/[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 1487 by ph10, Wed Jun 18 17:17:03 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 47  supporting internal functions that are n Line 47  supporting internal functions that are n
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing pattern start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing pattern end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
# Line 115  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 253  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 268  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 295  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 322  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 375  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 390  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 455  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 509  static const char error_texts[] = Line 540  static const char error_texts[] =
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542    "setting UTF is disabled by the application\0"    "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553    ;    ;
554    
555  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 648  static const pcre_uint8 ebcdic_chartab[] Line 689  static const pcre_uint8 ebcdic_chartab[]
689  #endif  #endif
690    
691    
692    /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730    /* This table is used to check whether auto-possessification is possible
731    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732    left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
785      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
790      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
870  /*************************************************  /*************************************************
871  *            Find an error text                  *  *            Find an error text                  *
# Line 675  return s; Line 893  return s;
893  }  }
894    
895    
896    
897  /*************************************************  /*************************************************
898  *           Expand the workspace                 *  *           Expand the workspace                 *
899  *************************************************/  *************************************************/
# Line 752  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 971  return (*p == CHAR_RIGHT_CURLY_BRACKET);
971  *************************************************/  *************************************************/
972    
973  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
974  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
975  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.
976  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
977  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982    chptr          points to the data character    chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
# Line 965  else Line 1183  else
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1191      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1192      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1193      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1194      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1195      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1202      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 1001  else Line 1223  else
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          escape = -s;          escape = -s;
1229          break;          break;
# Line 1009  else Line 1231  else
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1235      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1236      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1244      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1035  else Line 1255  else
1255  #endif  #endif
1256      break;      break;
1257    
1258      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1259      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1260      If not, { is treated as a data character. */  
1261        case CHAR_o:
1262        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264          {
1265          ptr += 2;
1266          c = 0;
1267          overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269            {
1270            register pcre_uint32 cc = *ptr++;
1271            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272    #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274    #endif
1275            c = (c << 3) + cc - CHAR_0 ;
1276    #if defined COMPILE_PCRE8
1277            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE16
1279            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282    #endif
1283            }
1284          if (overflow)
1285            {
1286            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287            *errorcodeptr = ERR34;
1288            }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290            {
1291            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294          }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      case CHAR_x:      case CHAR_x:
1301      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1303        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305          {          {
# Line 1060  else Line 1316  else
1316  #endif  #endif
1317            }            }
1318          }          }
1319        break;        }    /* End JavaScript handling */
       }  
1320    
1321      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        c = 0;      else
1329        overflow = FALSE;        {
1330        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331          {          {
1332          register pcre_uint32 cc = *pt++;          ptr += 2;
1333          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1346          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1347  #endif  #endif
1348    
1349  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1353          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1354          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355  #endif  #endif
1356    
1357  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1358          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1360          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1362          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363  #endif  #endif
1364          }            }
1365    
1366        if (overflow)          if (overflow)
1367          {            {
1368          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1370          }            }
1371    
1372        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373          {            {
1374          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375          ptr = pt;            }
         break;  
         }  
1376    
1377        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1378        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1379        }          \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1387        {        else
1388        pcre_uint32 cc;                          /* Some compilers don't like */          {
1389        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1398        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400  #endif  #endif
1401        }            }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1192  if ((options & PCRE_UCP) != 0 && escape Line 1466  if ((options & PCRE_UCP) != 0 && escape
1466  return escape;  return escape;
1467  }  }
1468    
1469    
1470    
1471  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1472  /*************************************************  /*************************************************
1473  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1289  return FALSE; Line 1565  return FALSE;
1565    
1566    
1567    
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 1315  read_repeat_counts(const pcre_uchar *p, Line 1590  read_repeat_counts(const pcre_uchar *p,
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597    }      {
1598        *errorcodeptr = ERR5;
1599  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1600  Also, max must not be less than min. */      }
1601      }
1602    
1603  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604    {    {
1605    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612        }          {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616          }
1617      if (max < min)      if (max < min)
1618        {        {
1619        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1347  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 1358  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / 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;  
 }  
   
   
   
   
 /*************************************************  
1633  *      Find first significant op code            *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
# Line 1696  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671      case OP_NCREF:      case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673      case OP_NRREF:      case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
# Line 1712  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
# Line 1836  for (;;) Line 1807  for (;;)
1807      case OP_COMMIT:      case OP_COMMIT:
1808      case OP_CREF:      case OP_CREF:
1809      case OP_DEF:      case OP_DEF:
1810        case OP_DNCREF:
1811        case OP_DNRREF:
1812      case OP_DOLL:      case OP_DOLL:
1813      case OP_DOLLM:      case OP_DOLLM:
1814      case OP_EOD:      case OP_EOD:
1815      case OP_EODN:      case OP_EODN:
1816      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1817      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1818      case OP_PRUNE:      case OP_PRUNE:
1819      case OP_REVERSE:      case OP_REVERSE:
# Line 1937  for (;;) Line 1908  for (;;)
1908    
1909      switch (*cc)      switch (*cc)
1910        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1911        case OP_CRSTAR:        case OP_CRSTAR:
1912        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRQUERY:        case OP_CRQUERY:
1916        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1917          case OP_CRPOSSTAR:
1918          case OP_CRPOSPLUS:
1919          case OP_CRPOSQUERY:
1920        return -1;        return -1;
1921    
1922        case OP_CRRANGE:        case OP_CRRANGE:
1923        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1924          case OP_CRPOSRANGE:
1925        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1927        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2015  for (;;) Line 1990  for (;;)
1990      case OP_QUERYI:      case OP_QUERYI:
1991      case OP_REF:      case OP_REF:
1992      case OP_REFI:      case OP_REFI:
1993        case OP_DNREF:
1994        case OP_DNREFI:
1995      case OP_SBRA:      case OP_SBRA:
1996      case OP_SBRAPOS:      case OP_SBRAPOS:
1997      case OP_SCBRA:      case OP_SCBRA:
# Line 2051  for (;;) Line 2028  for (;;)
2028    
2029    
2030    
   
2031  /*************************************************  /*************************************************
2032  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2033  *************************************************/  *************************************************/
# Line 2361  Returns:      TRUE if what is matched co Line 2337  Returns:      TRUE if what is matched co
2337  typedef struct recurse_check {  typedef struct recurse_check {
2338    struct recurse_check *prev;    struct recurse_check *prev;
2339    const pcre_uchar *group;    const pcre_uchar *group;
2340  } recurse_check;  } recurse_check;
2341    
2342  static BOOL  static BOOL
2343  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
# Line 2377  for (code = first_significant_code(code Line 2353  for (code = first_significant_code(code
2353    const pcre_uchar *ccode;    const pcre_uchar *ccode;
2354    
2355    c = *code;    c = *code;
2356    
2357    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2358    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2359    
# Line 2405  for (code = first_significant_code(code Line 2381  for (code = first_significant_code(code
2381      NULL. */      NULL. */
2382    
2383      if (cd->start_workspace != NULL)      if (cd->start_workspace != NULL)
2384        {        {
2385        const pcre_uchar *tcode;        const pcre_uchar *tcode;
2386        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2387          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2388        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2389        }        }
2390    
2391      /* If we are scanning a completed pattern, there are no forward references      /* If we are scanning a completed pattern, there are no forward references
2392      and all groups are complete. We need to detect whether this is a recursive      and all groups are complete. We need to detect whether this is a recursive
2393      call, as otherwise there will be an infinite loop. If it is a recursion,      call, as otherwise there will be an infinite loop. If it is a recursion,
2394      just skip over it. Simple recursions are easily detected. For mutual      just skip over it. Simple recursions are easily detected. For mutual
2395      recursions we keep a chain on the stack. */      recursions we keep a chain on the stack. */
2396    
2397      else      else
2398        {        {
2399        recurse_check *r = recurses;        recurse_check *r = recurses;
2400        const pcre_uchar *endgroup = scode;        const pcre_uchar *endgroup = scode;
2401    
2402        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2403        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2404    
2405        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2406          if (r->group == scode) break;          if (r->group == scode) break;
2407        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2436  for (code = first_significant_code(code Line 2412  for (code = first_significant_code(code
2412    
2413      empty_branch = FALSE;      empty_branch = FALSE;
2414      this_recurse.prev = recurses;      this_recurse.prev = recurses;
2415      this_recurse.group = scode;      this_recurse.group = scode;
2416    
2417      do      do
2418        {        {
2419        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 2514  for (code = first_significant_code(code
2514        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2515        case OP_CRQUERY:        case OP_CRQUERY:
2516        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2517          case OP_CRPOSSTAR:
2518          case OP_CRPOSQUERY:
2519        break;        break;
2520    
2521        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2522        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2523        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2524          case OP_CRPOSPLUS:
2525        return FALSE;        return FALSE;
2526    
2527        case OP_CRRANGE:        case OP_CRRANGE:
2528        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2529          case OP_CRPOSRANGE:
2530        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2531        break;        break;
2532        }        }
# Line 2557  for (code = first_significant_code(code Line 2537  for (code = first_significant_code(code
2537      case OP_ANY:      case OP_ANY:
2538      case OP_ALLANY:      case OP_ALLANY:
2539      case OP_ANYBYTE:      case OP_ANYBYTE:
2540    
2541      case OP_PROP:      case OP_PROP:
2542      case OP_NOTPROP:      case OP_NOTPROP:
2543      case OP_ANYNL:      case OP_ANYNL:
2544    
2545      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2546      case OP_HSPACE:      case OP_HSPACE:
2547      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2548      case OP_VSPACE:      case OP_VSPACE:
2549      case OP_EXTUNI:      case OP_EXTUNI:
2550    
2551      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2552      case OP_DIGIT:      case OP_DIGIT:
2553      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2554      case OP_WHITESPACE:      case OP_WHITESPACE:
2555      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2556      case OP_WORDCHAR:      case OP_WORDCHAR:
2557    
2558      case OP_CHAR:      case OP_CHAR:
2559      case OP_CHARI:      case OP_CHARI:
2560      case OP_NOT:      case OP_NOT:
2561      case OP_NOTI:      case OP_NOTI:
2562    
2563      case OP_PLUS:      case OP_PLUS:
2564      case OP_PLUSI:      case OP_PLUSI:
2565      case OP_MINPLUS:      case OP_MINPLUS:
# Line 2589  for (code = first_significant_code(code Line 2569  for (code = first_significant_code(code
2569      case OP_NOTPLUSI:      case OP_NOTPLUSI:
2570      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2571      case OP_NOTMINPLUSI:      case OP_NOTMINPLUSI:
2572    
2573      case OP_POSPLUS:      case OP_POSPLUS:
2574      case OP_POSPLUSI:      case OP_POSPLUSI:
2575      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2576      case OP_NOTPOSPLUSI:      case OP_NOTPOSPLUSI:
2577    
2578      case OP_EXACT:      case OP_EXACT:
2579      case OP_EXACTI:      case OP_EXACTI:
2580      case OP_NOTEXACT:      case OP_NOTEXACT:
2581      case OP_NOTEXACTI:      case OP_NOTEXACTI:
2582    
2583      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2584      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2585      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2586      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2587    
2588      return FALSE;      return FALSE;
2589    
2590      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2644  for (code = first_significant_code(code Line 2624  for (code = first_significant_code(code
2624  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2625      case OP_STAR:      case OP_STAR:
2626      case OP_STARI:      case OP_STARI:
2627      case OP_NOTSTAR:      case OP_NOTSTAR:
2628      case OP_NOTSTARI:      case OP_NOTSTARI:
2629    
2630      case OP_MINSTAR:      case OP_MINSTAR:
2631      case OP_MINSTARI:      case OP_MINSTARI:
2632      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
2633      case OP_NOTMINSTARI:      case OP_NOTMINSTARI:
2634    
2635      case OP_POSSTAR:      case OP_POSSTAR:
2636      case OP_POSSTARI:      case OP_POSSTARI:
2637      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
2638      case OP_NOTPOSSTARI:      case OP_NOTPOSSTARI:
2639    
2640      case OP_QUERY:      case OP_QUERY:
2641      case OP_QUERYI:      case OP_QUERYI:
2642      case OP_NOTQUERY:      case OP_NOTQUERY:
2643      case OP_NOTQUERYI:      case OP_NOTQUERYI:
2644    
2645      case OP_MINQUERY:      case OP_MINQUERY:
2646      case OP_MINQUERYI:      case OP_MINQUERYI:
2647      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
2648      case OP_NOTMINQUERYI:      case OP_NOTMINQUERYI:
2649    
2650      case OP_POSQUERY:      case OP_POSQUERY:
2651      case OP_POSQUERYI:      case OP_POSQUERYI:
2652      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
2653      case OP_NOTPOSQUERYI:      case OP_NOTPOSQUERYI:
2654    
2655      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2656      break;      break;
2657    
2658      case OP_UPTO:      case OP_UPTO:
2659      case OP_UPTOI:      case OP_UPTOI:
2660      case OP_NOTUPTO:      case OP_NOTUPTO:
2661      case OP_NOTUPTOI:      case OP_NOTUPTOI:
2662    
2663      case OP_MINUPTO:      case OP_MINUPTO:
2664      case OP_MINUPTOI:      case OP_MINUPTOI:
2665      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
2666      case OP_NOTMINUPTOI:      case OP_NOTMINUPTOI:
2667    
2668      case OP_POSUPTO:      case OP_POSUPTO:
2669      case OP_POSUPTOI:      case OP_POSUPTOI:
2670      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
2671      case OP_NOTPOSUPTOI:      case OP_NOTPOSUPTOI:
2672    
2673      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2674      break;      break;
2675  #endif  #endif
# Line 2753  return TRUE; Line 2733  return TRUE;
2733    
2734    
2735  /*************************************************  /*************************************************
2736  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2737  *************************************************/  *************************************************/
2738    
2739  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2740  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2741    
2742  Returns:   TRUE or FALSE  Arguments:  c opcode
2743    Returns:    base opcode for the type
2744  */  */
2745    
2746  static BOOL  static pcre_uchar
2747  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2748  {  {
2749  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2750  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2751  for (++ptr; *ptr != CHAR_NULL; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2752    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2753    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2754      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2755  }  }
2756    
2757    
2758    
2759    #ifdef SUPPORT_UCP
2760  /*************************************************  /*************************************************
2761  *          Check POSIX class name                *  *        Check a character and a property        *
2762  *************************************************/  *************************************************/
2763    
2764  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2765  such as [:alnum:].  is adjacent to a fixed character.
2766    
2767  Arguments:  Arguments:
2768    ptr        points to the first letter    c            the character
2769    len        the length of the name    ptype        the property type
2770      pdata        the data for the type
2771      negated      TRUE if it's a negated property (\P or \p{^)
2772    
2773  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2774  */  */
2775    
2776  static int  static BOOL
2777  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2778      BOOL negated)
2779  {  {
2780  const char *pn = posix_names;  const pcre_uint32 *p;
2781  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2782    
2783  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2784  that is referenced. This means that groups can be replicated for fixed    {
2785  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2786  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2787  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2788  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2789    
2790  This function has been extended with the possibility of forward references for    case PT_GC:
2791  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2792    
2793  Arguments:    case PT_PC:
2794    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2795    
2796  Returns:     nothing    case PT_SC:
2797  */    return (pdata == prop->script) == negated;
2798    
2799  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2800    
2801  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2802    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2803    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2804    
2805    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2806    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2807      at release 8.34. */
2808    
2809    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2810      case PT_PXSPACE:  /* POSIX space */
2811      switch(c)
2812      {      {
2813      offset = (int)GET(hc, 0);      HSPACE_CASES:
2814      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2815        {      return negated;
2816        PUT(hc, 0, offset + adjust);  
2817        break;      default:
2818        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2819      }      }
2820      break;  /* Control never reaches here */
2821    
2822    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2823    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2824              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2825              c == CHAR_UNDERSCORE) == negated;
2826    
2827    if (hc >= cd->hwm)    case PT_CLIST:
2828      p = PRIV(ucd_caseless_sets) + prop->caseset;
2829      for (;;)
2830      {      {
2831      offset = (int)GET(ptr, 1);      if (c < *p) return !negated;
2832      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2833      }      }
2834      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2835    }    }
2836    
2837    return FALSE;
2838  }  }
2839    #endif  /* SUPPORT_UCP */
2840    
2841    
2842    
2843  /*************************************************  /*************************************************
2844  *        Insert an automatic callout point       *  *        Fill the character property list        *
2845  *************************************************/  *************************************************/
2846    
2847  /* 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-
2848  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2849    
2850  Arguments:  Arguments:
2851    code           current code pointer    code        points to start of expression
2852    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2853    cd             pointers to tables etc    fcc         points to case-flipping table
2854      list        points to output list
2855                  list[0] will be filled with the opcode
2856                  list[1] will be non-zero if this opcode
2857                    can match an empty character string
2858                  list[2..7] depends on the opcode
2859    
2860  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2861                  NULL if *code is not accepted
2862  */  */
2863    
2864  static pcre_uchar *  static const pcre_uchar *
2865  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2866      const pcre_uint8 *fcc, pcre_uint32 *list)
2867  {  {
2868  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2869  *code++ = 255;  pcre_uchar base;
2870  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uchar *end;
2871  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 chr;
 return code + 2 * LINK_SIZE;  
 }  
2872    
2873    #ifdef SUPPORT_UCP
2874    pcre_uint32 *clist_dest;
2875    const pcre_uint32 *clist_src;
2876    #else
2877    utf = utf;  /* Suppress "unused parameter" compiler warning */
2878    #endif
2879    
2880    list[0] = c;
2881    list[1] = FALSE;
2882    code++;
2883    
2884  /*************************************************  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2885  *         Complete a callout item                *    {
2886  *************************************************/    base = get_repeat_base(c);
2887      c -= (base - OP_STAR);
2888    
2889  /* A callout item contains the length of the next item in the pattern, which    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2890  we can't fill in till after we have reached the relevant point. This is used      code += IMM2_SIZE;
 for both automatic and manual callouts.  
2891    
2892  Arguments:    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
   previous_callout   points to previous callout item  
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
2893    
2894  Returns:             nothing    switch(base)
2895  */      {
2896        case OP_STAR:
2897        list[0] = OP_CHAR;
2898        break;
2899    
2900  static void      case OP_STARI:
2901  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_CHARI;
2902  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
   
   
   
 #ifdef SUPPORT_UCP  
 /*************************************************  
 *           Get othercase range                  *  
 *************************************************/  
   
 /* This function is passed the start and end of a class range, in UTF-8 mode  
 with UCP support. It searches up the characters, looking for ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
   
 Arguments:  
   cptr        points to starting character value; updated  
   d           end value  
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
   
 Yield:        -1 when no more  
                0 when a range is returned  
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
 */  
2903    
2904  static int      case OP_NOTSTAR:
2905  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      list[0] = OP_NOT;
2906    pcre_uint32 *odptr)      break;
 {  
 pcre_uint32 c, othercase, next;  
 unsigned int co;  
2907    
2908  /* Find the first character that has an other case. If it has multiple other      case OP_NOTSTARI:
2909  cases, return its case offset value. */      list[0] = OP_NOTI;
2910        break;
2911    
2912  for (c = *cptr; c <= d; c++)      case OP_TYPESTAR:
2913    {      list[0] = *code;
2914    if ((co = UCD_CASESET(c)) != 0)      code++;
2915      {      break;
     *ocptr = c++;   /* Character that has the set */  
     *cptr = c;      /* Rest of input range */  
     return (int)co;  
2916      }      }
2917    if ((othercase = UCD_OTHERCASE(c)) != c) break;    c = list[0];
2918    }    }
2919    
2920  if (c > d) return -1;  /* Reached end of range */  switch(c)
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
2921    {    {
2922    if (UCD_OTHERCASE(c) != next) break;    case OP_NOT_DIGIT:
2923    next++;    case OP_DIGIT:
2924    }    case OP_NOT_WHITESPACE:
2925      case OP_WHITESPACE:
2926  *odptr = next - 1;     /* End of othercase range */    case OP_NOT_WORDCHAR:
2927  *cptr = c;             /* Rest of input range */    case OP_WORDCHAR:
2928  return 0;    case OP_ANY:
2929  }    case OP_ALLANY:
2930      case OP_ANYNL:
2931      case OP_NOT_HSPACE:
2932      case OP_HSPACE:
2933  /*************************************************    case OP_NOT_VSPACE:
2934  *        Check a character and a property        *    case OP_VSPACE:
2935  *************************************************/    case OP_EXTUNI:
2936      case OP_EODN:
2937  /* This function is called by check_auto_possessive() when a property item    case OP_EOD:
2938  is adjacent to a fixed character.    case OP_DOLL:
2939      case OP_DOLLM:
2940      return code;
2941    
2942  Arguments:    case OP_CHAR:
2943    c            the character    case OP_NOT:
2944    ptype        the property type    GETCHARINCTEST(chr, code);
2945    pdata        the data for the type    list[2] = chr;
2946    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2947      return code;
2948    
2949  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2950  */    case OP_NOTI:
2951      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2952      GETCHARINCTEST(chr, code);
2953      list[2] = chr;
2954    
 static BOOL  
 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  
 {  
2955  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2956  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2957        list[3] = fcc[chr];
2958      else
2959        list[3] = UCD_OTHERCASE(chr);
2960    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2961      list[3] = (chr < 256) ? fcc[chr] : chr;
2962    #else
2963      list[3] = fcc[chr];
2964  #endif  #endif
2965    
2966  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
   
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
2967    
2968    case PT_GC:    if (chr == list[3])
2969    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;      list[3] = NOTACHAR;
2970      else
2971        list[4] = NOTACHAR;
2972      return code;
2973    
2974    case PT_PC:  #ifdef SUPPORT_UCP
2975    return (pdata == prop->chartype) == negated;    case OP_PROP:
2976      case OP_NOTPROP:
2977      if (code[0] != PT_CLIST)
2978        {
2979        list[2] = code[0];
2980        list[3] = code[1];
2981        return code + 2;
2982        }
2983    
2984    case PT_SC:    /* Convert only if we have enough space. */
   return (pdata == prop->script) == negated;  
2985    
2986    /* These are specials */    clist_src = PRIV(ucd_caseless_sets) + code[1];
2987      clist_dest = list + 2;
2988      code += 2;
2989    
2990    case PT_ALNUM:    do {
2991    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||       if (clist_dest >= list + 8)
2992            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;         {
2993           /* Early return if there is not enough space. This should never
2994           happen, since all clists are shorter than 5 character now. */
2995           list[2] = code[0];
2996           list[3] = code[1];
2997           return code;
2998           }
2999         *clist_dest++ = *clist_src;
3000         }
3001      while(*clist_src++ != NOTACHAR);
3002    
3003    case PT_SPACE:    /* Perl space */    /* All characters are stored. The terminating NOTACHAR
3004    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    is copied form the clist itself. */
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3005    
3006    case PT_PXSPACE:  /* POSIX space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3007    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
3008            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #endif
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3009    
3010    case PT_WORD:    case OP_NCLASS:
3011    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    case OP_CLASS:
3012            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3013            c == CHAR_UNDERSCORE) == negated;    case OP_XCLASS:
3014      if (c == OP_XCLASS)
3015        end = code + GET(code, 0) - 1;
3016      else
3017    #endif
3018        end = code + 32 / sizeof(pcre_uchar);
3019    
3020  #ifdef SUPPORT_UCP    switch(*end)
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3021      {      {
3022      if (c < *p) return !negated;      case OP_CRSTAR:
3023      if (c == *p++) return negated;      case OP_CRMINSTAR:
3024        case OP_CRQUERY:
3025        case OP_CRMINQUERY:
3026        case OP_CRPOSSTAR:
3027        case OP_CRPOSQUERY:
3028        list[1] = TRUE;
3029        end++;
3030        break;
3031    
3032        case OP_CRPLUS:
3033        case OP_CRMINPLUS:
3034        case OP_CRPOSPLUS:
3035        end++;
3036        break;
3037    
3038        case OP_CRRANGE:
3039        case OP_CRMINRANGE:
3040        case OP_CRPOSRANGE:
3041        list[1] = (GET2(end, 1) == 0);
3042        end += 1 + 2 * IMM2_SIZE;
3043        break;
3044      }      }
3045    break;  /* Control never reaches here */    list[2] = (pcre_uint32)(end - code);
3046  #endif    return end;
3047    }    }
3048    return NULL;    /* Opcode not accepted */
 return FALSE;  
3049  }  }
 #endif  /* SUPPORT_UCP */  
3050    
3051    
3052    
3053  /*************************************************  /*************************************************
3054  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3055  *************************************************/  *************************************************/
3056    
3057  /* This function is called for unlimited repeats of certain items, to see  /* Checks whether the base and the current opcode have a common character, in
3058  whether the next thing could possibly match the repeated item. If not, it makes  which case the base cannot be possessified.
 sense to automatically possessify the repeated item.  
3059    
3060  Arguments:  Arguments:
3061    previous      pointer to the repeated opcode    code        points to the byte code
3062    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3063    ptr           next character in pattern    cd          static compile data
3064    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3065    
3066  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3067  */  */
3068    
3069  static BOOL  static BOOL
3070  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3071    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3072  {  {
3073  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3074  pcre_uint32 next;  pcre_uint32 list[8];
3075  int escape;  const pcre_uint32 *chr_ptr;
3076  pcre_uchar op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3077    const pcre_uint32 *list_ptr;
3078  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3079    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3080    const pcre_uchar *xclass_flags;
3081    #endif
3082    const pcre_uint8 *class_bitset;
3083    const pcre_uint8 *set1, *set2, *set_end;
3084    pcre_uint32 chr;
3085    BOOL accepted, invert_bits;
3086    BOOL entered_a_group = FALSE;
3087    
3088    /* Note: the base_list[1] contains whether the current opcode has greedy
3089    (represented by a non-zero value) quantifier. This is a different from
3090    other character type lists, which stores here that the character iterator
3091    matches to an empty string (also represented by a non-zero value). */
3092    
3093  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3094    {    {
3095    for (;;)    /* All operations move the code pointer forward.
3096      Therefore infinite recursions are not possible. */
3097    
3098      c = *code;
3099    
3100      /* Skip over callouts */
3101    
3102      if (c == OP_CALLOUT)
3103      {      {
3104      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3105      if (*ptr == CHAR_NUMBER_SIGN)      continue;
       {  
       ptr++;  
       while (*ptr != CHAR_NULL)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3106      }      }
   }  
3107    
3108  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3109  value is a character, a negative value is an escape value. */      {
3110        do code += GET(code, 1); while (*code == OP_ALT);
3111        c = *code;
3112        }
3113    
3114  if (*ptr == CHAR_BACKSLASH)    switch(c)
3115    {      {
3116    int temperrorcode = 0;      case OP_END:
3117    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,      case OP_KETRPOS:
3118      FALSE);      /* TRUE only in greedy case. The non-greedy case could be replaced by
3119    if (temperrorcode != 0) return FALSE;      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3120    ptr++;    /* Point after the escape sequence */      uses more memory, which we cannot get at this stage.) */
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3121    
3122  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3123    
3124  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3125    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3126    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3127      {      cannot be converted to a possessive form. */
3128      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3129      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[1] == 0) return FALSE;
3130    
3131        switch(*(code - GET(code, 1)))
3132        {        {
3133        ptr++;        case OP_ASSERT:
3134        while (*ptr != CHAR_NULL)        case OP_ASSERT_NOT:
3135          {        case OP_ASSERTBACK:
3136          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }        case OP_ASSERTBACK_NOT:
3137          ptr++;        case OP_ONCE:
3138  #ifdef SUPPORT_UTF        case OP_ONCE_NC:
3139          if (utf) FORWARDCHAR(ptr);        /* Atomic sub-patterns and assertions can always auto-possessify their
3140  #endif        last iterator. However, if the group was entered as a result of checking
3141          }        a previous iterator, this is not possible. */
3142    
3143          return !entered_a_group;
3144        }        }
     else break;  
     }  
   }  
3145    
3146  /* If the next thing is itself optional, we have to give up. */      code += PRIV(OP_lengths)[c];
3147        continue;
3148    
3149        case OP_ONCE:
3150        case OP_ONCE_NC:
3151        case OP_BRA:
3152        case OP_CBRA:
3153        next_code = code + GET(code, 1);
3154        code += PRIV(OP_lengths)[c];
3155    
3156  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      while (*next_code == OP_ALT)
3157    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)        {
3158      return FALSE;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3159          code = next_code + 1 + LINK_SIZE;
3160          next_code += GET(next_code, 1);
3161          }
3162    
3163  /* If the previous item is a character, get its value. */      entered_a_group = TRUE;
3164        continue;
3165    
3166  if (op_code == OP_CHAR || op_code == OP_CHARI ||      case OP_BRAZERO:
3167      op_code == OP_NOT || op_code == OP_NOTI)      case OP_BRAMINZERO:
   {  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3168    
3169  /* Now compare the next item with the previous opcode. First, handle cases when      next_code = code + 1;
3170  the next item is a character. */      if (*next_code != OP_BRA && *next_code != OP_CBRA
3171            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3172    
3173        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3174    
3175        /* The bracket content will be checked by the
3176        OP_BRA/OP_CBRA case above. */
3177        next_code += 1 + LINK_SIZE;
3178        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3179          return FALSE;
3180    
3181  if (escape == 0)      code += PRIV(OP_lengths)[c];
3182    {      continue;
   /* For a caseless UTF match, the next character may have more than one other  
   case, which maps to the special PT_CLIST property. Check this first. */  
3183    
3184  #ifdef SUPPORT_UCP      default:
3185    if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)      break;
     {  
     unsigned int ocs = UCD_CASESET(next);  
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
3186      }      }
 #endif  
3187    
3188    switch(op_code)    /* Check for a supported opcode, and load its properties. */
3189    
3190      code = get_chr_property_list(code, utf, cd->fcc, list);
3191      if (code == NULL) return FALSE;    /* Unsupported */
3192    
3193      /* If either opcode is a small character list, set pointers for comparing
3194      characters from that list with another list, or with a property. */
3195    
3196      if (base_list[0] == OP_CHAR)
3197      {      {
3198      case OP_CHAR:      chr_ptr = base_list + 2;
3199      return c != next;      list_ptr = list;
3200        }
3201      else if (list[0] == OP_CHAR)
3202        {
3203        chr_ptr = list + 2;
3204        list_ptr = base_list;
3205        }
3206    
3207      /* For CHARI (caseless character) we must check the other case. If we have    /* Character bitsets can also be compared to certain opcodes. */
     Unicode property support, we can use it to test the other case of  
     high-valued characters. We know that next can have only one other case,  
     because multi-other-case characters are dealt with above. */  
3208    
3209      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3210      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3211  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3212      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3213        {  #endif
3214        pcre_uint32 othercase;        )
3215        if (next < 128) othercase = cd->fcc[next]; else      {
3216  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3217        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3218  #else  #else
3219        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3220  #endif  #endif
3221        return c != othercase;        {
3222          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3223          list_ptr = list;
3224        }        }
3225      else      else
3226  #endif  /* SUPPORT_UTF */        {
3227      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */        set1 = (pcre_uint8 *)(code - list[2]);
3228          list_ptr = base_list;
3229          }
3230    
3231      case OP_NOT:      invert_bits = FALSE;
3232      return c == next;      switch(list_ptr[0])
3233          {
3234          case OP_CLASS:
3235          case OP_NCLASS:
3236          set2 = (pcre_uint8 *)
3237            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3238          break;
3239    
3240      case OP_NOTI:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3241      if (c == next) return TRUE;        case OP_XCLASS:
3242  #ifdef SUPPORT_UTF        xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3243      if (utf)        if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3244          if ((*xclass_flags & XCL_MAP) == 0)
3245            {
3246            /* No bits are set for characters < 256. */
3247            if (list[1] == 0) return TRUE;
3248            /* Might be an empty repeat. */
3249            continue;
3250            }
3251          set2 = (pcre_uint8 *)(xclass_flags + 1);
3252          break;
3253    #endif
3254    
3255          case OP_NOT_DIGIT:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_DIGIT:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3260          break;
3261    
3262          case OP_NOT_WHITESPACE:
3263          invert_bits = TRUE;
3264          /* Fall through */
3265          case OP_WHITESPACE:
3266          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3267          break;
3268    
3269          case OP_NOT_WORDCHAR:
3270          invert_bits = TRUE;
3271          /* Fall through */
3272          case OP_WORDCHAR:
3273          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3274          break;
3275    
3276          default:
3277          return FALSE;
3278          }
3279    
3280        /* Because the sets are unaligned, we need
3281        to perform byte comparison here. */
3282        set_end = set1 + 32;
3283        if (invert_bits)
3284        {        {
3285        pcre_uint32 othercase;        do
3286        if (next < 128) othercase = cd->fcc[next]; else          {
3287            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3288            }
3289          while (set1 < set_end);
3290          }
3291        else
3292          {
3293          do
3294            {
3295            if ((*set1++ & *set2++) != 0) return FALSE;
3296            }
3297          while (set1 < set_end);
3298          }
3299    
3300        if (list[1] == 0) return TRUE;
3301        /* Might be an empty repeat. */
3302        continue;
3303        }
3304    
3305      /* Some property combinations also acceptable. Unicode property opcodes are
3306      processed specially; the rest can be handled with a lookup table. */
3307    
3308      else
3309        {
3310        pcre_uint32 leftop, rightop;
3311    
3312        leftop = base_list[0];
3313        rightop = list[0];
3314    
3315  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3316        othercase = UCD_OTHERCASE(next);      accepted = FALSE; /* Always set in non-unicode case. */
3317  #else      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3318        othercase = NOTACHAR;        {
3319          if (rightop == OP_EOD)
3320            accepted = TRUE;
3321          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3322            {
3323            int n;
3324            const pcre_uint8 *p;
3325            BOOL same = leftop == rightop;
3326            BOOL lisprop = leftop == OP_PROP;
3327            BOOL risprop = rightop == OP_PROP;
3328            BOOL bothprop = lisprop && risprop;
3329    
3330            /* There's a table that specifies how each combination is to be
3331            processed:
3332              0   Always return FALSE (never auto-possessify)
3333              1   Character groups are distinct (possessify if both are OP_PROP)
3334              2   Check character categories in the same group (general or particular)
3335              3   Return TRUE if the two opcodes are not the same
3336              ... see comments below
3337            */
3338    
3339            n = propposstab[base_list[2]][list[2]];
3340            switch(n)
3341              {
3342              case 0: break;
3343              case 1: accepted = bothprop; break;
3344              case 2: accepted = (base_list[3] == list[3]) != same; break;
3345              case 3: accepted = !same; break;
3346    
3347              case 4:  /* Left general category, right particular category */
3348              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3349              break;
3350    
3351              case 5:  /* Right general category, left particular category */
3352              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3353              break;
3354    
3355              /* This code is logically tricky. Think hard before fiddling with it.
3356              The posspropstab table has four entries per row. Each row relates to
3357              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3358              Only WORD actually needs all four entries, but using repeats for the
3359              others means they can all use the same code below.
3360    
3361              The first two entries in each row are Unicode general categories, and
3362              apply always, because all the characters they include are part of the
3363              PCRE character set. The third and fourth entries are a general and a
3364              particular category, respectively, that include one or more relevant
3365              characters. One or the other is used, depending on whether the check
3366              is for a general or a particular category. However, in both cases the
3367              category contains more characters than the specials that are defined
3368              for the property being tested against. Therefore, it cannot be used
3369              in a NOTPROP case.
3370    
3371              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3372              Underscore is covered by ucp_P or ucp_Po. */
3373    
3374              case 6:  /* Left alphanum vs right general category */
3375              case 7:  /* Left space vs right general category */
3376              case 8:  /* Left word vs right general category */
3377              p = posspropstab[n-6];
3378              accepted = risprop && lisprop ==
3379                (list[3] != p[0] &&
3380                 list[3] != p[1] &&
3381                (list[3] != p[2] || !lisprop));
3382              break;
3383    
3384              case 9:   /* Right alphanum vs left general category */
3385              case 10:  /* Right space vs left general category */
3386              case 11:  /* Right word vs left general category */
3387              p = posspropstab[n-9];
3388              accepted = lisprop && risprop ==
3389                (base_list[3] != p[0] &&
3390                 base_list[3] != p[1] &&
3391                (base_list[3] != p[2] || !risprop));
3392              break;
3393    
3394              case 12:  /* Left alphanum vs right particular category */
3395              case 13:  /* Left space vs right particular category */
3396              case 14:  /* Left word vs right particular category */
3397              p = posspropstab[n-12];
3398              accepted = risprop && lisprop ==
3399                (catposstab[p[0]][list[3]] &&
3400                 catposstab[p[1]][list[3]] &&
3401                (list[3] != p[3] || !lisprop));
3402              break;
3403    
3404              case 15:  /* Right alphanum vs left particular category */
3405              case 16:  /* Right space vs left particular category */
3406              case 17:  /* Right word vs left particular category */
3407              p = posspropstab[n-15];
3408              accepted = lisprop && risprop ==
3409                (catposstab[p[0]][base_list[3]] &&
3410                 catposstab[p[1]][base_list[3]] &&
3411                (base_list[3] != p[3] || !risprop));
3412              break;
3413              }
3414            }
3415          }
3416    
3417        else
3418    #endif  /* SUPPORT_UCP */
3419    
3420        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3421               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3422               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3423    
3424        if (!accepted) return FALSE;
3425    
3426        if (list[1] == 0) return TRUE;
3427        /* Might be an empty repeat. */
3428        continue;
3429        }
3430    
3431      /* Control reaches here only if one of the items is a small character list.
3432      All characters are checked against the other side. */
3433    
3434      do
3435        {
3436        chr = *chr_ptr;
3437    
3438        switch(list_ptr[0])
3439          {
3440          case OP_CHAR:
3441          ochr_ptr = list_ptr + 2;
3442          do
3443            {
3444            if (chr == *ochr_ptr) return FALSE;
3445            ochr_ptr++;
3446            }
3447          while(*ochr_ptr != NOTACHAR);
3448          break;
3449    
3450          case OP_NOT:
3451          ochr_ptr = list_ptr + 2;
3452          do
3453            {
3454            if (chr == *ochr_ptr)
3455              break;
3456            ochr_ptr++;
3457            }
3458          while(*ochr_ptr != NOTACHAR);
3459          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3460          break;
3461    
3462          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3463          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3464    
3465          case OP_DIGIT:
3466          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3467          break;
3468    
3469          case OP_NOT_DIGIT:
3470          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3471          break;
3472    
3473          case OP_WHITESPACE:
3474          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3475          break;
3476    
3477          case OP_NOT_WHITESPACE:
3478          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3479          break;
3480    
3481          case OP_WORDCHAR:
3482          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3483          break;
3484    
3485          case OP_NOT_WORDCHAR:
3486          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3487          break;
3488    
3489          case OP_HSPACE:
3490          switch(chr)
3491            {
3492            HSPACE_CASES: return FALSE;
3493            default: break;
3494            }
3495          break;
3496    
3497          case OP_NOT_HSPACE:
3498          switch(chr)
3499            {
3500            HSPACE_CASES: break;
3501            default: return FALSE;
3502            }
3503          break;
3504    
3505          case OP_ANYNL:
3506          case OP_VSPACE:
3507          switch(chr)
3508            {
3509            VSPACE_CASES: return FALSE;
3510            default: break;
3511            }
3512          break;
3513    
3514          case OP_NOT_VSPACE:
3515          switch(chr)
3516            {
3517            VSPACE_CASES: break;
3518            default: return FALSE;
3519            }
3520          break;
3521    
3522          case OP_DOLL:
3523          case OP_EODN:
3524          switch (chr)
3525            {
3526            case CHAR_CR:
3527            case CHAR_LF:
3528            case CHAR_VT:
3529            case CHAR_FF:
3530            case CHAR_NEL:
3531    #ifndef EBCDIC
3532            case 0x2028:
3533            case 0x2029:
3534    #endif  /* Not EBCDIC */
3535            return FALSE;
3536            }
3537          break;
3538    
3539          case OP_EOD:    /* Can always possessify before \z */
3540          break;
3541    
3542    #ifdef SUPPORT_UCP
3543          case OP_PROP:
3544          case OP_NOTPROP:
3545          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3546                list_ptr[0] == OP_NOTPROP))
3547            return FALSE;
3548          break;
3549  #endif  #endif
3550        return c == othercase;  
3551          case OP_NCLASS:
3552          if (chr > 255) return FALSE;
3553          /* Fall through */
3554    
3555          case OP_CLASS:
3556          if (chr > 255) break;
3557          class_bitset = (pcre_uint8 *)
3558            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3559          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3560          break;
3561    
3562    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3563          case OP_XCLASS:
3564          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3565              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3566          break;
3567    #endif
3568    
3569          default:
3570          return FALSE;
3571        }        }
3572    
3573        chr_ptr++;
3574        }
3575      while(*chr_ptr != NOTACHAR);
3576    
3577      /* At least one character must be matched from this opcode. */
3578    
3579      if (list[1] == 0) return TRUE;
3580      }
3581    
3582    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3583    but some compilers complain about an unreachable statement. */
3584    
3585    }
3586    
3587    
3588    
3589    /*************************************************
3590    *    Scan compiled regex for auto-possession     *
3591    *************************************************/
3592    
3593    /* Replaces single character iterations with their possessive alternatives
3594    if appropriate. This function modifies the compiled opcode!
3595    
3596    Arguments:
3597      code        points to start of the byte code
3598      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3599      cd          static compile data
3600    
3601    Returns:      nothing
3602    */
3603    
3604    static void
3605    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3606    {
3607    register pcre_uchar c;
3608    const pcre_uchar *end;
3609    pcre_uchar *repeat_opcode;
3610    pcre_uint32 list[8];
3611    
3612    for (;;)
3613      {
3614      c = *code;
3615    
3616      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3617        {
3618        c -= get_repeat_base(c) - OP_STAR;
3619        end = (c <= OP_MINUPTO) ?
3620          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3621        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3622    
3623        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3624          {
3625          switch(c)
3626            {
3627            case OP_STAR:
3628            *code += OP_POSSTAR - OP_STAR;
3629            break;
3630    
3631            case OP_MINSTAR:
3632            *code += OP_POSSTAR - OP_MINSTAR;
3633            break;
3634    
3635            case OP_PLUS:
3636            *code += OP_POSPLUS - OP_PLUS;
3637            break;
3638    
3639            case OP_MINPLUS:
3640            *code += OP_POSPLUS - OP_MINPLUS;
3641            break;
3642    
3643            case OP_QUERY:
3644            *code += OP_POSQUERY - OP_QUERY;
3645            break;
3646    
3647            case OP_MINQUERY:
3648            *code += OP_POSQUERY - OP_MINQUERY;
3649            break;
3650    
3651            case OP_UPTO:
3652            *code += OP_POSUPTO - OP_UPTO;
3653            break;
3654    
3655            case OP_MINUPTO:
3656            *code += OP_POSUPTO - OP_MINUPTO;
3657            break;
3658            }
3659          }
3660        c = *code;
3661        }
3662      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3663        {
3664    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3665        if (c == OP_XCLASS)
3666          repeat_opcode = code + GET(code, 1);
3667      else      else
3668  #endif  /* SUPPORT_UTF */  #endif
3669      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */        repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3670    
3671        c = *repeat_opcode;
3672        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3673          {
3674          /* end must not be NULL. */
3675          end = get_chr_property_list(code, utf, cd->fcc, list);
3676    
3677      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.        list[1] = (c & 1) == 0;
     When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  
3678    
3679      case OP_DIGIT:        if (compare_opcodes(end, utf, cd, list, end))
3680      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;          {
3681            switch (c)
3682              {
3683              case OP_CRSTAR:
3684              case OP_CRMINSTAR:
3685              *repeat_opcode = OP_CRPOSSTAR;
3686              break;
3687    
3688      case OP_NOT_DIGIT:            case OP_CRPLUS:
3689      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;            case OP_CRMINPLUS:
3690              *repeat_opcode = OP_CRPOSPLUS;
3691              break;
3692    
3693              case OP_CRQUERY:
3694              case OP_CRMINQUERY:
3695              *repeat_opcode = OP_CRPOSQUERY;
3696              break;
3697    
3698              case OP_CRRANGE:
3699              case OP_CRMINRANGE:
3700              *repeat_opcode = OP_CRPOSRANGE;
3701              break;
3702              }
3703            }
3704          }
3705        c = *code;
3706        }
3707    
3708      switch(c)
3709        {
3710        case OP_END:
3711        return;
3712    
3713        case OP_TYPESTAR:
3714        case OP_TYPEMINSTAR:
3715        case OP_TYPEPLUS:
3716        case OP_TYPEMINPLUS:
3717        case OP_TYPEQUERY:
3718        case OP_TYPEMINQUERY:
3719        case OP_TYPEPOSSTAR:
3720        case OP_TYPEPOSPLUS:
3721        case OP_TYPEPOSQUERY:
3722        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3723        break;
3724    
3725        case OP_TYPEUPTO:
3726        case OP_TYPEMINUPTO:
3727        case OP_TYPEEXACT:
3728        case OP_TYPEPOSUPTO:
3729        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3730          code += 2;
3731        break;
3732    
3733    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3734        case OP_XCLASS:
3735        code += GET(code, 1);
3736        break;
3737    #endif
3738    
3739        case OP_MARK:
3740        case OP_PRUNE_ARG:
3741        case OP_SKIP_ARG:
3742        case OP_THEN_ARG:
3743        code += code[1];
3744        break;
3745        }
3746    
3747      /* Add in the fixed length from the table */
3748    
3749      code += PRIV(OP_lengths)[c];
3750    
3751      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3752      a multi-byte character. The length in the table is a minimum, so we have to
3753      arrange to skip the extra bytes. */
3754    
3755    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3756      if (utf) switch(c)
3757        {
3758        case OP_CHAR:
3759        case OP_CHARI:
3760        case OP_NOT:
3761        case OP_NOTI:
3762        case OP_STAR:
3763        case OP_MINSTAR:
3764        case OP_PLUS:
3765        case OP_MINPLUS:
3766        case OP_QUERY:
3767        case OP_MINQUERY:
3768        case OP_UPTO:
3769        case OP_MINUPTO:
3770        case OP_EXACT:
3771        case OP_POSSTAR:
3772        case OP_POSPLUS:
3773        case OP_POSQUERY:
3774        case OP_POSUPTO:
3775        case OP_STARI:
3776        case OP_MINSTARI:
3777        case OP_PLUSI:
3778        case OP_MINPLUSI:
3779        case OP_QUERYI:
3780        case OP_MINQUERYI:
3781        case OP_UPTOI:
3782        case OP_MINUPTOI:
3783        case OP_EXACTI:
3784        case OP_POSSTARI:
3785        case OP_POSPLUSI:
3786        case OP_POSQUERYI:
3787        case OP_POSUPTOI:
3788        case OP_NOTSTAR:
3789        case OP_NOTMINSTAR:
3790        case OP_NOTPLUS:
3791        case OP_NOTMINPLUS:
3792        case OP_NOTQUERY:
3793        case OP_NOTMINQUERY:
3794        case OP_NOTUPTO:
3795        case OP_NOTMINUPTO:
3796        case OP_NOTEXACT:
3797        case OP_NOTPOSSTAR:
3798        case OP_NOTPOSPLUS:
3799        case OP_NOTPOSQUERY:
3800        case OP_NOTPOSUPTO:
3801        case OP_NOTSTARI:
3802        case OP_NOTMINSTARI:
3803        case OP_NOTPLUSI:
3804        case OP_NOTMINPLUSI:
3805        case OP_NOTQUERYI:
3806        case OP_NOTMINQUERYI:
3807        case OP_NOTUPTOI:
3808        case OP_NOTMINUPTOI:
3809        case OP_NOTEXACTI:
3810        case OP_NOTPOSSTARI:
3811        case OP_NOTPOSPLUSI:
3812        case OP_NOTPOSQUERYI:
3813        case OP_NOTPOSUPTOI:
3814        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3815        break;
3816        }
3817    #else
3818      (void)(utf);  /* Keep compiler happy by referencing function argument */
3819    #endif
3820      }
3821    }
3822    
3823    
3824    
3825    /*************************************************
3826    *           Check for POSIX class syntax         *
3827    *************************************************/
3828    
3829    /* This function is called when the sequence "[:" or "[." or "[=" is
3830    encountered in a character class. It checks whether this is followed by a
3831    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3832    reach an unescaped ']' without the special preceding character, return FALSE.
3833    
3834    Originally, this function only recognized a sequence of letters between the
3835    terminators, but it seems that Perl recognizes any sequence of characters,
3836    though of course unknown POSIX names are subsequently rejected. Perl gives an
3837    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3838    didn't consider this to be a POSIX class. Likewise for [:1234:].
3839    
3840    The problem in trying to be exactly like Perl is in the handling of escapes. We
3841    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3842    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3843    below handles the special case of \], but does not try to do any other escape
3844    processing. This makes it different from Perl for cases such as [:l\ower:]
3845    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3846    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3847    I think.
3848    
3849    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3850    It seems that the appearance of a nested POSIX class supersedes an apparent
3851    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3852    a digit.
3853    
3854    In Perl, unescaped square brackets may also appear as part of class names. For
3855    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3856    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3857    seem right at all. PCRE does not allow closing square brackets in POSIX class
3858    names.
3859    
3860    Arguments:
3861      ptr      pointer to the initial [
3862      endptr   where to return the end pointer
3863    
3864    Returns:   TRUE or FALSE
3865    */
3866    
3867    static BOOL
3868    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3869    {
3870    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3871    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3872    for (++ptr; *ptr != CHAR_NULL; ptr++)
3873      {
3874      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3875        ptr++;
3876      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3877      else
3878        {
3879        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3880          {
3881          *endptr = ptr;
3882          return TRUE;
3883          }
3884        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3885             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3886              ptr[1] == CHAR_EQUALS_SIGN) &&
3887            check_posix_syntax(ptr, endptr))
3888          return FALSE;
3889        }
3890      }
3891    return FALSE;
3892    }
3893    
3894    
3895    
3896    
3897    /*************************************************
3898    *          Check POSIX class name                *
3899    *************************************************/
3900    
3901    /* This function is called to check the name given in a POSIX-style class entry
3902    such as [:alnum:].
3903    
3904    Arguments:
3905      ptr        points to the first letter
3906      len        the length of the name
3907    
3908    Returns:     a value representing the name, or -1 if unknown
3909    */
3910    
3911    static int
3912    check_posix_name(const pcre_uchar *ptr, int len)
3913    {
3914    const char *pn = posix_names;
3915    register int yield = 0;
3916    while (posix_name_lengths[yield] != 0)
3917      {
3918      if (len == posix_name_lengths[yield] &&
3919        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3920      pn += posix_name_lengths[yield] + 1;
3921      yield++;
3922      }
3923    return -1;
3924    }
3925    
3926    
3927    /*************************************************
3928    *    Adjust OP_RECURSE items in repeated group   *
3929    *************************************************/
3930    
3931    /* OP_RECURSE items contain an offset from the start of the regex to the group
3932    that is referenced. This means that groups can be replicated for fixed
3933    repetition simply by copying (because the recursion is allowed to refer to
3934    earlier groups that are outside the current group). However, when a group is
3935    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3936    inserted before it, after it has been compiled. This means that any OP_RECURSE
3937    items within it that refer to the group itself or any contained groups have to
3938    have their offsets adjusted. That one of the jobs of this function. Before it
3939    is called, the partially compiled regex must be temporarily terminated with
3940    OP_END.
3941    
3942      case OP_WHITESPACE:  This function has been extended with the possibility of forward references for
3943      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  recursions and subroutine calls. It must also check the list of such references
3944    for the group we are dealing with. If it finds that one of the recursions in
3945    the current group is on this list, it adjusts the offset in the list, not the
3946    value in the reference (which is a group number).
3947    
3948      case OP_NOT_WHITESPACE:  Arguments:
3949      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;    group      points to the start of the group
3950      adjust     the amount by which the group is to be moved
3951      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3952      cd         contains pointers to tables etc.
3953      save_hwm   the hwm forward reference pointer at the start of the group
3954    
3955      case OP_WORDCHAR:  Returns:     nothing
3956      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  */
3957    
3958      case OP_NOT_WORDCHAR:  static void
3959      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3960      pcre_uchar *save_hwm)
3961    {
3962    pcre_uchar *ptr = group;
3963    
3964      case OP_HSPACE:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3965      case OP_NOT_HSPACE:    {
3966      switch(next)    int offset;
3967        {    pcre_uchar *hc;
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3968    
3969        default:    /* See if this recursion is on the forward reference list. If so, adjust the
3970        return op_code != OP_NOT_HSPACE;    reference. */
       }  
3971    
3972      case OP_ANYNL:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3973      case OP_VSPACE:      {
3974      case OP_NOT_VSPACE:      offset = (int)GET(hc, 0);
3975      switch(next)      if (cd->start_code + offset == ptr + 1)
3976        {        {
3977        VSPACE_CASES:        PUT(hc, 0, offset + adjust);
3978        return op_code == OP_NOT_VSPACE;        break;
   
       default:  
       return op_code != OP_NOT_VSPACE;  
3979        }        }
   
 #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;  
3980      }      }
   }  
3981    
3982  /* 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
3983  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. */  
3984    
3985  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
3986      {      {
3987      case ESC_d:      offset = (int)GET(ptr, 1);
3988      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3989        }
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3990    
3991      case ESC_S:    ptr += 1 + LINK_SIZE;
3992      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    }
3993    }
3994    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3995    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3996    
3997      case ESC_h:  /*************************************************
3998      case ESC_H:  *        Insert an automatic callout point       *
3999      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
4000    
4001        default:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4002        return escape == ESC_h;  callout points before each pattern item.
       }  
4003    
4004      case ESC_v:  Arguments:
4005      case ESC_V:    code           current code pointer
4006      switch(c)    ptr            current pattern pointer
4007        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return escape != ESC_v;  
4008    
4009        default:  Returns:         new code pointer
4010        return escape == ESC_v;  */
       }  
4011    
4012      /* When PCRE_UCP is set, these values get generated for \d etc. Find  static pcre_uchar *
4013      their substitutions and process them. The result will always be either  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4014      ESC_p or ESC_P. Then fall through to process those values. */  {
4015    *code++ = OP_CALLOUT;
4016    *code++ = 255;
4017    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4018    PUT(code, LINK_SIZE, 0);                       /* Default length */
4019    return code + 2 * LINK_SIZE;
4020    }
4021    
 #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 */  
4022    
     case ESC_p:  
     case ESC_P:  
       {  
       unsigned int ptype = 0, pdata = 0;  
       int errorcodeptr;  
       BOOL negated;  
4023    
4024        ptr--;      /* Make ptr point at the p or P */  /*************************************************
4025        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  *         Complete a callout item                *
4026          return FALSE;  *************************************************/
       ptr++;      /* Point past the final curly ket */  
4027    
4028        /* 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
4029        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
4030        to the original \d etc. At this point, ptr will point to a zero byte. */  for both automatic and manual callouts.
4031    
4032        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  Arguments:
4033          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    previous_callout   points to previous callout item
4034            return FALSE;    ptr                current pattern pointer
4035      cd                 pointers to tables etc
4036    
4037        /* Do the property check. */  Returns:             nothing
4038    */
4039    
4040        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  static void
4041        }  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4042  #endif  {
4043    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4044    PUT(previous_callout, 2 + LINK_SIZE, length);
4045    }
4046    
     default:  
     return FALSE;  
     }  
4047    
   /* 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.) */  
4048    
4049    case OP_DIGIT:  #ifdef SUPPORT_UCP
4050    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  /*************************************************
4051           escape == ESC_h || escape == ESC_v || escape == ESC_R;  *           Get othercase range                  *
4052    *************************************************/
4053    
4054    case OP_NOT_DIGIT:  /* This function is passed the start and end of a class range, in UTF-8 mode
4055    return escape == ESC_d;  with UCP support. It searches up the characters, looking for ranges of
4056    characters in the "other" case. Each call returns the next one, updating the
4057    start address. A character with multiple other cases is returned on its own
4058    with a special return value.
4059    
4060    case OP_WHITESPACE:  Arguments:
4061    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4062      d           end value
4063      ocptr       where to put start of othercase range
4064      odptr       where to put end of othercase range
4065    
4066    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4067    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4068                  >0 the CASESET offset for char with multiple other cases
4069                    in this case, ocptr contains the original
4070    */
4071    
4072    case OP_HSPACE:  static int
4073    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4074           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4075    {
4076    pcre_uint32 c, othercase, next;
4077    unsigned int co;
4078    
4079    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4080    return escape == ESC_h;  cases, return its case offset value. */
4081    
4082    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4083    case OP_ANYNL:    {
4084    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4085    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4086        *ocptr = c++;   /* Character that has the set */
4087        *cptr = c;      /* Rest of input range */
4088        return (int)co;
4089        }
4090      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4091      }
4092    
4093    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4094    
4095    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4096    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
4097           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4098    
4099    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4100    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4101    
4102    default:  for (++c; c <= d; c++)
4103    return FALSE;    {
4104      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4105      next++;
4106    }    }
4107    
4108  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4109    *cptr = c;             /* Rest of input range */
4110    return 0;
4111  }  }
4112    #endif  /* SUPPORT_UCP */
4113    
4114    
4115    
# Line 3531  add_to_class(pcre_uint8 *classbits, pcre Line 4139  add_to_class(pcre_uint8 *classbits, pcre
4139    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4140  {  {
4141  pcre_uint32 c;  pcre_uint32 c;
4142    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4143  int n8 = 0;  int n8 = 0;
4144    
4145  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
# Line 3574  if ((options & PCRE_CASELESS) != 0) Line 4183  if ((options & PCRE_CASELESS) != 0)
4183    
4184    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4185    
4186    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4187      {      {
4188      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4189      n8++;      n8++;
# Line 3599  in all cases. */ Line 4208  in all cases. */
4208    
4209  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4210    
4211  /* 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. */  
4212    
4213  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4214    {    {
4215    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4216      {    SETBIT(classbits, c);
4217      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4218    }    }
4219    
4220  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4221    if (start <= 0xff) start = 0xff + 1;
4222    
4223    if (end >= start)
4224    {    {
4225    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4226  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4227    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4228      {      {
# Line 3654  else Line 4262  else
4262    
4263    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4264    }    }
4265    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4266    
4267  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4268  }  }
# Line 3754  to find out the amount of memory needed, Line 4363  to find out the amount of memory needed,
4363  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4364    
4365  Arguments:  Arguments:
4366    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4367    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4368    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4369    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4370    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
4371    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
4372    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
4373    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
4374    bcptr          points to current branch chain    bcptr             points to current branch chain
4375    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
4376    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
4377    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
4378                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
4379    
4380  Returns:         TRUE on success  Returns:            TRUE on success
4381                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4382  */  */
4383    
4384  static BOOL  static BOOL
# Line 3875  for (;; ptr++) Line 4484  for (;; ptr++)
4484    BOOL reset_bracount;    BOOL reset_bracount;
4485    int class_has_8bitchar;    int class_has_8bitchar;
4486    int class_one_char;    int class_one_char;
4487    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4488      BOOL xclass_has_prop;
4489    #endif
4490    int newoptions;    int newoptions;
4491    int recno;    int recno;
4492    int refsign;    int refsign;
# Line 3992  for (;; ptr++) Line 4604  for (;; ptr++)
4604          }          }
4605        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4606        }        }
4607        /* Control does not reach here. */
4608      }      }
4609    
4610    /* 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
4611    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. */  
4612    
4613    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4614      {      {
4615      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4616        {        {
4617          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4618          if (c != CHAR_NUMBER_SIGN) break;
4619        ptr++;        ptr++;
4620        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4621          {          {
4622          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4623              {                          /* IS_NEWLINE sets cd->nllen. */
4624              ptr += cd->nllen;
4625              break;
4626              }
4627          ptr++;          ptr++;
4628  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4629          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4630  #endif  #endif
4631          }          }
4632        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;  
4633        }        }
4634      }      }
4635    
4636    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4637    
4638      is_quantifier =
4639        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4640        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4641    
4642      /* Fill in length of a previous callout, except when the next thing is a
4643      quantifier or when processing a property substitution string in UCP mode. */
4644    
4645      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4646           after_manual_callout-- <= 0)
4647        {
4648        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4649          complete_callout(previous_callout, ptr, cd);
4650        previous_callout = NULL;
4651        }
4652    
4653      /* Create auto callout, except for quantifiers, or while processing property
4654      strings that are substituted for \w etc in UCP mode. */
4655    
4656    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4657      {      {
4658      previous_callout = code;      previous_callout = code;
4659      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4660      }      }
4661    
4662      /* Process the next pattern item. */
4663    
4664    switch(c)    switch(c)
4665      {      {
4666      /* ===================================================================*/      /* ===================================================================*/
4667      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4668      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4669      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4670      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4073  for (;; ptr++) Line 4694  for (;; ptr++)
4694      previous = NULL;      previous = NULL;
4695      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4696        {        {
4697        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;        if (firstcharflags == REQ_UNSET)
4698            zerofirstcharflags = firstcharflags = REQ_NONE;
4699        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4700        }        }
4701      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 4121  for (;; ptr++) Line 4743  for (;; ptr++)
4743        }        }
4744      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4745    
4746        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4747        used for "start of word" and "end of word". As these are otherwise illegal
4748        sequences, we don't break anything by recognizing them. They are replaced
4749        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4750        erroneous and are handled by the normal code below. */
4751    
4752      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4753        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4754          {
4755          nestptr = ptr + 7;
4756          ptr = sub_start_of_word - 1;
4757          continue;
4758          }
4759    
4760        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4761          {
4762          nestptr = ptr + 7;
4763          ptr = sub_end_of_word - 1;
4764          continue;
4765          }
4766    
4767        /* Handle a real character class. */
4768    
4769      previous = code;      previous = code;
4770    
4771      /* 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 4822  for (;; ptr++)
4822    
4823      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4824    
4825        /* Extended class (xclass) will be used when characters > 255
4826        might match. */
4827    
4828    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4829        xclass = FALSE;
4830        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4831        class_uchardata_base = class_uchardata;   /* Save the start */
4832    #endif
4833    
4834      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4835      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 <
4836      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
4837      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4838        are present in the class. */
4839    
4840      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4841      class_one_char = 0;      class_one_char = 0;
4842    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4843        xclass_has_prop = FALSE;
4844    #endif
4845    
4846      /* 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
4847      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 4850  for (;; ptr++)
4850    
4851      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4852    
 #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  
   
4853      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4854      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
4855      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4224  for (;; ptr++) Line 4875  for (;; ptr++)
4875        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4876          {          {
4877          xclass = TRUE;          xclass = TRUE;
4878          *lengthptr += class_uchardata - class_uchardata_base;          *lengthptr += (int)(class_uchardata - class_uchardata_base);
4879          class_uchardata = class_uchardata_base;          class_uchardata = class_uchardata_base;
4880          }          }
4881  #endif  #endif
# Line 4286  for (;; ptr++) Line 4937  for (;; ptr++)
4937            posix_class = 0;            posix_class = 0;
4938    
4939          /* 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
4940          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4941            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4942            directly. */
4943    
4944  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4945          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4946            {            {
4947              unsigned int ptype = 0;
4948            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4949    
4950              /* The posix_substitutes table specifies which POSIX classes can be
4951              converted to \p or \P items. */
4952    
4953            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4954              {              {
4955              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4956              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4957              continue;              continue;
4958              }              }
4959    
4960              /* There are three other classes that generate special property calls
4961              that are recognized only in an XCLASS. */
4962    
4963              else switch(posix_class)
4964                {
4965                case PC_GRAPH:
4966                ptype = PT_PXGRAPH;
4967                /* Fall through */
4968                case PC_PRINT:
4969                if (ptype == 0) ptype = PT_PXPRINT;
4970                /* Fall through */
4971                case PC_PUNCT:
4972                if (ptype == 0) ptype = PT_PXPUNCT;
4973                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4974                *class_uchardata++ = ptype;
4975                *class_uchardata++ = 0;
4976                xclass_has_prop = TRUE;
4977                ptr = tempptr + 1;
4978                continue;
4979    
4980                /* For all other POSIX classes, no special action is taken in UCP
4981                mode. Fall through to the non_UCP case. */
4982    
4983                default:
4984                break;
4985                }
4986            }            }
4987  #endif  #endif
4988          /* 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
4989          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
4990          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
4991          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
4992            bit map that is being built. */
4993    
4994          posix_class *= 3;          posix_class *= 3;
4995    
# Line 4420  for (;; ptr++) Line 5106  for (;; ptr++)
5106              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5107              continue;              continue;
5108    
5109              /* 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
5110              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
5111              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5112              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5113                we could just adjust the appropriate bit. From PCRE 8.34 we no
5114                longer treat \s and \S specially. */
5115    
5116              case ESC_s:              case ESC_s:
5117              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];  
5118              continue;              continue;
5119    
5120              case ESC_S:              case ESC_S:
5121              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5122              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 */  
5123              continue;              continue;
5124    
5125              /* 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 5156  for (;; ptr++)
5156                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5157                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5158                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5159                  xclass_has_prop = TRUE;
5160                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5161                continue;                continue;
5162                }                }
# Line 4556  for (;; ptr++) Line 5242  for (;; ptr++)
5242  #endif  #endif
5243          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5244    
5245          /* 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
5246          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
5247          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5248            warning is given, so PCRE now faults it as it is almost certainly a
5249            mistake on the user's part. */
5250    
5251          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5252            {            {
5253            int descape;            if (d == CHAR_BACKSLASH)
5254            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5255            if (*errorcodeptr != 0) goto FAILED;              int descape;
5256                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5257                if (*errorcodeptr != 0) goto FAILED;
5258    
5259            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5260                special causes an error. */
5261    
5262            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5263                {                {
5264                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5265                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5266                    *errorcodeptr = ERR83;
5267                    goto FAILED;
5268                    }
5269                }                }
5270              }              }
5271    
5272              /* A hyphen followed by a POSIX class is treated in the same way. */
5273    
5274              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5275                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5276                        ptr[1] == CHAR_EQUALS_SIGN) &&
5277                       check_posix_syntax(ptr, &tempptr))
5278                {
5279                *errorcodeptr = ERR83;
5280                goto FAILED;
5281                }
5282            }            }
5283    
5284          /* Check that the two values are in the correct order. Optimize          /* Check that the two values are in the correct order. Optimize
# Line 4622  for (;; ptr++) Line 5325  for (;; ptr++)
5325        whatever repeat count may follow. In the case of reqchar, save the        whatever repeat count may follow. In the case of reqchar, save the
5326        previous value for reinstating. */        previous value for reinstating. */
5327    
5328        if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)        if (!inescq && class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5329          {          {
5330          ptr++;          ptr++;
5331          zeroreqchar = reqchar;          zeroreqchar = reqchar;
# Line 4748  for (;; ptr++) Line 5451  for (;; ptr++)
5451        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5452        code += LINK_SIZE;        code += LINK_SIZE;
5453        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5454          if (xclass_has_prop) *code |= XCL_HASPROP;
5455    
5456        /* 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;
5457        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 5461  for (;; ptr++)
5461          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5462          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5463            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5464            if (negate_class && !xclass_has_prop)
5465              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5466          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5467          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5468          }          }
# Line 4839  for (;; ptr++) Line 5545  for (;; ptr++)
5545    
5546      tempcode = previous;      tempcode = previous;
5547    
5548        /* Before checking for a possessive quantifier, we must skip over
5549        whitespace and comments in extended mode because Perl allows white space at
5550        this point. */
5551    
5552        if ((options & PCRE_EXTENDED) != 0)
5553          {
5554          const pcre_uchar *p = ptr + 1;
5555          for (;;)
5556            {
5557            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5558            if (*p != CHAR_NUMBER_SIGN) break;
5559            p++;
5560            while (*p != CHAR_NULL)
5561              {
5562              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5563                {                        /* IS_NEWLINE sets cd->nllen. */
5564                p += cd->nllen;
5565                break;
5566                }
5567              p++;
5568    #ifdef SUPPORT_UTF
5569              if (utf) FORWARDCHAR(p);
5570    #endif
5571              }           /* Loop for comment characters */
5572            }             /* Loop for multiple comments */
5573          ptr = p - 1;    /* Character before the next significant one. */
5574          }
5575    
5576      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5577      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5578      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 5667  for (;; ptr++)
5667            }            }
5668          }          }
5669    
       /* 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;  
         }  
   
5670        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
5671        }        }
5672    
# Line 4963  for (;; ptr++) Line 5684  for (;; ptr++)
5684        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5685        c = *previous;        c = *previous;
5686    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5687        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5688        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5689          {          {
# Line 5119  for (;; ptr++) Line 5832  for (;; ptr++)
5832      /* 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
5833      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}. */
5834    
5835      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5836  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5837               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5838  #endif  #endif
5839               *previous == OP_REF ||               *previous == OP_REF   || *previous == OP_REFI ||
5840               *previous == OP_REFI)               *previous == OP_DNREF || *previous == OP_DNREFI)
5841        {        {
5842        if (repeat_max == 0)        if (repeat_max == 0)
5843          {          {
# Line 5153  for (;; ptr++) Line 5865  for (;; ptr++)
5865      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
5866      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 >=
5867      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5868      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5869      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5870        for Perl compatibility. */
5871    
5872      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5873        {        {
# Line 5172  for (;; ptr++) Line 5885  for (;; ptr++)
5885        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5886        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5887        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
5888        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5889    
5890        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5891          {          {
# Line 5270  for (;; ptr++) Line 5983  for (;; ptr++)
5983            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
5984            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5985            integer type when available, otherwise double. */            integer type when available, otherwise double. */
5986    
5987            if (lengthptr != NULL)            if (lengthptr != NULL)
5988              {              {
5989              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
# Line 5307  for (;; ptr++) Line 6020  for (;; ptr++)
6020                while (cd->hwm > cd->start_workspace + cd->workspace_size -                while (cd->hwm > cd->start_workspace + cd->workspace_size -
6021                       WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                       WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6022                  {                  {
6023                  int save_offset = save_hwm - cd->start_workspace;                  size_t save_offset = save_hwm - cd->start_workspace;
6024                  int this_offset = this_hwm - cd->start_workspace;                  size_t this_offset = this_hwm - cd->start_workspace;
6025                  *errorcodeptr = expand_workspace(cd);                  *errorcodeptr = expand_workspace(cd);
6026                  if (*errorcodeptr != 0) goto FAILED;                  if (*errorcodeptr != 0) goto FAILED;
6027                  save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;                  save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
# Line 5389  for (;; ptr++) Line 6102  for (;; ptr++)
6102            while (cd->hwm > cd->start_workspace + cd->workspace_size -            while (cd->hwm > cd->start_workspace + cd->workspace_size -
6103                   WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                   WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6104              {              {
6105              int save_offset = save_hwm - cd->start_workspace;              size_t save_offset = save_hwm - cd->start_workspace;
6106              int this_offset = this_hwm - cd->start_workspace;              size_t this_offset = this_hwm - cd->start_workspace;
6107              *errorcodeptr = expand_workspace(cd);              *errorcodeptr = expand_workspace(cd);
6108              if (*errorcodeptr != 0) goto FAILED;              if (*errorcodeptr != 0) goto FAILED;
6109              save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;              save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
# Line 5545  for (;; ptr++) Line 6258  for (;; ptr++)
6258        goto FAILED;        goto FAILED;
6259        }        }
6260    
6261      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6262      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6263      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
6264      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6265      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.  
6266    
6267      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6268      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6269      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6270        tempcode, not at previous, which might be the first part of a string whose
6271      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. */  
6272    
6273      if (possessive_quantifier)      if (possessive_quantifier)
6274        {        {
6275        int len;        int len;
6276    
6277        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6278          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6279          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6280          remains is greater than zero, there's a further opcode that can be
6281          handled. If not, do nothing, leaving the EXACT alone. */
6282    
6283          switch(*tempcode)
6284            {
6285            case OP_TYPEEXACT:
6286          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6287            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6288            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6289            break;
6290    
6291        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6292          {  
6293            case OP_CHAR:
6294            case OP_CHARI:
6295            case OP_NOT:
6296            case OP_NOTI:
6297            case OP_EXACT:
6298            case OP_EXACTI:
6299            case OP_NOTEXACT:
6300            case OP_NOTEXACTI:
6301          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6302  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6303          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6304            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6305  #endif  #endif
6306            break;
6307    
6308            /* For the class opcodes, the repeat operator appears at the end;
6309            adjust tempcode to point to it. */
6310    
6311            case OP_CLASS:
6312            case OP_NCLASS:
6313            tempcode += 1 + 32/sizeof(pcre_uchar);
6314            break;
6315    
6316    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6317            case OP_XCLASS:
6318            tempcode += GET(tempcode, 1);
6319            break;
6320    #endif
6321          }          }
6322    
6323          /* If tempcode is equal to code (which points to the end of the repeated
6324          item), it means we have skipped an EXACT item but there is no following
6325          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6326          all other cases, tempcode will be pointing to the repeat opcode, and will
6327          be less than code, so the value of len will be greater than 0. */
6328    
6329        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6330          if (len > 0)
6331            {
6332            unsigned int repcode = *tempcode;
6333    
6334            /* There is a table for possessifying opcodes, all of which are less
6335            than OP_CALLOUT. A zero entry means there is no possessified version.
6336            */
6337    
6338            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6339              *tempcode = opcode_possessify[repcode];
6340    
6341            /* For opcode without a special possessified version, wrap the item in
6342            ONCE brackets. Because we are moving code along, we must ensure that any
6343            pending recursive references are updated. */
6344    
6345            else
6346              {
6347              *code = OP_END;
6348              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6349              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6350              code += 1 + LINK_SIZE;
6351              len += 1 + LINK_SIZE;
6352              tempcode[0] = OP_ONCE;
6353              *code++ = OP_KET;
6354              PUTINC(code, 0, len);
6355              PUT(tempcode, 1, len);
6356              }
6357            }
6358    
6359    #ifdef NEVER
6360        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6361          {          {
6362          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5609  for (;; ptr++) Line 6384  for (;; ptr++)
6384          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6385          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6386    
6387            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6388            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6389            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6390            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6391    
6392          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6393          pending recursive references are updated. */          pending recursive references are updated. */
6394    
# Line 5624  for (;; ptr++) Line 6404  for (;; ptr++)
6404          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6405          break;          break;
6406          }          }
6407    #endif
6408        }        }
6409    
6410      /* 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 6593  for (;; ptr++)
6593          tempptr = ptr;          tempptr = ptr;
6594    
6595          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6596          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',
6597          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6598            recursion tests.
6599          There are several syntaxes for testing a named group: (?(name)) is used  
6600          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6601            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6602          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6603          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
6604          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
6605          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.  
6606    
6607          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6608          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 6626  for (;; ptr++)
6626                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6627            break;            break;
6628    
6629          /* 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
6630          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. */
6631    
6632          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6633          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6634          refsign = -1;          refsign = -1;     /* => not a number */
6635            namelen = -1;     /* => not a name; must set to avoid warning */
6636            name = NULL;      /* Always set to avoid warning */
6637            recno = 0;        /* Always set to avoid warning */
6638    
6639          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6640    
6641          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6642            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6643            {            {
6644            terminator = -1;            terminator = -1;
6645            ptr += 2;            ptr += 2;
# Line 5863  for (;; ptr++) Line 6647  for (;; ptr++)
6647            }            }
6648    
6649          /* 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
6650          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6651            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6652    
6653          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6654            {            {
6655            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6656            ptr++;            ptr++;
6657            }            }
6658          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6659            {            {
6660            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6661            ptr++;            ptr++;
# Line 5878  for (;; ptr++) Line 6663  for (;; ptr++)
6663          else          else
6664            {            {
6665            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6666            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6667                else if (IS_DIGIT(*ptr)) refsign = 0;
6668            }            }
6669    
6670          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6671    
6672          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6673            {            {
6674            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6675            *errorcodeptr = ERR28;              {
6676            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6677                ptr++;
6678                }
6679            }            }
6680    
6681          /* 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
6682            a name is one of a number of duplicates, a different opcode is used and
6683            it needs more memory. Unfortunately we cannot tell whether a name is a
6684            duplicate in the first pass, so we have to allow for more memory. */
6685    
6686          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6687            {            {
6688            if (recno >= 0)            if (IS_DIGIT(*ptr))
6689              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6690            ptr++;              *errorcodeptr = ERR84;
6691                goto FAILED;
6692                }
6693              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6694                {
6695                *errorcodeptr = ERR28;   /* Assertion expected */
6696                goto FAILED;
6697                }
6698              name = ptr++;
6699              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6700                {
6701                ptr++;
6702                }
6703              namelen = (int)(ptr - name);
6704              if (lengthptr != NULL && (options & PCRE_DUPNAMES) != 0)
6705                *lengthptr += IMM2_SIZE;
6706            }            }
6707          namelen = (int)(ptr - name);  
6708            /* Check the terminator */
6709