/[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 1072 by chpe, Tue Oct 16 15:54:40 2012 UTC revision 1408 by ph10, Tue Dec 3 16:27:00 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  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 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 250  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 265  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 292  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 319  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 372  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 387  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 452  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 518  static const char error_texts[] =
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 505  static const char error_texts[] = Line 539  static const char error_texts[] =
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550    ;    ;
551    
552  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 686  static const pcre_uint8 ebcdic_chartab[]
686  #endif  #endif
687    
688    
689    /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726    
727    /* This table is used to check whether auto-possessification is possible
728    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729    left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
782      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
783      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
784      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
867  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 883  find_error_text(int n)
883  const char *s = error_texts;  const char *s = error_texts;
884  for (; n > 0; n--)  for (; n > 0; n--)
885    {    {
886    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
887    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
888    }    }
889  return s;  return s;
890  }  }
891    
892    
893    
894  /*************************************************  /*************************************************
895  *           Expand the workspace                 *  *           Expand the workspace                 *
896  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 968  return (*p == CHAR_RIGHT_CURLY_BRACKET);
968  *************************************************/  *************************************************/
969    
970  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
971  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
972  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
973  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
974  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979    chptr          points to the data character    chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 989  Returns:         zero => a data characte
989  */  */
990    
991  static int  static int
992  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
994  {  {
995  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 1004  ptr--;                            /* Set
1004    
1005  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1006    
1007  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008    
1009  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 795  Otherwise further processing may be requ Line 1013  Otherwise further processing may be requ
1013  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1014  /* Not alphanumeric */  /* Not alphanumeric */
1015  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1016  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
1017      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018    
1019  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1020  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1064  else
1064            }            }
1065    
1066  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1067          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1068  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1069          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1070  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1071          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1072  #endif  #endif
1073            {            {
1074            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1115  else
1115      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116        {        {
1117        const pcre_uchar *p;        const pcre_uchar *p;
1118        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121          {          {
1122          escape = ESC_k;          escape = ESC_k;
1123          break;          break;
# Line 961  else Line 1180  else
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1188      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1189      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1190      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1191      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195        Inside a character class, \ followed by a digit is always either a literal
1196        8 or 9 or an octal number. */
1197    
1198      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1220  else
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          escape = -s;          escape = -s;
1226          break;          break;
# Line 1005  else Line 1228  else
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1232      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1241      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1252  else
1252  #endif  #endif
1253      break;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1257      If not, { is treated as a data character. */  
1258        case CHAR_o:
1259        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260          {
1261          ptr += 2;
1262          c = 0;
1263          overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265            {
1266            register pcre_uint32 cc = *ptr++;
1267            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268    #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270    #endif
1271            c = (c << 3) + cc - CHAR_0 ;
1272    #if defined COMPILE_PCRE8
1273            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274    #elif defined COMPILE_PCRE16
1275            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278    #endif
1279            }
1280          if (overflow)
1281            {
1282            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283            *errorcodeptr = ERR34;
1284            }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286            {
1287            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290          }
1291        break;
1292    
1293        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296      case CHAR_x:      case CHAR_x:
1297      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1299        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301          {          {
# Line 1056  else Line 1312  else
1312  #endif  #endif
1313            }            }
1314          }          }
1315        break;        }    /* End JavaScript handling */
       }  
1316    
1317      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1320        seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        c = 0;      else
1325        overflow = FALSE;        {
1326        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          register pcre_uint32 cc = *pt++;          ptr += 2;
1329          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1330            overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1337          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1338  #endif  #endif
1339    
1340  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1344          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1345          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346  #endif  #endif
1347    
1348  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1349          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1351          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1353          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354  #endif  #endif
1355          }            }
1356    
1357        if (overflow)          if (overflow)
1358          {            {
1359          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1361          }            }
1362    
1363        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364          {            {
1365          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366          ptr = pt;            }
         break;  
         }  
1367    
1368        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1369        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1370        }          \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1378        {        else
1379        pcre_uint32 cc;                          /* Some compilers don't like */          {
1380        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1389        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391  #endif  #endif
1392        }            }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1401  else
1401    
1402      case CHAR_c:      case CHAR_c:
1403      c = *(++ptr);      c = *(++ptr);
1404      if (c == 0)      if (c == CHAR_NULL)
1405        {        {
1406        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1407        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1457  if ((options & PCRE_UCP) != 0 && escape
1457  return escape;  return escape;
1458  }  }
1459    
1460    
1461    
1462  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1463  /*************************************************  /*************************************************
1464  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1472  escape sequence.
1472  Argument:  Argument:
1473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1474    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1475    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1476      pdataptr       points to an unsigned int that is set to the detailed property value
1477    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1478    
1479  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1480  */  */
1481    
1482  static int  static BOOL
1483  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484      unsigned int *pdataptr, int *errorcodeptr)
1485  {  {
1486  pcre_uchar c;  pcre_uchar c;
1487  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1489  const pcre_uchar *ptr = *ptrptr;
1489  pcre_uchar name[32];  pcre_uchar name[32];
1490    
1491  c = *(++ptr);  c = *(++ptr);
1492  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1493    
1494  *negptr = FALSE;  *negptr = FALSE;
1495    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1506  if (c == CHAR_LEFT_CURLY_BRACKET)
1506    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507      {      {
1508      c = *(++ptr);      c = *(++ptr);
1509      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1510      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511      name[i] = c;      name[i] = c;
1512      }      }
# Line 1263  while (bot < top) Line 1536  while (bot < top)
1536    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537    if (r == 0)    if (r == 0)
1538      {      {
1539      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1540      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1541        return TRUE;
1542      }      }
1543    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1544    }    }
1545    
1546  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1547  *ptrptr = ptr;  *ptrptr = ptr;
1548  return -1;  return FALSE;
1549    
1550  ERROR_RETURN:  ERROR_RETURN:
1551  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1552  *ptrptr = ptr;  *ptrptr = ptr;
1553  return -1;  return FALSE;
1554  }  }
1555  #endif  #endif
1556    
1557    
1558    
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1351  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; 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 != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) 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) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) 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 == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) 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 == 0) 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++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1627  *      Find first significant op code            *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
# Line 1684  for (;;) Line 1662  for (;;)
1662    
1663      case OP_CALLOUT:      case OP_CALLOUT:
1664      case OP_CREF:      case OP_CREF:
1665      case OP_NCREF:      case OP_DNCREF:
1666      case OP_RREF:      case OP_RREF:
1667      case OP_NRREF:      case OP_DNRREF:
1668      case OP_DEF:      case OP_DEF:
1669      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1670      break;      break;
# Line 1700  for (;;) Line 1678  for (;;)
1678    
1679    
1680    
   
1681  /*************************************************  /*************************************************
1682  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1683  *************************************************/  *************************************************/
# Line 1824  for (;;) Line 1801  for (;;)
1801      case OP_COMMIT:      case OP_COMMIT:
1802      case OP_CREF:      case OP_CREF:
1803      case OP_DEF:      case OP_DEF:
1804        case OP_DNCREF:
1805        case OP_DNRREF:
1806      case OP_DOLL:      case OP_DOLL:
1807      case OP_DOLLM:      case OP_DOLLM:
1808      case OP_EOD:      case OP_EOD:
1809      case OP_EODN:      case OP_EODN:
1810      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812      case OP_PRUNE:      case OP_PRUNE:
1813      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1829  for (;;)
1829      case OP_NOTI:      case OP_NOTI:
1830      branchlength++;      branchlength++;
1831      cc += 2;      cc += 2;
1832  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1833      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834  #endif  #endif
1835      break;      break;
# Line 1866  for (;;) Line 1843  for (;;)
1843      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1844      branchlength += (int)GET2(cc,1);      branchlength += (int)GET2(cc,1);
1845      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1846  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1847      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848  #endif  #endif
1849      break;      break;
1850    
1851      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1852      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1853      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854        cc += 2;        cc += 2;
1855      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1856      break;      break;
# Line 1909  for (;;) Line 1886  for (;;)
1886    
1887      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1888    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1889      case OP_CLASS:      case OP_CLASS:
1890      case OP_NCLASS:      case OP_NCLASS:
1891    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892        case OP_XCLASS:
1893        /* The original code caused an unsigned overflow in 64 bit systems,
1894        so now we use a conditional statement. */
1895        if (op == OP_XCLASS)
1896          cc += GET(cc, 1);
1897        else
1898          cc += PRIV(OP_lengths)[OP_CLASS];
1899    #else
1900      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1901    #endif
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918          case OP_CRPOSRANGE:
1919        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1921        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 1999  for (;;) Line 1984  for (;;)
1984      case OP_QUERYI:      case OP_QUERYI:
1985      case OP_REF:      case OP_REF:
1986      case OP_REFI:      case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989      case OP_SBRA:      case OP_SBRA:
1990      case OP_SBRAPOS:      case OP_SBRAPOS:
1991      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 2112  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102          code += 2;          code += 2;
2103        break;        break;
2104    
2105        case OP_MARK:        case OP_MARK:
2106        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2108        case OP_THEN_ARG:        case OP_THEN_ARG:
2109        code += code[1];        code += code[1];
2110        break;        break;
# Line 2232  for (;;) Line 2215  for (;;)
2215        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2216        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2217        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2218        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219          code += 2;          code += 2;
2220        break;        break;
2221    
2222        case OP_MARK:        case OP_MARK:
2223        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2224        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2225          case OP_THEN_ARG:
2226        code += code[1];        code += code[1];
2227        break;        break;
2228          }
       case OP_THEN_ARG:  
       code += code[1];  
       break;  
       }  
2229    
2230      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2231    
# Line 2343  Arguments: Line 2323  Arguments:
2323    endcode     points to where to stop    endcode     points to where to stop
2324    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2326      recurses    chain of recurse_check to catch mutual recursion
2327    
2328  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2329  */  */
2330    
2331    typedef struct recurse_check {
2332      struct recurse_check *prev;
2333      const pcre_uchar *group;
2334    } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2339  {  {
2340  register pcre_uchar c;  register pcre_uchar c;
2341    recurse_check this_recurse;
2342    
2343  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2379  for (code = first_significant_code(code Line 2367  for (code = first_significant_code(code
2367    
2368    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2369      {      {
2370      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371      BOOL empty_branch;      BOOL empty_branch;
2372    
2373      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2374        when called to scan a completed pattern by setting cd->start_workspace to
2375        NULL. */
2376    
2377        if (cd->start_workspace != NULL)
2378          {
2379          const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383          }
2384    
2385        /* If we are scanning a completed pattern, there are no forward references
2386        and all groups are complete. We need to detect whether this is a recursive
2387        call, as otherwise there will be an infinite loop. If it is a recursion,
2388        just skip over it. Simple recursions are easily detected. For mutual
2389        recursions we keep a chain on the stack. */
2390    
2391        else
2392          {
2393          recurse_check *r = recurses;
2394          const pcre_uchar *endgroup = scode;
2395    
2396      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397        if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399      /* Not a forward reference, test for completed backward reference */        for (r = recurses; r != NULL; r = r->prev)
2400            if (r->group == scode) break;
2401          if (r != NULL) continue;   /* Mutual recursion */
2402          }
2403    
2404      empty_branch = FALSE;      /* Completed reference; scan the referenced group, remembering it on the
2405      scode = cd->start_code + GET(code, 1);      stack chain to detect mutual recursions. */
     if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */  
2406    
2407      /* Completed backwards reference */      empty_branch = FALSE;
2408        this_recurse.prev = recurses;
2409        this_recurse.group = scode;
2410    
2411      do      do
2412        {        {
2413        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414          {          {
2415          empty_branch = TRUE;          empty_branch = TRUE;
2416          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2466  for (code = first_significant_code(code
2466        empty_branch = FALSE;        empty_branch = FALSE;
2467        do        do
2468          {          {
2469          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470            empty_branch = TRUE;            empty_branch = TRUE;
2471          code += GET(code, 1);          code += GET(code, 1);
2472          }          }
# Line 2495  for (code = first_significant_code(code Line 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 2511  for (code = first_significant_code(code Line 2528  for (code = first_significant_code(code
2528    
2529      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      case OP_NOTPROP:
2537        case OP_ANYNL:
2538    
2539        case OP_NOT_HSPACE:
2540        case OP_HSPACE:
2541        case OP_NOT_VSPACE:
2542        case OP_VSPACE:
2543      case OP_EXTUNI:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARI:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555      case OP_NOTI:      case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558        case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
2560      case OP_POSPLUS:      case OP_MINPLUSI:
2561      case OP_EXACT:  
2562      case OP_NOTPLUS:      case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2558  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2599      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2600      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2601      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602        code += 2;        code += 2;
2603      break;      break;
2604    
# Line 2572  for (code = first_significant_code(code Line 2612  for (code = first_significant_code(code
2612      return TRUE;      return TRUE;
2613    
2614      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619      case OP_STAR:      case OP_STAR:
2620      case OP_STARI:      case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624      case OP_MINSTAR:      case OP_MINSTAR:
2625      case OP_MINSTARI:      case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629      case OP_POSSTAR:      case OP_POSSTAR:
2630      case OP_POSSTARI:      case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634      case OP_QUERY:      case OP_QUERY:
2635      case OP_QUERYI:      case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639      case OP_MINQUERY:      case OP_MINQUERY:
2640      case OP_MINQUERYI:      case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644      case OP_POSQUERY:      case OP_POSQUERY:
2645      case OP_POSQUERYI:      case OP_POSQUERYI:
2646        case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650      break;      break;
2651    
2652      case OP_UPTO:      case OP_UPTO:
2653      case OP_UPTOI:      case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657      case OP_MINUPTO:      case OP_MINUPTO:
2658      case OP_MINUPTOI:      case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662      case OP_POSUPTO:      case OP_POSUPTO:
2663      case OP_POSUPTOI:      case OP_POSUPTOI:
2664        case OP_NOTPOSUPTO:
2665        case OP_NOTPOSUPTOI:
2666    
2667      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668      break;      break;
2669  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2674  for (code = first_significant_code(code
2674      case OP_MARK:      case OP_MARK:
2675      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2676      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2677      case OP_THEN_ARG:      case OP_THEN_ARG:
2678      code += code[1];      code += code[1];
2679      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2717  could_be_empty(const pcre_uchar *code, c
2717  {  {
2718  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2719    {    {
2720    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721      return FALSE;      return FALSE;
2722    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2723    }    }
# Line 2662  return TRUE; Line 2727  return TRUE;
2727    
2728    
2729  /*************************************************  /*************************************************
2730  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2731  *************************************************/  *************************************************/
2732    
2733  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2734  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2735    
2736  Returns:   TRUE or FALSE  Arguments:  c opcode
2737    Returns:    base opcode for the type
2738  */  */
2739    
2740  static BOOL  static pcre_uchar
2741  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2742  {  {
2743  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2744  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2748      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2749  }  }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754  /*************************************************  /*************************************************
2755  *          Check POSIX class name                *  *        Check a character and a property        *
2756  *************************************************/  *************************************************/
2757    
2758  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2759  such as [:alnum:].  is adjacent to a fixed character.
2760    
2761  Arguments:  Arguments:
2762    ptr        points to the first letter    c            the character
2763    len        the length of the name    ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2768  */  */
2769    
2770  static int  static BOOL
2771  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773  {  {
2774  const char *pn = posix_names;  const pcre_uint32 *p;
2775  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2776    
2777  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2778  that is referenced. This means that groups can be replicated for fixed    {
2779  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2780  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2781  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2782  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2783    
2784  This function has been extended with the possibility of forward references for    case PT_GC:
2785  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2786    
2787  Arguments:    case PT_PC:
2788    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2789    
2790  Returns:     nothing    case PT_SC:
2791  */    return (pdata == prop->script) == negated;
2792    
2793  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2794    
2795  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2796    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2798    
2799    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806      {      {
2807      offset = GET(hc, 0);      HSPACE_CASES:
2808      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2809        {      return negated;
2810        PUT(hc, 0, offset + adjust);  
2811        break;      default:
2812        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813      }      }
2814      break;  /* Control never reaches here */
2815    
2816    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2817    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821    if (hc >= cd->hwm)    case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824      {      {
2825      offset = GET(ptr, 1);      if (c < *p) return !negated;
2826      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2827      }      }
2828      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2829    }    }
2830    
2831    return FALSE;
2832  }  }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837  /*************************************************  /*************************************************
2838  *        Insert an automatic callout point       *  *        Fill the character property list        *
2839  *************************************************/  *************************************************/
2840    
2841  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2842  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2843    
2844  Arguments:  Arguments:
2845    code           current code pointer    code        points to start of expression
2846    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847    cd             pointers to tables etc    fcc         points to case-flipping table
2848      list        points to output list
2849  Returns:         new code pointer                list[0] will be filled with the opcode
2850  */                list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852  static pcre_uchar *                list[2..7] depends on the opcode
 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  
 {  
 *code++ = OP_CALLOUT;  
 *code++ = 255;  
 PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  
 PUT(code, LINK_SIZE, 0);                       /* Default length */  
 return code + 2 * LINK_SIZE;  
 }  
   
   
   
 /*************************************************  
 *         Complete a callout item                *  
 *************************************************/  
   
 /* A callout item contains the length of the next item in the pattern, which  
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
   
 Arguments:  
   previous_callout   points to previous callout item  
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
2853    
2854  Returns:             nothing  Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856  */  */
2857    
2858  static void  static const pcre_uchar *
2859  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861  {  {
2862  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  pcre_uchar c = *code;
2863  PUT(previous_callout, 2 + LINK_SIZE, length);  const pcre_uchar *end;
2864  }  const pcre_uint32 *clist_src;
2865    pcre_uint32 *clist_dest;
2866    pcre_uint32 chr;
2867    pcre_uchar base;
2868    
2869    list[0] = c;
2870    list[1] = FALSE;
2871    code++;
2872    
2873    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2874      {
2875      base = get_repeat_base(c);
2876      c -= (base - OP_STAR);
2877    
2878  #ifdef SUPPORT_UCP    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2879  /*************************************************      code += IMM2_SIZE;
 *           Get othercase range                  *  
 *************************************************/  
2880    
2881  /* This function is passed the start and end of a class range, in UTF-8 mode    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 with UCP support. It searches up the characters, looking for ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
2882    
2883  Arguments:    switch(base)
2884    cptr        points to starting character value; updated      {
2885    d           end value      case OP_STAR:
2886    ocptr       where to put start of othercase range      list[0] = OP_CHAR;
2887    odptr       where to put end of othercase range      break;
2888    
2889  Yield:        -1 when no more      case OP_STARI:
2890                 0 when a range is returned      list[0] = OP_CHARI;
2891                >0 the CASESET offset for char with multiple other cases      break;
                 in this case, ocptr contains the original  
 */  
2892    
2893  static int      case OP_NOTSTAR:
2894  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      list[0] = OP_NOT;
2895    pcre_uint32 *odptr)      break;
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2896    
2897  /* Find the first character that has an other case. If it has multiple other      case OP_NOTSTARI:
2898  cases, return its case offset value. */      list[0] = OP_NOTI;
2899        break;
2900    
2901  for (c = *cptr; c <= d; c++)      case OP_TYPESTAR:
2902    {      list[0] = *code;
2903    if ((co = UCD_CASESET(c)) != 0)      code++;
2904      {      break;
2905      *ocptr = c++;   /* Character that has the set */      }
2906      *cptr = c;      /* Rest of input range */    c = list[0];
     return co;  
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
2907    }    }
2908    
2909  if (c > d) return -1;  /* Reached end of range */  switch(c)
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
2910    {    {
2911    if (UCD_OTHERCASE(c) != next) break;    case OP_NOT_DIGIT:
2912    next++;    case OP_DIGIT:
2913    }    case OP_NOT_WHITESPACE:
2914      case OP_WHITESPACE:
2915  *odptr = next - 1;     /* End of othercase range */    case OP_NOT_WORDCHAR:
2916  *cptr = c;             /* Rest of input range */    case OP_WORDCHAR:
2917  return 0;    case OP_ANY:
2918  }    case OP_ALLANY:
2919      case OP_ANYNL:
2920      case OP_NOT_HSPACE:
2921      case OP_HSPACE:
2922  /*************************************************    case OP_NOT_VSPACE:
2923  *        Check a character and a property        *    case OP_VSPACE:
2924  *************************************************/    case OP_EXTUNI:
2925      case OP_EODN:
2926  /* This function is called by check_auto_possessive() when a property item    case OP_EOD:
2927  is adjacent to a fixed character.    case OP_DOLL:
2928      case OP_DOLLM:
2929      return code;
2930    
2931  Arguments:    case OP_CHAR:
2932    c            the character    case OP_NOT:
2933    ptype        the property type    GETCHARINCTEST(chr, code);
2934    pdata        the data for the type    list[2] = chr;
2935    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2936      return code;
2937    
2938  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2939  */    case OP_NOTI:
2940      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2941      GETCHARINCTEST(chr, code);
2942      list[2] = chr;
2943    
 static BOOL  
 check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  
 {  
2944  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2945  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2946        list[3] = fcc[chr];
2947      else
2948        list[3] = UCD_OTHERCASE(chr);
2949    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2950      list[3] = (chr < 256) ? fcc[chr] : chr;
2951    #else
2952      list[3] = fcc[chr];
2953  #endif  #endif
2954    
2955  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
2956    
2957  switch(ptype)    if (chr == list[3])
2958    {      list[3] = NOTACHAR;
2959    case PT_LAMP:    else
2960    return (prop->chartype == ucp_Lu ||      list[4] = NOTACHAR;
2961            prop->chartype == ucp_Ll ||    return code;
           prop->chartype == ucp_Lt) == negated;  
2962    
2963    case PT_GC:  #ifdef SUPPORT_UCP
2964    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;    case OP_PROP:
2965      case OP_NOTPROP:
2966      if (code[0] != PT_CLIST)
2967        {
2968        list[2] = code[0];
2969        list[3] = code[1];
2970        return code + 2;
2971        }
2972    
2973    case PT_PC:    /* Convert only if we have enough space. */
   return (pdata == prop->chartype) == negated;  
2974    
2975    case PT_SC:    clist_src = PRIV(ucd_caseless_sets) + code[1];
2976    return (pdata == prop->script) == negated;    clist_dest = list + 2;
2977      code += 2;
2978    
2979    /* These are specials */    do {
2980         if (clist_dest >= list + 8)
2981           {
2982           /* Early return if there is not enough space. This should never
2983           happen, since all clists are shorter than 5 character now. */
2984           list[2] = code[0];
2985           list[3] = code[1];
2986           return code;
2987           }
2988         *clist_dest++ = *clist_src;
2989         }
2990      while(*clist_src++ != NOTACHAR);
2991    
2992    case PT_ALNUM:    /* All characters are stored. The terminating NOTACHAR
2993    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    is copied form the clist itself. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
2994    
2995    case PT_SPACE:    /* Perl space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2996    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
2997            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  #endif
           == negated;  
2998    
2999    case PT_PXSPACE:  /* POSIX space */    case OP_NCLASS:
3000    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    case OP_CLASS:
3001            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3002            c == CHAR_FF || c == CHAR_CR)    case OP_XCLASS:
3003            == negated;    if (c == OP_XCLASS)
3004        end = code + GET(code, 0) - 1;
3005      else
3006    #endif
3007        end = code + 32 / sizeof(pcre_uchar);
3008    
3009    case PT_WORD:    switch(*end)
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   
 #ifdef SUPPORT_UCP  
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3010      {      {
3011      if ((unsigned int)c < *p) return !negated;      case OP_CRSTAR:
3012      if ((unsigned int)c == *p++) return negated;      case OP_CRMINSTAR:
3013        case OP_CRQUERY:
3014        case OP_CRMINQUERY:
3015        case OP_CRPOSSTAR:
3016        case OP_CRPOSQUERY:
3017        list[1] = TRUE;
3018        end++;
3019        break;
3020    
3021        case OP_CRPLUS:
3022        case OP_CRMINPLUS:
3023        case OP_CRPOSPLUS:
3024        end++;
3025        break;
3026    
3027        case OP_CRRANGE:
3028        case OP_CRMINRANGE:
3029        case OP_CRPOSRANGE:
3030        list[1] = (GET2(end, 1) == 0);
3031        end += 1 + 2 * IMM2_SIZE;
3032        break;
3033      }      }
3034    break;  /* Control never reaches here */    list[2] = end - code;
3035  #endif    return end;
3036    }    }
3037    return NULL;    /* Opcode not accepted */
 return FALSE;  
3038  }  }
 #endif  /* SUPPORT_UCP */  
3039    
3040    
3041    
3042  /*************************************************  /*************************************************
3043  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3044  *************************************************/  *************************************************/
3045    
3046  /* 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
3047  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.  
3048    
3049  Arguments:  Arguments:
3050    previous      pointer to the repeated opcode    code        points to the byte code
3051    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3052    ptr           next character in pattern    cd          static compile data
3053    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3054    
3055  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3056  */  */
3057    
3058  static BOOL  static BOOL
3059  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3060    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3061  {  {
3062  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3063  pcre_uint32 next;  pcre_uint32 list[8];
3064  int escape;  const pcre_uint32 *chr_ptr;
3065  int op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3066    const pcre_uint32 *list_ptr;
3067  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3068    const pcre_uint8 *class_bitset;
3069    const pcre_uint32 *set1, *set2, *set_end;
3070    pcre_uint32 chr;
3071    BOOL accepted, invert_bits;
3072    
3073    /* Note: the base_list[1] contains whether the current opcode has greedy
3074    (represented by a non-zero value) quantifier. This is a different from
3075    other character type lists, which stores here that the character iterator
3076    matches to an empty string (also represented by a non-zero value). */
3077    
3078  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3079    {    {
3080    for (;;)    /* All operations move the code pointer forward.
3081      Therefore infinite recursions are not possible. */
3082    
3083      c = *code;
3084    
3085      /* Skip over callouts */
3086    
3087      if (c == OP_CALLOUT)
3088      {      {
3089      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3090      if (*ptr == CHAR_NUMBER_SIGN)      continue;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3091      }      }
   }  
3092    
3093  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3094  value is a character, a negative value is an escape value. */      {
3095        do code += GET(code, 1); while (*code == OP_ALT);
3096        c = *code;
3097        }
3098    
3099  if (*ptr == CHAR_BACKSLASH)    switch(c)
3100    {      {
3101    int temperrorcode = 0;      case OP_END:
3102    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);      case OP_KETRPOS:
3103    if (temperrorcode != 0) return FALSE;      /* TRUE only in greedy case. The non-greedy case could be replaced by
3104    ptr++;    /* Point after the escape sequence */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3105    }      uses more memory, which we cannot get at this stage.) */
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3106    
3107  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3108    
3109  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3110    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3111    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3112      {      cannot be converted to a possessive form. */
3113      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3114      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[1] == 0) return FALSE;
3115    
3116        switch(*(code - GET(code, 1)))
3117          {
3118          case OP_ASSERT:
3119          case OP_ASSERT_NOT:
3120          case OP_ASSERTBACK:
3121          case OP_ASSERTBACK_NOT:
3122          case OP_ONCE:
3123          case OP_ONCE_NC:
3124          /* Atomic sub-patterns and assertions can always auto-possessify their
3125          last iterator. */
3126          return TRUE;
3127          }
3128    
3129        code += PRIV(OP_lengths)[c];
3130        continue;
3131    
3132        case OP_ONCE:
3133        case OP_ONCE_NC:
3134        case OP_BRA:
3135        case OP_CBRA:
3136        next_code = code + GET(code, 1);
3137        code += PRIV(OP_lengths)[c];
3138    
3139        while (*next_code == OP_ALT)
3140        {        {
3141        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3142        while (*ptr != 0)        code = next_code + 1 + LINK_SIZE;
3143          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3144        }        }
3145      else break;      continue;
     }  
   }  
3146    
3147  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3148        case OP_BRAMINZERO:
3149    
3150  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3151    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3152      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3153    
3154        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3155    
3156        /* The bracket content will be checked by the
3157        OP_BRA/OP_CBRA case above. */
3158        next_code += 1 + LINK_SIZE;
3159        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3160          return FALSE;
3161    
3162  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3163        continue;
3164        }
3165    
3166  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)  
   //if (escape == 0) switch(op_code)  
   {  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3167    
3168  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3169  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3170    
3171  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3172    {    characters from that list with another list, or with a property. */
3173    /* For a caseless UTF match, the next character may have more than one other  
3174    case, which maps to the special PT_CLIST property. Check this first. */    if (base_list[0] == OP_CHAR)
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3175      {      {
3176      int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3177      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3178      }      }
3179  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3180      {      {
3181      case OP_CHAR:      chr_ptr = list + 2;
3182      return c != next;      list_ptr = base_list;
3183        }
3184    
3185      /* 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. */  
3186    
3187      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3188      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3189  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3190      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3191        {  #endif
3192        pcre_uint32 othercase;        )
3193        if (next < 128) othercase = cd->fcc[next]; else      {
3194  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3195        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3196  #else  #else
3197        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3198  #endif  #endif
3199        return c != othercase;        {
3200          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3201          list_ptr = list;
3202        }        }
3203      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)  
3204        {        {
3205        pcre_uint32 othercase;        set1 = (pcre_uint32 *)(code - list[2]);
3206        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;  
3207        }        }
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3208    
3209      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      invert_bits = FALSE;
3210      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      switch(list_ptr[0])
3211          {
3212          case OP_CLASS:
3213          case OP_NCLASS:
3214          set2 = (pcre_uint32 *)
3215            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3216          break;
3217    
3218      case OP_DIGIT:        /* OP_XCLASS cannot be supported here, because its bitset
3219      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        is not necessarily complete. E.g: [a-\0x{200}] is stored
3220          as a character range, and the appropriate bits are not set. */
3221    
3222      case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
3223      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;          invert_bits = TRUE;
3224            /* Fall through */
3225          case OP_DIGIT:
3226            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3227            break;
3228    
3229      case OP_WHITESPACE:        case OP_NOT_WHITESPACE:
3230      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;          invert_bits = TRUE;
3231            /* Fall through */
3232          case OP_WHITESPACE:
3233            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3234            break;
3235    
3236      case OP_NOT_WHITESPACE:        case OP_NOT_WORDCHAR:
3237      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;          invert_bits = TRUE;
3238            /* Fall through */
3239          case OP_WORDCHAR:
3240            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3241            break;
3242    
3243      case OP_WORDCHAR:        default:
3244      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        return FALSE;
3245          }
3246    
3247      case OP_NOT_WORDCHAR:      /* Compare 4 bytes to improve speed. */
3248      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;      set_end = set1 + (32 / 4);
3249        if (invert_bits)
3250          {
3251          do
3252            {
3253            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3254            }
3255          while (set1 < set_end);
3256          }
3257        else
3258          {
3259          do
3260            {
3261            if ((*set1++ & *set2++) != 0) return FALSE;
3262            }
3263          while (set1 < set_end);
3264          }
3265    
3266      case OP_HSPACE:      if (list[1] == 0) return TRUE;
3267      case OP_NOT_HSPACE:      /* Might be an empty repeat. */
3268      switch(next)      continue;
3269        }
3270    
3271      /* Some property combinations also acceptable. Unicode property opcodes are
3272      processed specially; the rest can be handled with a lookup table. */
3273    
3274      else
3275        {
3276        pcre_uint32 leftop, rightop;
3277    
3278        leftop = base_list[0];
3279        rightop = list[0];
3280    
3281    #ifdef SUPPORT_UCP
3282        accepted = FALSE; /* Always set in non-unicode case. */
3283        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3284        {        {
3285        HSPACE_CASES:        if (rightop == OP_EOD)
3286        return op_code == OP_NOT_HSPACE;          accepted = TRUE;
3287          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3288            {
3289            int n;
3290            const pcre_uint8 *p;
3291            BOOL same = leftop == rightop;
3292            BOOL lisprop = leftop == OP_PROP;
3293            BOOL risprop = rightop == OP_PROP;
3294            BOOL bothprop = lisprop && risprop;
3295    
3296            /* There's a table that specifies how each combination is to be
3297            processed:
3298              0   Always return FALSE (never auto-possessify)
3299              1   Character groups are distinct (possessify if both are OP_PROP)
3300              2   Check character categories in the same group (general or particular)
3301              3   Return TRUE if the two opcodes are not the same
3302              ... see comments below
3303            */
3304    
3305            n = propposstab[base_list[2]][list[2]];
3306            switch(n)
3307              {
3308              case 0: break;
3309              case 1: accepted = bothprop; break;
3310              case 2: accepted = (base_list[3] == list[3]) != same; break;
3311              case 3: accepted = !same; break;
3312    
3313        default:            case 4:  /* Left general category, right particular category */
3314        return op_code != OP_NOT_HSPACE;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3315              break;
3316    
3317              case 5:  /* Right general category, left particular category */
3318              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3319              break;
3320    
3321              /* This code is logically tricky. Think hard before fiddling with it.
3322              The posspropstab table has four entries per row. Each row relates to
3323              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3324              Only WORD actually needs all four entries, but using repeats for the
3325              others means they can all use the same code below.
3326    
3327              The first two entries in each row are Unicode general categories, and
3328              apply always, because all the characters they include are part of the
3329              PCRE character set. The third and fourth entries are a general and a
3330              particular category, respectively, that include one or more relevant
3331              characters. One or the other is used, depending on whether the check
3332              is for a general or a particular category. However, in both cases the
3333              category contains more characters than the specials that are defined
3334              for the property being tested against. Therefore, it cannot be used
3335              in a NOTPROP case.
3336    
3337              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3338              Underscore is covered by ucp_P or ucp_Po. */
3339    
3340              case 6:  /* Left alphanum vs right general category */
3341              case 7:  /* Left space vs right general category */
3342              case 8:  /* Left word vs right general category */
3343              p = posspropstab[n-6];
3344              accepted = risprop && lisprop ==
3345                (list[3] != p[0] &&
3346                 list[3] != p[1] &&
3347                (list[3] != p[2] || !lisprop));
3348              break;
3349    
3350              case 9:   /* Right alphanum vs left general category */
3351              case 10:  /* Right space vs left general category */
3352              case 11:  /* Right word vs left general category */
3353              p = posspropstab[n-9];
3354              accepted = lisprop && risprop ==
3355                (base_list[3] != p[0] &&
3356                 base_list[3] != p[1] &&
3357                (base_list[3] != p[2] || !risprop));
3358              break;
3359    
3360              case 12:  /* Left alphanum vs right particular category */
3361              case 13:  /* Left space vs right particular category */
3362              case 14:  /* Left word vs right particular category */
3363              p = posspropstab[n-12];
3364              accepted = risprop && lisprop ==
3365                (catposstab[p[0]][list[3]] &&
3366                 catposstab[p[1]][list[3]] &&
3367                (list[3] != p[3] || !lisprop));
3368              break;
3369    
3370              case 15:  /* Right alphanum vs left particular category */
3371              case 16:  /* Right space vs left particular category */
3372              case 17:  /* Right word vs left particular category */
3373              p = posspropstab[n-15];
3374              accepted = lisprop && risprop ==
3375                (catposstab[p[0]][base_list[3]] &&
3376                 catposstab[p[1]][base_list[3]] &&
3377                (base_list[3] != p[3] || !risprop));
3378              break;
3379              }
3380            }
3381        }        }
3382    
3383      case OP_ANYNL:      else
3384      case OP_VSPACE:  #endif  /* SUPPORT_UCP */
3385      case OP_NOT_VSPACE:  
3386      switch(next)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3387               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3388               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3389    
3390        if (!accepted)
3391          return FALSE;
3392    
3393        if (list[1] == 0) return TRUE;
3394        /* Might be an empty repeat. */
3395        continue;
3396        }
3397    
3398      /* Control reaches here only if one of the items is a small character list.
3399      All characters are checked against the other side. */
3400    
3401      do
3402        {
3403        chr = *chr_ptr;
3404    
3405        switch(list_ptr[0])
3406        {        {
3407        VSPACE_CASES:        case OP_CHAR:
3408        return op_code == OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3409          do
3410            {
3411            if (chr == *ochr_ptr) return FALSE;
3412            ochr_ptr++;
3413            }
3414          while(*ochr_ptr != NOTACHAR);
3415          break;
3416    
3417        default:        case OP_NOT:
3418        return op_code != OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3419        }        do
3420            {
3421            if (chr == *ochr_ptr)
3422              break;
3423            ochr_ptr++;
3424            }
3425          while(*ochr_ptr != NOTACHAR);
3426          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3427          break;
3428    
3429          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3430          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3431    
3432          case OP_DIGIT:
3433          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3434          break;
3435    
3436          case OP_NOT_DIGIT:
3437          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3438          break;
3439    
3440          case OP_WHITESPACE:
3441          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3442          break;
3443    
3444          case OP_NOT_WHITESPACE:
3445          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3446          break;
3447    
3448          case OP_WORDCHAR:
3449          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3450          break;
3451    
3452          case OP_NOT_WORDCHAR:
3453          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3454          break;
3455    
3456          case OP_HSPACE:
3457          switch(chr)
3458            {
3459            HSPACE_CASES: return FALSE;
3460            default: break;
3461            }
3462          break;
3463    
3464          case OP_NOT_HSPACE:
3465          switch(chr)
3466            {
3467            HSPACE_CASES: break;
3468            default: return FALSE;
3469            }
3470          break;
3471    
3472          case OP_ANYNL:
3473          case OP_VSPACE:
3474          switch(chr)
3475            {
3476            VSPACE_CASES: return FALSE;
3477            default: break;
3478            }
3479          break;
3480    
3481          case OP_NOT_VSPACE:
3482          switch(chr)
3483            {
3484            VSPACE_CASES: break;
3485            default: return FALSE;
3486            }
3487          break;
3488    
3489          case OP_DOLL:
3490          case OP_EODN:
3491          switch (chr)
3492            {
3493            case CHAR_CR:
3494            case CHAR_LF:
3495            case CHAR_VT:
3496            case CHAR_FF:
3497            case CHAR_NEL:
3498    #ifndef EBCDIC
3499            case 0x2028:
3500            case 0x2029:
3501    #endif  /* Not EBCDIC */
3502            return FALSE;
3503            }
3504          break;
3505    
3506          case OP_EOD:    /* Can always possessify before \z */
3507          break;
3508    
3509  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3510      case OP_PROP:        case OP_PROP:
3511      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3512          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3513                list_ptr[0] == OP_NOTPROP))
3514            return FALSE;
3515          break;
3516    #endif
3517    
3518      case OP_NOTPROP:        case OP_NCLASS:
3519      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3520          /* Fall through */
3521    
3522          case OP_CLASS:
3523          if (chr > 255) break;
3524          class_bitset = (pcre_uint8 *)
3525            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3526          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3527          break;
3528    
3529    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3530          case OP_XCLASS:
3531          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3532              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3533          break;
3534  #endif  #endif
3535    
3536      default:        default:
3537      return FALSE;        return FALSE;
3538          }
3539    
3540        chr_ptr++;
3541      }      }
3542      while(*chr_ptr != NOTACHAR);
3543    
3544      /* At least one character must be matched from this opcode. */
3545    
3546      if (list[1] == 0) return TRUE;
3547    }    }
3548    
3549  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  return FALSE;
3550  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  }
3551  generated only when PCRE_UCP is *not* set, that is, when only ASCII  
3552  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
3553  replaced by OP_PROP codes when PCRE_UCP is set. */  
3554    /*************************************************
3555    *    Scan compiled regex for auto-possession     *
3556    *************************************************/
3557    
3558    /* Replaces single character iterations with their possessive alternatives
3559    if appropriate. This function modifies the compiled opcode!
3560    
3561    Arguments:
3562      code        points to start of the byte code
3563      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3564      cd          static compile data
3565    
3566    Returns:      nothing
3567    */
3568    
3569    static void
3570    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3571    {
3572    register pcre_uchar c;
3573    const pcre_uchar *end;
3574    pcre_uchar *repeat_opcode;
3575    pcre_uint32 list[8];
3576    
3577  switch(op_code)  for (;;)
3578    {    {
3579    case OP_CHAR:    c = *code;
3580    case OP_CHARI:  
3581    switch(escape)    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3582      {      {
3583      case ESC_d:      c -= get_repeat_base(c) - OP_STAR;
3584      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      end = (c <= OP_MINUPTO) ?
3585          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3586        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3587    
3588      case ESC_D:      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3589      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;        {
3590          switch(c)
3591            {
3592            case OP_STAR:
3593            *code += OP_POSSTAR - OP_STAR;
3594            break;
3595    
3596      case ESC_s:          case OP_MINSTAR:
3597      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;          *code += OP_POSSTAR - OP_MINSTAR;
3598            break;
3599    
3600      case ESC_S:          case OP_PLUS:
3601      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;          *code += OP_POSPLUS - OP_PLUS;
3602            break;
3603    
3604      case ESC_w:          case OP_MINPLUS:
3605      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;          *code += OP_POSPLUS - OP_MINPLUS;
3606            break;
3607    
3608      case ESC_W:          case OP_QUERY:
3609      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;          *code += OP_POSQUERY - OP_QUERY;
3610            break;
3611    
3612      case ESC_h:          case OP_MINQUERY:
3613      case ESC_H:          *code += OP_POSQUERY - OP_MINQUERY;
3614      switch(c)          break;
3615        {  
3616        HSPACE_CASES:          case OP_UPTO:
3617        return escape != ESC_h;          *code += OP_POSUPTO - OP_UPTO;
3618            break;
3619        default:  
3620        return escape == ESC_h;          case OP_MINUPTO:
3621            *code += OP_MINUPTO - OP_UPTO;
3622            break;
3623            }
3624        }        }
3625        c = *code;
3626        }
3627      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3628        {
3629    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3630        if (c == OP_XCLASS)
3631          repeat_opcode = code + GET(code, 1);
3632        else
3633    #endif
3634          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3635    
3636      case ESC_v:      c = *repeat_opcode;
3637      case ESC_V:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
     switch(c)  
3638        {        {
3639        VSPACE_CASES:        /* end must not be NULL. */
3640        return escape != ESC_v;        end = get_chr_property_list(code, utf, cd->fcc, list);
3641    
3642        default:        list[1] = (c & 1) == 0;
3643        return escape == ESC_v;  
3644        }        if (compare_opcodes(end, utf, cd, list, end))
3645            {
3646            switch (c)
3647              {
3648              case OP_CRSTAR:
3649              case OP_CRMINSTAR:
3650              *repeat_opcode = OP_CRPOSSTAR;
3651              break;
3652    
3653      /* When PCRE_UCP is set, these values get generated for \d etc. Find            case OP_CRPLUS:
3654      their substitutions and process them. The result will always be either            case OP_CRMINPLUS:
3655      ESC_p or ESC_P. Then fall through to process those values. */            *repeat_opcode = OP_CRPOSPLUS;
3656              break;
3657    
3658  #ifdef SUPPORT_UCP            case OP_CRQUERY:
3659      case ESC_du:            case OP_CRMINQUERY:
3660      case ESC_DU:            *repeat_opcode = OP_CRPOSQUERY;
3661      case ESC_wu:            break;
3662      case ESC_WU:  
3663      case ESC_su:            case OP_CRRANGE:
3664      case ESC_SU:            case OP_CRMINRANGE:
3665        {            *repeat_opcode = OP_CRPOSRANGE;
3666        int temperrorcode = 0;            break;
3667        ptr = substitutes[escape - ESC_DU];            }
3668        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);          }
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
3669        }        }
3670      /* Fall through */      c = *code;
3671        }
3672    
3673      case ESC_p:    switch(c)
3674      case ESC_P:      {
3675        {      case OP_END:
3676        int ptype, pdata, errorcodeptr;      return;
3677        BOOL negated;  
3678        case OP_TYPESTAR:
3679        case OP_TYPEMINSTAR:
3680        case OP_TYPEPLUS:
3681        case OP_TYPEMINPLUS:
3682        case OP_TYPEQUERY:
3683        case OP_TYPEMINQUERY:
3684        case OP_TYPEPOSSTAR:
3685        case OP_TYPEPOSPLUS:
3686        case OP_TYPEPOSQUERY:
3687        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3688        break;
3689    
3690        case OP_TYPEUPTO:
3691        case OP_TYPEMINUPTO:
3692        case OP_TYPEEXACT:
3693        case OP_TYPEPOSUPTO:
3694        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3695          code += 2;
3696        break;
3697    
3698    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3699        case OP_XCLASS:
3700        code += GET(code, 1);
3701        break;
3702    #endif
3703    
3704        case OP_MARK:
3705        case OP_PRUNE_ARG:
3706        case OP_SKIP_ARG:
3707        case OP_THEN_ARG:
3708        code += code[1];
3709        break;
3710        }
3711    
3712      /* Add in the fixed length from the table */
3713    
3714        ptr--;      /* Make ptr point at the p or P */    code += PRIV(OP_lengths)[c];
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3715    
3716        /* If the property item is optional, we have to give up. (When generated    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3717        from \d etc by PCRE_UCP, this test will have been applied much earlier,    a multi-byte character. The length in the table is a minimum, so we have to
3718        to the original \d etc. At this point, ptr will point to a zero byte. */    arrange to skip the extra bytes. */
3719    
3720        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3721          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    if (utf) switch(c)
3722            return FALSE;      {
3723        case OP_CHAR:
3724        case OP_CHARI:
3725        case OP_NOT:
3726        case OP_NOTI:
3727        case OP_STAR:
3728        case OP_MINSTAR:
3729        case OP_PLUS:
3730        case OP_MINPLUS:
3731        case OP_QUERY:
3732        case OP_MINQUERY:
3733        case OP_UPTO:
3734        case OP_MINUPTO:
3735        case OP_EXACT:
3736        case OP_POSSTAR:
3737        case OP_POSPLUS:
3738        case OP_POSQUERY:
3739        case OP_POSUPTO:
3740        case OP_STARI:
3741        case OP_MINSTARI:
3742        case OP_PLUSI:
3743        case OP_MINPLUSI:
3744        case OP_QUERYI:
3745        case OP_MINQUERYI:
3746        case OP_UPTOI:
3747        case OP_MINUPTOI:
3748        case OP_EXACTI:
3749        case OP_POSSTARI:
3750        case OP_POSPLUSI:
3751        case OP_POSQUERYI:
3752        case OP_POSUPTOI:
3753        case OP_NOTSTAR:
3754        case OP_NOTMINSTAR:
3755        case OP_NOTPLUS:
3756        case OP_NOTMINPLUS:
3757        case OP_NOTQUERY:
3758        case OP_NOTMINQUERY:
3759        case OP_NOTUPTO:
3760        case OP_NOTMINUPTO:
3761        case OP_NOTEXACT:
3762        case OP_NOTPOSSTAR:
3763        case OP_NOTPOSPLUS:
3764        case OP_NOTPOSQUERY:
3765        case OP_NOTPOSUPTO:
3766        case OP_NOTSTARI:
3767        case OP_NOTMINSTARI:
3768        case OP_NOTPLUSI:
3769        case OP_NOTMINPLUSI:
3770        case OP_NOTQUERYI:
3771        case OP_NOTMINQUERYI:
3772        case OP_NOTUPTOI:
3773        case OP_NOTMINUPTOI:
3774        case OP_NOTEXACTI:
3775        case OP_NOTPOSSTARI:
3776        case OP_NOTPOSPLUSI:
3777        case OP_NOTPOSQUERYI:
3778        case OP_NOTPOSUPTOI:
3779        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3780        break;
3781        }
3782    #else
3783      (void)(utf);  /* Keep compiler happy by referencing function argument */
3784    #endif
3785      }
3786    }
3787    
3788    
3789    
3790    /*************************************************
3791    *           Check for POSIX class syntax         *
3792    *************************************************/
3793    
3794    /* This function is called when the sequence "[:" or "[." or "[=" is
3795    encountered in a character class. It checks whether this is followed by a
3796    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3797    reach an unescaped ']' without the special preceding character, return FALSE.
3798    
3799    Originally, this function only recognized a sequence of letters between the
3800    terminators, but it seems that Perl recognizes any sequence of characters,
3801    though of course unknown POSIX names are subsequently rejected. Perl gives an
3802    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3803    didn't consider this to be a POSIX class. Likewise for [:1234:].
3804    
3805    The problem in trying to be exactly like Perl is in the handling of escapes. We
3806    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3807    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3808    below handles the special case of \], but does not try to do any other escape
3809    processing. This makes it different from Perl for cases such as [:l\ower:]
3810    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3811    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3812    I think.
3813    
3814    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3815    It seems that the appearance of a nested POSIX class supersedes an apparent
3816    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3817    a digit.
3818    
3819    In Perl, unescaped square brackets may also appear as part of class names. For
3820    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3821    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3822    seem right at all. PCRE does not allow closing square brackets in POSIX class
3823    names.
3824    
3825    Arguments:
3826      ptr      pointer to the initial [
3827      endptr   where to return the end pointer
3828    
3829    Returns:   TRUE or FALSE
3830    */
3831    
3832    static BOOL
3833    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3834    {
3835    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3836    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3837    for (++ptr; *ptr != CHAR_NULL; ptr++)
3838      {
3839      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3840        ptr++;
3841      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3842      else
3843        {
3844        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3845          {
3846          *endptr = ptr;
3847          return TRUE;
3848          }
3849        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3850             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3851              ptr[1] == CHAR_EQUALS_SIGN) &&
3852            check_posix_syntax(ptr, endptr))
3853          return FALSE;
3854        }
3855      }
3856    return FALSE;
3857    }
3858    
3859    
3860    
3861    
3862    /*************************************************
3863    *          Check POSIX class name                *
3864    *************************************************/
3865    
3866    /* This function is called to check the name given in a POSIX-style class entry
3867    such as [:alnum:].
3868    
3869    Arguments:
3870      ptr        points to the first letter
3871      len        the length of the name
3872    
3873    Returns:     a value representing the name, or -1 if unknown
3874    */
3875    
3876    static int
3877    check_posix_name(const pcre_uchar *ptr, int len)
3878    {
3879    const char *pn = posix_names;
3880    register int yield = 0;
3881    while (posix_name_lengths[yield] != 0)
3882      {
3883      if (len == posix_name_lengths[yield] &&
3884        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3885      pn += posix_name_lengths[yield] + 1;
3886      yield++;
3887      }
3888    return -1;
3889    }
3890    
3891    
3892    /*************************************************
3893    *    Adjust OP_RECURSE items in repeated group   *
3894    *************************************************/
3895    
3896    /* OP_RECURSE items contain an offset from the start of the regex to the group
3897    that is referenced. This means that groups can be replicated for fixed
3898    repetition simply by copying (because the recursion is allowed to refer to
3899    earlier groups that are outside the current group). However, when a group is
3900    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3901    inserted before it, after it has been compiled. This means that any OP_RECURSE
3902    items within it that refer to the group itself or any contained groups have to
3903    have their offsets adjusted. That one of the jobs of this function. Before it
3904    is called, the partially compiled regex must be temporarily terminated with
3905    OP_END.
3906    
3907    This function has been extended with the possibility of forward references for
3908    recursions and subroutine calls. It must also check the list of such references
3909    for the group we are dealing with. If it finds that one of the recursions in
3910    the current group is on this list, it adjusts the offset in the list, not the
3911    value in the reference (which is a group number).
3912    
3913    Arguments:
3914      group      points to the start of the group
3915      adjust     the amount by which the group is to be moved
3916      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3917      cd         contains pointers to tables etc.
3918      save_hwm   the hwm forward reference pointer at the start of the group
3919    
3920    Returns:     nothing
3921    */
3922    
3923    static void
3924    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3925      pcre_uchar *save_hwm)
3926    {
3927    pcre_uchar *ptr = group;
3928    
3929    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3930      {
3931      int offset;
3932      pcre_uchar *hc;
3933    
3934      /* See if this recursion is on the forward reference list. If so, adjust the
3935      reference. */
3936    
3937      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3938        {
3939        offset = (int)GET(hc, 0);
3940        if (cd->start_code + offset == ptr + 1)
3941          {
3942          PUT(hc, 0, offset + adjust);
3943          break;
3944          }
3945        }
3946    
3947      /* Otherwise, adjust the recursion offset if it's after the start of this
3948      group. */
3949    
3950      if (hc >= cd->hwm)
3951        {
3952        offset = (int)GET(ptr, 1);
3953        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3954        }
3955    
3956      ptr += 1 + LINK_SIZE;
3957      }
3958    }
3959    
3960    
3961    
3962    /*************************************************
3963    *        Insert an automatic callout point       *
3964    *************************************************/
3965    
3966    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3967    callout points before each pattern item.
3968    
3969    Arguments:
3970      code           current code pointer
3971      ptr            current pattern pointer
3972      cd             pointers to tables etc
3973    
3974    Returns:         new code pointer
3975    */
3976    
3977    static pcre_uchar *
3978    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3979    {
3980    *code++ = OP_CALLOUT;
3981    *code++ = 255;
3982    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3983    PUT(code, LINK_SIZE, 0);                       /* Default length */
3984    return code + 2 * LINK_SIZE;
3985    }
3986    
3987    
3988    
3989    /*************************************************
3990    *         Complete a callout item                *
3991    *************************************************/
3992    
3993    /* A callout item contains the length of the next item in the pattern, which
3994    we can't fill in till after we have reached the relevant point. This is used
3995    for both automatic and manual callouts.
3996    
3997        /* Do the property check. */  Arguments:
3998      previous_callout   points to previous callout item
3999      ptr                current pattern pointer
4000      cd                 pointers to tables etc
4001    
4002        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
4003        }  */
 #endif  
4004    
4005      default:  static void
4006      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4007      }  {
4008    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4009    PUT(previous_callout, 2 + LINK_SIZE, length);
4010    }
4011    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
4012    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
4013    
4014    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4015    return escape == ESC_d;  /*************************************************
4016    *           Get othercase range                  *
4017    *************************************************/
4018    
4019    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4020    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4021    characters in the "other" case. Each call returns the next one, updating the
4022    start address. A character with multiple other cases is returned on its own
4023    with a special return value.
4024    
4025    case OP_NOT_WHITESPACE:  Arguments:
4026    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
4027      d           end value
4028      ocptr       where to put start of othercase range
4029      odptr       where to put end of othercase range
4030    
4031    case OP_HSPACE:  Yield:        -1 when no more
4032    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
4033           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
4034                    in this case, ocptr contains the original
4035    */
4036    
4037    case OP_NOT_HSPACE:  static int
4038    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4039      pcre_uint32 *odptr)
4040    {
4041    pcre_uint32 c, othercase, next;
4042    unsigned int co;
4043    
4044    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
4045    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
4046    
4047    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4048    return escape == ESC_v || escape == ESC_R;    {
4049      if ((co = UCD_CASESET(c)) != 0)
4050        {
4051        *ocptr = c++;   /* Character that has the set */
4052        *cptr = c;      /* Rest of input range */
4053        return (int)co;
4054        }
4055      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4056      }
4057    
4058    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;  
4059    
4060    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4061    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4062    
4063    default:  for (++c; c <= d; c++)
4064    return FALSE;    {
4065      if (UCD_OTHERCASE(c) != next) break;
4066      next++;
4067    }    }
4068    
4069  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4070    *cptr = c;             /* Rest of input range */
4071    return 0;
4072  }  }
4073    #endif  /* SUPPORT_UCP */
4074    
4075    
4076    
# Line 3418  switch(op_code) Line 4079  switch(op_code)
4079  *************************************************/  *************************************************/
4080    
4081  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
4082  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
4083  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4084  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4085    
4086  Arguments:  Arguments:
4087    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4088    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4089    options       the options word    options       the options word
4090    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4091    start         start of range character    start         start of range character
4092    end           end of range character    end           end of range character
4093    
4094  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4095                  the pointer to extra data is updated                  the pointer to extra data is updated
4096  */  */
# Line 3441  add_to_class(pcre_uint8 *classbits, pcre Line 4102  add_to_class(pcre_uint8 *classbits, pcre
4102  pcre_uint32 c;  pcre_uint32 c;
4103  int n8 = 0;  int n8 = 0;
4104    
4105  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4106  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
4107  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
4108  range. */  range. */
4109    
4110  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4111    {    {
4112  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4113    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4114      {      {
4115      int rc;      int rc;
4116      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4117    
4118      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4119      c = start;      c = start;
4120    
4121      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4122        {        {
4123        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4124    
4125        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4126          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4127    
4128        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4129    
4130        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4131    
4132        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
4133        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
4134        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4135    
4136        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4137        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4138        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3481  if ((options & PCRE_CASELESS) != 0) Line 4142  if ((options & PCRE_CASELESS) != 0)
4142  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4143    
4144    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4145    
4146    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4147      {      {
4148      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4149      n8++;      n8++;
4150      }      }
4151    }    }
4152    
4153  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
4154  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
4155  in all cases. */  in all cases. */
# Line 3514  if (end < 0x100) Line 4175  if (end < 0x100)
4175    {    {
4176    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4177      {      {
4178      n8++;      n8++;
4179      SETBIT(classbits, c);      SETBIT(classbits, c);
4180      }      }
4181    }    }
4182    
4183  else  else
4184    {    {
4185    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4186    
4187  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4188    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4189      {      {
4190      if (start < end)      if (start < end)
4191        {        {
4192        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4193        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4194        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4195        }        }
4196      else if (start == end)      else if (start == end)
4197        {        {
4198        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4199        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4200        }        }
4201      }      }
4202    else    else
4203  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4204    
4205    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4206    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4207    
4208  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4209      {}      {}
4210  #else  #else
4211    if (start < end)    if (start < end)
4212      {      {
4213      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4218  else
4218      {      {
4219      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4220      *uchardata++ = start;      *uchardata++ = start;
4221      }      }
4222  #endif  #endif
4223    
4224    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4225    }    }
4226    
4227  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4228  }  }
4229    
4230    
4231    
4232    
4233  /*************************************************  /*************************************************
4234  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4235  *************************************************/  *************************************************/
4236    
4237  /* This function is used for adding a list of case-equivalent characters to a  /* This function is used for adding a list of case-equivalent characters to a
4238  class, and also for adding a list of horizontal or vertical whitespace. If the  class, and also for adding a list of horizontal or vertical whitespace. If the
4239  list is in order (which it should be), ranges of characters are detected and  list is in order (which it should be), ranges of characters are detected and
4240  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4244  Arguments:
4244    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4245    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4246    options       the options word    options       the options word
4247    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4248    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4249    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4250                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4251                    already know about                    already know about
4252    
4253  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4254                  the pointer to extra data is updated                  the pointer to extra data is updated
4255  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4263  while (p[0] < NOTACHAR)
4263    {    {
4264    int n = 0;    int n = 0;
4265    if (p[0] != except)    if (p[0] != except)
4266      {      {
4267      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4268      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4269      }      }
4270    p += n + 1;    p += n + 1;
4271    }    }
4272  return n8;  return n8;
4273  }  }
4274    
4275    
4276    
# Line 3624  Arguments: Line 4285  Arguments:
4285    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4286    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4287    options       the options word    options       the options word
4288    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4289    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4290    
4291  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4292                  the pointer to extra data is updated                  the pointer to extra data is updated
4293  */  */
4294    
4295  static int  static int
4296  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4297    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4298  {  {
4299  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4305  while (p[0] < NOTACHAR)
4305    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4306    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4307      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4308    p++;    p++;
4309    }    }
4310  return n8;  return n8;
4311  }  }
4312    
4313    
4314    
# Line 3662  to find out the amount of memory needed, Line 4323  to find out the amount of memory needed,
4323  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4324    
4325  Arguments:  Arguments:
4326    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4327    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4328    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4329    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4330    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4331    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4332    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4333    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4334    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4335    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4336                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4337      lengthptr         NULL during the real compile phase
4338                        points to length accumulator during pre-compile phase
4339    
4340  Returns:         TRUE on success  Returns:            TRUE on success
4341                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4342  */  */
4343    
4344  static BOOL  static BOOL
4345  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4346    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4347    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4348      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4349      branch_chain *bcptr, int cond_depth,
4350    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4351  {  {
4352  int repeat_type, op_type;  int repeat_type, op_type;
4353  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4354  int bravalue = 0;  int bravalue = 0;
4355  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4356  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4357  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4358    pcre_uint32 zeroreqchar, zerofirstchar;
4359    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4360  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4361  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4362  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4384  dynamically as we process the pattern. *
4384  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4385  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4386  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4387    #ifndef COMPILE_PCRE32
4388  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4389    #endif
4390  #else  #else
4391  BOOL utf = FALSE;  BOOL utf = FALSE;
4392  #endif  #endif
4393    
4394  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4395  class_uchardata always so that it can be passed to add_to_class() always,  class_uchardata always so that it can be passed to add_to_class() always,
4396  though it will not be used in non-UTF 8-bit cases. This avoids having to supply  though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4397  alternative calls for the different cases. */  alternative calls for the different cases. */
4398    
4399  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4421  to take the zero repeat into account. Th
4421  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4422  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4423    
4424  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4425    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4426    
4427  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4428  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3778  for (;; ptr++) Line 4448  for (;; ptr++)
4448    int recno;    int recno;
4449    int refsign;    int refsign;
4450    int skipbytes;    int skipbytes;
4451    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4452    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4453    int terminator;    int terminator;
4454    int mclength;    unsigned int mclength;
4455    int tempbracount;    unsigned int tempbracount;
4456    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4457    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4458    
4459    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4463  for (;; ptr++)
4463    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
4464    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4465    
4466    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4467      {      {
4468      ptr = nestptr;      ptr = nestptr;
4469      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4538  for (;; ptr++)
4538    
4539    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
4540    
4541    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4542      {      {
4543      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4544        {        {
# Line 3891  for (;; ptr++) Line 4561  for (;; ptr++)
4561          }          }
4562        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4563        }        }
4564        /* Control does not reach here. */
4565      }      }
4566    
4567    /* Fill in length of a previous callout, except when the next thing is    /* In extended mode, skip white space and comments. We need a loop in order
4568    a quantifier. */    to check for more white space and more comments after a comment. */
   
   is_quantifier =  
     c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||  
     (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));  
   
   if (!is_quantifier && previous_callout != NULL &&  
        after_manual_callout-- <= 0)  
     {  
     if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */  
       complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
   
   /* In extended mode, skip white space and comments. */  
4569    
4570    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4571      {      {
4572      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4573        {        {
4574          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4575          if (c != CHAR_NUMBER_SIGN) break;
4576        ptr++;        ptr++;
4577        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4578          {          {
4579          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4580              {                          /* IS_NEWLINE sets cd->nllen. */
4581              ptr += cd->nllen;
4582              break;
4583              }
4584          ptr++;          ptr++;
4585  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4586          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4587  #endif  #endif
4588          }          }
4589        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4590        }        }
4591      }      }
4592    
4593    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4594    
4595      is_quantifier =
4596        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4597        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4598    
4599      /* Fill in length of a previous callout, except when the next thing is a
4600      quantifier or when processing a property substitution string in UCP mode. */
4601    
4602      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4603           after_manual_callout-- <= 0)
4604        {
4605        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4606          complete_callout(previous_callout, ptr, cd);
4607        previous_callout = NULL;
4608        }
4609    
4610      /* Create auto callout, except for quantifiers, or while processing property
4611      strings that are substituted for \w etc in UCP mode. */
4612    
4613    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4614      {      {
4615      previous_callout = code;      previous_callout = code;
4616      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4617      }      }
4618    
4619      /* Process the next pattern item. */
4620    
4621    switch(c)    switch(c)
4622      {      {
4623      /* ===================================================================*/      /* ===================================================================*/
4624      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4625      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4626      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4627      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4628        *firstcharflagsptr = firstcharflags;
4629      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4630        *reqcharflagsptr = reqcharflags;
4631      *codeptr = code;      *codeptr = code;
4632      *ptrptr = ptr;      *ptrptr = ptr;
4633      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4651  for (;; ptr++)
4651      previous = NULL;      previous = NULL;
4652      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4653        {        {
4654        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4655        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4656        }        }
4657      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4666  for (;; ptr++)
4666      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4667    
4668      case CHAR_DOT:      case CHAR_DOT:
4669      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4670      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4671        zerofirstcharflags = firstcharflags;
4672      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4673        zeroreqcharflags = reqcharflags;
4674      previous = code;      previous = code;
4675      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4676      break;      break;
# Line 4015  for (;; ptr++) Line 4698  for (;; ptr++)
4698        goto FAILED;        goto FAILED;
4699        }        }
4700      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4701    
4702        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4703        used for "start of word" and "end of word". As these are otherwise illegal
4704        sequences, we don't break anything by recognizing them. They are replaced
4705        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4706        erroneous and are handled by the normal code below. */
4707    
4708      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4709        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4710          {
4711          nestptr = ptr + 7;
4712          ptr = sub_start_of_word - 1;
4713          continue;
4714          }
4715    
4716        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4717          {
4718          nestptr = ptr + 7;
4719          ptr = sub_end_of_word - 1;
4720          continue;
4721          }
4722    
4723        /* Handle a real character class. */
4724    
4725      previous = code;      previous = code;
4726    
4727      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
# Line 4061  for (;; ptr++) Line 4766  for (;; ptr++)
4766          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4767        {        {
4768        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4769        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4770        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4771          zerofirstcharflags = firstcharflags;
4772        break;        break;
4773        }        }
4774    
# Line 4097  for (;; ptr++) Line 4803  for (;; ptr++)
4803      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4804      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4805    
4806      if (c != 0) do      if (c != CHAR_NULL) do
4807        {        {
4808        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4809    
4810  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF