/[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 1394 by ph10, Sat Nov 9 09:17:20 2013 UTC
# 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 257  static const int verbcount = sizeof(verb Line 264  static const int verbcount = sizeof(verb
264  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
265  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
266  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
267  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270  static const char posix_names[] =  static const char posix_names[] =
271    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 268  static const char posix_names[] = Line 276  static const char posix_names[] =
276  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    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 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* 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
285  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
286  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 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311  /* 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
312  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. */  
313    
314  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
315  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 322  static const pcre_uchar string_pXwd[] = Line 334  static const pcre_uchar string_pXwd[] =
334  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
335    string_PNd,           /* \D */    string_PNd,           /* \D */
336    string_pNd,           /* \d */    string_pNd,           /* \d */
337    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339    string_PXwd,          /* \W */    string_PXwd,          /* \W */
340    string_pXwd           /* \w */    string_pXwd           /* \w */
341  };  };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
350    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 375  static const pcre_uchar *posix_substitut Line 393  static const pcre_uchar *posix_substitut
393    NULL,                 /* graph */    NULL,                 /* graph */
394    NULL,                 /* print */    NULL,                 /* print */
395    NULL,                 /* punct */    NULL,                 /* punct */
396    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398    NULL,                 /* xdigit */    NULL,                 /* xdigit */
399    /* Negated cases */    /* Negated cases */
400    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 390  static const pcre_uchar *posix_substitut Line 408  static const pcre_uchar *posix_substitut
408    NULL,                 /* ^graph */    NULL,                 /* ^graph */
409    NULL,                 /* ^print */    NULL,                 /* ^print */
410    NULL,                 /* ^punct */    NULL,                 /* ^punct */
411    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
414  };  };
415  #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 473  static const char error_texts[] =
473    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
474    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
475    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 509  static const char error_texts[] = Line 527  static const char error_texts[] =
527    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
528    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
529    "setting UTF is disabled by the application\0"    "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      "group name must start with a non-digit\0"
537    ;    ;
538    
539  /* 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 673  static const pcre_uint8 ebcdic_chartab[]
673  #endif  #endif
674    
675    
676    /* This table is used to check whether auto-possessification is possible
677    between adjacent character-type opcodes. The left-hand (repeated) opcode is
678    used to select the row, and the right-hand opcode is use to select the column.
679    A value of 1 means that auto-possessification is OK. For example, the second
680    value in the first row means that \D+\d can be turned into \D++\d.
681    
682    The Unicode property types (\P and \p) have to be present to fill out the table
683    because of what their opcode values are, but the table values should always be
684    zero because property types are handled separately in the code. The last four
685    columns apply to items that cannot be repeated, so there is no need to have
686    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
687    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
688    
689    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
690    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
691    
692    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
693    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
694      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
695      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
696      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
697      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
698      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
699      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
704      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
705      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
706      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
707      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
708      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
709      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
710      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
711    };
712    
713    
714    /* This table is used to check whether auto-possessification is possible
715    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
716    left-hand (repeated) opcode is used to select the row, and the right-hand
717    opcode is used to select the column. The values are as follows:
718    
719      0   Always return FALSE (never auto-possessify)
720      1   Character groups are distinct (possessify if both are OP_PROP)
721      2   Check character categories in the same group (general or particular)
722      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
723    
724      4   Check left general category vs right particular category
725      5   Check right general category vs left particular category
726    
727      6   Left alphanum vs right general category
728      7   Left space vs right general category
729      8   Left word vs right general category
730    
731      9   Right alphanum vs left general category
732     10   Right space vs left general category
733     11   Right word vs left general category
734    
735     12   Left alphanum vs right particular category
736     13   Left space vs right particular category
737     14   Left word vs right particular category
738    
739     15   Right alphanum vs left particular category
740     16   Right space vs left particular category
741     17   Right word vs left particular category
742    */
743    
744    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
745    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
746      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
747      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
748      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
749      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
750      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
751      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
753      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
754      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
756      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
757    };
758    
759    /* This table is used to check whether auto-possessification is possible
760    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
761    specifies a general category and the other specifies a particular category. The
762    row is selected by the general category and the column by the particular
763    category. The value is 1 if the particular category is not part of the general
764    category. */
765    
766    static const pcre_uint8 catposstab[7][30] = {
767    /* 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772      { 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 */
773      { 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 */
774      { 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 */
775    };
776    
777    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
778    a general or particular category. The properties in each row are those
779    that apply to the character set in question. Duplication means that a little
780    unnecessary work is done when checking, but this keeps things much simpler
781    because they can all use the same code. For more details see the comment where
782    this table is used.
783    
784    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
785    "space", but from Perl 5.18 it's included, so both categories are treated the
786    same here. */
787    
788    static const pcre_uint8 posspropstab[3][4] = {
789      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
790      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
791      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
792    };
793    
794    /* This table is used when converting repeating opcodes into possessified
795    versions as a result of an explicit possessive quantifier such as ++. A zero
796    value means there is no possessified version - in those cases the item in
797    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
798    because all relevant opcodes are less than that. */
799    
800    static const pcre_uint8 opcode_possessify[] = {
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
802      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
803    
804      0,                       /* NOTI */
805      OP_POSSTAR, 0,           /* STAR, MINSTAR */
806      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
807      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
808      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
809      0,                       /* EXACT */
810      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_POSSTARI, 0,          /* STARI, MINSTARI */
813      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
814      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
815      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
816      0,                       /* EXACTI */
817      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
820      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
821      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
822      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
823      0,                       /* NOTEXACT */
824      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
827      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
828      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
829      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
830      0,                       /* NOTEXACTI */
831      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
832    
833      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
834      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
835      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
836      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
837      0,                       /* TYPEEXACT */
838      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
839    
840      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
841      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
842      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
843      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
844      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
845    
846      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
847      0, 0,                    /* REF, REFI */
848      0, 0,                    /* DNREF, DNREFI */
849      0, 0                     /* RECURSE, CALLOUT */
850    };
851    
852    
853    
854  /*************************************************  /*************************************************
855  *            Find an error text                  *  *            Find an error text                  *
# Line 675  return s; Line 877  return s;
877  }  }
878    
879    
880    
881  /*************************************************  /*************************************************
882  *           Expand the workspace                 *  *           Expand the workspace                 *
883  *************************************************/  *************************************************/
# Line 752  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 955  return (*p == CHAR_RIGHT_CURLY_BRACKET);
955  *************************************************/  *************************************************/
956    
957  /* 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
958  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
959  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.
960  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
961  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
962  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
963    
964  Arguments:  Arguments:
965    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
966    chptr          points to the data character    chptr          points to a returned data character
967    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
968    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
969    options        the options bits    options        the options bits
# Line 965  else Line 1167  else
1167      break;      break;
1168    
1169      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1170      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
1171      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1172        recommended to avoid the ambiguities in the old syntax.
1173    
1174      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
1175      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
1176      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
1177      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
1178      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
1179      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
1180      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1181    
1182        Inside a character class, \ followed by a digit is always either a literal
1183        8 or 9 or an octal number. */
1184    
1185      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:
1186      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 1207  else
1207          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1208          break;          break;
1209          }          }
1210        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1211          {          {
1212          escape = -s;          escape = -s;
1213          break;          break;
# Line 1009  else Line 1215  else
1215        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1216        }        }
1217    
1218      /* 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
1219      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
1220      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
1221        changed so as not to insert the binary zero. */
1222    
1223      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1224        {  
1225        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1226    
1227      /* \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
1228      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 1239  else
1239  #endif  #endif
1240      break;      break;
1241    
1242      /* \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
1243      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}. */
1244      If not, { is treated as a data character. */  
1245        case CHAR_o:
1246        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1247          {
1248          ptr += 2;
1249          c = 0;
1250          overflow = FALSE;
1251          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1252            {
1253            register pcre_uint32 cc = *ptr++;
1254            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1255    #ifdef COMPILE_PCRE32
1256            if (c >= 0x20000000l) { overflow = TRUE; break; }
1257    #endif
1258            c = (c << 3) + cc - CHAR_0 ;
1259    #if defined COMPILE_PCRE8
1260            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1261    #elif defined COMPILE_PCRE16
1262            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1263    #elif defined COMPILE_PCRE32
1264            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1265    #endif
1266            }
1267          if (overflow)
1268            {
1269            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1270            *errorcodeptr = ERR34;
1271            }
1272          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1273            {
1274            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1275            }
1276          else *errorcodeptr = ERR80;
1277          }
1278        break;
1279    
1280        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1281        numbers. Otherwise it is a lowercase x letter. */
1282    
1283      case CHAR_x:      case CHAR_x:
1284      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1285        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1286        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1287          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1288          {          {
# Line 1060  else Line 1299  else
1299  #endif  #endif
1300            }            }
1301          }          }
1302        break;        }    /* End JavaScript handling */
       }  
1303    
1304      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1305        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1306        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1307        seems to read hex digits up to the first non-such, and ignore the rest, so
1308        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1309        now gives an error. */
1310    
1311        c = 0;      else
1312        overflow = FALSE;        {
1313        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1314          {          {
1315          register pcre_uint32 cc = *pt++;          ptr += 2;
1316          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1317            overflow = FALSE;
1318            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1319              {
1320              register pcre_uint32 cc = *ptr++;
1321              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1322    
1323  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1324          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1325  #endif  #endif
1326    
1327  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1328          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1329          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1330  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1331          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 */
1332          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1333  #endif  #endif
1334    
1335  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1336          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1337  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1338          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1339  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1340          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1341  #endif  #endif
1342          }            }
1343    
1344        if (overflow)          if (overflow)
1345          {            {
1346          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1347          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1348          }            }
1349    
1350        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1351          {            {
1352          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1353          ptr = pt;            }
         break;  
         }  
1354    
1355        /* 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.
1356        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1357        }          \x handling, but nowadays Perl gives an error, which seems much more
1358            sensible, so we do too. */
1359    
1360      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1361            }   /* End of \x{} processing */
1362    
1363      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1364      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1365        {        else
1366        pcre_uint32 cc;                          /* Some compilers don't like */          {
1367        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1368            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1369              {
1370              pcre_uint32 cc;                          /* Some compilers don't like */
1371              cc = *(++ptr);                           /* ++ in initializers */
1372  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1373        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1374        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1375  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1376        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1377        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1378  #endif  #endif
1379        }            }
1380            }     /* End of \xdd handling */
1381          }       /* End of Perl-style \x handling */
1382      break;      break;
1383    
1384      /* 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 1444  if ((options & PCRE_UCP) != 0 && escape
1444  return escape;  return escape;
1445  }  }
1446    
1447    
1448    
1449  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1450  /*************************************************  /*************************************************
1451  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1289  return FALSE; Line 1543  return FALSE;
1543    
1544    
1545    
   
1546  /*************************************************  /*************************************************
1547  *         Read repeat counts                     *  *         Read repeat counts                     *
1548  *************************************************/  *************************************************/
# Line 1358  return p; Line 1611  return p;
1611    
1612    
1613  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1614  *      Find first significant op code            *  *      Find first significant op code            *
1615  *************************************************/  *************************************************/
1616    
# Line 1696  for (;;) Line 1649  for (;;)
1649    
1650      case OP_CALLOUT:      case OP_CALLOUT:
1651      case OP_CREF:      case OP_CREF:
1652      case OP_NCREF:      case OP_DNCREF:
1653      case OP_RREF:      case OP_RREF:
1654      case OP_NRREF:      case OP_DNRREF:
1655      case OP_DEF:      case OP_DEF:
1656      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1657      break;      break;
# Line 1712  for (;;) Line 1665  for (;;)
1665    
1666    
1667    
   
1668  /*************************************************  /*************************************************
1669  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1670  *************************************************/  *************************************************/
# Line 1836  for (;;) Line 1788  for (;;)
1788      case OP_COMMIT:      case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
1790      case OP_DEF:      case OP_DEF:
1791        case OP_DNCREF:
1792        case OP_DNRREF:
1793      case OP_DOLL:      case OP_DOLL:
1794      case OP_DOLLM:      case OP_DOLLM:
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799      case OP_PRUNE:      case OP_PRUNE:
1800      case OP_REVERSE:      case OP_REVERSE:
# Line 1937  for (;;) Line 1889  for (;;)
1889    
1890      switch (*cc)      switch (*cc)
1891        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1892        case OP_CRSTAR:        case OP_CRSTAR:
1893        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1894          case OP_CRPLUS:
1895          case OP_CRMINPLUS:
1896        case OP_CRQUERY:        case OP_CRQUERY:
1897        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1898          case OP_CRPOSSTAR:
1899          case OP_CRPOSPLUS:
1900          case OP_CRPOSQUERY:
1901        return -1;        return -1;
1902    
1903        case OP_CRRANGE:        case OP_CRRANGE:
1904        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1905          case OP_CRPOSRANGE:
1906        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1907        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1908        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2015  for (;;) Line 1971  for (;;)
1971      case OP_QUERYI:      case OP_QUERYI:
1972      case OP_REF:      case OP_REF:
1973      case OP_REFI:      case OP_REFI:
1974        case OP_DNREF:
1975        case OP_DNREFI:
1976      case OP_SBRA:      case OP_SBRA:
1977      case OP_SBRAPOS:      case OP_SBRAPOS:
1978      case OP_SCBRA:      case OP_SCBRA:
# Line 2051  for (;;) Line 2009  for (;;)
2009    
2010    
2011    
   
2012  /*************************************************  /*************************************************
2013  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2014  *************************************************/  *************************************************/
# Line 2361  Returns:      TRUE if what is matched co Line 2318  Returns:      TRUE if what is matched co
2318  typedef struct recurse_check {  typedef struct recurse_check {
2319    struct recurse_check *prev;    struct recurse_check *prev;
2320    const pcre_uchar *group;    const pcre_uchar *group;
2321  } recurse_check;  } recurse_check;
2322    
2323  static BOOL  static BOOL
2324  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 2334  for (code = first_significant_code(code
2334    const pcre_uchar *ccode;    const pcre_uchar *ccode;
2335    
2336    c = *code;    c = *code;
2337    
2338    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2339    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2340    
# Line 2405  for (code = first_significant_code(code Line 2362  for (code = first_significant_code(code
2362      NULL. */      NULL. */
2363    
2364      if (cd->start_workspace != NULL)      if (cd->start_workspace != NULL)
2365        {        {
2366        const pcre_uchar *tcode;        const pcre_uchar *tcode;
2367        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2368          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;
2369        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2370        }        }
2371    
2372      /* If we are scanning a completed pattern, there are no forward references      /* If we are scanning a completed pattern, there are no forward references
2373      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
2374      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,
2375      just skip over it. Simple recursions are easily detected. For mutual      just skip over it. Simple recursions are easily detected. For mutual
2376      recursions we keep a chain on the stack. */      recursions we keep a chain on the stack. */
2377    
2378      else      else
2379        {        {
2380        recurse_check *r = recurses;        recurse_check *r = recurses;
2381        const pcre_uchar *endgroup = scode;        const pcre_uchar *endgroup = scode;
2382    
2383        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2384        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2385    
2386        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2387          if (r->group == scode) break;          if (r->group == scode) break;
2388        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2436  for (code = first_significant_code(code Line 2393  for (code = first_significant_code(code
2393    
2394      empty_branch = FALSE;      empty_branch = FALSE;
2395      this_recurse.prev = recurses;      this_recurse.prev = recurses;
2396      this_recurse.group = scode;      this_recurse.group = scode;
2397    
2398      do      do
2399        {        {
2400        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 2495  for (code = first_significant_code(code
2495        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2496        case OP_CRQUERY:        case OP_CRQUERY:
2497        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2498          case OP_CRPOSSTAR:
2499          case OP_CRPOSQUERY:
2500        break;        break;
2501    
2502        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2503        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2504        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2505          case OP_CRPOSPLUS:
2506        return FALSE;        return FALSE;
2507    
2508        case OP_CRRANGE:        case OP_CRRANGE:
2509        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2510          case OP_CRPOSRANGE:
2511        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2512        break;        break;
2513        }        }
# Line 2557  for (code = first_significant_code(code Line 2518  for (code = first_significant_code(code
2518      case OP_ANY:      case OP_ANY:
2519      case OP_ALLANY:      case OP_ALLANY:
2520      case OP_ANYBYTE:      case OP_ANYBYTE:
2521    
2522      case OP_PROP:      case OP_PROP:
2523      case OP_NOTPROP:      case OP_NOTPROP:
2524      case OP_ANYNL:      case OP_ANYNL:
2525    
2526      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2527      case OP_HSPACE:      case OP_HSPACE:
2528      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2529      case OP_VSPACE:      case OP_VSPACE:
2530      case OP_EXTUNI:      case OP_EXTUNI:
2531    
2532      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2533      case OP_DIGIT:      case OP_DIGIT:
2534      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2535      case OP_WHITESPACE:      case OP_WHITESPACE:
2536      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2537      case OP_WORDCHAR:      case OP_WORDCHAR:
2538    
2539      case OP_CHAR:      case OP_CHAR:
2540      case OP_CHARI:      case OP_CHARI:
2541      case OP_NOT:      case OP_NOT:
2542      case OP_NOTI:      case OP_NOTI:
2543    
2544      case OP_PLUS:      case OP_PLUS:
2545      case OP_PLUSI:      case OP_PLUSI:
2546      case OP_MINPLUS:      case OP_MINPLUS:
# Line 2589  for (code = first_significant_code(code Line 2550  for (code = first_significant_code(code
2550      case OP_NOTPLUSI:      case OP_NOTPLUSI:
2551      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2552      case OP_NOTMINPLUSI:      case OP_NOTMINPLUSI:
2553    
2554      case OP_POSPLUS:      case OP_POSPLUS:
2555      case OP_POSPLUSI:      case OP_POSPLUSI:
2556      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2557      case OP_NOTPOSPLUSI:      case OP_NOTPOSPLUSI:
2558    
2559      case OP_EXACT:      case OP_EXACT:
2560      case OP_EXACTI:      case OP_EXACTI:
2561      case OP_NOTEXACT:      case OP_NOTEXACT:
2562      case OP_NOTEXACTI:      case OP_NOTEXACTI:
2563    
2564      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2565      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2566      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2567      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2568    
2569      return FALSE;      return FALSE;
2570    
2571      /* 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 2605  for (code = first_significant_code(code
2605  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2606      case OP_STAR:      case OP_STAR:
2607      case OP_STARI:      case OP_STARI:
2608      case OP_NOTSTAR:      case OP_NOTSTAR:
2609      case OP_NOTSTARI:      case OP_NOTSTARI:
2610    
2611      case OP_MINSTAR:      case OP_MINSTAR:
2612      case OP_MINSTARI:      case OP_MINSTARI:
2613      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
2614      case OP_NOTMINSTARI:      case OP_NOTMINSTARI:
2615    
2616      case OP_POSSTAR:      case OP_POSSTAR:
2617      case OP_POSSTARI:      case OP_POSSTARI:
2618      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
2619      case OP_NOTPOSSTARI:      case OP_NOTPOSSTARI:
2620    
2621      case OP_QUERY:      case OP_QUERY:
2622      case OP_QUERYI:      case OP_QUERYI:
2623      case OP_NOTQUERY:      case OP_NOTQUERY:
2624      case OP_NOTQUERYI:      case OP_NOTQUERYI:
2625    
2626      case OP_MINQUERY:      case OP_MINQUERY:
2627      case OP_MINQUERYI:      case OP_MINQUERYI:
2628      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
2629      case OP_NOTMINQUERYI:      case OP_NOTMINQUERYI:
2630    
2631      case OP_POSQUERY:      case OP_POSQUERY:
2632      case OP_POSQUERYI:      case OP_POSQUERYI:
2633      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
2634      case OP_NOTPOSQUERYI:      case OP_NOTPOSQUERYI:
2635    
2636      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2637      break;      break;
2638    
2639      case OP_UPTO:      case OP_UPTO:
2640      case OP_UPTOI:      case OP_UPTOI:
2641      case OP_NOTUPTO:      case OP_NOTUPTO:
2642      case OP_NOTUPTOI:      case OP_NOTUPTOI:
2643    
2644      case OP_MINUPTO:      case OP_MINUPTO:
2645      case OP_MINUPTOI:      case OP_MINUPTOI:
2646      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
2647      case OP_NOTMINUPTOI:      case OP_NOTMINUPTOI:
2648    
2649      case OP_POSUPTO:      case OP_POSUPTO:
2650      case OP_POSUPTOI:      case OP_POSUPTOI:
2651      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
2652      case OP_NOTPOSUPTOI:      case OP_NOTPOSUPTOI:
2653    
2654      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]);
2655      break;      break;
2656  #endif  #endif
# Line 2753  return TRUE; Line 2714  return TRUE;
2714    
2715    
2716  /*************************************************  /*************************************************
2717  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2718  *************************************************/  *************************************************/
2719    
2720  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2721  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  
2722    
2723  Returns:   TRUE or FALSE  Arguments:  c opcode
2724    Returns:    base opcode for the type
2725  */  */
2726    
2727  static BOOL  static pcre_uchar
2728  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2729  {  {
2730  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2731  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2732  for (++ptr; *ptr != CHAR_NULL; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2733    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2734    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2735      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;  
2736  }  }
2737    
2738    
2739    
2740    #ifdef SUPPORT_UCP
2741  /*************************************************  /*************************************************
2742  *          Check POSIX class name                *  *        Check a character and a property        *
2743  *************************************************/  *************************************************/
2744    
2745  /* 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
2746  such as [:alnum:].  is adjacent to a fixed character.
2747    
2748  Arguments:  Arguments:
2749    ptr        points to the first letter    c            the character
2750    len        the length of the name    ptype        the property type
2751      pdata        the data for the type
2752      negated      TRUE if it's a negated property (\P or \p{^)
2753    
2754  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2755  */  */
2756    
2757  static int  static BOOL
2758  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2759      BOOL negated)
2760  {  {
2761  const char *pn = posix_names;  const pcre_uint32 *p;
2762  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   *  
 *************************************************/  
2763    
2764  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2765  that is referenced. This means that groups can be replicated for fixed    {
2766  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2767  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2768  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2769  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.  
2770    
2771  This function has been extended with the possibility of forward references for    case PT_GC:
2772  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).  
2773    
2774  Arguments:    case PT_PC:
2775    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  
2776    
2777  Returns:     nothing    case PT_SC:
2778  */    return (pdata == prop->script) == negated;
2779    
2780  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;  
2781    
2782  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2783    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2784    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2785    
2786    /* 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
2787    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2788      at release 8.34. */
2789    
2790    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2791      case PT_PXSPACE:  /* POSIX space */
2792      switch(c)
2793      {      {
2794      offset = (int)GET(hc, 0);      HSPACE_CASES:
2795      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2796        {      return negated;
2797        PUT(hc, 0, offset + adjust);  
2798        break;      default:
2799        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2800      }      }
2801      break;  /* Control never reaches here */
2802    
2803    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2804    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2805              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2806              c == CHAR_UNDERSCORE) == negated;
2807    
2808    if (hc >= cd->hwm)    case PT_CLIST:
2809      p = PRIV(ucd_caseless_sets) + prop->caseset;
2810      for (;;)
2811      {      {
2812      offset = (int)GET(ptr, 1);      if (c < *p) return !negated;
2813      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2814      }      }
2815      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2816    }    }
2817    
2818    return FALSE;
2819  }  }
2820    #endif  /* SUPPORT_UCP */
2821    
2822    
2823    
2824  /*************************************************  /*************************************************
2825  *        Insert an automatic callout point       *  *        Fill the character property list        *
2826  *************************************************/  *************************************************/
2827    
2828  /* 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-
2829  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2830    
2831  Arguments:  Arguments:
2832    code           current code pointer    code        points to start of expression
2833    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2834    cd             pointers to tables etc    fcc         points to case-flipping table
2835      list        points to output list
2836                  list[0] will be filled with the opcode
2837                  list[1] will be non-zero if this opcode
2838                    can match an empty character string
2839                  list[2..7] depends on the opcode
2840    
2841  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2842                  NULL if *code is not accepted
2843  */  */
2844    
2845  static pcre_uchar *  static const pcre_uchar *
2846  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2847      const pcre_uint8 *fcc, pcre_uint32 *list)
2848  {  {
2849  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2850  *code++ = 255;  const pcre_uchar *end;
2851  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2852  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2853  return code + 2 * LINK_SIZE;  pcre_uint32 chr;
2854  }  pcre_uchar base;
   
2855    
2856    list[0] = c;
2857    list[1] = FALSE;
2858    code++;
2859    
2860  /*************************************************  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2861  *         Complete a callout item                *    {
2862  *************************************************/    base = get_repeat_base(c);
2863      c -= (base - OP_STAR);
2864    
2865  /* 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)
2866  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.  
2867    
2868  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  
2869    
2870  Returns:             nothing    switch(base)
2871  */      {
2872        case OP_STAR:
2873        list[0] = OP_CHAR;
2874        break;
2875    
2876  static void      case OP_STARI:
2877  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_CHARI;
2878  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2879    
2880        case OP_NOTSTAR:
2881        list[0] = OP_NOT;
2882        break;
2883    
2884        case OP_NOTSTARI:
2885        list[0] = OP_NOTI;
2886        break;
2887    
2888  #ifdef SUPPORT_UCP      case OP_TYPESTAR:
2889  /*************************************************      list[0] = *code;
2890  *           Get othercase range                  *      code++;
2891  *************************************************/      break;
2892        }
2893      c = list[0];
2894      }
2895    
2896  /* This function is passed the start and end of a class range, in UTF-8 mode  switch(c)
 with UCP support. It searches up the characters, looking for ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
   
 Arguments:  
   cptr        points to starting character value; updated  
   d           end value  
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
   
 Yield:        -1 when no more  
                0 when a range is returned  
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
 */  
   
 static int  
 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 unsigned int co;  
   
 /* Find the first character that has an other case. If it has multiple other  
 cases, return its case offset value. */  
   
 for (c = *cptr; c <= d; c++)  
   {  
   if ((co = UCD_CASESET(c)) != 0)  
     {  
     *ocptr = c++;   /* Character that has the set */  
     *cptr = c;      /* Rest of input range */  
     return (int)co;  
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
   
 if (c > d) return -1;  /* Reached end of range */  
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
2897    {    {
2898    if (UCD_OTHERCASE(c) != next) break;    case OP_NOT_DIGIT:
2899    next++;    case OP_DIGIT:
2900    }    case OP_NOT_WHITESPACE:
2901      case OP_WHITESPACE:
2902  *odptr = next - 1;     /* End of othercase range */    case OP_NOT_WORDCHAR:
2903  *cptr = c;             /* Rest of input range */    case OP_WORDCHAR:
2904  return 0;    case OP_ANY:
2905  }    case OP_ALLANY:
2906      case OP_ANYNL:
2907      case OP_NOT_HSPACE:
2908      case OP_HSPACE:
2909  /*************************************************    case OP_NOT_VSPACE:
2910  *        Check a character and a property        *    case OP_VSPACE:
2911  *************************************************/    case OP_EXTUNI:
2912      case OP_EODN:
2913  /* This function is called by check_auto_possessive() when a property item    case OP_EOD:
2914  is adjacent to a fixed character.    case OP_DOLL:
2915      case OP_DOLLM:
2916      return code;
2917    
2918  Arguments:    case OP_CHAR:
2919    c            the character    case OP_NOT:
2920    ptype        the property type    GETCHARINCTEST(chr, code);
2921    pdata        the data for the type    list[2] = chr;
2922    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2923      return code;
2924    
2925  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2926  */    case OP_NOTI:
2927      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2928      GETCHARINCTEST(chr, code);
2929      list[2] = chr;
2930    
 static BOOL  
 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  
 {  
2931  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2932  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2933        list[3] = fcc[chr];
2934      else
2935        list[3] = UCD_OTHERCASE(chr);
2936    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2937      list[3] = (chr < 256) ? fcc[chr] : chr;
2938    #else
2939      list[3] = fcc[chr];
2940  #endif  #endif
2941    
2942  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;  
2943    
2944    case PT_GC:    if (chr == list[3])
2945    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;      list[3] = NOTACHAR;
2946      else
2947        list[4] = NOTACHAR;
2948      return code;
2949    
2950    case PT_PC:  #ifdef SUPPORT_UCP
2951    return (pdata == prop->chartype) == negated;    case OP_PROP:
2952      case OP_NOTPROP:
2953      if (code[0] != PT_CLIST)
2954        {
2955        list[2] = code[0];
2956        list[3] = code[1];
2957        return code + 2;
2958        }
2959    
2960    case PT_SC:    /* Convert only if we have enough space. */
   return (pdata == prop->script) == negated;  
2961    
2962    /* These are specials */    clist_src = PRIV(ucd_caseless_sets) + code[1];
2963      clist_dest = list + 2;
2964      code += 2;
2965    
2966    case PT_ALNUM:    do {
2967    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||       if (clist_dest >= list + 8)
2968            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;         {
2969           /* Early return if there is not enough space. This should never
2970           happen, since all clists are shorter than 5 character now. */
2971           list[2] = code[0];
2972           list[3] = code[1];
2973           return code;
2974           }
2975         *clist_dest++ = *clist_src;
2976         }
2977      while(*clist_src++ != NOTACHAR);
2978    
2979    case PT_SPACE:    /* Perl space */    /* All characters are stored. The terminating NOTACHAR
2980    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;  
2981    
2982    case PT_PXSPACE:  /* POSIX space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2983    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
2984            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #endif
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
2985    
2986    case PT_WORD:    case OP_NCLASS:
2987    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    case OP_CLASS:
2988            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2989            c == CHAR_UNDERSCORE) == negated;    case OP_XCLASS:
2990      if (c == OP_XCLASS)
2991        end = code + GET(code, 0) - 1;
2992      else
2993    #endif
2994        end = code + 32 / sizeof(pcre_uchar);
2995    
2996  #ifdef SUPPORT_UCP    switch(*end)
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
2997      {      {
2998      if (c < *p) return !negated;      case OP_CRSTAR:
2999      if (c == *p++) return negated;      case OP_CRMINSTAR:
3000        case OP_CRQUERY:
3001        case OP_CRMINQUERY:
3002        case OP_CRPOSSTAR:
3003        case OP_CRPOSQUERY:
3004        list[1] = TRUE;
3005        end++;
3006        break;
3007    
3008        case OP_CRPLUS:
3009        case OP_CRMINPLUS:
3010        case OP_CRPOSPLUS:
3011        end++;
3012        break;
3013    
3014        case OP_CRRANGE:
3015        case OP_CRMINRANGE:
3016        case OP_CRPOSRANGE:
3017        list[1] = (GET2(end, 1) == 0);
3018        end += 1 + 2 * IMM2_SIZE;
3019        break;
3020      }      }
3021    break;  /* Control never reaches here */    list[2] = end - code;
3022  #endif    return end;
3023    }    }
3024    return NULL;    /* Opcode not accepted */
 return FALSE;  
3025  }  }
 #endif  /* SUPPORT_UCP */  
3026    
3027    
3028    
3029  /*************************************************  /*************************************************
3030  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3031  *************************************************/  *************************************************/
3032    
3033  /* 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
3034  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.  
3035    
3036  Arguments:  Arguments:
3037    previous      pointer to the repeated opcode    code        points to the byte code
3038    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3039    ptr           next character in pattern    cd          static compile data
3040    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3041    
3042  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3043  */  */
3044    
3045  static BOOL  static BOOL
3046  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3047    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3048  {  {
3049  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3050  pcre_uint32 next;  pcre_uint32 list[8];
3051  int escape;  const pcre_uint32 *chr_ptr;
3052  pcre_uchar op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3053    const pcre_uint32 *list_ptr;
3054  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3055    const pcre_uint8 *class_bitset;
3056    const pcre_uint32 *set1, *set2, *set_end;
3057    pcre_uint32 chr;
3058    BOOL accepted, invert_bits;
3059    
3060    /* Note: the base_list[1] contains whether the current opcode has greedy
3061    (represented by a non-zero value) quantifier. This is a different from
3062    other character type lists, which stores here that the character iterator
3063    matches to an empty string (also represented by a non-zero value). */
3064    
3065  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3066    {    {
3067    for (;;)    /* All operations move the code pointer forward.
3068      Therefore infinite recursions are not possible. */
3069    
3070      c = *code;
3071    
3072      /* Skip over callouts */
3073    
3074      if (c == OP_CALLOUT)
3075      {      {
3076      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3077      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;  
3078      }      }
   }  
3079    
3080  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3081  value is a character, a negative value is an escape value. */      {
3082        do code += GET(code, 1); while (*code == OP_ALT);
3083        c = *code;
3084        }
3085    
3086      switch(c)
3087        {
3088        case OP_END:
3089        case OP_KETRPOS:
3090        /* TRUE only in greedy case. The non-greedy case could be replaced by
3091        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3092        uses more memory, which we cannot get at this stage.) */
3093    
3094        return base_list[1] != 0;
3095    
3096  if (*ptr == CHAR_BACKSLASH)      case OP_KET:
3097    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3098    int temperrorcode = 0;      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3099    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,      cannot be converted to a possessive form. */
3100      FALSE);  
3101    if (temperrorcode != 0) return FALSE;      if (base_list[1] == 0) return FALSE;
3102    ptr++;    /* Point after the escape sequence */  
3103    }      switch(*(code - GET(code, 1)))
3104  else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)        {
3105    {        case OP_ASSERT:
3106    escape = 0;        case OP_ASSERT_NOT:
3107  #ifdef SUPPORT_UTF        case OP_ASSERTBACK:
3108    if (utf) { GETCHARINC(next, ptr); } else        case OP_ASSERTBACK_NOT:
3109  #endif        case OP_ONCE:
3110    next = *ptr++;        case OP_ONCE_NC:
3111    }        /* Atomic sub-patterns and assertions can always auto-possessify their
3112  else return FALSE;        last iterator. */
3113          return TRUE;
3114          }
3115    
3116  /* Skip whitespace and comments in extended mode */      code += PRIV(OP_lengths)[c];
3117        continue;
3118    
3119  if ((options & PCRE_EXTENDED) != 0)      case OP_ONCE:
3120    {      case OP_ONCE_NC:
3121    for (;;)      case OP_BRA:
3122      {      case OP_CBRA:
3123      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      next_code = code + GET(code, 1);
3124      if (*ptr == CHAR_NUMBER_SIGN)      code += PRIV(OP_lengths)[c];
3125    
3126        while (*next_code == OP_ALT)
3127        {        {
3128        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3129        while (*ptr != CHAR_NULL)        code = next_code + 1 + LINK_SIZE;
3130          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3131        }        }
3132      else break;      continue;
     }  
   }  
3133    
3134  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3135        case OP_BRAMINZERO:
3136    
3137  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3138    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3139      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3140    
3141        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3142    
3143        /* The bracket content will be checked by the
3144        OP_BRA/OP_CBRA case above. */
3145        next_code += 1 + LINK_SIZE;
3146        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3147          return FALSE;
3148    
3149  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3150        continue;
3151        }
3152    
3153  if (op_code == OP_CHAR || op_code == OP_CHARI ||    /* Check for a supported opcode, and load its properties. */
     op_code == OP_NOT || op_code == OP_NOTI)  
   {  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3154    
3155  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3156  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3157    
3158  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3159    {    characters from that list with another list, or with a property. */
   /* For a caseless UTF match, the next character may have more than one other  
   case, which maps to the special PT_CLIST property. Check this first. */  
3160    
3161  #ifdef SUPPORT_UCP    if (base_list[0] == OP_CHAR)
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3162      {      {
3163      unsigned int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3164      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3165      }      }
3166  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3167      {      {
3168      case OP_CHAR:      chr_ptr = list + 2;
3169      return c != next;      list_ptr = base_list;
3170        }
3171    
3172      /* 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. */  
3173    
3174      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3175      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3176  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3177      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3178        {  #endif
3179        pcre_uint32 othercase;        )
3180        if (next < 128) othercase = cd->fcc[next]; else      {
3181  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3182        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3183  #else  #else
3184        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3185  #endif  #endif
3186        return c != othercase;        {
3187          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3188          list_ptr = list;
3189        }        }
3190      else      else
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
   
     case OP_NOT:  
     return c == next;  
   
     case OP_NOTI:  
     if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
3191        {        {
3192        pcre_uint32 othercase;        set1 = (pcre_uint32 *)(code - list[2]);
3193        if (next < 128) othercase = cd->fcc[next]; else        list_ptr = base_list;
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
3194        }        }
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3195    
3196      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      invert_bits = FALSE;
3197      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      switch(list_ptr[0])
3198          {
3199          case OP_CLASS:
3200          case OP_NCLASS:
3201          set2 = (pcre_uint32 *)
3202            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3203          break;
3204    
3205      case OP_DIGIT:        /* OP_XCLASS cannot be supported here, because its bitset
3206      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        is not necessarily complete. E.g: [a-\0x{200}] is stored
3207          as a character range, and the appropriate bits are not set. */
3208    
3209      case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
3210      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;          invert_bits = TRUE;
3211            /* Fall through */
3212          case OP_DIGIT:
3213            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3214            break;
3215    
3216      case OP_WHITESPACE:        case OP_NOT_WHITESPACE:
3217      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;          invert_bits = TRUE;
3218            /* Fall through */
3219          case OP_WHITESPACE:
3220            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3221            break;
3222    
3223      case OP_NOT_WHITESPACE:        case OP_NOT_WORDCHAR:
3224      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;          invert_bits = TRUE;
3225            /* Fall through */
3226          case OP_WORDCHAR:
3227            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3228            break;
3229    
3230      case OP_WORDCHAR:        default:
3231      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        return FALSE;
3232          }
3233    
3234      case OP_NOT_WORDCHAR:      /* Compare 4 bytes to improve speed. */
3235      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;      set_end = set1 + (32 / 4);
3236        if (invert_bits)
3237          {
3238          do
3239            {
3240            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3241            }
3242          while (set1 < set_end);
3243          }
3244        else
3245          {
3246          do
3247            {
3248            if ((*set1++ & *set2++) != 0) return FALSE;
3249            }
3250          while (set1 < set_end);
3251          }
3252    
3253      case OP_HSPACE:      if (list[1] == 0) return TRUE;
3254      case OP_NOT_HSPACE:      /* Might be an empty repeat. */
3255      switch(next)      continue;
3256        }
3257    
3258      /* Some property combinations also acceptable. Unicode property opcodes are
3259      processed specially; the rest can be handled with a lookup table. */
3260    
3261      else
3262        {
3263        pcre_uint32 leftop, rightop;
3264    
3265        leftop = base_list[0];
3266        rightop = list[0];
3267    
3268    #ifdef SUPPORT_UCP
3269        accepted = FALSE; /* Always set in non-unicode case. */
3270        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3271        {        {
3272        HSPACE_CASES:        if (rightop == OP_EOD)
3273        return op_code == OP_NOT_HSPACE;          accepted = TRUE;
3274          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3275            {
3276            int n;
3277            const pcre_uint8 *p;
3278            BOOL same = leftop == rightop;
3279            BOOL lisprop = leftop == OP_PROP;
3280            BOOL risprop = rightop == OP_PROP;
3281            BOOL bothprop = lisprop && risprop;
3282    
3283            /* There's a table that specifies how each combination is to be
3284            processed:
3285              0   Always return FALSE (never auto-possessify)
3286              1   Character groups are distinct (possessify if both are OP_PROP)
3287              2   Check character categories in the same group (general or particular)
3288              3   Return TRUE if the two opcodes are not the same
3289              ... see comments below
3290            */
3291    
3292            n = propposstab[base_list[2]][list[2]];
3293            switch(n)
3294              {
3295              case 0: break;
3296              case 1: accepted = bothprop; break;
3297              case 2: accepted = (base_list[3] == list[3]) != same; break;
3298              case 3: accepted = !same; break;
3299    
3300        default:            case 4:  /* Left general category, right particular category */
3301        return op_code != OP_NOT_HSPACE;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3302              break;
3303    
3304              case 5:  /* Right general category, left particular category */
3305              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3306              break;
3307    
3308              /* This code is logically tricky. Think hard before fiddling with it.
3309              The posspropstab table has four entries per row. Each row relates to
3310              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3311              Only WORD actually needs all four entries, but using repeats for the
3312              others means they can all use the same code below.
3313    
3314              The first two entries in each row are Unicode general categories, and
3315              apply always, because all the characters they include are part of the
3316              PCRE character set. The third and fourth entries are a general and a
3317              particular category, respectively, that include one or more relevant
3318              characters. One or the other is used, depending on whether the check
3319              is for a general or a particular category. However, in both cases the
3320              category contains more characters than the specials that are defined
3321              for the property being tested against. Therefore, it cannot be used
3322              in a NOTPROP case.
3323    
3324              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3325              Underscore is covered by ucp_P or ucp_Po. */
3326    
3327              case 6:  /* Left alphanum vs right general category */
3328              case 7:  /* Left space vs right general category */
3329              case 8:  /* Left word vs right general category */
3330              p = posspropstab[n-6];
3331              accepted = risprop && lisprop ==
3332                (list[3] != p[0] &&
3333                 list[3] != p[1] &&
3334                (list[3] != p[2] || !lisprop));
3335              break;
3336    
3337              case 9:   /* Right alphanum vs left general category */
3338              case 10:  /* Right space vs left general category */
3339              case 11:  /* Right word vs left general category */
3340              p = posspropstab[n-9];
3341              accepted = lisprop && risprop ==
3342                (base_list[3] != p[0] &&
3343                 base_list[3] != p[1] &&
3344                (base_list[3] != p[2] || !risprop));
3345              break;
3346    
3347              case 12:  /* Left alphanum vs right particular category */
3348              case 13:  /* Left space vs right particular category */
3349              case 14:  /* Left word vs right particular category */
3350              p = posspropstab[n-12];
3351              accepted = risprop && lisprop ==
3352                (catposstab[p[0]][list[3]] &&
3353                 catposstab[p[1]][list[3]] &&
3354                (list[3] != p[3] || !lisprop));
3355              break;
3356    
3357              case 15:  /* Right alphanum vs left particular category */
3358              case 16:  /* Right space vs left particular category */
3359              case 17:  /* Right word vs left particular category */
3360              p = posspropstab[n-15];
3361              accepted = lisprop && risprop ==
3362                (catposstab[p[0]][base_list[3]] &&
3363                 catposstab[p[1]][base_list[3]] &&
3364                (base_list[3] != p[3] || !risprop));
3365              break;
3366              }
3367            }
3368        }        }
3369    
3370      case OP_ANYNL:      else
3371      case OP_VSPACE:  #endif  /* SUPPORT_UCP */
3372      case OP_NOT_VSPACE:  
3373      switch(next)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3374               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3375               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3376    
3377        if (!accepted)
3378          return FALSE;
3379    
3380        if (list[1] == 0) return TRUE;
3381        /* Might be an empty repeat. */
3382        continue;
3383        }
3384    
3385      /* Control reaches here only if one of the items is a small character list.
3386      All characters are checked against the other side. */
3387    
3388      do
3389        {
3390        chr = *chr_ptr;
3391    
3392        switch(list_ptr[0])
3393        {        {
3394        VSPACE_CASES:        case OP_CHAR:
3395        return op_code == OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3396          do
3397            {
3398            if (chr == *ochr_ptr) return FALSE;
3399            ochr_ptr++;
3400            }
3401          while(*ochr_ptr != NOTACHAR);
3402          break;
3403    
3404        default:        case OP_NOT:
3405        return op_code != OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3406        }        do
3407            {
3408            if (chr == *ochr_ptr)
3409              break;
3410            ochr_ptr++;
3411            }
3412          while(*ochr_ptr != NOTACHAR);
3413          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3414          break;
3415    
3416          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3417          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3418    
3419          case OP_DIGIT:
3420          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3421          break;
3422    
3423          case OP_NOT_DIGIT:
3424          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3425          break;
3426    
3427          case OP_WHITESPACE:
3428          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3429          break;
3430    
3431          case OP_NOT_WHITESPACE:
3432          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3433          break;
3434    
3435          case OP_WORDCHAR:
3436          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3437          break;
3438    
3439          case OP_NOT_WORDCHAR:
3440          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3441          break;
3442    
3443          case OP_HSPACE:
3444          switch(chr)
3445            {
3446            HSPACE_CASES: return FALSE;
3447            default: break;
3448            }
3449          break;
3450    
3451          case OP_NOT_HSPACE:
3452          switch(chr)
3453            {
3454            HSPACE_CASES: break;
3455            default: return FALSE;
3456            }
3457          break;
3458    
3459          case OP_ANYNL:
3460          case OP_VSPACE:
3461          switch(chr)
3462            {
3463            VSPACE_CASES: return FALSE;
3464            default: break;
3465            }
3466          break;
3467    
3468          case OP_NOT_VSPACE:
3469          switch(chr)
3470            {
3471            VSPACE_CASES: break;
3472            default: return FALSE;
3473            }
3474          break;
3475    
3476          case OP_DOLL:
3477          case OP_EODN:
3478          switch (chr)
3479            {
3480            case CHAR_CR:
3481            case CHAR_LF:
3482            case CHAR_VT:
3483            case CHAR_FF:
3484            case CHAR_NEL:
3485    #ifndef EBCDIC
3486            case 0x2028:
3487            case 0x2029:
3488    #endif  /* Not EBCDIC */
3489            return FALSE;
3490            }
3491          break;
3492    
3493          case OP_EOD:    /* Can always possessify before \z */
3494          break;
3495    
3496  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3497      case OP_PROP:        case OP_PROP:
3498      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3499          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3500                list_ptr[0] == OP_NOTPROP))
3501            return FALSE;
3502          break;
3503    #endif
3504    
3505      case OP_NOTPROP:        case OP_NCLASS:
3506      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3507          /* Fall through */
3508    
3509          case OP_CLASS:
3510          if (chr > 255) break;
3511          class_bitset = (pcre_uint8 *)
3512            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3513          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3514          break;
3515    
3516    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3517          case OP_XCLASS:
3518          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3519              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3520          break;
3521  #endif  #endif
3522    
3523      default:        default:
3524      return FALSE;        return FALSE;
3525          }
3526    
3527        chr_ptr++;
3528      }      }
3529      while(*chr_ptr != NOTACHAR);
3530    
3531      /* At least one character must be matched from this opcode. */
3532    
3533      if (list[1] == 0) return TRUE;
3534    }    }
3535    
3536  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  return FALSE;
3537  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  }
3538  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. */  
3539    
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3540    
3541      case ESC_D:  /*************************************************
3542      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  *    Scan compiled regex for auto-possession     *
3543    *************************************************/
3544    
3545    /* Replaces single character iterations with their possessive alternatives
3546    if appropriate. This function modifies the compiled opcode!
3547    
3548      case ESC_s:  Arguments:
3549      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;    code        points to start of the byte code
3550      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3551      cd          static compile data
3552    
3553    Returns:      nothing
3554    */
3555    
3556      case ESC_S:  static void
3557      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3558    {
3559    register pcre_uchar c;
3560    const pcre_uchar *end;
3561    pcre_uchar *repeat_opcode;
3562    pcre_uint32 list[8];
3563    
3564      case ESC_w:  for (;;)
3565      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;    {
3566      c = *code;
3567    
3568      case ESC_W:    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3569      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;      {
3570        c -= get_repeat_base(c) - OP_STAR;
3571        end = (c <= OP_MINUPTO) ?
3572          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3573        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3574    
3575      case ESC_h:      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
     case ESC_H:  
     switch(c)  
3576        {        {
3577        HSPACE_CASES:        switch(c)
3578        return escape != ESC_h;          {
3579            case OP_STAR:
3580            *code += OP_POSSTAR - OP_STAR;
3581            break;
3582    
3583        default:          case OP_MINSTAR:
3584        return escape == ESC_h;          *code += OP_POSSTAR - OP_MINSTAR;
3585        }          break;
3586    
3587      case ESC_v:          case OP_PLUS:
3588      case ESC_V:          *code += OP_POSPLUS - OP_PLUS;
3589      switch(c)          break;
       {  
       VSPACE_CASES:  
       return escape != ESC_v;  
3590    
3591        default:          case OP_MINPLUS:
3592        return escape == ESC_v;          *code += OP_POSPLUS - OP_MINPLUS;
3593        }          break;
3594    
3595            case OP_QUERY:
3596            *code += OP_POSQUERY - OP_QUERY;
3597            break;
3598    
3599      /* When PCRE_UCP is set, these values get generated for \d etc. Find          case OP_MINQUERY:
3600      their substitutions and process them. The result will always be either          *code += OP_POSQUERY - OP_MINQUERY;
3601      ESC_p or ESC_P. Then fall through to process those values. */          break;
3602    
3603  #ifdef SUPPORT_UCP          case OP_UPTO:
3604      case ESC_du:          *code += OP_POSUPTO - OP_UPTO;
3605      case ESC_DU:          break;
3606      case ESC_wu:  
3607      case ESC_WU:          case OP_MINUPTO:
3608      case ESC_su:          *code += OP_MINUPTO - OP_UPTO;
3609      case ESC_SU:          break;
3610        {          }
       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 */  
3611        }        }
3612      /* Fall through */      c = *code;
3613        }
3614      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3615        {
3616    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3617        if (c == OP_XCLASS)
3618          repeat_opcode = code + GET(code, 1);
3619        else
3620    #endif
3621          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3622    
3623      case ESC_p:      c = *repeat_opcode;
3624      case ESC_P:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3625        {        {
3626        unsigned int ptype = 0, pdata = 0;        /* end must not be NULL. */
3627        int errorcodeptr;        end = get_chr_property_list(code, utf, cd->fcc, list);
       BOOL negated;  
3628    
3629        ptr--;      /* Make ptr point at the p or P */        list[1] = (c & 1) == 0;
       if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  
         return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3630    
3631        /* If the property item is optional, we have to give up. (When generated        if (compare_opcodes(end, utf, cd, list, end))
3632        from \d etc by PCRE_UCP, this test will have been applied much earlier,          {
3633        to the original \d etc. At this point, ptr will point to a zero byte. */          switch (c)
3634              {
3635              case OP_CRSTAR:
3636              case OP_CRMINSTAR:
3637              *repeat_opcode = OP_CRPOSSTAR;
3638              break;
3639    
3640        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||            case OP_CRPLUS:
3641          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)            case OP_CRMINPLUS:
3642            return FALSE;            *repeat_opcode = OP_CRPOSPLUS;
3643              break;
3644    
3645        /* Do the property check. */            case OP_CRQUERY:
3646              case OP_CRMINQUERY:
3647              *repeat_opcode = OP_CRPOSQUERY;
3648              break;
3649    
3650        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);            case OP_CRRANGE:
3651              case OP_CRMINRANGE:
3652              *repeat_opcode = OP_CRPOSRANGE;
3653              break;
3654              }
3655            }
3656        }        }
3657  #endif      c = *code;
   
     default:  
     return FALSE;  
3658      }      }
3659    
3660    /* In principle, support for Unicode properties should be integrated here as    switch(c)
3661    well. It means re-organizing the above code so as to get hold of the property      {
3662    values before switching on the op-code. However, I wonder how many patterns      case OP_END:
3663    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,      return;
   these op-codes are never generated.) */  
3664    
3665    case OP_DIGIT:      case OP_TYPESTAR:
3666    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||      case OP_TYPEMINSTAR:
3667           escape == ESC_h || escape == ESC_v || escape == ESC_R;      case OP_TYPEPLUS:
3668        case OP_TYPEMINPLUS:
3669        case OP_TYPEQUERY:
3670        case OP_TYPEMINQUERY:
3671        case OP_TYPEPOSSTAR:
3672        case OP_TYPEPOSPLUS:
3673        case OP_TYPEPOSQUERY:
3674        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3675        break;
3676    
3677    case OP_NOT_DIGIT:      case OP_TYPEUPTO:
3678    return escape == ESC_d;      case OP_TYPEMINUPTO:
3679        case OP_TYPEEXACT:
3680        case OP_TYPEPOSUPTO:
3681        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3682          code += 2;
3683        break;
3684    
3685    case OP_WHITESPACE:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3686    return escape == ESC_S || escape == ESC_d || escape == ESC_w;      case OP_XCLASS:
3687        code += GET(code, 1);
3688        break;
3689    #endif
3690    
3691    case OP_NOT_WHITESPACE:      case OP_MARK:
3692    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;      case OP_PRUNE_ARG:
3693        case OP_SKIP_ARG:
3694        case OP_THEN_ARG:
3695        code += code[1];
3696        break;
3697        }
3698    
3699    case OP_HSPACE:    /* Add in the fixed length from the table */
   return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  
          escape == ESC_w || escape == ESC_v || escape == ESC_R;  
3700    
3701    case OP_NOT_HSPACE:    code += PRIV(OP_lengths)[c];
   return escape == ESC_h;  
3702    
3703    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3704    case OP_ANYNL:    a multi-byte character. The length in the table is a minimum, so we have to
3705    case OP_VSPACE:    arrange to skip the extra bytes. */
3706    return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
3707    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3708      if (utf) switch(c)
3709        {
3710        case OP_CHAR:
3711        case OP_CHARI:
3712        case OP_NOT:
3713        case OP_NOTI:
3714        case OP_STAR:
3715        case OP_MINSTAR:
3716        case OP_PLUS:
3717        case OP_MINPLUS:
3718        case OP_QUERY:
3719        case OP_MINQUERY:
3720        case OP_UPTO:
3721        case OP_MINUPTO:
3722        case OP_EXACT:
3723        case OP_POSSTAR:
3724        case OP_POSPLUS:
3725        case OP_POSQUERY:
3726        case OP_POSUPTO:
3727        case OP_STARI:
3728        case OP_MINSTARI:
3729        case OP_PLUSI:
3730        case OP_MINPLUSI:
3731        case OP_QUERYI:
3732        case OP_MINQUERYI:
3733        case OP_UPTOI:
3734        case OP_MINUPTOI:
3735        case OP_EXACTI:
3736        case OP_POSSTARI:
3737        case OP_POSPLUSI:
3738        case OP_POSQUERYI:
3739        case OP_POSUPTOI:
3740        case OP_NOTSTAR:
3741        case OP_NOTMINSTAR:
3742        case OP_NOTPLUS:
3743        case OP_NOTMINPLUS:
3744        case OP_NOTQUERY:
3745        case OP_NOTMINQUERY:
3746        case OP_NOTUPTO:
3747        case OP_NOTMINUPTO:
3748        case OP_NOTEXACT:
3749        case OP_NOTPOSSTAR:
3750        case OP_NOTPOSPLUS:
3751        case OP_NOTPOSQUERY:
3752        case OP_NOTPOSUPTO:
3753        case OP_NOTSTARI:
3754        case OP_NOTMINSTARI:
3755        case OP_NOTPLUSI:
3756        case OP_NOTMINPLUSI:
3757        case OP_NOTQUERYI:
3758        case OP_NOTMINQUERYI:
3759        case OP_NOTUPTOI:
3760        case OP_NOTMINUPTOI:
3761        case OP_NOTEXACTI:
3762        case OP_NOTPOSSTARI:
3763        case OP_NOTPOSPLUSI:
3764        case OP_NOTPOSQUERYI:
3765        case OP_NOTPOSUPTOI:
3766        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3767        break;
3768        }
3769    #else
3770      (void)(utf);  /* Keep compiler happy by referencing function argument */
3771    #endif
3772      }
3773    }
3774    
3775    
3776    
3777    /*************************************************
3778    *           Check for POSIX class syntax         *
3779    *************************************************/
3780    
3781    /* This function is called when the sequence "[:" or "[." or "[=" is
3782    encountered in a character class. It checks whether this is followed by a
3783    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3784    reach an unescaped ']' without the special preceding character, return FALSE.
3785    
3786    Originally, this function only recognized a sequence of letters between the
3787    terminators, but it seems that Perl recognizes any sequence of characters,
3788    though of course unknown POSIX names are subsequently rejected. Perl gives an
3789    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3790    didn't consider this to be a POSIX class. Likewise for [:1234:].
3791    
3792    The problem in trying to be exactly like Perl is in the handling of escapes. We
3793    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3794    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3795    below handles the special case of \], but does not try to do any other escape
3796    processing. This makes it different from Perl for cases such as [:l\ower:]
3797    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3798    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3799    I think.
3800    
3801    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3802    It seems that the appearance of a nested POSIX class supersedes an apparent
3803    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3804    a digit.
3805    
3806    In Perl, unescaped square brackets may also appear as part of class names. For
3807    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3808    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3809    seem right at all. PCRE does not allow closing square brackets in POSIX class
3810    names.
3811    
3812    Arguments:
3813      ptr      pointer to the initial [
3814      endptr   where to return the end pointer
3815    
3816    Returns:   TRUE or FALSE
3817    */
3818    
3819    static BOOL
3820    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3821    {
3822    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3823    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3824    for (++ptr; *ptr != CHAR_NULL; ptr++)
3825      {
3826      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3827        ptr++;
3828      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3829      else
3830        {
3831        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3832          {
3833          *endptr = ptr;
3834          return TRUE;
3835          }
3836        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3837             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3838              ptr[1] == CHAR_EQUALS_SIGN) &&
3839            check_posix_syntax(ptr, endptr))
3840          return FALSE;
3841        }
3842      }
3843    return FALSE;
3844    }
3845    
3846    
3847    
3848    
3849    /*************************************************
3850    *          Check POSIX class name                *
3851    *************************************************/
3852    
3853    /* This function is called to check the name given in a POSIX-style class entry
3854    such as [:alnum:].
3855    
3856    Arguments:
3857      ptr        points to the first letter
3858      len        the length of the name
3859    
3860    Returns:     a value representing the name, or -1 if unknown
3861    */
3862    
3863    static int
3864    check_posix_name(const pcre_uchar *ptr, int len)
3865    {
3866    const char *pn = posix_names;
3867    register int yield = 0;
3868    while (posix_name_lengths[yield] != 0)
3869      {
3870      if (len == posix_name_lengths[yield] &&
3871        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3872      pn += posix_name_lengths[yield] + 1;
3873      yield++;
3874      }
3875    return -1;
3876    }
3877    
3878    
3879    /*************************************************
3880    *    Adjust OP_RECURSE items in repeated group   *
3881    *************************************************/
3882    
3883    /* OP_RECURSE items contain an offset from the start of the regex to the group
3884    that is referenced. This means that groups can be replicated for fixed
3885    repetition simply by copying (because the recursion is allowed to refer to
3886    earlier groups that are outside the current group). However, when a group is
3887    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3888    inserted before it, after it has been compiled. This means that any OP_RECURSE
3889    items within it that refer to the group itself or any contained groups have to
3890    have their offsets adjusted. That one of the jobs of this function. Before it
3891    is called, the partially compiled regex must be temporarily terminated with
3892    OP_END.
3893    
3894    This function has been extended with the possibility of forward references for
3895    recursions and subroutine calls. It must also check the list of such references
3896    for the group we are dealing with. If it finds that one of the recursions in
3897    the current group is on this list, it adjusts the offset in the list, not the
3898    value in the reference (which is a group number).
3899    
3900    Arguments:
3901      group      points to the start of the group
3902      adjust     the amount by which the group is to be moved
3903      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3904      cd         contains pointers to tables etc.
3905      save_hwm   the hwm forward reference pointer at the start of the group
3906    
3907    Returns:     nothing
3908    */
3909    
3910    static void
3911    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3912      pcre_uchar *save_hwm)
3913    {
3914    pcre_uchar *ptr = group;
3915    
3916    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3917      {
3918      int offset;
3919      pcre_uchar *hc;
3920    
3921      /* See if this recursion is on the forward reference list. If so, adjust the
3922      reference. */
3923    
3924      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3925        {
3926        offset = (int)GET(hc, 0);
3927        if (cd->start_code + offset == ptr + 1)
3928          {
3929          PUT(hc, 0, offset + adjust);
3930          break;
3931          }
3932        }
3933    
3934      /* Otherwise, adjust the recursion offset if it's after the start of this
3935      group. */
3936    
3937      if (hc >= cd->hwm)
3938        {
3939        offset = (int)GET(ptr, 1);
3940        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3941        }
3942    
3943      ptr += 1 + LINK_SIZE;
3944      }
3945    }
3946    
3947    
3948    
3949    /*************************************************
3950    *        Insert an automatic callout point       *
3951    *************************************************/
3952    
3953    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3954    callout points before each pattern item.
3955    
3956    Arguments:
3957      code           current code pointer
3958      ptr            current pattern pointer
3959      cd             pointers to tables etc
3960    
3961    Returns:         new code pointer
3962    */
3963    
3964    static pcre_uchar *
3965    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3966    {
3967    *code++ = OP_CALLOUT;
3968    *code++ = 255;
3969    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3970    PUT(code, LINK_SIZE, 0);                       /* Default length */
3971    return code + 2 * LINK_SIZE;
3972    }
3973    
3974    
3975    
3976    /*************************************************
3977    *         Complete a callout item                *
3978    *************************************************/
3979    
3980    /* A callout item contains the length of the next item in the pattern, which
3981    we can't fill in till after we have reached the relevant point. This is used
3982    for both automatic and manual callouts.
3983    
3984    Arguments:
3985      previous_callout   points to previous callout item
3986      ptr                current pattern pointer
3987      cd                 pointers to tables etc
3988    
3989    Returns:             nothing
3990    */
3991    
3992    static void
3993    complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3994    {
3995    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3996    PUT(previous_callout, 2 + LINK_SIZE, length);
3997    }
3998    
3999    
4000    
4001    #ifdef SUPPORT_UCP
4002    /*************************************************
4003    *           Get othercase range                  *
4004    *************************************************/
4005    
4006    /* This function is passed the start and end of a class range, in UTF-8 mode
4007    with UCP support. It searches up the characters, looking for ranges of
4008    characters in the "other" case. Each call returns the next one, updating the
4009    start address. A character with multiple other cases is returned on its own
4010    with a special return value.
4011    
4012    Arguments:
4013      cptr        points to starting character value; updated
4014      d           end value
4015      ocptr       where to put start of othercase range
4016      odptr       where to put end of othercase range
4017    
4018    Yield:        -1 when no more
4019                   0 when a range is returned
4020                  >0 the CASESET offset for char with multiple other cases
4021                    in this case, ocptr contains the original
4022    */
4023    
4024    static int
4025    get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4026      pcre_uint32 *odptr)
4027    {
4028    pcre_uint32 c, othercase, next;
4029    unsigned int co;
4030    
4031    /* Find the first character that has an other case. If it has multiple other
4032    cases, return its case offset value. */
4033    
4034    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4035    return escape == ESC_v || escape == ESC_R;    {
4036      if ((co = UCD_CASESET(c)) != 0)
4037        {
4038        *ocptr = c++;   /* Character that has the set */
4039        *cptr = c;      /* Rest of input range */
4040        return (int)co;
4041        }
4042      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4043      }
4044    
4045    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  
          escape == ESC_v || escape == ESC_R;  
4046    
4047    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4048    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4049    
4050    default:  for (++c; c <= d; c++)
4051    return FALSE;    {
4052      if (UCD_OTHERCASE(c) != next) break;
4053      next++;
4054    }    }
4055    
4056  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4057    *cptr = c;             /* Rest of input range */
4058    return 0;
4059  }  }
4060    #endif  /* SUPPORT_UCP */
4061    
4062    
4063    
# Line 3754  to find out the amount of memory needed, Line 4310  to find out the amount of memory needed,
4310  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4311    
4312  Arguments:  Arguments:
4313    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4314    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4315    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4316    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4317    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
4318    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
4319    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
4320    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
4321    bcptr          points to current branch chain    bcptr             points to current branch chain
4322    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
4323    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
4324    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
4325                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
4326    
4327  Returns:         TRUE on success  Returns:            TRUE on success
4328                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4329  */  */
4330    
4331  static BOOL  static BOOL
# Line 3994  for (;; ptr++) Line 4550  for (;; ptr++)
4550        }        }
4551      }      }
4552    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4553    is_quantifier =    is_quantifier =
4554      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4555      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4556    
4557    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4558      quantifier or when processing a property substitution string in UCP mode. */
4559    
4560      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4561         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4562      {      {
4563      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
# Line 4032  for (;; ptr++) Line 4588  for (;; ptr++)
4588        }        }
4589      }      }
4590    
4591    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4592      are substituted for \w etc in UCP mode. */
4593    
4594    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4595      {      {
4596      previous_callout = code;      previous_callout = code;
4597      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 4286  for (;; ptr++) Line 4843  for (;; ptr++)
4843            posix_class = 0;            posix_class = 0;
4844    
4845          /* 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
4846          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4847            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4848            directly. */
4849    
4850  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4851          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4852            {            {
4853              unsigned int ptype = 0;
4854            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4855    
4856              /* The posix_substitutes table specifies which POSIX classes can be
4857              converted to \p or \P items. */
4858    
4859            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4860              {              {
4861              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4862              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4863              continue;              continue;
4864              }              }
4865    
4866              /* There are three other classes that generate special property calls
4867              that are recognized only in an XCLASS. */
4868    
4869              else switch(posix_class)
4870                {
4871                case PC_GRAPH:
4872                ptype = PT_PXGRAPH;
4873                /* Fall through */
4874                case PC_PRINT:
4875                if (ptype == 0) ptype = PT_PXPRINT;
4876                /* Fall through */
4877                case PC_PUNCT:
4878                if (ptype == 0) ptype = PT_PXPUNCT;
4879                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4880                *class_uchardata++ = ptype;
4881                *class_uchardata++ = 0;
4882                ptr = tempptr + 1;
4883                continue;
4884    
4885                /* For all other POSIX classes, no special action is taken in UCP
4886                mode. Fall through to the non_UCP case. */
4887    
4888                default:
4889                break;
4890                }
4891            }            }
4892  #endif  #endif
4893          /* 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
4894          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
4895          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
4896          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
4897            bit map that is being built. */
4898    
4899          posix_class *= 3;          posix_class *= 3;
4900    
# Line 4420  for (;; ptr++) Line 5011  for (;; ptr++)
5011              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5012              continue;              continue;
5013    
5014              /* 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
5015              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
5016              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5017              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5018                we could just adjust the appropriate bit. From PCRE 8.34 we no
5019                longer treat \s and \S specially. */
5020    
5021              case ESC_s:              case ESC_s:
5022              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];  
5023              continue;              continue;
5024    
5025              case ESC_S:              case ESC_S:
5026              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5027              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 */  
5028              continue;              continue;
5029    
5030              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
# Line 4555  for (;; ptr++) Line 5145  for (;; ptr++)
5145          else          else
5146  #endif  #endif
5147          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5148    
5149          /* 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
5150          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
5151          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5152            warning is given, so PCRE now faults it as it is almost certainly a
5153          if (!inescq && d == CHAR_BACKSLASH)          mistake on the user's part. */
5154            {  
5155            int descape;          if (!inescq)
5156            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);            {
5157            if (*errorcodeptr != 0) goto FAILED;            if (d == CHAR_BACKSLASH)
   
           /* \b is backspace; any other special means the '-' was literal. */  
   
           if (descape != 0)  
5158              {              {
5159              if (descape == ESC_b) d = CHAR_BS; else              int descape;
5160                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5161                if (*errorcodeptr != 0) goto FAILED;
5162    
5163                /* 0 means a character was put into d; \b is backspace; any other
5164                special causes an error. */
5165    
5166                if (descape != 0)
5167                {                {
5168                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5169                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5170                    *errorcodeptr = ERR83;
5171                    goto FAILED;
5172                    }
5173                }                }
5174              }              }
5175            }  
5176              /* A hyphen followed by a POSIX class is treated in the same way. */
5177    
5178              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5179                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5180                        ptr[1] == CHAR_EQUALS_SIGN) &&
5181                       check_posix_syntax(ptr, &tempptr))
5182                {
5183                *errorcodeptr = ERR83;
5184                goto FAILED;
5185                }
5186              }
5187    
5188          /* Check that the two values are in the correct order. Optimize          /* Check that the two values are in the correct order. Optimize
5189          one-character ranges. */          one-character ranges. */
# Line 4933  for (;; ptr++) Line 5540  for (;; ptr++)
5540            }            }
5541          }          }
5542    
       /* 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;  
         }  
   
5543        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
5544        }        }
5545    
# Line 4963  for (;; ptr++) Line 5557  for (;; ptr++)
5557        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5558        c = *previous;        c = *previous;
5559    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5560        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5561        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5562          {          {
# Line 5119  for (;; ptr++) Line 5705  for (;; ptr++)
5705      /* 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
5706      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}. */
5707    
5708      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5709  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5710               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5711  #endif  #endif
5712               *previous == OP_REF ||               *previous == OP_REF   || *previous == OP_REFI ||
5713               *previous == OP_REFI)               *previous == OP_DNREF || *previous == OP_DNREFI)
5714        {        {
5715        if (repeat_max == 0)        if (repeat_max == 0)
5716          {          {
# Line 5153  for (;; ptr++) Line 5738  for (;; ptr++)
5738      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
5739      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 >=
5740      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5741      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5742      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5743        for Perl compatibility. */
5744    
5745      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5746        {        {
# Line 5172  for (;; ptr++) Line 5758  for (;; ptr++)
5758        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5759        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5760        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
5761        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5762    
5763        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5764          {          {
# Line 5545  for (;; ptr++) Line 6131  for (;; ptr++)
6131        goto FAILED;        goto FAILED;
6132        }        }
6133    
6134      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6135      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6136      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
6137      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6138      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.  
6139    
6140      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6141      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6142      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6143        tempcode, not at previous, which might be the first part of a string whose
6144      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. */  
6145    
6146      if (possessive_quantifier)      if (possessive_quantifier)
6147        {        {
6148        int len;        int len;
6149    
6150        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6151          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6152          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6153          remains is greater than zero, there's a further opcode that can be
6154          handled. If not, do nothing, leaving the EXACT alone. */
6155    
6156          switch(*tempcode)
6157            {
6158            case OP_TYPEEXACT:
6159          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6160            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6161            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6162            break;
6163    
6164        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6165          {  
6166            case OP_CHAR:
6167            case OP_CHARI:
6168            case OP_NOT:
6169            case OP_NOTI:
6170            case OP_EXACT:
6171            case OP_EXACTI:
6172            case OP_NOTEXACT:
6173            case OP_NOTEXACTI:
6174          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6175  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6176          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6177            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6178  #endif  #endif
6179            break;
6180    
6181            /* For the class opcodes, the repeat operator appears at the end;
6182            adjust tempcode to point to it. */
6183    
6184            case OP_CLASS:
6185            case OP_NCLASS:
6186            tempcode += 1 + 32/sizeof(pcre_uchar);
6187            break;
6188    
6189    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6190            case OP_XCLASS:
6191            tempcode += GET(tempcode, 1);
6192            break;
6193    #endif
6194          }          }
6195    
6196          /* If tempcode is equal to code (which points to the end of the repeated
6197          item), it means we have skipped an EXACT item but there is no following
6198          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6199          all other cases, tempcode will be pointing to the repeat opcode, and will
6200          be less than code, so the value of len will be greater than 0. */
6201    
6202        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6203          if (len > 0)
6204            {
6205            unsigned int repcode = *tempcode;
6206    
6207            /* There is a table for possessifying opcodes, all of which are less
6208            than OP_CALLOUT. A zero entry means there is no possessified version.
6209            */
6210    
6211            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6212              *tempcode = opcode_possessify[repcode];
6213    
6214            /* For opcode without a special possessified version, wrap the item in
6215            ONCE brackets. Because we are moving code along, we must ensure that any
6216            pending recursive references are updated. */
6217    
6218            else
6219              {
6220              *code = OP_END;
6221              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6222              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6223              code += 1 + LINK_SIZE;
6224              len += 1 + LINK_SIZE;
6225              tempcode[0] = OP_ONCE;
6226              *code++ = OP_KET;
6227              PUTINC(code, 0, len);
6228              PUT(tempcode, 1, len);
6229              }
6230            }
6231    
6232    #ifdef NEVER
6233        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6234          {          {
6235          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5609  for (;; ptr++) Line 6257  for (;; ptr++)
6257          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6258          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6259    
6260            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6261            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6262            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6263            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6264    
6265          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6266          pending recursive references are updated. */          pending recursive references are updated. */
6267    
# Line 5624  for (;; ptr++) Line 6277  for (;; ptr++)
6277          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6278          break;          break;
6279          }          }
6280    #endif
6281        }        }
6282    
6283      /* 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 6466  for (;; ptr++)
6466          tempptr = ptr;          tempptr = ptr;
6467    
6468          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6469          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',
6470          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6471            recursion tests.
6472          There are several syntaxes for testing a named group: (?(name)) is used  
6473          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6474            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6475          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6476          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
6477          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
6478          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.  
6479    
6480          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6481          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 6499  for (;; ptr++)
6499                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6500            break;            break;
6501    
6502          /* 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
6503          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. */
6504    
6505          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6506          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
# Line 5855  for (;; ptr++) Line 6508  for (;; ptr++)
6508    
6509          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6510    
6511          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6512            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6513            {            {
6514            terminator = -1;            terminator = -1;
6515            ptr += 2;            ptr += 2;
# Line 5863  for (;; ptr++) Line 6517  for (;; ptr++)
6517            }            }
6518    
6519          /* 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
6520          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6521            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6522    
6523          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6524            {            {
6525            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6526            ptr++;            ptr++;
6527            }            }
6528          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6529            {            {
6530            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6531            ptr++;            ptr++;
# Line 5878  for (;; ptr++) Line 6533  for (;; ptr++)
6533          else          else
6534            {            {
6535            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6536            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6537                else if (IS_DIGIT(*ptr)) refsign = 0;
6538            }            }
6539    
6540          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6541    
6542          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6543            {            {
6544            ptr += 1;  /* To get the right offset */            recno = 0;
6545            *errorcodeptr = ERR28;            while (IS_DIGIT(*ptr))
6546            goto FAILED;              {
6547            }              recno = recno * 10 + (int)(*ptr - CHAR_0);
6548                ptr++;
6549          /* Read the name, but also get it as a number if it's all digits */              }
6550              }
6551    
6552            /* Otherwise we expect to read a name; anything else is an error. When
6553            a name is one of a number of duplicates, a different opcode is used and
6554            it needs more memory. Unfortunately we cannot tell whether a name is a
6555            duplicate in the first pass, so we have to allow for more memory. */
6556    
6557          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6558            {            {
6559            if (recno >= 0)            if (IS_DIGIT(*ptr))
6560              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6561            ptr++;              *errorcodeptr = ERR84;
6562                goto FAILED;
6563                }
6564              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6565                {
6566                *errorcodeptr = ERR28;   /* Assertion expected */
6567                goto FAILED;
6568                }
6569              name = ptr++;
6570              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6571                {
6572                ptr++;
6573                }
6574              namelen = (int)(ptr - name);
6575              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6576            }            }
6577          namelen = (int)(ptr - name);  
6578            /* Check the terminator */
6579    
6580          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6581              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6582            {            {
6583            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6584            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6585            goto FAILED;            goto FAILED;
6586            }            }
6587    
# Line 5915  for (;; ptr++) Line 6590  for (;; ptr++)
6590          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6591    
6592          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
6593          reference. If the string started with "+" or "-" we require the rest to          reference. If refsign is not negative, it means we have a number in
6594          be digits, in which case recno will be set. */          recno. */
6595    
6596          if (refsign > 0)          if (refsign >= 0)
6597            {            {
6598            if (recno <= 0)            if (recno <= 0)
6599              {              {
6600              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6601              goto FAILED;              goto FAILED;
6602              }              }
6603            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6604              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6605            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6606              {              {
6607              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 5936  for (;; ptr++) Line 6611  for (;; ptr++)
6611            break;            break;
6612            }            }
6613    
6614          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise look for the name. */
         name. If we find a name, add one to the opcode to change OP_CREF or  
         OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,  
         except they record that the reference was originally to a name. The  
         information is used to check duplicate names. */  
6615    
6616          slot = cd->name_table;          slot = cd->name_table;
6617          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 5949  for (;; ptr++) Line 6620  for (;; ptr++)
6620            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6621            }            }
6622    
6623          /* Found a previous named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6624            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6625            appropriate data values. Otherwise, just insert the unique subpattern
6626            number. */
6627    
6628          if (i < cd->names_found)          if (i < cd->names_found)
6629            {            {
6630            recno = GET2(slot, 0);            int offset = i++;
6631            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6632            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6633            }            for (; i < cd->names_found; i++)
6634                {
6635          /* Search the pattern for a forward reference */              slot += cd->name_entry_size;
6636                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6637          else if ((i = find_parens(cd, name, namelen,              count++;
6638                          (options & PCRE_EXTENDED) != 0, utf)) > 0)              }
6639            {            if (count > 1)
6640            PUT2(code, 2+LINK_SIZE, i);              {
6641            code[1+LINK_SIZE]++;              PUT2(code, 2+LINK_SIZE, offset);
6642                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6643                skipbytes += IMM2_SIZE;
6644                code[1+LINK_SIZE]++;
6645                }
6646              else  /* Not a duplicated name */
6647                {
6648                PUT2(code, 2+LINK_SIZE, recno);
6649                }
6650            }            }
6651    
6652          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
6653          after the opening parenthesis [e.g. (?(abc)...] and in this case there          after the opening parenthesis [e.g. (?(abc)...] and in this case there
6654          are some further alternatives to try. For the cases where terminator !=          are some further alternatives to try. For the cases where terminator !=
6655          0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have          CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6656          now checked all the possibilities, so give an error. */          we have now checked all the possibilities, so give an error. */
6657    
6658          else if (terminator != CHAR_NULL)          else if (terminator != CHAR_NULL)
6659            {            {
# Line 6008  for (;; ptr++) Line 6690  for (;; ptr++)
6690            skipbytes = 1;            skipbytes = 1;
6691            }            }
6692    
6693          /* Check for the "name" actually being a subpattern number. We are          /* Reference to an unidentified subpattern. */
         in the second pass here, so final_bracount is set. */  
   
         else if (recno > 0 && recno <= cd->final_bracount)  
           {  
           PUT2(code, 2+LINK_SIZE, recno);  
           }  
   
         /* Either an unidentified subpattern, or a reference to (?(0) */  
6694    
6695          else          else
6696            {            {
6697            *errorcodeptr = (recno == 0)? ERR35: ERR15;            *errorcodeptr = ERR15;
6698            goto FAILED;            goto FAILED;
6699            }            }
6700          break;          break;
# Line 6033  for (;; ptr++) Line 6707  for (;; ptr++)
6707          ptr++;          ptr++;
6708          break;          break;
6709    
6710            /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6711            thing to do, but Perl allows all assertions to be quantified, and when
6712            they contain capturing parentheses there may be a potential use for
6713            this feature. Not that that applies to a quantified (?!) but we allow
6714            it for uniformity. */
6715    
6716          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6717          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
6718          ptr++;          ptr++;
6719          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6720                 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6721                (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6722            {            {
6723            *code++ = OP_FAIL;            *code++ = OP_FAIL;
6724            previous = NULL;            previous = NULL;
# Line 6130  for (;; ptr++) Line 6811  for (;; ptr++)
6811          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6812          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
6813          case CHAR_APOSTROPHE:          case CHAR_APOSTROPHE:
6814            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6815              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6816            name = ++ptr;
6817            if (IS_DIGIT(*ptr))
6818            {            {
6819            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?            *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6820              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            goto FAILED;
6821            name = ++ptr;            }
6822            while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6823            namelen = (int)(ptr - name);
6824    
6825            while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          /* In the pre-compile phase, do a syntax check, remember the longest
6826            namelen = (int)(ptr - name);          name, and then remember the group in a vector, expanding it if
6827            necessary. Duplicates for the same number are skipped; other duplicates
6828            are checked for validity. In the actual compile, there is nothing to
6829            do. */
6830    
6831            /* In the pre-compile phase, just do a syntax check. */          if (lengthptr != NULL)
6832              {
6833              named_group *ng;
6834              pcre_uint32 number = cd->bracount + 1;
6835    
6836            if (lengthptr != NULL)            if (*ptr != (pcre_uchar)terminator)
6837              {              {
6838              if (*ptr != (pcre_uchar)terminator)              *errorcodeptr = ERR42;
6839                {              goto FAILED;
6840                *errorcodeptr = ERR42;              }
6841                goto FAILED;  
6842                }            if (cd->names_found >= MAX_NAME_COUNT)
6843              if (cd->names_found >= MAX_NAME_COUNT)              {
6844                *errorcodeptr = ERR49;
6845                goto FAILED;
6846                }
6847    
6848              if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
6849                {
6850                cd->name_entry_size = namelen + IMM2_SIZE + 1;
6851                if (namelen > MAX_NAME_SIZE)
6852                {                {
6853                *errorcodeptr = ERR49;                *errorcodeptr = ERR48;
6854                goto FAILED;                goto FAILED;
6855                }                }
6856              if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)              }
6857    
6858              /* Scan the list to check for duplicates. For duplicate names, if the
6859              number is the same, break the loop, which causes the name to be
6860              discarded; otherwise, if DUPNAMES is not set, give an error.
6861              If it is set, allow the name with a different number, but continue
6862              scanning in case this is a duplicate with the same number. For
6863              non-duplicate names, give an error if the number is duplicated. */
6864    
6865              ng = cd->named_groups;
6866              for (i = 0; i < cd->names_found; i++, ng++)
6867                {
6868                if (namelen == ng->length &&
6869                    STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6870                {                {
6871                cd->name_entry_size = namelen + IMM2_SIZE + 1;                if (ng->number == number) break;
6872                if (namelen > MAX_NAME_SIZE)                if ((options & PCRE_DUPNAMES) == 0)
6873                  {                  {
6874                  *errorcodeptr = ERR48;                  *errorcodeptr = ERR43;
6875                  goto FAILED;                  goto FAILED;
6876                  }                  }
6877                  cd->dupnames = TRUE;  /* Duplicate names exist */
6878                  }
6879                else if (ng->number == number)
6880                  {
6881                  *errorcodeptr = ERR65;
6882                  goto FAILED;
6883                }                }
6884              }              }
6885    
6886            /* In the real compile, create the entry in the table, maintaining            if (i >= cd->names_found)     /* Not a duplicate with same number */
           alphabetical order. Duplicate names for different numbers are  
           permitted only if PCRE_DUPNAMES is set. Duplicate names for the same  
           number are always OK. (An existing number can be re-used if (?|  
           appears in the pattern.) In either event, a duplicate name results in  
           a duplicate entry in the table, even if the number is the same. This  
           is because the number of names, and hence the table size, is computed  
           in the pre-compile, and it affects various numbers and pointers which  
           would all have to be modified, and the compiled code moved down, if  
           duplicates with the same number were omitted from the table. This  
           doesn't seem worth the hassle. However, *different* names for the  
           same number are not permitted. */  
   
           else  
6887              {              {
6888              BOOL dupname = FALSE;              /* Increase the list size if necessary */
             slot = cd->name_table;  
6889    
6890              for (i = 0; i < cd->names_found; i++)              if (cd->names_found >= cd->named_group_list_size)
6891                {                {
6892                int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(namelen));                int newsize = cd->named_group_list_size * 2;
6893                if (crc == 0)                named_group *newspace = (PUBL(malloc))
6894                  {                  (newsize * sizeof(named_group));
                 if (slot[IMM2_SIZE+namelen] == 0)  
                   {  
                   if (GET2(slot, 0) != cd->bracount + 1 &&  
                       (options & PCRE_DUPNAMES) == 0)  
                     {  
                     *errorcodeptr = ERR43;  
                     goto FAILED;  
                     }  
                   else dupname = TRUE;  
                   }  
                 else crc = -1;      /* Current name is a substring */  
                 }  
   
               /* Make space in the table and break the loop for an earlier  
               name. For a duplicate or later name, carry on. We do this for  
               duplicates so that in the simple case (when ?(| is not used) they  
               are in order of their numbers. */  
6895    
6896                if (crc < 0)                if (newspace == NULL)
6897                  {                  {
6898                  memmove(slot + cd->name_entry_size, slot,                  *errorcodeptr = ERR21;
6899                    IN_UCHARS((cd->names_found - i) * cd->name_entry_size));                  goto FAILED;
                 break;  
6900                  }                  }
6901    
6902                /* Continue the loop for a later or duplicate name */                memcpy(newspace, cd->named_groups,
6903                    cd->named_group_list_size * sizeof(named_group));
6904                slot += cd->name_entry_size;                if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
6905                }                  (PUBL(free))((void *)cd->named_groups);
6906                  cd->named_groups = newspace;
6907              /* For non-duplicate names, check for a duplicate number before                cd->named_group_list_size = newsize;
             adding the new name. */  
   
             if (!dupname)  
               {  
               pcre_uchar *cslot = cd->name_table;  
               for (i = 0; i < cd->names_found; i++)  
                 {  
                 if (cslot != slot)  
                   {  
                   if (GET2(cslot, 0) == cd->bracount + 1)  
                     {  
                     *errorcodeptr = ERR65;  
                     goto FAILED;  
                     }  
                   }  
                 else i--;  
                 cslot += cd->name_entry_size;  
                 }  
6908                }                }
6909    
6910              PUT2(slot, 0, cd->bracount + 1);              cd->named_groups[cd->names_found].name = name;
6911              memcpy(slot + IMM2_SIZE, name, IN_UCHARS(namelen));              cd->named_groups[cd->names_found].length = namelen;
6912              slot[IMM2_SIZE + namelen] = 0;              cd->named_groups[cd->names_found].number = number;
6913                cd->names_found++;
6914              }              }
6915            }            }
6916    
6917          /* In both pre-compile and compile, count the number of names we've          ptr++;                    /* Move past > or ' in both passes. */
         encountered. */  
   
         cd->names_found++;  
         ptr++;                    /* Move past > or ' */  
6918          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
6919    
6920    
# Line 6265  for (;; ptr++) Line 6932  for (;; ptr++)
6932    
6933          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
6934          name = ++ptr;          name = ++ptr;
6935            if (IS_DIGIT(*ptr))
6936              {
6937              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6938              goto FAILED;
6939              }
6940          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6941          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
6942    
# Line 6277  for (;; ptr++) Line 6949  for (;; ptr++)
6949    
6950          if (lengthptr != NULL)          if (lengthptr != NULL)
6951            {            {
6952            const pcre_uchar *temp;            named_group *ng;
6953    
6954            if (namelen == 0)            if (namelen == 0)
6955              {              {
# Line 6295  for (;; ptr++) Line 6967  for (;; ptr++)
6967              goto FAILED;              goto FAILED;
6968              }              }
6969    
6970            /* The name table does not exist in the first pass, so we cannot            /* The name table does not exist in the first pass; instead we must
6971            do a simple search as in the code below. Instead, we have to scan the            scan the list of names encountered so far in order to get the
6972            pattern to find the number. It is important that we scan it only as            number. If the name is not found, set the value to 0 for a forward
6973            far as we have got because the syntax of named subpatterns has not            reference. */
6974            been checked for the rest of the pattern, and find_parens() assumes  
6975            correct syntax. In any case, it's a waste of resources to scan            ng = cd->named_groups;
6976            further. We stop the scan at the current point by temporarily            for (i = 0; i < cd->names_found; i++, ng++)
6977            adjusting the value of cd->endpattern. */              {
6978                if (namelen == ng->length &&
6979            temp = cd->end_pattern;                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6980            cd->end_pattern = ptr;                break;
6981            recno = find_parens(cd, name, namelen,              }
6982              (options & PCRE_EXTENDED) != 0, utf);            recno = (i < cd->names_found)? ng->number : 0;
6983            cd->end_pattern = temp;  
6984            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            /* Count named back references. */
6985    
6986              if (!is_recurse) cd->namedrefcount++;
6987            }            }
6988    
6989          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, search the name table. We check the name
6990          first, and then check that we have reached the end of the name in the          first, and then check that we have reached the end of the name in the
6991          table. That way, if the name that is longer than any in the table,          table. That way, if the name is longer than any in the table, the
6992          the comparison will fail without reading beyond the table entry. */          comparison will fail without reading beyond the table entry. */
6993    
6994          else          else
6995            {            {
# Line 6328  for (;; ptr++) Line 7002  for (;; ptr++)
7002              slot += cd->name_entry_size;              slot += cd->name_entry_size;
7003              }              }
7004    
7005            if (i < cd->names_found)         /* Back reference */            if (i < cd->names_found)
7006              {              {
7007              recno = GET2(slot, 0);              recno = GET2(slot, 0);
7008              }              }
7009            else if ((recno =                /* Forward back reference */            else
                     find_parens(cd, name, namelen,  
                       (options & PCRE_EXTENDED) != 0, utf)) <= 0)  
7010              {              {
7011              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
7012              goto FAILED;              goto FAILED;
7013              }              }
7014            }            }
7015    
7016          /* In both phases, we can now go to the code than handles numerical          /* In both phases, for recursions, we can now go to the code than
7017          recursion or backreferences. */          handles numerical recursion. */
7018    
7019          if (is_recurse) goto HANDLE_RECURSION;          if (is_recurse) goto HANDLE_RECURSION;
7020            else goto HANDLE_REFERENCE;  
7021            /* In the second pass we must see if the name is duplicated. If so, we
7022            generate a different opcode. */
7023    
7024            if (lengthptr == NULL && cd->dupnames)
7025              {
7026              int count = 1;
7027              unsigned int index = i;
7028              pcre_uchar *cslot = slot + cd->name_entry_size;
7029    
7030              for (i++; i < cd->names_found; i++)
7031                {
7032                if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
7033                count++;
7034                cslot += cd->name_entry_size;
7035                }
7036    
7037              if (count > 1)
7038                {
7039                if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7040                previous = code;
7041                *code++ = ((options & PCRE_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
7042                PUT2INC(code, 0, index);
7043                PUT2INC(code, 0, count);
7044    
7045                /* Process each potentially referenced group. */
7046    
7047                for (; slot < cslot; slot += cd->name_entry_size)
7048                  {
7049                  open_capitem *oc;
7050                  recno = GET2(slot, 0);
7051                  cd->backref_map |= (recno < 32)? (1 << recno) : 1;
7052                  if (recno > cd->top_backref) cd->top_backref = recno;
7053    
7054                  /* Check to see if this back reference is recursive, that it, it
7055                  is inside the group that it references. A flag is set so that the
7056                  group can be made atomic. */
7057    
7058                  for (oc = cd->open_caps; oc != NULL; oc = oc->next)
7059                    {
7060                    if (oc->number == recno)
7061                      {
7062                      oc->flag = TRUE;
7063                      break;
7064                      }
7065                    }
7066                  }
7067