3308  #endif          {
3309    }          int n;
3310            const pcre_uint8 *p;
3311            BOOL same = leftop == rightop;
3312            BOOL lisprop = leftop == OP_PROP;
3313            BOOL risprop = rightop == OP_PROP;
3314            BOOL bothprop = lisprop && risprop;
3315    
3316            /* There's a table that specifies how each combination is to be
3317            processed:
3318              0   Always return FALSE (never auto-possessify)
3319              1   Character groups are distinct (possessify if both are OP_PROP)
3320              2   Check character categories in the same group (general or particular)
3321              3   Return TRUE if the two opcodes are not the same
3322              ... see comments below
3323            */
3324    
3325            n = propposstab[base_list[2]][list[2]];
3326            switch(n)
3327              {
3328              case 0: break;
3329              case 1: accepted = bothprop; break;
3330              case 2: accepted = (base_list[3] == list[3]) != same; break;
3331              case 3: accepted = !same; break;
3332    
3333  /* Now compare the next item with the previous opcode. First, handle cases when            case 4:  /* Left general category, right particular category */
3334  the next item is a character. */            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3335              break;
3336    
3337  if (escape == 0)            case 5:  /* Right general category, left particular category */
3338    {            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3339    /* For a caseless UTF match, the next character may have more than one other            break;
   case, which maps to the special PT_CLIST property. Check this first. */  
3340    
3341  #ifdef SUPPORT_UCP            /* This code is logically tricky. Think hard before fiddling with it.
3342    if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)            The posspropstab table has four entries per row. Each row relates to
3343      {            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3344      unsigned int ocs = UCD_CASESET(next);            Only WORD actually needs all four entries, but using repeats for the
3345      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);            others means they can all use the same code below.
3346      }  
3347  #endif            The first two entries in each row are Unicode general categories, and
3348              apply always, because all the characters they include are part of the
3349              PCRE character set. The third and fourth entries are a general and a
3350              particular category, respectively, that include one or more relevant
3351              characters. One or the other is used, depending on whether the check
3352              is for a general or a particular category. However, in both cases the
3353              category contains more characters than the specials that are defined
3354              for the property being tested against. Therefore, it cannot be used
3355              in a NOTPROP case.
3356    
3357              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3358              Underscore is covered by ucp_P or ucp_Po. */
3359    
3360              case 6:  /* Left alphanum vs right general category */
3361              case 7:  /* Left space vs right general category */
3362              case 8:  /* Left word vs right general category */
3363              p = posspropstab[n-6];
3364              accepted = risprop && lisprop ==
3365                (list[3] != p[0] &&
3366                 list[3] != p[1] &&
3367                (list[3] != p[2] || !lisprop));
3368              break;
3369    
3370    switch(op_code)            case 9:   /* Right alphanum vs left general category */
3371      {            case 10:  /* Right space vs left general category */
3372      case OP_CHAR:            case 11:  /* Right word vs left general category */
3373      return c != next;            p = posspropstab[n-9];
3374              accepted = lisprop && risprop ==
3375                (base_list[3] != p[0] &&
3376                 base_list[3] != p[1] &&
3377                (base_list[3] != p[2] || !risprop));
3378              break;
3379    
3380      /* For CHARI (caseless character) we must check the other case. If we have            case 12:  /* Left alphanum vs right particular category */
3381      Unicode property support, we can use it to test the other case of            case 13:  /* Left space vs right particular category */
3382      high-valued characters. We know that next can have only one other case,            case 14:  /* Left word vs right particular category */
3383      because multi-other-case characters are dealt with above. */            p = posspropstab[n-12];
3384              accepted = risprop && lisprop ==
3385                (catposstab[p[0]][list[3]] &&
3386                 catposstab[p[1]][list[3]] &&
3387                (list[3] != p[3] || !lisprop));
3388              break;
3389    
3390      case OP_CHARI:            case 15:  /* Right alphanum vs left particular category */
3391      if (c == next) return FALSE;            case 16:  /* Right space vs left particular category */
3392  #ifdef SUPPORT_UTF            case 17:  /* Right word vs left particular category */
3393      if (utf)            p = posspropstab[n-15];
3394              accepted = lisprop && risprop ==
3395                (catposstab[p[0]][base_list[3]] &&
3396                 catposstab[p[1]][base_list[3]] &&
3397                (base_list[3] != p[3] || !risprop));
3398              break;
3399              }
3400            }
3401          }
3402    
3403        else
3404    #endif  /* SUPPORT_UCP */
3405    
3406        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3407               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3408               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3409    
3410        if (!accepted)
3411          return FALSE;
3412    
3413        if (list[1] == 0) return TRUE;
3414        /* Might be an empty repeat. */
3415        continue;
3416        }
3417    
3418      /* Control reaches here only if one of the items is a small character list.
3419      All characters are checked against the other side. */
3420    
3421      do
3422        {
3423        chr = *chr_ptr;
3424    
3425        switch(list_ptr[0])
3426        {        {
3427        pcre_uint32 othercase;        case OP_CHAR:
3428        if (next < 128) othercase = cd->fcc[next]; else        ochr_ptr = list_ptr + 2;
3429          do
3430            {
3431            if (chr == *ochr_ptr) return FALSE;
3432            ochr_ptr++;
3433            }
3434          while(*ochr_ptr != NOTACHAR);
3435          break;
3436    
3437          case OP_NOT:
3438          ochr_ptr = list_ptr + 2;
3439          do
3440            {
3441            if (chr == *ochr_ptr)
3442              break;
3443            ochr_ptr++;
3444            }
3445          while(*ochr_ptr != NOTACHAR);
3446          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3447          break;
3448    
3449          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3450          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3451    
3452          case OP_DIGIT:
3453          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3454          break;
3455    
3456          case OP_NOT_DIGIT:
3457          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3458          break;
3459    
3460          case OP_WHITESPACE:
3461          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3462          break;
3463    
3464          case OP_NOT_WHITESPACE:
3465          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3466          break;
3467    
3468          case OP_WORDCHAR:
3469          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3470          break;
3471    
3472          case OP_NOT_WORDCHAR:
3473          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3474          break;
3475    
3476          case OP_HSPACE:
3477          switch(chr)
3478            {
3479            HSPACE_CASES: return FALSE;
3480            default: break;
3481            }
3482          break;
3483    
3484          case OP_NOT_HSPACE:
3485          switch(chr)
3486            {
3487            HSPACE_CASES: break;
3488            default: return FALSE;
3489            }
3490          break;
3491    
3492          case OP_ANYNL:
3493          case OP_VSPACE:
3494          switch(chr)
3495            {
3496            VSPACE_CASES: return FALSE;
3497            default: break;
3498            }
3499          break;
3500    
3501          case OP_NOT_VSPACE:
3502          switch(chr)
3503            {
3504            VSPACE_CASES: break;
3505            default: return FALSE;
3506            }
3507          break;
3508    
3509          case OP_DOLL:
3510          case OP_EODN:
3511          switch (chr)
3512            {
3513            case CHAR_CR:
3514            case CHAR_LF:
3515            case CHAR_VT:
3516            case CHAR_FF:
3517            case CHAR_NEL:
3518    #ifndef EBCDIC
3519            case 0x2028:
3520            case 0x2029:
3521    #endif  /* Not EBCDIC */
3522            return FALSE;
3523            }
3524          break;
3525    
3526          case OP_EOD:    /* Can always possessify before \z */
3527          break;
3528    
3529  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3530        othercase = UCD_OTHERCASE(next);        case OP_PROP:
3531  #else        case OP_NOTPROP:
3532        othercase = NOTACHAR;        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3533                list_ptr[0] == OP_NOTPROP))
3534            return FALSE;
3535          break;
3536    #endif
3537    
3538          case OP_NCLASS:
3539          if (chr > 255) return FALSE;
3540          /* Fall through */
3541    
3542          case OP_CLASS:
3543          if (chr > 255) break;
3544          class_bitset = (pcre_uint8 *)
3545            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3546          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3547          break;
3548    
3549    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3550          case OP_XCLASS:
3551          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3552              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3553          break;
3554  #endif  #endif
3555        return c != othercase;  
3556          default:
3557          return FALSE;
3558          }
3559    
3560        chr_ptr++;
3561        }
3562      while(*chr_ptr != NOTACHAR);
3563    
3564      /* At least one character must be matched from this opcode. */
3565    
3566      if (list[1] == 0) return TRUE;
3567      }
3568    
3569    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3570    but some compilers complain about an unreachable statement. */
3571    
3572    }
3573    
3574    
3575    
3576    /*************************************************
3577    *    Scan compiled regex for auto-possession     *
3578    *************************************************/
3579    
3580    /* Replaces single character iterations with their possessive alternatives
3581    if appropriate. This function modifies the compiled opcode!
3582    
3583    Arguments:
3584      code        points to start of the byte code
3585      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3586      cd          static compile data
3587    
3588    Returns:      nothing
3589    */
3590    
3591    static void
3592    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3593    {
3594    register pcre_uchar c;
3595    const pcre_uchar *end;
3596    pcre_uchar *repeat_opcode;
3597    pcre_uint32 list[8];
3598    
3599    for (;;)
3600      {
3601      c = *code;
3602    
3603      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3604        {
3605        c -= get_repeat_base(c) - OP_STAR;
3606        end = (c <= OP_MINUPTO) ?
3607          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3608        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3609    
3610        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3611          {
3612          switch(c)
3613            {
3614            case OP_STAR:
3615            *code += OP_POSSTAR - OP_STAR;
3616            break;
3617    
3618            case OP_MINSTAR:
3619            *code += OP_POSSTAR - OP_MINSTAR;
3620            break;
3621    
3622            case OP_PLUS:
3623            *code += OP_POSPLUS - OP_PLUS;
3624            break;
3625    
3626            case OP_MINPLUS:
3627            *code += OP_POSPLUS - OP_MINPLUS;
3628            break;
3629    
3630            case OP_QUERY:
3631            *code += OP_POSQUERY - OP_QUERY;
3632            break;
3633    
3634            case OP_MINQUERY:
3635            *code += OP_POSQUERY - OP_MINQUERY;
3636            break;
3637    
3638            case OP_UPTO:
3639            *code += OP_POSUPTO - OP_UPTO;
3640            break;
3641    
3642            case OP_MINUPTO:
3643            *code += OP_MINUPTO - OP_UPTO;
3644            break;
3645            }
3646        }        }
3647        c = *code;
3648        }
3649      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3650        {
3651    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3652        if (c == OP_XCLASS)
3653          repeat_opcode = code + GET(code, 1);
3654      else      else
3655  #endif  /* SUPPORT_UTF */  #endif
3656      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */        repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3657    
3658      case OP_NOT:      c = *repeat_opcode;
3659      return c == next;      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3660          {
3661          /* end must not be NULL. */
3662          end = get_chr_property_list(code, utf, cd->fcc, list);
3663    
3664          list[1] = (c & 1) == 0;
3665    
3666          if (compare_opcodes(end, utf, cd, list, end))
3667            {
3668            switch (c)
3669              {
3670              case OP_CRSTAR:
3671              case OP_CRMINSTAR:
3672              *repeat_opcode = OP_CRPOSSTAR;
3673              break;
3674    
3675              case OP_CRPLUS:
3676              case OP_CRMINPLUS:
3677              *repeat_opcode = OP_CRPOSPLUS;
3678              break;
3679    
3680              case OP_CRQUERY:
3681              case OP_CRMINQUERY:
3682              *repeat_opcode = OP_CRPOSQUERY;
3683              break;
3684    
3685              case OP_CRRANGE:
3686              case OP_CRMINRANGE:
3687              *repeat_opcode = OP_CRPOSRANGE;
3688              break;
3689              }
3690            }
3691          }
3692        c = *code;
3693        }
3694    
3695      switch(c)
3696        {
3697        case OP_END:
3698        return;
3699    
3700        case OP_TYPESTAR:
3701        case OP_TYPEMINSTAR:
3702        case OP_TYPEPLUS:
3703        case OP_TYPEMINPLUS:
3704        case OP_TYPEQUERY:
3705        case OP_TYPEMINQUERY:
3706        case OP_TYPEPOSSTAR:
3707        case OP_TYPEPOSPLUS:
3708        case OP_TYPEPOSQUERY:
3709        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3710        break;
3711    
3712        case OP_TYPEUPTO:
3713        case OP_TYPEMINUPTO:
3714        case OP_TYPEEXACT:
3715        case OP_TYPEPOSUPTO:
3716        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3717          code += 2;
3718        break;
3719    
3720    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3721        case OP_XCLASS:
3722        code += GET(code, 1);
3723        break;
3724    #endif
3725    
3726        case OP_MARK:
3727        case OP_PRUNE_ARG:
3728        case OP_SKIP_ARG:
3729        case OP_THEN_ARG:
3730        code += code[1];
3731        break;
3732        }
3733    
3734      /* Add in the fixed length from the table */
3735    
3736      code += PRIV(OP_lengths)[c];
3737    
3738      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3739      a multi-byte character. The length in the table is a minimum, so we have to
3740      arrange to skip the extra bytes. */
3741    
3742    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3743      if (utf) switch(c)
3744        {
3745        case OP_CHAR:
3746        case OP_CHARI:
3747        case OP_NOT:
3748      case OP_NOTI:      case OP_NOTI:
3749      if (c == next) return TRUE;      case OP_STAR:
3750  #ifdef SUPPORT_UTF      case OP_MINSTAR:
3751      if (utf)      case OP_PLUS:
3752        {      case OP_MINPLUS:
3753        pcre_uint32 othercase;      case OP_QUERY:
3754        if (next < 128) othercase = cd->fcc[next]; else      case OP_MINQUERY:
3755  #ifdef SUPPORT_UCP      case OP_UPTO:
3756        othercase = UCD_OTHERCASE(next);      case OP_MINUPTO:
3757        case OP_EXACT:
3758        case OP_POSSTAR:
3759        case OP_POSPLUS:
3760        case OP_POSQUERY:
3761        case OP_POSUPTO:
3762        case OP_STARI:
3763        case OP_MINSTARI:
3764        case OP_PLUSI:
3765        case OP_MINPLUSI:
3766        case OP_QUERYI:
3767        case OP_MINQUERYI:
3768        case OP_UPTOI:
3769        case OP_MINUPTOI:
3770        case OP_EXACTI:
3771        case OP_POSSTARI:
3772        case OP_POSPLUSI:
3773        case OP_POSQUERYI:
3774        case OP_POSUPTOI:
3775        case OP_NOTSTAR:
3776        case OP_NOTMINSTAR:
3777        case OP_NOTPLUS:
3778        case OP_NOTMINPLUS:
3779        case OP_NOTQUERY:
3780        case OP_NOTMINQUERY:
3781        case OP_NOTUPTO:
3782        case OP_NOTMINUPTO:
3783        case OP_NOTEXACT:
3784        case OP_NOTPOSSTAR:
3785        case OP_NOTPOSPLUS:
3786        case OP_NOTPOSQUERY:
3787        case OP_NOTPOSUPTO:
3788        case OP_NOTSTARI:
3789        case OP_NOTMINSTARI:
3790        case OP_NOTPLUSI:
3791        case OP_NOTMINPLUSI:
3792        case OP_NOTQUERYI:
3793        case OP_NOTMINQUERYI:
3794        case OP_NOTUPTOI:
3795        case OP_NOTMINUPTOI:
3796        case OP_NOTEXACTI:
3797        case OP_NOTPOSSTARI:
3798        case OP_NOTPOSPLUSI:
3799        case OP_NOTPOSQUERYI:
3800        case OP_NOTPOSUPTOI:
3801        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3802        break;
3803        }
3804  #else  #else
3805        othercase = NOTACHAR;    (void)(utf);  /* Keep compiler happy by referencing function argument */
3806  #endif  #endif
3807        return c == othercase;    }
3808    }
3809    
3810    
3811    
3812    /*************************************************
3813    *           Check for POSIX class syntax         *
3814    *************************************************/
3815    
3816    /* This function is called when the sequence "[:" or "[." or "[=" is
3817    encountered in a character class. It checks whether this is followed by a
3818    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3819    reach an unescaped ']' without the special preceding character, return FALSE.
3820    
3821    Originally, this function only recognized a sequence of letters between the
3822    terminators, but it seems that Perl recognizes any sequence of characters,
3823    though of course unknown POSIX names are subsequently rejected. Perl gives an
3824    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3825    didn't consider this to be a POSIX class. Likewise for [:1234:].
3826    
3827    The problem in trying to be exactly like Perl is in the handling of escapes. We
3828    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3829    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3830    below handles the special case of \], but does not try to do any other escape
3831    processing. This makes it different from Perl for cases such as [:l\ower:]
3832    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3833    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3834    I think.
3835    
3836    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3837    It seems that the appearance of a nested POSIX class supersedes an apparent
3838    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3839    a digit.
3840    
3841    In Perl, unescaped square brackets may also appear as part of class names. For
3842    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3843    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3844    seem right at all. PCRE does not allow closing square brackets in POSIX class
3845    names.
3846    
3847    Arguments:
3848      ptr      pointer to the initial [
3849      endptr   where to return the end pointer
3850    
3851    Returns:   TRUE or FALSE
3852    */
3853    
3854    static BOOL
3855    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3856    {
3857    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3858    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3859    for (++ptr; *ptr != CHAR_NULL; ptr++)
3860      {
3861      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3862        ptr++;
3863      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3864      else
3865        {
3866        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3867          {
3868          *endptr = ptr;
3869          return TRUE;
3870        }        }
3871      else      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3872  #endif  /* SUPPORT_UTF */           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3873      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */            ptr[1] == CHAR_EQUALS_SIGN) &&
3874            check_posix_syntax(ptr, endptr))
3875          return FALSE;
3876        }
3877      }
3878    return FALSE;
3879    }
3880    
3881    
3882    
3883    
3884    /*************************************************
3885    *          Check POSIX class name                *
3886    *************************************************/
3887    
3888    /* This function is called to check the name given in a POSIX-style class entry
3889    such as [:alnum:].
3890    
3891    Arguments:
3892      ptr        points to the first letter
3893      len        the length of the name
3894    
3895    Returns:     a value representing the name, or -1 if unknown
3896    */
3897    
3898    static int
3899    check_posix_name(const pcre_uchar *ptr, int len)
3900    {
3901    const char *pn = posix_names;
3902    register int yield = 0;
3903    while (posix_name_lengths[yield] != 0)
3904      {
3905      if (len == posix_name_lengths[yield] &&
3906        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3907      pn += posix_name_lengths[yield] + 1;
3908      yield++;
3909      }
3910    return -1;
3911    }
3912    
     /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  
     When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  
3913    
3914      case OP_DIGIT:  /*************************************************
3915      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  *    Adjust OP_RECURSE items in repeated group   *
3916    *************************************************/
3917    
3918      case OP_NOT_DIGIT:  /* OP_RECURSE items contain an offset from the start of the regex to the group
3919      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  that is referenced. This means that groups can be replicated for fixed
3920    repetition simply by copying (because the recursion is allowed to refer to
3921    earlier groups that are outside the current group). However, when a group is
3922    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3923    inserted before it, after it has been compiled. This means that any OP_RECURSE
3924    items within it that refer to the group itself or any contained groups have to
3925    have their offsets adjusted. That one of the jobs of this function. Before it
3926    is called, the partially compiled regex must be temporarily terminated with
3927    OP_END.
3928    
3929      case OP_WHITESPACE:  This function has been extended with the possibility of forward references for
3930      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  recursions and subroutine calls. It must also check the list of such references
3931    for the group we are dealing with. If it finds that one of the recursions in
3932    the current group is on this list, it adjusts the offset in the list, not the
3933    value in the reference (which is a group number).
3934    
3935      case OP_NOT_WHITESPACE:  Arguments:
3936      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;    group      points to the start of the group
3937      adjust     the amount by which the group is to be moved
3938      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3939      cd         contains pointers to tables etc.
3940      save_hwm   the hwm forward reference pointer at the start of the group
3941    
3942      case OP_WORDCHAR:  Returns:     nothing
3943      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  */
3944    
3945      case OP_NOT_WORDCHAR:  static void
3946      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3947      pcre_uchar *save_hwm)
3948    {
3949    pcre_uchar *ptr = group;
3950    
3951      case OP_HSPACE:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3952      case OP_NOT_HSPACE:    {
3953      switch(next)    int offset;
3954        {    pcre_uchar *hc;
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3955    
3956        default:    /* See if this recursion is on the forward reference list. If so, adjust the
3957        return op_code != OP_NOT_HSPACE;    reference. */
       }  
3958    
3959      case OP_ANYNL:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3960      case OP_VSPACE:      {
3961      case OP_NOT_VSPACE:      offset = (int)GET(hc, 0);
3962      switch(next)      if (cd->start_code + offset == ptr + 1)
3963        {        {
3964        VSPACE_CASES:        PUT(hc, 0, offset + adjust);
3965        return op_code == OP_NOT_VSPACE;        break;
   
       default:  
       return op_code != OP_NOT_VSPACE;  
3966        }        }
   
 #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;  
3967      }      }
   }  
3968    
3969  /* 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
3970  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. */  
3971    
3972  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
3973      {      {
3974      case ESC_d:      offset = (int)GET(ptr, 1);
3975      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3976        }
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3977    
3978      case ESC_S:    ptr += 1 + LINK_SIZE;
3979      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    }
3980    }
3981    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3982    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3983    
3984      case ESC_h:  /*************************************************
3985      case ESC_H:  *        Insert an automatic callout point       *
3986      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
3987    
3988        default:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3989        return escape == ESC_h;  callout points before each pattern item.
       }  
3990    
3991      case ESC_v:  Arguments:
3992      case ESC_V:    code           current code pointer
3993      switch(c)    ptr            current pattern pointer
3994        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return escape != ESC_v;  
3995    
3996        default:  Returns:         new code pointer
3997        return escape == ESC_v;  */
       }  
3998    
3999      /* When PCRE_UCP is set, these values get generated for \d etc. Find  static pcre_uchar *
4000      their substitutions and process them. The result will always be either  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4001      ESC_p or ESC_P. Then fall through to process those values. */  {
4002    *code++ = OP_CALLOUT;
4003    *code++ = 255;
4004    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4005    PUT(code, LINK_SIZE, 0);                       /* Default length */
4006    return code + 2 * LINK_SIZE;
4007    }
4008    
 #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 */  
4009    
     case ESC_p:  
     case ESC_P:  
       {  
       unsigned int ptype = 0, pdata = 0;  
       int errorcodeptr;  
       BOOL negated;  
4010    
4011        ptr--;      /* Make ptr point at the p or P */  /*************************************************
4012        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  *         Complete a callout item                *
4013          return FALSE;  *************************************************/
       ptr++;      /* Point past the final curly ket */  
4014    
4015        /* 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
4016        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
4017        to the original \d etc. At this point, ptr will point to a zero byte. */  for both automatic and manual callouts.
4018    
4019        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  Arguments:
4020          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    previous_callout   points to previous callout item
4021            return FALSE;    ptr                current pattern pointer
4022      cd                 pointers to tables etc
4023    
4024        /* Do the property check. */  Returns:             nothing
4025    */
4026    
4027        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  static void
4028        }  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4029  #endif  {
4030    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4031    PUT(previous_callout, 2 + LINK_SIZE, length);
4032    }
4033    
     default:  
     return FALSE;  
     }  
4034    
   /* 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.) */  
4035    
4036    case OP_DIGIT:  #ifdef SUPPORT_UCP
4037    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  /*************************************************
4038           escape == ESC_h || escape == ESC_v || escape == ESC_R;  *           Get othercase range                  *
4039    *************************************************/
4040    
4041    case OP_NOT_DIGIT:  /* This function is passed the start and end of a class range, in UTF-8 mode
4042    return escape == ESC_d;  with UCP support. It searches up the characters, looking for ranges of
4043    characters in the "other" case. Each call returns the next one, updating the
4044    start address. A character with multiple other cases is returned on its own
4045    with a special return value.
4046    
4047    case OP_WHITESPACE:  Arguments:
4048    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4049      d           end value
4050      ocptr       where to put start of othercase range
4051      odptr       where to put end of othercase range
4052    
4053    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4054    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4055                  >0 the CASESET offset for char with multiple other cases
4056                    in this case, ocptr contains the original
4057    */
4058    
4059    case OP_HSPACE:  static int
4060    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4061           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4062    {
4063    pcre_uint32 c, othercase, next;
4064    unsigned int co;
4065    
4066    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4067    return escape == ESC_h;  cases, return its case offset value. */
4068    
4069    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4070    case OP_ANYNL:    {
4071    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4072    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4073        *ocptr = c++;   /* Character that has the set */
4074        *cptr = c;      /* Rest of input range */
4075        return (int)co;
4076        }
4077      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4078      }
4079    
4080    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4081    
4082    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4083    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
4084           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4085    
4086    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4087    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4088    
4089    default:  for (++c; c <= d; c++)
4090    return FALSE;    {
4091      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4092      next++;
4093    }    }
4094    
4095  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4096    *cptr = c;             /* Rest of input range */
4097    return 0;
4098  }  }
4099    #endif  /* SUPPORT_UCP */
4100    
4101    
4102    
# Line 3531  add_to_class(pcre_uint8 *classbits, pcre Line 4126  add_to_class(pcre_uint8 *classbits, pcre
4126    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4127  {  {
4128  pcre_uint32 c;  pcre_uint32 c;
4129    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4130  int n8 = 0;  int n8 = 0;
4131    
4132  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
# Line 3548  if ((options & PCRE_CASELESS) != 0) Line 4144  if ((options & PCRE_CASELESS) != 0)
4144    
4145      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4146      c = start;      c = start;
4147    
4148      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4149        {        {
4150        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
# Line 3574  if ((options & PCRE_CASELESS) != 0) Line 4170  if ((options & PCRE_CASELESS) != 0)
4170    
4171    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4172    
4173    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4174      {      {
4175      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4176      n8++;      n8++;
# Line 3599  in all cases. */ Line 4195  in all cases. */
4195    
4196  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4197    
4198  /* 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. */  
4199    
4200  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4201    {    {
4202    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4203      {    SETBIT(classbits, c);
4204      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4205    }    }
4206    
4207  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4208    if (start <= 0xff) start = 0xff + 1;
4209    
4210    if (end >= start)
4211    {    {
4212    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4213  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4214    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4215      {      {
# Line 3654  else Line 4249  else
4249    
4250    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4251    }    }
4252    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4253    
4254  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4255  }  }
# Line 3754  to find out the amount of memory needed, Line 4350  to find out the amount of memory needed,
4350  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4351    
4352  Arguments:  Arguments:
4353    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4354    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4355    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4356    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4357    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
4358    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
4359    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
4360    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
4361    bcptr          points to current branch chain    bcptr             points to current branch chain
4362    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
4363    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
4364    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
4365                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
4366    
4367  Returns:         TRUE on success  Returns:            TRUE on success
4368                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4369  */  */
4370    
4371  static BOOL  static BOOL
# Line 3875  for (;; ptr++) Line 4471  for (;; ptr++)
4471    BOOL reset_bracount;    BOOL reset_bracount;
4472    int class_has_8bitchar;    int class_has_8bitchar;
4473    int class_one_char;    int class_one_char;
4474    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4475      BOOL xclass_has_prop;
4476    #endif
4477    int newoptions;    int newoptions;
4478    int recno;    int recno;
4479    int refsign;    int refsign;
# Line 3992  for (;; ptr++) Line 4591  for (;; ptr++)
4591          }          }
4592        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4593        }        }
4594        /* Control does not reach here. */
4595      }      }
4596    
4597    /* 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
4598    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. */  
4599    
4600    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4601      {      {
4602      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4603        {        {
4604          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4605          if (c != CHAR_NUMBER_SIGN) break;
4606        ptr++;        ptr++;
4607        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4608          {          {
4609          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4610              {                          /* IS_NEWLINE sets cd->nllen. */
4611              ptr += cd->nllen;
4612              break;
4613              }
4614          ptr++;          ptr++;
4615  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4616          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4617  #endif  #endif
4618          }          }
4619        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;  
4620        }        }
4621      }      }
4622    
4623    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4624    
4625      is_quantifier =
4626        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4627        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4628    
4629      /* Fill in length of a previous callout, except when the next thing is a
4630      quantifier or when processing a property substitution string in UCP mode. */
4631    
4632      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4633           after_manual_callout-- <= 0)
4634        {
4635        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4636          complete_callout(previous_callout, ptr, cd);
4637        previous_callout = NULL;
4638        }
4639    
4640      /* Create auto callout, except for quantifiers, or while processing property
4641      strings that are substituted for \w etc in UCP mode. */
4642    
4643    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4644      {      {
4645      previous_callout = code;      previous_callout = code;
4646      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4647      }      }
4648    
4649      /* Process the next pattern item. */
4650    
4651    switch(c)    switch(c)
4652      {      {
4653      /* ===================================================================*/      /* ===================================================================*/
4654      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4655      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4656      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4657      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4121  for (;; ptr++) Line 4729  for (;; ptr++)
4729        }        }
4730      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4731    
4732        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4733        used for "start of word" and "end of word". As these are otherwise illegal
4734        sequences, we don't break anything by recognizing them. They are replaced
4735        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4736        erroneous and are handled by the normal code below. */
4737    
4738      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4739        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4740          {
4741          nestptr = ptr + 7;
4742          ptr = sub_start_of_word - 1;
4743          continue;
4744          }
4745    
4746        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4747          {
4748          nestptr = ptr + 7;
4749          ptr = sub_end_of_word - 1;
4750          continue;
4751          }
4752    
4753        /* Handle a real character class. */
4754    
4755      previous = code;      previous = code;
4756    
4757      /* 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 4808  for (;; ptr++)
4808    
4809      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4810    
4811        /* Extended class (xclass) will be used when characters > 255
4812        might match. */
4813    
4814    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4815        xclass = FALSE;
4816        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4817        class_uchardata_base = class_uchardata;   /* Save the start */
4818    #endif
4819    
4820      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4821      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 <
4822      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
4823      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4824        are present in the class. */
4825    
4826      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4827      class_one_char = 0;      class_one_char = 0;
4828    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4829        xclass_has_prop = FALSE;
4830    #endif
4831    
4832      /* 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
4833      temporary bit of memory, in case the class contains fewer than two      temporary bit of memory, in case the class contains fewer than two
# Line 4193  for (;; ptr++) Line 4836  for (;; ptr++)
4836    
4837      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4838    
 #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  
   
4839      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4840      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
4841      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4286  for (;; ptr++) Line 4923  for (;; ptr++)
4923            posix_class = 0;            posix_class = 0;
4924    
4925          /* 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
4926          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4927            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4928            directly. */
4929    
4930  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4931          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4932            {            {
4933              unsigned int ptype = 0;
4934            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4935    
4936              /* The posix_substitutes table specifies which POSIX classes can be
4937              converted to \p or \P items. */
4938    
4939            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4940              {              {
4941              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4942              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4943              continue;              continue;
4944              }              }
4945    
4946              /* There are three other classes that generate special property calls
4947              that are recognized only in an XCLASS. */
4948    
4949              else switch(posix_class)
4950                {
4951                case PC_GRAPH:
4952                ptype = PT_PXGRAPH;
4953                /* Fall through */
4954                case PC_PRINT:
4955                if (ptype == 0) ptype = PT_PXPRINT;
4956                /* Fall through */
4957                case PC_PUNCT:
4958                if (ptype == 0) ptype = PT_PXPUNCT;
4959                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4960                *class_uchardata++ = ptype;
4961                *class_uchardata++ = 0;
4962                xclass_has_prop = TRUE;
4963                ptr = tempptr + 1;
4964                continue;
4965    
4966                /* For all other POSIX classes, no special action is taken in UCP
4967                mode. Fall through to the non_UCP case. */
4968    
4969                default:
4970                break;
4971                }
4972            }            }
4973  #endif  #endif
4974          /* 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
4975          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
4976          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
4977          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
4978            bit map that is being built. */
4979    
4980          posix_class *= 3;          posix_class *= 3;
4981    
# Line 4420  for (;; ptr++) Line 5092  for (;; ptr++)
5092              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5093              continue;              continue;
5094    
5095              /* 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
5096              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
5097              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5098              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5099                we could just adjust the appropriate bit. From PCRE 8.34 we no
5100                longer treat \s and \S specially. */
5101    
5102              case ESC_s:              case ESC_s:
5103              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];  
5104              continue;              continue;
5105    
5106              case ESC_S:              case ESC_S:
5107              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5108              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 */  
5109              continue;              continue;
5110    
5111              /* 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 5142  for (;; ptr++)
5142                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5143                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5144                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5145                  xclass_has_prop = TRUE;
5146                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5147                continue;                continue;
5148                }                }
# Line 4556  for (;; ptr++) Line 5228  for (;; ptr++)
5228  #endif  #endif
5229          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5230    
5231          /* 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
5232          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
5233          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5234            warning is given, so PCRE now faults it as it is almost certainly a
5235            mistake on the user's part. */
5236    
5237          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5238            {            {
5239            int descape;            if (d == CHAR_BACKSLASH)
5240            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5241            if (*errorcodeptr != 0) goto FAILED;              int descape;
5242                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5243                if (*errorcodeptr != 0) goto FAILED;
5244    
5245            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5246                special causes an error. */
5247    
5248            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5249                {                {
5250                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5251                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5252                    *errorcodeptr = ERR83;
5253                    goto FAILED;
5254                    }
5255                }                }
5256              }              }
5257    
5258              /* A hyphen followed by a POSIX class is treated in the same way. */
5259    
5260              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5261                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5262                        ptr[1] == CHAR_EQUALS_SIGN) &&
5263                       check_posix_syntax(ptr, &tempptr))
5264                {
5265                *errorcodeptr = ERR83;
5266                goto FAILED;
5267                }
5268            }            }
5269    
5270          /* 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 5437  for (;; ptr++)
5437        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5438        code += LINK_SIZE;        code += LINK_SIZE;
5439        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5440          if (xclass_has_prop) *code |= XCL_HASPROP;
5441    
5442        /* 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;
5443        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 5447  for (;; ptr++)
5447          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5448          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5449            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5450            if (negate_class && !xclass_has_prop)
5451              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5452          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5453          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5454          }          }
# Line 4839  for (;; ptr++) Line 5531  for (;; ptr++)
5531    
5532      tempcode = previous;      tempcode = previous;
5533    
5534        /* Before checking for a possessive quantifier, we must skip over
5535        whitespace and comments in extended mode because Perl allows white space at
5536        this point. */
5537    
5538        if ((options & PCRE_EXTENDED) != 0)
5539          {
5540          const pcre_uchar *p = ptr + 1;
5541          for (;;)
5542            {
5543            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5544            if (*p != CHAR_NUMBER_SIGN) break;
5545            p++;
5546            while (*p != CHAR_NULL)
5547              {
5548              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5549                {                        /* IS_NEWLINE sets cd->nllen. */
5550                p += cd->nllen;
5551                break;
5552                }
5553              p++;
5554    #ifdef SUPPORT_UTF
5555              if (utf) FORWARDCHAR(p);
5556    #endif
5557              }           /* Loop for comment characters */
5558            }             /* Loop for multiple comments */
5559          ptr = p - 1;    /* Character before the next significant one. */
5560          }
5561    
5562      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5563      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5564      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 5653  for (;; ptr++)
5653            }            }
5654          }          }
5655    
       /* 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;  
         }  
   
5656        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
5657        }        }
5658    
# Line 4963  for (;; ptr++) Line 5670  for (;; ptr++)
5670        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5671        c = *previous;        c = *previous;
5672    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5673        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5674        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5675          {          {
# Line 5119  for (;; ptr++) Line 5818  for (;; ptr++)
5818      /* 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
5819      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}. */
5820    
5821      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5822  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5823               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5824  #endif  #endif
5825               *previous == OP_REF ||               *previous == OP_REF   || *previous == OP_REFI ||
5826               *previous == OP_REFI)               *previous == OP_DNREF || *previous == OP_DNREFI)
5827        {        {
5828        if (repeat_max == 0)        if (repeat_max == 0)
5829          {          {
# Line 5153  for (;; ptr++) Line 5851  for (;; ptr++)
5851      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
5852      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 >=
5853      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5854      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5855      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5856        for Perl compatibility. */
5857    
5858      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5859        {        {
# Line 5172  for (;; ptr++) Line 5871  for (;; ptr++)
5871        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5872        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5873        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
5874        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5875    
5876        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5877          {          {
# Line 5545  for (;; ptr++) Line 6244  for (;; ptr++)
6244        goto FAILED;        goto FAILED;
6245        }        }
6246    
6247      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6248      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6249      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
6250      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6251      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.  
6252    
6253      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6254      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6255      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6256        tempcode, not at previous, which might be the first part of a string whose
6257      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. */  
6258    
6259      if (possessive_quantifier)      if (possessive_quantifier)
6260        {        {
6261        int len;        int len;
6262    
6263        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6264          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6265          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6266          remains is greater than zero, there's a further opcode that can be
6267          handled. If not, do nothing, leaving the EXACT alone. */
6268    
6269          switch(*tempcode)
6270            {
6271            case OP_TYPEEXACT:
6272          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6273            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6274            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6275            break;
6276    
6277        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6278          {  
6279            case OP_CHAR:
6280            case OP_CHARI:
6281            case OP_NOT:
6282            case OP_NOTI:
6283            case OP_EXACT:
6284            case OP_EXACTI:
6285            case OP_NOTEXACT:
6286            case OP_NOTEXACTI:
6287          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6288  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6289          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6290            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6291  #endif  #endif
6292            break;
6293    
6294            /* For the class opcodes, the repeat operator appears at the end;
6295            adjust tempcode to point to it. */
6296    
6297            case OP_CLASS:
6298            case OP_NCLASS:
6299            tempcode += 1 + 32/sizeof(pcre_uchar);
6300            break;
6301    
6302    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6303            case OP_XCLASS:
6304            tempcode += GET(tempcode, 1);
6305            break;
6306    #endif
6307          }          }
6308    
6309          /* If tempcode is equal to code (which points to the end of the repeated
6310          item), it means we have skipped an EXACT item but there is no following
6311          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6312          all other cases, tempcode will be pointing to the repeat opcode, and will
6313          be less than code, so the value of len will be greater than 0. */
6314    
6315        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6316          if (len > 0)
6317            {
6318            unsigned int repcode = *tempcode;
6319    
6320            /* There is a table for possessifying opcodes, all of which are less
6321            than OP_CALLOUT. A zero entry means there is no possessified version.
6322            */
6323    
6324            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6325              *tempcode = opcode_possessify[repcode];
6326    
6327            /* For opcode without a special possessified version, wrap the item in
6328            ONCE brackets. Because we are moving code along, we must ensure that any
6329            pending recursive references are updated. */
6330    
6331            else
6332              {
6333              *code = OP_END;
6334              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6335              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6336              code += 1 + LINK_SIZE;
6337              len += 1 + LINK_SIZE;
6338              tempcode[0] = OP_ONCE;
6339              *code++ = OP_KET;
6340              PUTINC(code, 0, len);
6341              PUT(tempcode, 1, len);
6342              }
6343            }
6344    
6345    #ifdef NEVER
6346        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6347          {          {
6348          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5609  for (;; ptr++) Line 6370  for (;; ptr++)
6370          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6371          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6372    
6373            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6374            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6375            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6376            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6377    
6378          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6379          pending recursive references are updated. */          pending recursive references are updated. */
6380    
# Line 5624  for (;; ptr++) Line 6390  for (;; ptr++)
6390          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6391          break;          break;
6392          }          }
6393    #endif
6394        }        }
6395    
6396      /* 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 6579  for (;; ptr++)
6579          tempptr = ptr;          tempptr = ptr;
6580    
6581          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6582          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',
6583          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6584            recursion tests.
6585          There are several syntaxes for testing a named group: (?(name)) is used  
6586          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6587            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6588          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6589          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
6590          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
6591          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.  
6592    
6593          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6594          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 6612  for (;; ptr++)
6612                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6613            break;            break;
6614    
6615          /* 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
6616          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. */
6617    
6618          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6619          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6620          refsign = -1;          refsign = -1;     /* => not a number */
6621            namelen = -1;     /* => not a name; must set to avoid warning */
6622            name = NULL;      /* Always set to avoid warning */
6623            recno = 0;        /* Always set to avoid warning */
6624    
6625          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6626    
6627          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6628            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6629            {            {
6630            terminator = -1;            terminator = -1;
6631            ptr += 2;            ptr += 2;
# Line 5863  for (;; ptr++) Line 6633  for (;; ptr++)
6633            }            }
6634    
6635          /* 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
6636          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6637            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6638    
6639          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6640            {            {
6641            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6642            ptr++;            ptr++;
6643            }            }
6644          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6645            {            {
6646            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6647            ptr++;            ptr++;
# Line 5878  for (;; ptr++) Line 6649  for (;; ptr++)
6649          else          else
6650            {            {
6651            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6652            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6653                else if (IS_DIGIT(*ptr)) refsign = 0;
6654            }            }
6655    
6656          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6657    
6658          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6659            {            {
6660            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6661            *errorcodeptr = ERR28;              {
6662            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6663                ptr++;
6664                }
6665            }            }
6666    
6667          /* 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
6668            a name is one of a number of duplicates, a different opcode is used and
6669            it needs more memory. Unfortunately we cannot tell whether a name is a
6670            duplicate in the first pass, so we have to allow for more memory. */
6671    
6672          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6673            {            {
6674            if (recno >= 0)            if (IS_DIGIT(*ptr))
6675              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6676            ptr++;              *errorcodeptr = ERR84;
6677                goto FAILED;
6678                }
6679              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6680                {
6681                *errorcodeptr = ERR28;   /* Assertion expected */
6682                goto FAILED;
6683                }
6684              name = ptr++;
6685              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6686                {
6687                ptr++;
6688                }
6689              namelen = (int)(ptr - name);
6690              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6691            }            }
6692          namelen = (int)(ptr - name);  
6693            /* Check the terminator */
6694    
6695          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6696              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6697            {            {
6698            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6699            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6700            goto FAILED;            goto FAILED;
6701            }            }
6702    
# Line 5915  for (;; ptr++) Line 6705  for (;; ptr++)
6705          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6706    
6707          /* 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
6708          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
6709          be digits, in which case recno will be set. */          recno. */
6710    
6711          if (refsign > 0)          if (refsign >= 0)
6712            {            {
6713            if (recno <= 0)            if (recno <= 0)
6714              {              {
6715              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6716              goto FAILED;              goto FAILED;
6717              }              }
6718            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6719              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6720            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6721              {              {
6722              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 5936  for (;; ptr++) Line 6726  for (;; ptr++)
6726            break;            break;
6727            }            }
6728    
6729          /* 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  
         OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,  
         except they record that the reference was originally to a name. The  
         information is used to check duplicate names. */  
6730    
6731          slot = cd->name_table;          slot = cd->name_table;
6732          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 5949  for (;; ptr++) Line 6735  for (;; ptr++)
6735            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6736            }            }
6737    
6738          /* Found a previous named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6739            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6740            appropriate data values. Otherwise, just insert the unique subpattern
6741            number. */
6742    
6743          if (i < cd->names_found)          if (i < cd->names_found)
6744            {            {
6745            recno = GET2(slot, 0);            int offset = i++;
6746            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6747            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6748            }            for (; i < cd->names_found; i++)
6749                {
6750          /* Search the pattern for a forward reference */              slot += cd->name_entry_size;
6751                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6752          else if ((i = find_parens(cd, name, namelen,              count++;
6753                          (options & PCRE_EXTENDED) != 0, utf)) > 0)              }
6754            {            if (count > 1)
6755            PUT2(code, 2+LINK_SIZE, i);              {
6756            code[1+LINK_SIZE]++;              PUT2(code, 2+LINK_SIZE, offset);
6757                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6758                skipbytes += IMM2_SIZE;
6759                code[1+LINK_SIZE]++;
6760                }
6761              else  /* Not a duplicated name */
6762                {
6763                PUT2(code, 2+LINK_SIZE, recno);
6764                }
6765            }            }
6766    
6767          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
6768          after the opening parenthesis [e.g. (?(abc)...] and in this case there          after the opening parenthesis [e.g. (?(abc)...] and in this case there
6769          are some further alternatives to try. For the cases where terminator !=          are some further alternatives to try. For the cases where terminator !=
6770          0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have          CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6771          now checked all the possibilities, so give an error. */          we have now checked all the possibilities, so give an error. */
6772    
6773          else if (terminator != CHAR_NULL)          else if (terminator != CHAR_NULL)
6774            {            {
# Line 6008  for (;; ptr++) Line 6805  for (;; ptr++)
6805            skipbytes = 1;            skipbytes = 1;
6806            }            }
6807    
6808          /* Check for the "name" actually being a subpattern number. We are          /* Reference to an unidentified subpattern. */
         in the second pass here, so final_bracount is set. */  
   
         else if (recno > 0 && recno <= cd->final_bracount)  
           {  
           PUT2(code, 2+LINK_SIZE, recno);  
           }  
   
         /* Either an unidentified subpattern, or a reference to (?(0) */  
6809    
6810          else          else
6811            {            {
6812            *errorcodeptr = (recno == 0)? ERR35: ERR15;            *errorcodeptr = ERR15;
6813            goto FAILED;            goto FAILED;
6814            }            }
6815          break;          break;
# Line 6033  for (;; ptr++) Line 6822  for (;; ptr++)
6822          ptr++;          ptr++;
6823          break;          break;
6824    
6825            /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6826            thing to do, but Perl allows all assertions to be quantified, and when
6827            they contain capturing parentheses there may be a potential use for
6828            this feature. Not that that applies to a quantified (?!) but we allow
6829            it for uniformity. */
6830    
6831          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6832          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
6833          ptr++;          ptr++;
6834          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6835                 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6836                (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6837            {            {
6838            *code++ = OP_FAIL;            *code++ = OP_FAIL;
6839            previous = NULL;            previous = NULL;
# Line 6130  for (;; ptr++) Line 6926  for (;; ptr++)
6926          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6927          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
6928          case CHAR_APOSTROPHE:          case CHAR_APOSTROPHE:
6929            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6930              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6931            name = ++ptr;
6932            if (IS_DIGIT(*ptr))
6933            {            {
6934            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?            *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6935              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            goto FAILED;