/[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 1067 by chpe, Tue Oct 16 15:54:22 2012 UTC revision 1462 by ph10, Mon Mar 10 16:28:54 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552    ;    ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 688  static const pcre_uint8 ebcdic_chartab[]
688  #endif  #endif
689    
690    
691    /* This table is used to check whether auto-possessification is possible
692    between adjacent character-type opcodes. The left-hand (repeated) opcode is
693    used to select the row, and the right-hand opcode is use to select the column.
694    A value of 1 means that auto-possessification is OK. For example, the second
695    value in the first row means that \D+\d can be turned into \D++\d.
696    
697    The Unicode property types (\P and \p) have to be present to fill out the table
698    because of what their opcode values are, but the table values should always be
699    zero because property types are handled separately in the code. The last four
700    columns apply to items that cannot be repeated, so there is no need to have
701    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
702    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
703    
704    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
705    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
706    
707    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
708    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
709      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
710      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
711      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
712      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
713      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
714      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
720      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
724      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
725      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
726    };
727    
728    
729    /* This table is used to check whether auto-possessification is possible
730    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
731    left-hand (repeated) opcode is used to select the row, and the right-hand
732    opcode is used to select the column. The values are as follows:
733    
734      0   Always return FALSE (never auto-possessify)
735      1   Character groups are distinct (possessify if both are OP_PROP)
736      2   Check character categories in the same group (general or particular)
737      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
738    
739      4   Check left general category vs right particular category
740      5   Check right general category vs left particular category
741    
742      6   Left alphanum vs right general category
743      7   Left space vs right general category
744      8   Left word vs right general category
745    
746      9   Right alphanum vs left general category
747     10   Right space vs left general category
748     11   Right word vs left general category
749    
750     12   Left alphanum vs right particular category
751     13   Left space vs right particular category
752     14   Left word vs right particular category
753    
754     15   Right alphanum vs left particular category
755     16   Right space vs left particular category
756     17   Right word vs left particular category
757    */
758    
759    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
760    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
761      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
762      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
763      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
764      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
765      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
766      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
767      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
769      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
770      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
772    };
773    
774    /* This table is used to check whether auto-possessification is possible
775    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
776    specifies a general category and the other specifies a particular category. The
777    row is selected by the general category and the column by the particular
778    category. The value is 1 if the particular category is not part of the general
779    category. */
780    
781    static const pcre_uint8 catposstab[7][30] = {
782    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
783      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
784      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
790    };
791    
792    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
793    a general or particular category. The properties in each row are those
794    that apply to the character set in question. Duplication means that a little
795    unnecessary work is done when checking, but this keeps things much simpler
796    because they can all use the same code. For more details see the comment where
797    this table is used.
798    
799    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
800    "space", but from Perl 5.18 it's included, so both categories are treated the
801    same here. */
802    
803    static const pcre_uint8 posspropstab[3][4] = {
804      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
805      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
806      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807    };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
869  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 885  find_error_text(int n)
885  const char *s = error_texts;  const char *s = error_texts;
886  for (; n > 0; n--)  for (; n > 0; n--)
887    {    {
888    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
889    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
890    }    }
891  return s;  return s;
892  }  }
893    
894    
895    
896  /*************************************************  /*************************************************
897  *           Expand the workspace                 *  *           Expand the workspace                 *
898  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 970  return (*p == CHAR_RIGHT_CURLY_BRACKET);
970  *************************************************/  *************************************************/
971    
972  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
973  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
974  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
975  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
976  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
977  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981    chptr          points to the data character    chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 991  Returns:         zero => a data characte
991  */  */
992    
993  static int  static int
994  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
995    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
996  {  {
997  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 1006  ptr--;                            /* Set
1006    
1007  /* 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. */
1008    
1009  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1010    
1011  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1012  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 1015  Otherwise further processing may be requ
1015  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1016  /* Not alphanumeric */  /* Not alphanumeric */
1017  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1018  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)
1019      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1020    
1021  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1022  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1066  else
1066            }            }
1067    
1068  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1069          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1070  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1071          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1072  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1073          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1074  #endif  #endif
1075            {            {
1076            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1117  else
1117      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1118        {        {
1119        const pcre_uchar *p;        const pcre_uchar *p;
1120        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++)
1121          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1122        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1123          {          {
1124          escape = ESC_k;          escape = ESC_k;
1125          break;          break;
# Line 961  else Line 1182  else
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1190      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1191      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1192      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1193      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1194      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1222  else
1222          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1223          break;          break;
1224          }          }
1225        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          escape = -s;          escape = -s;
1228          break;          break;
# Line 1005  else Line 1230  else
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1234      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1235      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1243      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1254  else
1254  #endif  #endif
1255      break;      break;
1256    
1257      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1258      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1259      If not, { is treated as a data character. */  
1260        case CHAR_o:
1261        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262          {
1263          ptr += 2;
1264          c = 0;
1265          overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267            {
1268            register pcre_uint32 cc = *ptr++;
1269            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270    #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272    #endif
1273            c = (c << 3) + cc - CHAR_0 ;
1274    #if defined COMPILE_PCRE8
1275            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE16
1277            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280    #endif
1281            }
1282          if (overflow)
1283            {
1284            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285            *errorcodeptr = ERR34;
1286            }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288            {
1289            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292          }
1293        break;
1294    
1295        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298      case CHAR_x:      case CHAR_x:
1299      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1301        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303          {          {
# Line 1056  else Line 1314  else
1314  #endif  #endif
1315            }            }
1316          }          }
1317        break;        }    /* End JavaScript handling */
       }  
1318    
1319      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        c = 0;      else
1327        overflow = FALSE;        {
1328        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329          {          {
1330          register pcre_uint32 cc = *pt++;          ptr += 2;
1331          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1332            overflow = FALSE;
1333            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1339          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1340  #endif  #endif
1341    
1342  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1346          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1347          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348  #endif  #endif
1349    
1350  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1351          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1353          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1355          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356  #endif  #endif
1357          }            }
1358    
1359        if (overflow)          if (overflow)
1360          {            {
1361          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1363          }            }
1364    
1365        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366          {            {
1367          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368          ptr = pt;            }
         break;  
         }  
1369    
1370        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1371        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1372        }          \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1380        {        else
1381        pcre_uint32 cc;                          /* Some compilers don't like */          {
1382        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1391        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393  #endif  #endif
1394        }            }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1403  else
1403    
1404      case CHAR_c:      case CHAR_c:
1405      c = *(++ptr);      c = *(++ptr);
1406      if (c == 0)      if (c == CHAR_NULL)
1407        {        {
1408        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1409        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1459  if ((options & PCRE_UCP) != 0 && escape
1459  return escape;  return escape;
1460  }  }
1461    
1462    
1463    
1464  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1465  /*************************************************  /*************************************************
1466  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1474  escape sequence.
1474  Argument:  Argument:
1475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1476    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
1477    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
1478      pdataptr       points to an unsigned int that is set to the detailed property value
1479    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1480    
1481  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
1482  */  */
1483    
1484  static int  static BOOL
1485  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1486      unsigned int *pdataptr, int *errorcodeptr)
1487  {  {
1488  pcre_uchar c;  pcre_uchar c;
1489  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1491  const pcre_uchar *ptr = *ptrptr;
1491  pcre_uchar name[32];  pcre_uchar name[32];
1492    
1493  c = *(++ptr);  c = *(++ptr);
1494  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1495    
1496  *negptr = FALSE;  *negptr = FALSE;
1497    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1508  if (c == CHAR_LEFT_CURLY_BRACKET)
1508    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1509      {      {
1510      c = *(++ptr);      c = *(++ptr);
1511      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1512      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1513      name[i] = c;      name[i] = c;
1514      }      }
# Line 1263  while (bot < top) Line 1538  while (bot < top)
1538    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);
1539    if (r == 0)    if (r == 0)
1540      {      {
1541      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1542      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1543        return TRUE;
1544      }      }
1545    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1546    }    }
1547    
1548  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1549  *ptrptr = ptr;  *ptrptr = ptr;
1550  return -1;  return FALSE;
1551    
1552  ERROR_RETURN:  ERROR_RETURN:
1553  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1554  *ptrptr = ptr;  *ptrptr = ptr;
1555  return -1;  return FALSE;
1556  }  }
1557  #endif  #endif
1558    
1559    
1560    
   
1561  /*************************************************  /*************************************************
1562  *         Read repeat counts                     *  *         Read repeat counts                     *
1563  *************************************************/  *************************************************/
# Line 1351  return p; Line 1626  return p;
1626    
1627    
1628  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1629  *      Find first significant op code            *  *      Find first significant op code            *
1630  *************************************************/  *************************************************/
1631    
# Line 1684  for (;;) Line 1664  for (;;)
1664    
1665      case OP_CALLOUT:      case OP_CALLOUT:
1666      case OP_CREF:      case OP_CREF:
1667      case OP_NCREF:      case OP_DNCREF:
1668      case OP_RREF:      case OP_RREF:
1669      case OP_NRREF:      case OP_DNRREF:
1670      case OP_DEF:      case OP_DEF:
1671      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1672      break;      break;
# Line 1700  for (;;) Line 1680  for (;;)
1680    
1681    
1682    
   
1683  /*************************************************  /*************************************************
1684  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1685  *************************************************/  *************************************************/
# Line 1744  for (;;) Line 1723  for (;;)
1723    {    {
1724    int d;    int d;
1725    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1726    register int op = *cc;    register pcre_uchar op = *cc;
1727    
1728    switch (op)    switch (op)
1729      {      {
# Line 1824  for (;;) Line 1803  for (;;)
1803      case OP_COMMIT:      case OP_COMMIT:
1804      case OP_CREF:      case OP_CREF:
1805      case OP_DEF:      case OP_DEF:
1806        case OP_DNCREF:
1807        case OP_DNRREF:
1808      case OP_DOLL:      case OP_DOLL:
1809      case OP_DOLLM:      case OP_DOLLM:
1810      case OP_EOD:      case OP_EOD:
1811      case OP_EODN:      case OP_EODN:
1812      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1813      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1814      case OP_PRUNE:      case OP_PRUNE:
1815      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1831  for (;;)
1831      case OP_NOTI:      case OP_NOTI:
1832      branchlength++;      branchlength++;
1833      cc += 2;      cc += 2;
1834  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1835      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1836  #endif  #endif
1837      break;      break;
# Line 1864  for (;;) Line 1843  for (;;)
1843      case OP_EXACTI:      case OP_EXACTI:
1844      case OP_NOTEXACT:      case OP_NOTEXACT:
1845      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1846      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1847      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1848  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1849      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1850  #endif  #endif
1851      break;      break;
1852    
1853      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1854      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1855      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)
1856        cc += 2;        cc += 2;
1857      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1858      break;      break;
# Line 1909  for (;;) Line 1888  for (;;)
1888    
1889      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1890    
 #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  
   
1891      case OP_CLASS:      case OP_CLASS:
1892      case OP_NCLASS:      case OP_NCLASS:
1893    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1894        case OP_XCLASS:
1895        /* The original code caused an unsigned overflow in 64 bit systems,
1896        so now we use a conditional statement. */
1897        if (op == OP_XCLASS)
1898          cc += GET(cc, 1);
1899        else
1900          cc += PRIV(OP_lengths)[OP_CLASS];
1901    #else
1902      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1903    #endif
1904    
1905      switch (*cc)      switch (*cc)
1906        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1907        case OP_CRSTAR:        case OP_CRSTAR:
1908        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1909          case OP_CRPLUS:
1910          case OP_CRMINPLUS:
1911        case OP_CRQUERY:        case OP_CRQUERY:
1912        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1913          case OP_CRPOSSTAR:
1914          case OP_CRPOSPLUS:
1915          case OP_CRPOSQUERY:
1916        return -1;        return -1;
1917    
1918        case OP_CRRANGE:        case OP_CRRANGE:
1919        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1920          case OP_CRPOSRANGE:
1921        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1922        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1923        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1924        break;        break;
1925    
# Line 1999  for (;;) Line 1986  for (;;)
1986      case OP_QUERYI:      case OP_QUERYI:
1987      case OP_REF:      case OP_REF:
1988      case OP_REFI:      case OP_REFI:
1989        case OP_DNREF:
1990        case OP_DNREFI:
1991      case OP_SBRA:      case OP_SBRA:
1992      case OP_SBRAPOS:      case OP_SBRAPOS:
1993      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2024  for (;;)
2024    
2025    
2026    
   
2027  /*************************************************  /*************************************************
2028  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2029  *************************************************/  *************************************************/
# Line 2059  PRIV(find_bracket)(const pcre_uchar *cod Line 2047  PRIV(find_bracket)(const pcre_uchar *cod
2047  {  {
2048  for (;;)  for (;;)
2049    {    {
2050    register int c = *code;    register pcre_uchar c = *code;
2051    
2052    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2053    
# Line 2082  for (;;) Line 2070  for (;;)
2070    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2071             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2072      {      {
2073      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2074      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2075      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2076      }      }
# Line 2112  for (;;) Line 2100  for (;;)
2100        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2101        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2102        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2103        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)
2104          code += 2;          code += 2;
2105        break;        break;
2106    
2107        case OP_MARK:        case OP_MARK:
2108        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2109        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2110        case OP_THEN_ARG:        case OP_THEN_ARG:
2111        code += code[1];        code += code[1];
2112        break;        break;
# Line 2197  find_recurse(const pcre_uchar *code, BOO Line 2182  find_recurse(const pcre_uchar *code, BOO
2182  {  {
2183  for (;;)  for (;;)
2184    {    {
2185    register int c = *code;    register pcre_uchar c = *code;
2186    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2187    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2188    
# Line 2232  for (;;) Line 2217  for (;;)
2217        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2218        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2219        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2220        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)
2221          code += 2;          code += 2;
2222        break;        break;
2223    
2224        case OP_MARK:        case OP_MARK:
2225        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2226        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2227        code += code[1];        case OP_THEN_ARG:
       break;  
   
       case OP_THEN_ARG:  
2228        code += code[1];        code += code[1];
2229        break;        break;
2230        }        }
# Line 2343  Arguments: Line 2325  Arguments:
2325    endcode     points to where to stop    endcode     points to where to stop
2326    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2327    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2328      recurses    chain of recurse_check to catch mutual recursion
2329    
2330  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2331  */  */
2332    
2333    typedef struct recurse_check {
2334      struct recurse_check *prev;
2335      const pcre_uchar *group;
2336    } recurse_check;
2337    
2338  static BOOL  static BOOL
2339  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2340    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2341  {  {
2342  register int c;  register pcre_uchar c;
2343    recurse_check this_recurse;
2344    
2345  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2346       code < endcode;       code < endcode;
2347       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 2369  for (code = first_significant_code(code
2369    
2370    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2371      {      {
2372      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2373      BOOL empty_branch;      BOOL empty_branch;
2374    
2375      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2376        when called to scan a completed pattern by setting cd->start_workspace to
2377        NULL. */
2378    
2379        if (cd->start_workspace != NULL)
2380          {
2381          const pcre_uchar *tcode;
2382          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2383            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2384          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2385          }
2386    
2387        /* If we are scanning a completed pattern, there are no forward references
2388        and all groups are complete. We need to detect whether this is a recursive
2389        call, as otherwise there will be an infinite loop. If it is a recursion,
2390        just skip over it. Simple recursions are easily detected. For mutual
2391        recursions we keep a chain on the stack. */
2392    
2393        else
2394          {
2395          recurse_check *r = recurses;
2396          const pcre_uchar *endgroup = scode;
2397    
2398          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2399          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2400    
2401      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        for (r = recurses; r != NULL; r = r->prev)
2402        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;          if (r->group == scode) break;
2403          if (r != NULL) continue;   /* Mutual recursion */
2404          }
2405    
2406      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2407        stack chain to detect mutual recursions. */
2408    
2409      empty_branch = FALSE;      empty_branch = FALSE;
2410      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2411      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
2412    
2413      do      do
2414        {        {
2415        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2416          {          {
2417          empty_branch = TRUE;          empty_branch = TRUE;
2418          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2468  for (code = first_significant_code(code
2468        empty_branch = FALSE;        empty_branch = FALSE;
2469        do        do
2470          {          {
2471          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2472            empty_branch = TRUE;            empty_branch = TRUE;
2473          code += GET(code, 1);          code += GET(code, 1);
2474          }          }
# Line 2495  for (code = first_significant_code(code Line 2510  for (code = first_significant_code(code
2510        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2511        case OP_CRQUERY:        case OP_CRQUERY:
2512        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2513          case OP_CRPOSSTAR:
2514          case OP_CRPOSQUERY:
2515        break;        break;
2516    
2517        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2518        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2519        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2520          case OP_CRPOSPLUS:
2521        return FALSE;        return FALSE;
2522    
2523        case OP_CRRANGE:        case OP_CRRANGE:
2524        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2525          case OP_CRPOSRANGE:
2526        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2527        break;        break;
2528        }        }
# Line 2511  for (code = first_significant_code(code Line 2530  for (code = first_significant_code(code
2530    
2531      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2532    
2533        case OP_ANY:
2534        case OP_ALLANY:
2535        case OP_ANYBYTE:
2536    
2537      case OP_PROP:      case OP_PROP:
2538      case OP_NOTPROP:      case OP_NOTPROP:
2539        case OP_ANYNL:
2540    
2541        case OP_NOT_HSPACE:
2542        case OP_HSPACE:
2543        case OP_NOT_VSPACE:
2544        case OP_VSPACE:
2545      case OP_EXTUNI:      case OP_EXTUNI:
2546    
2547      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2548      case OP_DIGIT:      case OP_DIGIT:
2549      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2550      case OP_WHITESPACE:      case OP_WHITESPACE:
2551      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2552      case OP_WORDCHAR:      case OP_WORDCHAR:
2553      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2554      case OP_CHAR:      case OP_CHAR:
2555      case OP_CHARI:      case OP_CHARI:
2556      case OP_NOT:      case OP_NOT:
2557      case OP_NOTI:      case OP_NOTI:
2558    
2559      case OP_PLUS:      case OP_PLUS:
2560        case OP_PLUSI:
2561      case OP_MINPLUS:      case OP_MINPLUS:
2562      case OP_POSPLUS:      case OP_MINPLUSI:
2563      case OP_EXACT:  
2564      case OP_NOTPLUS:      case OP_NOTPLUS:
2565        case OP_NOTPLUSI:
2566      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2567        case OP_NOTMINPLUSI:
2568    
2569        case OP_POSPLUS:
2570        case OP_POSPLUSI:
2571      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2572        case OP_NOTPOSPLUSI:
2573    
2574        case OP_EXACT:
2575        case OP_EXACTI:
2576      case OP_NOTEXACT:      case OP_NOTEXACT:
2577        case OP_NOTEXACTI:
2578    
2579      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2580      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2581      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2582      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2583    
2584      return FALSE;      return FALSE;
2585    
2586      /* 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 2600  for (code = first_significant_code(code
2600      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2601      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2602      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2603      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)
2604        code += 2;        code += 2;
2605      break;      break;
2606    
# Line 2572  for (code = first_significant_code(code Line 2614  for (code = first_significant_code(code
2614      return TRUE;      return TRUE;
2615    
2616      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2617      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2618        followed by a multibyte character. */
2619    
2620  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2621      case OP_STAR:      case OP_STAR:
2622      case OP_STARI:      case OP_STARI:
2623        case OP_NOTSTAR:
2624        case OP_NOTSTARI:
2625    
2626      case OP_MINSTAR:      case OP_MINSTAR:
2627      case OP_MINSTARI:      case OP_MINSTARI:
2628        case OP_NOTMINSTAR:
2629        case OP_NOTMINSTARI:
2630    
2631      case OP_POSSTAR:      case OP_POSSTAR:
2632      case OP_POSSTARI:      case OP_POSSTARI:
2633        case OP_NOTPOSSTAR:
2634        case OP_NOTPOSSTARI:
2635    
2636      case OP_QUERY:      case OP_QUERY:
2637      case OP_QUERYI:      case OP_QUERYI:
2638        case OP_NOTQUERY:
2639        case OP_NOTQUERYI:
2640    
2641      case OP_MINQUERY:      case OP_MINQUERY:
2642      case OP_MINQUERYI:      case OP_MINQUERYI:
2643        case OP_NOTMINQUERY:
2644        case OP_NOTMINQUERYI:
2645    
2646      case OP_POSQUERY:      case OP_POSQUERY:
2647      case OP_POSQUERYI:      case OP_POSQUERYI:
2648        case OP_NOTPOSQUERY:
2649        case OP_NOTPOSQUERYI:
2650    
2651      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2652      break;      break;
2653    
2654      case OP_UPTO:      case OP_UPTO:
2655      case OP_UPTOI:      case OP_UPTOI:
2656        case OP_NOTUPTO:
2657        case OP_NOTUPTOI:
2658    
2659      case OP_MINUPTO:      case OP_MINUPTO:
2660      case OP_MINUPTOI:      case OP_MINUPTOI:
2661        case OP_NOTMINUPTO:
2662        case OP_NOTMINUPTOI:
2663    
2664      case OP_POSUPTO:      case OP_POSUPTO:
2665      case OP_POSUPTOI:      case OP_POSUPTOI:
2666        case OP_NOTPOSUPTO:
2667        case OP_NOTPOSUPTOI:
2668    
2669      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]);
2670      break;      break;
2671  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2676  for (code = first_significant_code(code
2676      case OP_MARK:      case OP_MARK:
2677      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2678      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2679      case OP_THEN_ARG:      case OP_THEN_ARG:
2680      code += code[1];      code += code[1];
2681      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2719  could_be_empty(const pcre_uchar *code, c
2719  {  {
2720  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2721    {    {
2722    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2723      return FALSE;      return FALSE;
2724    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2725    }    }
# Line 2662  return TRUE; Line 2729  return TRUE;
2729    
2730    
2731  /*************************************************  /*************************************************
2732  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2733  *************************************************/  *************************************************/
2734    
2735  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2736  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  
2737    
2738  Returns:   TRUE or FALSE  Arguments:  c opcode
2739    Returns:    base opcode for the type
2740  */  */
2741    
2742  static BOOL  static pcre_uchar
2743  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2744  {  {
2745  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2746  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2747  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2748    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2749    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2750      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;  
2751  }  }
2752    
2753    
2754    
2755    #ifdef SUPPORT_UCP
2756  /*************************************************  /*************************************************
2757  *          Check POSIX class name                *  *        Check a character and a property        *
2758  *************************************************/  *************************************************/
2759    
2760  /* 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
2761  such as [:alnum:].  is adjacent to a fixed character.
2762    
2763  Arguments:  Arguments:
2764    ptr        points to the first letter    c            the character
2765    len        the length of the name    ptype        the property type
2766      pdata        the data for the type
2767      negated      TRUE if it's a negated property (\P or \p{^)
2768    
2769  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2770  */  */
2771    
2772  static int  static BOOL
2773  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2774      BOOL negated)
2775  {  {
2776  const char *pn = posix_names;  const pcre_uint32 *p;
2777  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   *  
 *************************************************/  
2778    
2779  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2780  that is referenced. This means that groups can be replicated for fixed    {
2781  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2782  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2783  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2784  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.  
2785    
2786  This function has been extended with the possibility of forward references for    case PT_GC:
2787  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).  
2788    
2789  Arguments:    case PT_PC:
2790    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  
2791    
2792  Returns:     nothing    case PT_SC:
2793  */    return (pdata == prop->script) == negated;
2794    
2795  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;  
2796    
2797  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2798    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2799    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2800    
2801    /* 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
2802    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2803      at release 8.34. */
2804    
2805    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2806      case PT_PXSPACE:  /* POSIX space */
2807      switch(c)
2808      {      {
2809      offset = GET(hc, 0);      HSPACE_CASES:
2810      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2811        {      return negated;
2812        PUT(hc, 0, offset + adjust);  
2813        break;      default:
2814        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2815      }      }
2816      break;  /* Control never reaches here */
2817    
2818    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2819    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2820              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2821              c == CHAR_UNDERSCORE) == negated;
2822    
2823    if (hc >= cd->hwm)    case PT_CLIST:
2824      p = PRIV(ucd_caseless_sets) + prop->caseset;
2825      for (;;)
2826      {      {
2827      offset = GET(ptr, 1);      if (c < *p) return !negated;
2828      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2829      }      }
2830      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2831    }    }
2832    
2833    return FALSE;
2834  }  }
2835    #endif  /* SUPPORT_UCP */
2836    
2837    
2838    
2839  /*************************************************  /*************************************************
2840  *        Insert an automatic callout point       *  *        Fill the character property list        *
2841  *************************************************/  *************************************************/
2842    
2843  /* 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-
2844  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2845    
2846  Arguments:  Arguments:
2847    code           current code pointer    code        points to start of expression
2848    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2849    cd             pointers to tables etc    fcc         points to case-flipping table
2850      list        points to output list
2851                  list[0] will be filled with the opcode
2852                  list[1] will be non-zero if this opcode
2853                    can match an empty character string
2854                  list[2..7] depends on the opcode
2855    
2856  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2857                  NULL if *code is not accepted
2858  */  */
2859    
2860  static pcre_uchar *  static const pcre_uchar *
2861  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2862  {    const pcre_uint8 *fcc, pcre_uint32 *list)
 *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  
   
 Returns:             nothing  
 */  
   
 static void  
 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  
2863  {  {
2864  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  pcre_uchar c = *code;
2865  PUT(previous_callout, 2 + LINK_SIZE, length);  pcre_uchar base;
2866  }  const pcre_uchar *end;
2867    pcre_uint32 chr;
   
2868    
2869  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2870  /*************************************************  pcre_uint32 *clist_dest;
2871  *           Get othercase range                  *  const pcre_uint32 *clist_src;
2872  *************************************************/  #else
2873    utf = utf;  /* Suppress "unused parameter" compiler warning */
2874  /* This function is passed the start and end of a class range, in UTF-8 mode  #endif
 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.  
2875    
2876  Arguments:  list[0] = c;
2877    cptr        points to starting character value; updated  list[1] = FALSE;
2878    d           end value  code++;
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
2879    
2880  Yield:        -1 when no more  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2881                 0 when a range is returned    {
2882                >0 the CASESET offset for char with multiple other cases    base = get_repeat_base(c);
2883                  in this case, ocptr contains the original    c -= (base - OP_STAR);
 */  
2884    
2885  static int    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2886  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      code += IMM2_SIZE;
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2887    
2888  /* Find the first character that has an other case. If it has multiple other    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 cases, return its case offset value. */  
2889    
2890  for (c = *cptr; c <= d; c++)    switch(base)
   {  
   if ((co = UCD_CASESET(c)) != 0)  
2891      {      {
2892      *ocptr = c++;   /* Character that has the set */      case OP_STAR:
2893      *cptr = c;      /* Rest of input range */      list[0] = OP_CHAR;
2894      return co;      break;
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
   
 if (c > d) return -1;  /* Reached end of range */  
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
   {  
   if (UCD_OTHERCASE(c) != next) break;  
   next++;  
   }  
2895    
2896  *odptr = next - 1;     /* End of othercase range */      case OP_STARI:
2897  *cptr = c;             /* Rest of input range */      list[0] = OP_CHARI;
2898  return 0;      break;
 }  
2899    
2900        case OP_NOTSTAR:
2901        list[0] = OP_NOT;
2902        break;
2903    
2904        case OP_NOTSTARI:
2905        list[0] = OP_NOTI;
2906        break;
2907    
2908  /*************************************************      case OP_TYPESTAR:
2909  *        Check a character and a property        *      list[0] = *code;
2910  *************************************************/      code++;
2911        break;
2912        }
2913      c = list[0];
2914      }
2915    
2916  /* This function is called by check_auto_possessive() when a property item  switch(c)
2917  is adjacent to a fixed character.    {
2918      case OP_NOT_DIGIT:
2919      case OP_DIGIT:
2920      case OP_NOT_WHITESPACE:
2921      case OP_WHITESPACE:
2922      case OP_NOT_WORDCHAR:
2923      case OP_WORDCHAR:
2924      case OP_ANY:
2925      case OP_ALLANY:
2926      case OP_ANYNL:
2927      case OP_NOT_HSPACE:
2928      case OP_HSPACE:
2929      case OP_NOT_VSPACE:
2930      case OP_VSPACE:
2931      case OP_EXTUNI:
2932      case OP_EODN:
2933      case OP_EOD:
2934      case OP_DOLL:
2935      case OP_DOLLM:
2936      return code;
2937    
2938  Arguments:    case OP_CHAR:
2939    c            the character    case OP_NOT:
2940    ptype        the property type    GETCHARINCTEST(chr, code);
2941    pdata        the data for the type    list[2] = chr;
2942    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2943      return code;
2944    
2945  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2946  */    case OP_NOTI:
2947      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2948      GETCHARINCTEST(chr, code);
2949      list[2] = chr;
2950    
 static BOOL  
 check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  
 {  
2951  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2952  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2953        list[3] = fcc[chr];
2954      else
2955        list[3] = UCD_OTHERCASE(chr);
2956    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2957      list[3] = (chr < 256) ? fcc[chr] : chr;
2958    #else
2959      list[3] = fcc[chr];
2960  #endif  #endif
2961    
2962  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
2963    
2964  switch(ptype)    if (chr == list[3])
2965    {      list[3] = NOTACHAR;
2966    case PT_LAMP:    else
2967    return (prop->chartype == ucp_Lu ||      list[4] = NOTACHAR;
2968            prop->chartype == ucp_Ll ||    return code;
           prop->chartype == ucp_Lt) == negated;  
2969    
2970    case PT_GC:  #ifdef SUPPORT_UCP
2971    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;    case OP_PROP:
2972      case OP_NOTPROP:
2973      if (code[0] != PT_CLIST)
2974        {
2975        list[2] = code[0];
2976        list[3] = code[1];
2977        return code + 2;
2978        }
2979    
2980    case PT_PC:    /* Convert only if we have enough space. */
   return (pdata == prop->chartype) == negated;  
2981    
2982    case PT_SC:    clist_src = PRIV(ucd_caseless_sets) + code[1];
2983    return (pdata == prop->script) == negated;    clist_dest = list + 2;
2984      code += 2;
2985    
2986    /* These are specials */    do {
2987         if (clist_dest >= list + 8)
2988           {
2989           /* Early return if there is not enough space. This should never
2990           happen, since all clists are shorter than 5 character now. */
2991           list[2] = code[0];
2992           list[3] = code[1];
2993           return code;
2994           }
2995         *clist_dest++ = *clist_src;
2996         }
2997      while(*clist_src++ != NOTACHAR);
2998    
2999    case PT_ALNUM:    /* All characters are stored. The terminating NOTACHAR
3000    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    is copied form the clist itself. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3001    
3002    case PT_SPACE:    /* Perl space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3003    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
3004            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  #endif
           == negated;  
3005    
3006    case PT_PXSPACE:  /* POSIX space */    case OP_NCLASS:
3007    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    case OP_CLASS:
3008            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3009            c == CHAR_FF || c == CHAR_CR)    case OP_XCLASS:
3010            == negated;    if (c == OP_XCLASS)
3011        end = code + GET(code, 0) - 1;
3012      else
3013    #endif
3014        end = code + 32 / sizeof(pcre_uchar);
3015    
3016    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 (;;)  
3017      {      {
3018      if ((unsigned int)c < *p) return !negated;      case OP_CRSTAR:
3019      if ((unsigned int)c == *p++) return negated;      case OP_CRMINSTAR:
3020        case OP_CRQUERY:
3021        case OP_CRMINQUERY:
3022        case OP_CRPOSSTAR:
3023        case OP_CRPOSQUERY:
3024        list[1] = TRUE;
3025        end++;
3026        break;
3027    
3028        case OP_CRPLUS:
3029        case OP_CRMINPLUS:
3030        case OP_CRPOSPLUS:
3031        end++;
3032        break;
3033    
3034        case OP_CRRANGE:
3035        case OP_CRMINRANGE:
3036        case OP_CRPOSRANGE:
3037        list[1] = (GET2(end, 1) == 0);
3038        end += 1 + 2 * IMM2_SIZE;
3039        break;
3040      }      }
3041    break;  /* Control never reaches here */    list[2] = end - code;
3042  #endif    return end;
3043    }    }
3044    return NULL;    /* Opcode not accepted */
 return FALSE;  
3045  }  }
 #endif  /* SUPPORT_UCP */  
3046    
3047    
3048    
3049  /*************************************************  /*************************************************
3050  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3051  *************************************************/  *************************************************/
3052    
3053  /* 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
3054  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.  
3055    
3056  Arguments:  Arguments:
3057    previous      pointer to the repeated opcode    code        points to the byte code
3058    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3059    ptr           next character in pattern    cd          static compile data
3060    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3061    
3062  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3063  */  */
3064    
3065  static BOOL  static BOOL
3066  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3067    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3068  {  {
3069  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3070  pcre_uint32 next;  pcre_uint32 list[8];
3071  int escape;  const pcre_uint32 *chr_ptr;
3072  int op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3073    const pcre_uint32 *list_ptr;
3074  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3075    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3076    const pcre_uchar *xclass_flags;
3077    #endif
3078    const pcre_uint8 *class_bitset;
3079    const pcre_uint8 *set1, *set2, *set_end;
3080    pcre_uint32 chr;
3081    BOOL accepted, invert_bits;
3082    
3083    /* Note: the base_list[1] contains whether the current opcode has greedy
3084    (represented by a non-zero value) quantifier. This is a different from
3085    other character type lists, which stores here that the character iterator
3086    matches to an empty string (also represented by a non-zero value). */
3087    
3088  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3089    {    {
3090    for (;;)    /* All operations move the code pointer forward.
3091      Therefore infinite recursions are not possible. */
3092    
3093      c = *code;
3094    
3095      /* Skip over callouts */
3096    
3097      if (c == OP_CALLOUT)
3098      {      {
3099      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3100      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;  
3101      }      }
   }  
3102    
3103  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3104  value is a character, a negative value is an escape value. */      {
3105        do code += GET(code, 1); while (*code == OP_ALT);
3106        c = *code;
3107        }
3108    
3109  if (*ptr == CHAR_BACKSLASH)    switch(c)
3110    {      {
3111    int temperrorcode = 0;      case OP_END:
3112    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);      case OP_KETRPOS:
3113    if (temperrorcode != 0) return FALSE;      /* TRUE only in greedy case. The non-greedy case could be replaced by
3114    ptr++;    /* Point after the escape sequence */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3115    }      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;  
3116    
3117  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3118    
3119  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3120    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3121    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3122      {      cannot be converted to a possessive form. */
3123      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3124      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[1] == 0) return FALSE;
3125    
3126        switch(*(code - GET(code, 1)))
3127          {
3128          case OP_ASSERT:
3129          case OP_ASSERT_NOT:
3130          case OP_ASSERTBACK:
3131          case OP_ASSERTBACK_NOT:
3132          case OP_ONCE:
3133          case OP_ONCE_NC:
3134          /* Atomic sub-patterns and assertions can always auto-possessify their
3135          last iterator. */
3136          return TRUE;
3137          }
3138    
3139        code += PRIV(OP_lengths)[c];
3140        continue;
3141    
3142        case OP_ONCE:
3143        case OP_ONCE_NC:
3144        case OP_BRA:
3145        case OP_CBRA:
3146        next_code = code + GET(code, 1);
3147        code += PRIV(OP_lengths)[c];
3148    
3149        while (*next_code == OP_ALT)
3150        {        {
3151        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3152        while (*ptr != 0)        code = next_code + 1 + LINK_SIZE;
3153          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3154        }        }
3155      else break;      continue;
     }  
   }  
3156    
3157  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3158        case OP_BRAMINZERO:
3159    
3160  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3161    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3162      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3163    
3164        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3165    
3166        /* The bracket content will be checked by the
3167        OP_BRA/OP_CBRA case above. */
3168        next_code += 1 + LINK_SIZE;
3169        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3170          return FALSE;
3171    
3172  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3173        continue;
3174        }
3175    
3176  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  
   }  
3177    
3178  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3179  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3180    
3181  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3182    {    characters from that list with another list, or with a property. */
3183    /* For a caseless UTF match, the next character may have more than one other  
3184    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)  
3185      {      {
3186      int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3187      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3188      }      }
3189  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3190      {      {
3191      case OP_CHAR:      chr_ptr = list + 2;
3192      return c != next;      list_ptr = base_list;
3193        }
3194    
3195      /* 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. */  
3196    
3197      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3198      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3199  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3200      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3201        {  #endif
3202        pcre_uint32 othercase;        )
3203        if (next < 128) othercase = cd->fcc[next]; else      {
3204  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3205        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3206  #else  #else
3207        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3208  #endif  #endif
3209        return c != othercase;        {
3210          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3211          list_ptr = list;
3212        }        }
3213      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)  
3214        {        {
3215        pcre_uint32 othercase;        set1 = (pcre_uint8 *)(code - list[2]);
3216        if (next < 128) othercase = cd->fcc[next]; else        list_ptr = base_list;
3217  #ifdef SUPPORT_UCP        }
3218        othercase = UCD_OTHERCASE(next);  
3219  #else      invert_bits = FALSE;
3220        othercase = NOTACHAR;      switch(list_ptr[0])
3221          {
3222          case OP_CLASS:
3223          case OP_NCLASS:
3224          set2 = (pcre_uint8 *)
3225            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3226          break;
3227    
3228    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3229          case OP_XCLASS:
3230          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3231          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3232          if ((*xclass_flags & XCL_MAP) == 0)
3233            {
3234            /* No bits are set for characters < 256. */
3235            if (list[1] == 0) return TRUE;
3236            /* Might be an empty repeat. */
3237            continue;
3238            }
3239          set2 = (pcre_uint8 *)(xclass_flags + 1);
3240          break;
3241  #endif  #endif
3242        return c == othercase;  
3243          case OP_NOT_DIGIT:
3244          invert_bits = TRUE;
3245          /* Fall through */
3246          case OP_DIGIT:
3247          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3248          break;
3249    
3250          case OP_NOT_WHITESPACE:
3251          invert_bits = TRUE;
3252          /* Fall through */
3253          case OP_WHITESPACE:
3254          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3255          break;
3256    
3257          case OP_NOT_WORDCHAR:
3258          invert_bits = TRUE;
3259          /* Fall through */
3260          case OP_WORDCHAR:
3261          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3262          break;
3263    
3264          default:
3265          return FALSE;
3266          }
3267    
3268        /* Because the sets are unaligned, we need
3269        to perform byte comparison here. */
3270        set_end = set1 + 32;
3271        if (invert_bits)
3272          {
3273          do
3274            {
3275            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3276            }
3277          while (set1 < set_end);
3278        }        }
3279      else      else
3280  #endif  /* SUPPORT_UTF */        {
3281      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */        do
3282            {
3283            if ((*set1++ & *set2++) != 0) return FALSE;
3284            }
3285          while (set1 < set_end);
3286          }
3287    
3288      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      if (list[1] == 0) return TRUE;
3289      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      /* Might be an empty repeat. */
3290        continue;
3291        }
3292    
3293      case OP_DIGIT:    /* Some property combinations also acceptable. Unicode property opcodes are
3294      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;    processed specially; the rest can be handled with a lookup table. */
3295    
3296      case OP_NOT_DIGIT:    else
3297      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;      {
3298        pcre_uint32 leftop, rightop;
3299    
3300      case OP_WHITESPACE:      leftop = base_list[0];
3301      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;      rightop = list[0];
3302    
3303      case OP_NOT_WHITESPACE:  #ifdef SUPPORT_UCP
3304      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;      accepted = FALSE; /* Always set in non-unicode case. */
3305        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3306          {
3307          if (rightop == OP_EOD)
3308            accepted = TRUE;
3309          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3310            {
3311            int n;
3312            const pcre_uint8 *p;
3313            BOOL same = leftop == rightop;
3314            BOOL lisprop = leftop == OP_PROP;
3315            BOOL risprop = rightop == OP_PROP;
3316            BOOL bothprop = lisprop && risprop;
3317    
3318            /* There's a table that specifies how each combination is to be
3319            processed:
3320              0   Always return FALSE (never auto-possessify)
3321              1   Character groups are distinct (possessify if both are OP_PROP)
3322              2   Check character categories in the same group (general or particular)
3323              3   Return TRUE if the two opcodes are not the same
3324              ... see comments below
3325            */
3326    
3327            n = propposstab[base_list[2]][list[2]];
3328            switch(n)
3329              {
3330              case 0: break;
3331              case 1: accepted = bothprop; break;
3332              case 2: accepted = (base_list[3] == list[3]) != same; break;
3333              case 3: accepted = !same; break;
3334    
3335      case OP_WORDCHAR:            case 4:  /* Left general category, right particular category */
3336      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3337              break;
3338    
3339      case OP_NOT_WORDCHAR:            case 5:  /* Right general category, left particular category */
3340      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3341              break;
3342    
3343      case OP_HSPACE:            /* This code is logically tricky. Think hard before fiddling with it.
3344      case OP_NOT_HSPACE:            The posspropstab table has four entries per row. Each row relates to
3345      switch(next)            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3346        {            Only WORD actually needs all four entries, but using repeats for the
3347        HSPACE_CASES:            others means they can all use the same code below.
3348        return op_code == OP_NOT_HSPACE;  
3349              The first two entries in each row are Unicode general categories, and
3350              apply always, because all the characters they include are part of the
3351              PCRE character set. The third and fourth entries are a general and a
3352              particular category, respectively, that include one or more relevant
3353              characters. One or the other is used, depending on whether the check
3354              is for a general or a particular category. However, in both cases the
3355              category contains more characters than the specials that are defined
3356              for the property being tested against. Therefore, it cannot be used
3357              in a NOTPROP case.
3358    
3359              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3360              Underscore is covered by ucp_P or ucp_Po. */
3361    
3362              case 6:  /* Left alphanum vs right general category */
3363              case 7:  /* Left space vs right general category */
3364              case 8:  /* Left word vs right general category */
3365              p = posspropstab[n-6];
3366              accepted = risprop && lisprop ==
3367                (list[3] != p[0] &&
3368                 list[3] != p[1] &&
3369                (list[3] != p[2] || !lisprop));
3370              break;
3371    
3372        default:            case 9:   /* Right alphanum vs left general category */
3373        return op_code != OP_NOT_HSPACE;            case 10:  /* Right space vs left general category */
3374              case 11:  /* Right word vs left general category */
3375              p = posspropstab[n-9];
3376              accepted = lisprop && risprop ==
3377                (base_list[3] != p[0] &&
3378                 base_list[3] != p[1] &&
3379                (base_list[3] != p[2] || !risprop));
3380              break;
3381    
3382              case 12:  /* Left alphanum vs right particular category */
3383              case 13:  /* Left space vs right particular category */
3384              case 14:  /* Left word vs right particular category */
3385              p = posspropstab[n-12];
3386              accepted = risprop && lisprop ==
3387                (catposstab[p[0]][list[3]] &&
3388                 catposstab[p[1]][list[3]] &&
3389                (list[3] != p[3] || !lisprop));
3390              break;
3391    
3392              case 15:  /* Right alphanum vs left particular category */
3393              case 16:  /* Right space vs left particular category */
3394              case 17:  /* Right word vs left particular category */
3395              p = posspropstab[n-15];
3396              accepted = lisprop && risprop ==
3397                (catposstab[p[0]][base_list[3]] &&
3398                 catposstab[p[1]][base_list[3]] &&
3399                (base_list[3] != p[3] || !risprop));
3400              break;
3401              }
3402            }
3403        }        }
3404    
3405      case OP_ANYNL:      else
3406      case OP_VSPACE:  #endif  /* SUPPORT_UCP */
3407      case OP_NOT_VSPACE:  
3408      switch(next)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3409               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3410               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3411    
3412        if (!accepted)
3413          return FALSE;
3414    
3415        if (list[1] == 0) return TRUE;
3416        /* Might be an empty repeat. */
3417        continue;
3418        }
3419    
3420      /* Control reaches here only if one of the items is a small character list.
3421      All characters are checked against the other side. */
3422    
3423      do
3424        {
3425        chr = *chr_ptr;
3426    
3427        switch(list_ptr[0])
3428        {        {
3429        VSPACE_CASES:        case OP_CHAR:
3430        return op_code == OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3431          do
3432            {
3433            if (chr == *ochr_ptr) return FALSE;
3434            ochr_ptr++;
3435            }
3436          while(*ochr_ptr != NOTACHAR);
3437          break;
3438    
3439        default:        case OP_NOT:
3440        return op_code != OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3441        }        do
3442            {
3443            if (chr == *ochr_ptr)
3444              break;
3445            ochr_ptr++;
3446            }
3447          while(*ochr_ptr != NOTACHAR);
3448          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3449          break;
3450    
3451          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3452          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3453    
3454          case OP_DIGIT:
3455          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3456          break;
3457    
3458          case OP_NOT_DIGIT:
3459          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3460          break;
3461    
3462          case OP_WHITESPACE:
3463          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3464          break;
3465    
3466          case OP_NOT_WHITESPACE:
3467          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3468          break;
3469    
3470          case OP_WORDCHAR:
3471          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3472          break;
3473    
3474          case OP_NOT_WORDCHAR:
3475          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3476          break;
3477    
3478          case OP_HSPACE:
3479          switch(chr)
3480            {
3481            HSPACE_CASES: return FALSE;
3482            default: break;
3483            }
3484          break;
3485    
3486          case OP_NOT_HSPACE:
3487          switch(chr)
3488            {
3489            HSPACE_CASES: break;
3490            default: return FALSE;
3491            }
3492          break;
3493    
3494          case OP_ANYNL:
3495          case OP_VSPACE:
3496          switch(chr)
3497            {
3498            VSPACE_CASES: return FALSE;
3499            default: break;
3500            }
3501          break;
3502    
3503          case OP_NOT_VSPACE:
3504          switch(chr)
3505            {
3506            VSPACE_CASES: break;
3507            default: return FALSE;
3508            }
3509          break;
3510    
3511          case OP_DOLL:
3512          case OP_EODN:
3513          switch (chr)
3514            {
3515            case CHAR_CR:
3516            case CHAR_LF:
3517            case CHAR_VT:
3518            case CHAR_FF:
3519            case CHAR_NEL:
3520    #ifndef EBCDIC
3521            case 0x2028:
3522            case 0x2029:
3523    #endif  /* Not EBCDIC */
3524            return FALSE;
3525            }
3526          break;
3527    
3528          case OP_EOD:    /* Can always possessify before \z */
3529          break;
3530    
3531  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3532      case OP_PROP:        case OP_PROP:
3533      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3534          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3535                list_ptr[0] == OP_NOTPROP))
3536            return FALSE;
3537          break;
3538    #endif
3539    
3540      case OP_NOTPROP:        case OP_NCLASS:
3541      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3542          /* Fall through */
3543    
3544          case OP_CLASS:
3545          if (chr > 255) break;
3546          class_bitset = (pcre_uint8 *)
3547            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3548          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3549          break;
3550    
3551    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3552          case OP_XCLASS:
3553          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3554              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3555          break;
3556  #endif  #endif
3557    
3558      default:        default:
3559      return FALSE;        return FALSE;
3560          }
3561    
3562        chr_ptr++;
3563      }      }
3564      while(*chr_ptr != NOTACHAR);
3565    
3566      /* At least one character must be matched from this opcode. */
3567    
3568      if (list[1] == 0) return TRUE;
3569    }    }
3570    
3571  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3572  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  but some compilers complain about an unreachable statement. */
3573  generated only when PCRE_UCP is *not* set, that is, when only ASCII  
3574  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  }
3575  replaced by OP_PROP codes when PCRE_UCP is set. */  
3576    
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3577    
3578      case ESC_D:  /*************************************************
3579      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  *    Scan compiled regex for auto-possession     *
3580    *************************************************/
3581    
3582      case ESC_s:  /* Replaces single character iterations with their possessive alternatives
3583      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  if appropriate. This function modifies the compiled opcode!
3584    
3585      case ESC_S:  Arguments:
3586      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    code        points to start of the byte code
3587      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3588      cd          static compile data
3589    
3590      case ESC_w:  Returns:      nothing
3591      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  */
3592    
3593      case ESC_W:  static void
3594      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3595    {
3596    register pcre_uchar c;
3597    const pcre_uchar *end;
3598    pcre_uchar *repeat_opcode;
3599    pcre_uint32 list[8];
3600    
3601      case ESC_h:  for (;;)
3602      case ESC_H:    {
3603      switch(c)    c = *code;
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
3604    
3605      case ESC_v:    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3606      case ESC_V:      {
3607      switch(c)      c -= get_repeat_base(c) - OP_STAR;
3608        end = (c <= OP_MINUPTO) ?
3609          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3610        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3611    
3612        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3613        {        {
3614        VSPACE_CASES:        switch(c)
3615        return escape != ESC_v;          {
3616            case OP_STAR:
3617            *code += OP_POSSTAR - OP_STAR;
3618            break;
3619    
3620        default:          case OP_MINSTAR:
3621        return escape == ESC_v;          *code += OP_POSSTAR - OP_MINSTAR;
3622        }          break;
3623    
3624      /* When PCRE_UCP is set, these values get generated for \d etc. Find          case OP_PLUS:
3625      their substitutions and process them. The result will always be either          *code += OP_POSPLUS - OP_PLUS;
3626      ESC_p or ESC_P. Then fall through to process those values. */          break;
3627    
3628  #ifdef SUPPORT_UCP          case OP_MINPLUS:
3629      case ESC_du:          *code += OP_POSPLUS - OP_MINPLUS;
3630      case ESC_DU:          break;
3631      case ESC_wu:  
3632      case ESC_WU:          case OP_QUERY:
3633      case ESC_su:          *code += OP_POSQUERY - OP_QUERY;
3634      case ESC_SU:          break;
3635        {  
3636        int temperrorcode = 0;          case OP_MINQUERY:
3637        ptr = substitutes[escape - ESC_DU];          *code += OP_POSQUERY - OP_MINQUERY;
3638        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);          break;
3639        if (temperrorcode != 0) return FALSE;  
3640        ptr++;    /* For compatibility */          case OP_UPTO:
3641            *code += OP_POSUPTO - OP_UPTO;
3642            break;
3643    
3644            case OP_MINUPTO:
3645            *code += OP_POSUPTO - OP_MINUPTO;
3646            break;
3647            }
3648        }        }
3649      /* Fall through */      c = *code;
3650        }
3651      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3652        {
3653    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3654        if (c == OP_XCLASS)
3655          repeat_opcode = code + GET(code, 1);
3656        else
3657    #endif
3658          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3659    
3660      case ESC_p:      c = *repeat_opcode;
3661      case ESC_P:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3662        {        {
3663        int ptype, pdata, errorcodeptr;        /* end must not be NULL. */
3664        BOOL negated;        end = get_chr_property_list(code, utf, cd->fcc, list);
3665    
3666        ptr--;      /* Make ptr point at the p or P */        list[1] = (c & 1) == 0;
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3667    
3668        /* If the property item is optional, we have to give up. (When generated        if (compare_opcodes(end, utf, cd, list, end))
3669        from \d etc by PCRE_UCP, this test will have been applied much earlier,          {
3670        to the original \d etc. At this point, ptr will point to a zero byte. */          switch (c)
3671              {
3672              case OP_CRSTAR:
3673              case OP_CRMINSTAR:
3674              *repeat_opcode = OP_CRPOSSTAR;
3675              break;
3676    
3677        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||            case OP_CRPLUS:
3678          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)            case OP_CRMINPLUS:
3679            return FALSE;            *repeat_opcode = OP_CRPOSPLUS;
3680              break;
3681    
3682        /* Do the property check. */            case OP_CRQUERY:
3683              case OP_CRMINQUERY:
3684              *repeat_opcode = OP_CRPOSQUERY;
3685              break;
3686    
3687        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);            case OP_CRRANGE:
3688              case OP_CRMINRANGE:
3689              *repeat_opcode = OP_CRPOSRANGE;
3690              break;
3691              }
3692            }
3693        }        }
3694        c = *code;
3695        }
3696    
3697      switch(c)
3698        {
3699        case OP_END:
3700        return;
3701    
3702        case OP_TYPESTAR:
3703        case OP_TYPEMINSTAR:
3704        case OP_TYPEPLUS:
3705        case OP_TYPEMINPLUS:
3706        case OP_TYPEQUERY:
3707        case OP_TYPEMINQUERY:
3708        case OP_TYPEPOSSTAR:
3709        case OP_TYPEPOSPLUS:
3710        case OP_TYPEPOSQUERY:
3711        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3712        break;
3713    
3714        case OP_TYPEUPTO:
3715        case OP_TYPEMINUPTO:
3716        case OP_TYPEEXACT:
3717        case OP_TYPEPOSUPTO:
3718        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3719          code += 2;
3720        break;
3721    
3722    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3723        case OP_XCLASS:
3724        code += GET(code, 1);
3725        break;
3726  #endif  #endif
3727    
3728      default:      case OP_MARK:
3729      return FALSE;      case OP_PRUNE_ARG:
3730        case OP_SKIP_ARG:
3731        case OP_THEN_ARG:
3732        code += code[1];
3733        break;
3734      }      }
3735    
3736    /* In principle, support for Unicode properties should be integrated here as    /* Add in the fixed length from the table */
3737    well. It means re-organizing the above code so as to get hold of the property  
3738    values before switching on the op-code. However, I wonder how many patterns    code += PRIV(OP_lengths)[c];
3739    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
3740    these op-codes are never generated.) */    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3741      a multi-byte character. The length in the table is a minimum, so we have to
3742      arrange to skip the extra bytes. */
3743    
3744    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3745      if (utf) switch(c)
3746        {
3747        case OP_CHAR:
3748        case OP_CHARI:
3749        case OP_NOT:
3750        case OP_NOTI:
3751        case OP_STAR:
3752        case OP_MINSTAR:
3753        case OP_PLUS:
3754        case OP_MINPLUS:
3755        case OP_QUERY:
3756        case OP_MINQUERY:
3757        case OP_UPTO:
3758        case OP_MINUPTO:
3759        case OP_EXACT:
3760        case OP_POSSTAR:
3761        case OP_POSPLUS:
3762        case OP_POSQUERY:
3763        case OP_POSUPTO:
3764        case OP_STARI:
3765        case OP_MINSTARI:
3766        case OP_PLUSI:
3767        case OP_MINPLUSI:
3768        case OP_QUERYI:
3769        case OP_MINQUERYI:
3770        case OP_UPTOI:
3771        case OP_MINUPTOI:
3772        case OP_EXACTI:
3773        case OP_POSSTARI:
3774        case OP_POSPLUSI:
3775        case OP_POSQUERYI:
3776        case OP_POSUPTOI:
3777        case OP_NOTSTAR:
3778        case OP_NOTMINSTAR:
3779        case OP_NOTPLUS:
3780        case OP_NOTMINPLUS:
3781        case OP_NOTQUERY:
3782        case OP_NOTMINQUERY:
3783        case OP_NOTUPTO:
3784        case OP_NOTMINUPTO:
3785        case OP_NOTEXACT:
3786        case OP_NOTPOSSTAR:
3787        case OP_NOTPOSPLUS:
3788        case OP_NOTPOSQUERY:
3789        case OP_NOTPOSUPTO:
3790        case OP_NOTSTARI:
3791        case OP_NOTMINSTARI:
3792        case OP_NOTPLUSI:
3793        case OP_NOTMINPLUSI:
3794        case OP_NOTQUERYI:
3795        case OP_NOTMINQUERYI:
3796        case OP_NOTUPTOI:
3797        case OP_NOTMINUPTOI:
3798        case OP_NOTEXACTI:
3799        case OP_NOTPOSSTARI:
3800        case OP_NOTPOSPLUSI:
3801        case OP_NOTPOSQUERYI:
3802        case OP_NOTPOSUPTOI:
3803        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3804        break;
3805        }
3806    #else
3807      (void)(utf);  /* Keep compiler happy by referencing function argument */
3808    #endif
3809      }
3810    }
3811    
3812    
3813    
3814    /*************************************************
3815    *           Check for POSIX class syntax         *
3816    *************************************************/
3817    
3818    /* This function is called when the sequence "[:" or "[." or "[=" is
3819    encountered in a character class. It checks whether this is followed by a
3820    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3821    reach an unescaped ']' without the special preceding character, return FALSE.
3822    
3823    Originally, this function only recognized a sequence of letters between the
3824    terminators, but it seems that Perl recognizes any sequence of characters,
3825    though of course unknown POSIX names are subsequently rejected. Perl gives an
3826    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3827    didn't consider this to be a POSIX class. Likewise for [:1234:].
3828    
3829    The problem in trying to be exactly like Perl is in the handling of escapes. We
3830    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3831    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3832    below handles the special case of \], but does not try to do any other escape
3833    processing. This makes it different from Perl for cases such as [:l\ower:]
3834    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3835    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3836    I think.
3837    
3838    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3839    It seems that the appearance of a nested POSIX class supersedes an apparent
3840    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3841    a digit.
3842    
3843    In Perl, unescaped square brackets may also appear as part of class names. For
3844    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3845    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3846    seem right at all. PCRE does not allow closing square brackets in POSIX class
3847    names.
3848    
3849    Arguments:
3850      ptr      pointer to the initial [
3851      endptr   where to return the end pointer
3852    
3853    Returns:   TRUE or FALSE
3854    */
3855    
3856    static BOOL
3857    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3858    {
3859    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3860    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3861    for (++ptr; *ptr != CHAR_NULL; ptr++)
3862      {
3863      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3864        ptr++;
3865      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3866      else
3867        {
3868        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3869          {
3870          *endptr = ptr;
3871          return TRUE;
3872          }
3873        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3874             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3875              ptr[1] == CHAR_EQUALS_SIGN) &&
3876            check_posix_syntax(ptr, endptr))
3877          return FALSE;
3878        }
3879      }
3880    return FALSE;
3881    }
3882    
3883    
3884    
3885    
3886    /*************************************************
3887    *          Check POSIX class name                *
3888    *************************************************/
3889    
3890    /* This function is called to check the name given in a POSIX-style class entry
3891    such as [:alnum:].
3892    
3893    Arguments:
3894      ptr        points to the first letter
3895      len        the length of the name
3896    
3897    Returns:     a value representing the name, or -1 if unknown
3898    */
3899    
3900    static int
3901    check_posix_name(const pcre_uchar *ptr, int len)
3902    {
3903    const char *pn = posix_names;
3904    register int yield = 0;
3905    while (posix_name_lengths[yield] != 0)
3906      {
3907      if (len == posix_name_lengths[yield] &&
3908        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3909      pn += posix_name_lengths[yield] + 1;
3910      yield++;
3911      }
3912    return -1;
3913    }
3914    
3915    
3916    /*************************************************
3917    *    Adjust OP_RECURSE items in repeated group   *
3918    *************************************************/
3919    
3920    /* OP_RECURSE items contain an offset from the start of the regex to the group
3921    that is referenced. This means that groups can be replicated for fixed
3922    repetition simply by copying (because the recursion is allowed to refer to
3923    earlier groups that are outside the current group). However, when a group is
3924    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3925    inserted before it, after it has been compiled. This means that any OP_RECURSE
3926    items within it that refer to the group itself or any contained groups have to
3927    have their offsets adjusted. That one of the jobs of this function. Before it
3928    is called, the partially compiled regex must be temporarily terminated with
3929    OP_END.
3930    
3931    This function has been extended with the possibility of forward references for
3932    recursions and subroutine calls. It must also check the list of such references
3933    for the group we are dealing with. If it finds that one of the recursions in
3934    the current group is on this list, it adjusts the offset in the list, not the
3935    value in the reference (which is a group number).
3936    
3937    Arguments:
3938      group      points to the start of the group
3939      adjust     the amount by which the group is to be moved
3940      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3941      cd         contains pointers to tables etc.
3942      save_hwm   the hwm forward reference pointer at the start of the group
3943    
3944    Returns:     nothing
3945    */
3946    
3947    static void
3948    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3949      pcre_uchar *save_hwm)
3950    {
3951    pcre_uchar *ptr = group;
3952    
3953    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3954      {
3955      int offset;
3956      pcre_uchar *hc;
3957    
3958      /* See if this recursion is on the forward reference list. If so, adjust the
3959      reference. */
3960    
3961      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3962        {
3963        offset = (int)GET(hc, 0);
3964        if (cd->start_code + offset == ptr + 1)
3965          {
3966          PUT(hc, 0, offset + adjust);
3967          break;
3968          }
3969        }
3970    
3971      /* Otherwise, adjust the recursion offset if it's after the start of this
3972      group. */
3973    
3974      if (hc >= cd->hwm)
3975        {
3976        offset = (int)GET(ptr, 1);
3977        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3978        }
3979    
3980      ptr += 1 + LINK_SIZE;
3981      }
3982    }
3983    
3984    
3985    
3986    /*************************************************
3987    *        Insert an automatic callout point       *
3988    *************************************************/
3989    
3990    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3991    callout points before each pattern item.
3992    
3993    Arguments:
3994      code           current code pointer
3995      ptr            current pattern pointer
3996      cd             pointers to tables etc
3997    
3998    Returns:         new code pointer
3999    */
4000    
4001    static pcre_uchar *
4002    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4003    {
4004    *code++ = OP_CALLOUT;
4005    *code++ = 255;
4006    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4007    PUT(code, LINK_SIZE, 0);                       /* Default length */
4008    return code + 2 * LINK_SIZE;
4009    }
4010    
4011    
4012    
4013    /*************************************************
4014    *         Complete a callout item                *
4015    *************************************************/
4016    
4017    /* A callout item contains the length of the next item in the pattern, which
4018    we can't fill in till after we have reached the relevant point. This is used
4019    for both automatic and manual callouts.
4020    
4021    Arguments:
4022      previous_callout   points to previous callout item
4023      ptr                current pattern pointer
4024      cd                 pointers to tables etc
4025    
4026    Returns:             nothing
4027    */
4028    
4029    static void
4030    complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4031    {
4032    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4033    PUT(previous_callout, 2 + LINK_SIZE, length);
4034    }
4035    
4036    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
4037    
4038    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4039    return escape == ESC_d;  /*************************************************
4040    *           Get othercase range                  *
4041    *************************************************/
4042    
4043    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4044    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4045    characters in the "other" case. Each call returns the next one, updating the
4046    start address. A character with multiple other cases is returned on its own
4047    with a special return value.
4048    
4049    case OP_NOT_WHITESPACE:  Arguments:
4050    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
4051      d           end value
4052      ocptr       where to put start of othercase range
4053      odptr       where to put end of othercase range
4054    
4055    case OP_HSPACE:  Yield:        -1 when no more
4056    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
4057           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
4058                    in this case, ocptr contains the original
4059    */
4060    
4061    case OP_NOT_HSPACE:  static int
4062    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4063      pcre_uint32 *odptr)
4064    {
4065    pcre_uint32 c, othercase, next;
4066    unsigned int co;
4067    
4068    /* 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
4069    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
4070    
4071    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4072    return escape == ESC_v || escape == ESC_R;    {
4073      if ((co = UCD_CASESET(c)) != 0)
4074        {
4075        *ocptr = c++;   /* Character that has the set */
4076        *cptr = c;      /* Rest of input range */
4077        return (int)co;
4078        }
4079      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4080      }
4081    
4082    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;  
4083    
4084    case OP_NOT_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4085    return escape == ESC_w || escape == ESC_d;  range, which is either the end of the input range, or a character that has zero
4086    or more than one other cases. */
4087    
4088    default:  *ocptr = othercase;
4089    return FALSE;  next = othercase + 1;
4090    
4091    for (++c; c <= d; c++)
4092      {
4093      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4094      next++;
4095    }    }
4096    
4097  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4098    *cptr = c;             /* Rest of input range */
4099    return 0;
4100  }  }
4101    #endif  /* SUPPORT_UCP */
4102    
4103    
4104    
# Line 3418  switch(op_code) Line 4107  switch(op_code)
4107  *************************************************/  *************************************************/
4108    
4109  /* 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
4110  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
4111  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
4112  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4113    
4114  Arguments:  Arguments:
4115    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4116    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4117    options       the options word    options       the options word
4118    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4119    start         start of range character    start         start of range character
4120    end           end of range character    end           end of range character
4121    
4122  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4123                  the pointer to extra data is updated                  the pointer to extra data is updated
4124  */  */
# Line 3439  add_to_class(pcre_uint8 *classbits, pcre Line 4128  add_to_class(pcre_uint8 *classbits, pcre
4128    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4129  {  {
4130  pcre_uint32 c;  pcre_uint32 c;
4131    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4132  int n8 = 0;  int n8 = 0;
4133    
4134  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4135  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
4136  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
4137  range. */  range. */
4138    
4139  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4140    {    {
4141  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4142    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4143      {      {
4144      int rc;      int rc;
4145      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4146    
4147      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4148      c = start;      c = start;
4149    
4150      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4151        {        {
4152        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4153    
4154        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4155          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4156    
4157        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4158    
4159        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4160    
4161        /* 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
4162        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
4163        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4164    
4165        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4166        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4167        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 4171  if ((options & PCRE_CASELESS) != 0)
4171  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4172    
4173    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4174    
4175    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4176      {      {
4177      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4178      n8++;      n8++;
4179      }      }
4180    }    }
4181    
4182  /* 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
4183  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
4184  in all cases. */  in all cases. */
# Line 3507  in all cases. */ Line 4197  in all cases. */
4197    
4198  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4199    
4200  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
 data. */  
4201    
4202  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4203    {    {
4204    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4205      {    SETBIT(classbits, c);
4206      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4207    }    }
4208    
4209  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4210    {  if (start <= 0xff) start = 0xff + 1;
4211    
4212    if (end >= start)
4213      {
4214    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4215  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4216    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4217      {      {
4218      if (start < end)      if (start < end)
4219        {        {
4220        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4221        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4222        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4223        }        }
4224      else if (start == end)      else if (start == end)
4225        {        {
4226        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4227        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4228        }        }
4229      }      }
4230    else    else
4231  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4232    
4233    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4234    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4235    
4236  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4237      {}      {}
4238  #else  #else
4239    if (start < end)    if (start < end)
4240      {      {
4241      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4246  else
4246      {      {
4247      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4248      *uchardata++ = start;      *uchardata++ = start;
4249      }      }
4250  #endif  #endif
4251    
4252    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4253    }    }
4254    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4255    
4256  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4257  }  }
4258    
4259    
4260    
4261    
4262  /*************************************************  /*************************************************
4263  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4264  *************************************************/  *************************************************/
4265    
4266  /* 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
4267  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
4268  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
4269  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4273  Arguments:
4273    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4274    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4275    options       the options word    options       the options word
4276    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4277    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4278    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4279                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4280                    already know about                    already know about
4281    
4282  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4283                  the pointer to extra data is updated                  the pointer to extra data is updated
4284  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4292  while (p[0] < NOTACHAR)
4292    {    {
4293    int n = 0;    int n = 0;
4294    if (p[0] != except)    if (p[0] != except)
4295      {      {
4296      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4297      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4298      }      }
4299    p += n + 1;    p += n + 1;
4300    }    }
4301  return n8;  return n8;
4302  }  }
4303    
4304    
4305    
# Line 3624  Arguments: Line 4314  Arguments:
4314    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4315    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4316    options       the options word    options       the options word
4317    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4318    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4319    
4320  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4321                  the pointer to extra data is updated                  the pointer to extra data is updated
4322  */  */
4323    
4324  static int  static int
4325  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4326    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4327  {  {
4328  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4334  while (p[0] < NOTACHAR)
4334    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4335    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4336      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4337    p++;    p++;
4338    }    }
4339  return n8;  return n8;
4340  }  }
4341    
4342    
4343    
# Line 3662  to find out the amount of memory needed, Line 4352  to find out the amount of memory needed,
4352  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4353    
4354  Arguments:  Arguments:
4355    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4356    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4357    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4358    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4359    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4360    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4361    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4362    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4363    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4364    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4365                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4366      lengthptr         NULL during the real compile phase
4367                        points to length accumulator during pre-compile phase
4368    
4369  Returns:         TRUE on success  Returns:            TRUE on success
4370                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4371  */  */
4372    
4373  static BOOL  static BOOL
4374  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4375    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4376    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4377      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4378      branch_chain *bcptr, int cond_depth,
4379    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4380  {  {
4381  int repeat_type, op_type;  int repeat_type, op_type;
4382  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4383  int bravalue = 0;  int bravalue = 0;
4384  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4385  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4386  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4387    pcre_uint32 zeroreqchar, zerofirstchar;
4388    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4389  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4390  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4391  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4413  dynamically as we process the pattern. *
4413  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4414  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4415  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4416    #ifndef COMPILE_PCRE32
4417  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4418    #endif
4419  #else  #else
4420  BOOL utf = FALSE;  BOOL utf = FALSE;
4421  #endif  #endif
4422    
4423  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4424  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,
4425  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
4426  alternative calls for the different cases. */  alternative calls for the different cases. */
4427    
4428  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4450  to take the zero repeat into account. Th
4450  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4451  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4452    
4453  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4454    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4455    
4456  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4457  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 3774  for (;; ptr++) Line 4473  for (;; ptr++)
4473    BOOL reset_bracount;    BOOL reset_bracount;
4474    int class_has_8bitchar;    int class_has_8bitchar;
4475    int class_one_char;    int class_one_char;
4476    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4477      BOOL xclass_has_prop;
4478    #endif
4479    int newoptions;    int newoptions;
4480    int recno;    int recno;
4481    int refsign;    int refsign;
4482    int skipbytes;    int skipbytes;
4483    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4484    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4485    int terminator;    int terminator;
4486    int mclength;    unsigned int mclength;
4487    int tempbracount;    unsigned int tempbracount;
4488    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4489    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4490    
4491    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4495  for (;; ptr++)
4495    /* 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
4496    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4497    
4498    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4499      {      {
4500      ptr = nestptr;      ptr = nestptr;
4501      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4570  for (;; ptr++)
4570    
4571    /* 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 */
4572    
4573    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4574      {      {
4575      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4576        {        {
# Line 3891  for (;; ptr++) Line 4593  for (;; ptr++)
4593          }          }
4594        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4595        }        }
4596        /* Control does not reach here. */
4597      }      }
4598    
4599    /* 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
4600    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. */  
4601    
4602    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4603      {      {
4604      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4605        {        {
4606          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4607          if (c != CHAR_NUMBER_SIGN) break;
4608        ptr++;        ptr++;
4609        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4610          {          {
4611          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4612              {                          /* IS_NEWLINE sets cd->nllen. */
4613              ptr += cd->nllen;
4614              break;
4615              }
4616          ptr++;          ptr++;
4617  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4618          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4619  #endif  #endif
4620          }          }
4621        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4622        }        }
4623      }      }
4624    
4625    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4626    
4627      is_quantifier =
4628        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4629        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4630    
4631      /* Fill in length of a previous callout, except when the next thing is a
4632      quantifier or when processing a property substitution string in UCP mode. */
4633    
4634      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4635           after_manual_callout-- <= 0)
4636        {
4637        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4638          complete_callout(previous_callout, ptr, cd);
4639        previous_callout = NULL;
4640        }
4641    
4642      /* Create auto callout, except for quantifiers, or while processing property
4643      strings that are substituted for \w etc in UCP mode. */
4644    
4645    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4646      {      {
4647      previous_callout = code;      previous_callout = code;
4648      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4649      }      }
4650    
4651      /* Process the next pattern item. */
4652    
4653    switch(c)    switch(c)
4654      {      {
4655      /* ===================================================================*/      /* ===================================================================*/
4656      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4657      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4658      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4659      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4660        *firstcharflagsptr = firstcharflags;
4661      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4662        *reqcharflagsptr = reqcharflags;
4663      *codeptr = code;      *codeptr = code;
4664      *ptrptr = ptr;      *ptrptr = ptr;
4665      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4683  for (;; ptr++)
4683      previous = NULL;      previous = NULL;
4684      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4685        {        {
4686        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4687        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4688        }        }
4689      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4698  for (;; ptr++)
4698      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4699    
4700      case CHAR_DOT:      case CHAR_DOT:
4701      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4702      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4703        zerofirstcharflags = firstcharflags;
4704      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4705        zeroreqcharflags = reqcharflags;
4706      previous = code;      previous = code;
4707      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4708      break;      break;
# Line 4016  for (;; ptr++) Line 4731  for (;; ptr++)
4731        }        }
4732      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4733    
4734        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4735        used for "start of word" and "end of word". As these are otherwise illegal
4736        sequences, we don't break anything by recognizing them. They are replaced
4737        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4738        erroneous and are handled by the normal code below. */
4739    
4740      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4741        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4742          {
4743          nestptr = ptr + 7;
4744          ptr = sub_start_of_word - 1;
4745          continue;
4746          }
4747    
4748        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4749          {
4750          nestptr = ptr + 7;
4751          ptr = sub_end_of_word - 1;
4752          continue;
4753          }
4754    
4755        /* Handle a real character class. */
4756    
4757      previous = code;      previous = code;
4758    
4759      /* 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 4798  for (;; ptr++)
4798          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4799        {        {
4800        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4801        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4802        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4803          zerofirstcharflags = firstcharflags;
4804        break;        break;
4805        }        }
4806    
# Line 4072  for (;; ptr++) Line 4810  for (;; ptr++)
4810    
4811      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4812    
4813        /* Extended class (xclass) will be used when characters > 255
4814        might match. */
4815    
4816    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4817        xclass = FALSE;
4818        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4819        class_uchardata_base = class_uchardata;   /* Save the start */
4820    #endif
4821    
4822      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4823      class_has_8bitchar will be non-zero if the class contains at least one <      class_has_8bitchar will be non-zero if the class contains at least one <
4824      256 character; class_one_char will be 1 if the class contains just one      256 character; class_one_char will be 1 if the class contains just one
4825      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4826        are present in the class. */
4827    
4828      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4829      class_one_char = 0;      class_one_char = 0;
4830    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4831        xclass_has_prop = FALSE;
4832    #endif
4833    
4834      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
4835      temporary bit of memory, in case the class contains fewer than two      temporary bit of memory, in case the class contains fewer than two
# Line 4087  for (;; ptr++) Line 4838  for (;; ptr++)
4838    
4839      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4840    
 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  
     xclass = FALSE;  
     class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */  
     class_uchardata_base = class_uchardata;   /* Save the start */  
 #endif  
   
4841      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4842      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
4843      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4844    
4845      if (c != 0) do      if (c != CHAR_NULL) do
4846        {        {
4847        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4848    
4849  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
4850        if (utf && HAS_EXTRALEN(c))        if (utf && HAS_EXTRALEN(c))
4851          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4852          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
# Line 4112  for (;; ptr++) Line 4857  for (;; ptr++)
4857        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4858        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
4859        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4860        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4861        however. */        however. */
4862    
4863        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4176  for (;; ptr++) Line 4921  for (;; ptr++)
4921          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4922          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4923    
4924          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4925            posix_class = 0;            posix_class = 0;
4926    
4927          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
4928          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4929            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4930            directly. */
4931    
4932  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4933          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4934            {            {
4935              unsigned int ptype = 0;
4936            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4937    
4938              /* The posix_substitutes table specifies which POSIX classes can be