/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1072 by chpe, Tue Oct 16 15:54:40 2012 UTC revision 1480 by ph10, Tue May 27 18:02:51 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 1308  read_repeat_counts(const pcre_uchar *p, Line 1583  read_repeat_counts(const pcre_uchar *p,
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
1585    
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1587    {    {
1588    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1589    return p;    if (min > 65535)
1590    }      {
1591        *errorcodeptr = ERR5;
1592  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1593  Also, max must not be less than min. */      }
1594      }
1595    
1596  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1597    {    {
1598    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1599      {      {
1600      max = 0;      max = 0;
1601      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1602        {        {
1603        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1604        return p;        if (max > 65535)
1605        }          {
1606            *errorcodeptr = ERR5;
1607            return p;
1608            }
1609          }
1610      if (max < min)      if (max < min)
1611        {        {
1612        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1340  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1615  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1615      }      }
1616    }    }
1617    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1618  *minp = min;  *minp = min;
1619  *maxp = max;  *maxp = max;
1620  return p;  return p;
# Line 1351  return p; Line 1623  return p;
1623    
1624    
1625  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1626  *      Find first significant op code            *  *      Find first significant op code            *
1627  *************************************************/  *************************************************/
1628    
# Line 1684  for (;;) Line 1661  for (;;)
1661    
1662      case OP_CALLOUT:      case OP_CALLOUT:
1663      case OP_CREF:      case OP_CREF:
1664      case OP_NCREF:      case OP_DNCREF:
1665      case OP_RREF:      case OP_RREF:
1666      case OP_NRREF:      case OP_DNRREF:
1667      case OP_DEF:      case OP_DEF:
1668      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1669      break;      break;
# Line 1700  for (;;) Line 1677  for (;;)
1677    
1678    
1679    
   
1680  /*************************************************  /*************************************************
1681  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1682  *************************************************/  *************************************************/
# Line 1824  for (;;) Line 1800  for (;;)
1800      case OP_COMMIT:      case OP_COMMIT:
1801      case OP_CREF:      case OP_CREF:
1802      case OP_DEF:      case OP_DEF:
1803        case OP_DNCREF:
1804        case OP_DNRREF:
1805      case OP_DOLL:      case OP_DOLL:
1806      case OP_DOLLM:      case OP_DOLLM:
1807      case OP_EOD:      case OP_EOD:
1808      case OP_EODN:      case OP_EODN:
1809      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1810      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1811      case OP_PRUNE:      case OP_PRUNE:
1812      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1828  for (;;)
1828      case OP_NOTI:      case OP_NOTI:
1829      branchlength++;      branchlength++;
1830      cc += 2;      cc += 2;
1831  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1832      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1833  #endif  #endif
1834      break;      break;
# Line 1866  for (;;) Line 1842  for (;;)
1842      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1843      branchlength += (int)GET2(cc,1);      branchlength += (int)GET2(cc,1);
1844      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1845  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1846      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1847  #endif  #endif
1848      break;      break;
1849    
1850      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1851      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1852      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)
1853        cc += 2;        cc += 2;
1854      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1855      break;      break;
# Line 1909  for (;;) Line 1885  for (;;)
1885    
1886      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1887    
1888        case OP_CLASS:
1889        case OP_NCLASS:
1890  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1891      case OP_XCLASS:      case OP_XCLASS:
1892      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];      /* The original code caused an unsigned overflow in 64 bit systems,
1893      /* Fall through */      so now we use a conditional statement. */
1894        if (op == OP_XCLASS)
1895          cc += GET(cc, 1);
1896        else
1897          cc += PRIV(OP_lengths)[OP_CLASS];
1898    #else
1899        cc += PRIV(OP_lengths)[OP_CLASS];
1900  #endif  #endif
1901    
     case OP_CLASS:  
     case OP_NCLASS:  
     cc += PRIV(OP_lengths)[OP_CLASS];  
   
1902      switch (*cc)      switch (*cc)
1903        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1904        case OP_CRSTAR:        case OP_CRSTAR:
1905        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1906          case OP_CRPLUS:
1907          case OP_CRMINPLUS:
1908        case OP_CRQUERY:        case OP_CRQUERY:
1909        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1910          case OP_CRPOSSTAR:
1911          case OP_CRPOSPLUS:
1912          case OP_CRPOSQUERY:
1913        return -1;        return -1;
1914    
1915        case OP_CRRANGE:        case OP_CRRANGE:
1916        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1917          case OP_CRPOSRANGE:
1918        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1919        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1920        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 1999  for (;;) Line 1983  for (;;)
1983      case OP_QUERYI:      case OP_QUERYI:
1984      case OP_REF:      case OP_REF:
1985      case OP_REFI:      case OP_REFI:
1986        case OP_DNREF:
1987        case OP_DNREFI:
1988      case OP_SBRA:      case OP_SBRA:
1989      case OP_SBRAPOS:      case OP_SBRAPOS:
1990      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2021  for (;;)
2021    
2022    
2023    
   
2024  /*************************************************  /*************************************************
2025  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2026  *************************************************/  *************************************************/
# Line 2112  for (;;) Line 2097  for (;;)
2097        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2098        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2099        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2100        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)
2101          code += 2;          code += 2;
2102        break;        break;
2103    
2104        case OP_MARK:        case OP_MARK:
2105        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2106        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2107        case OP_THEN_ARG:        case OP_THEN_ARG:
2108        code += code[1];        code += code[1];
2109        break;        break;
# Line 2232  for (;;) Line 2214  for (;;)
2214        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2215        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2216        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2217        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)
2218          code += 2;          code += 2;
2219        break;        break;
2220    
2221        case OP_MARK:        case OP_MARK:
2222        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2223        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2224        case OP_THEN_ARG:        case OP_THEN_ARG:
2225        code += code[1];        code += code[1];
2226        break;        break;
# Line 2343  Arguments: Line 2322  Arguments:
2322    endcode     points to where to stop    endcode     points to where to stop
2323    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2324    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2325      recurses    chain of recurse_check to catch mutual recursion
2326    
2327  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2328  */  */
2329    
2330    typedef struct recurse_check {
2331      struct recurse_check *prev;
2332      const pcre_uchar *group;
2333    } recurse_check;
2334    
2335  static BOOL  static BOOL
2336  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2337    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2338  {  {
2339  register pcre_uchar c;  register pcre_uchar c;
2340    recurse_check this_recurse;
2341    
2342  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2343       code < endcode;       code < endcode;
2344       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 2366  for (code = first_significant_code(code
2366    
2367    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2368      {      {
2369      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2370      BOOL empty_branch;      BOOL empty_branch;
2371    
2372      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2373        when called to scan a completed pattern by setting cd->start_workspace to
2374        NULL. */
2375    
2376        if (cd->start_workspace != NULL)
2377          {
2378          const pcre_uchar *tcode;
2379          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2380            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2381          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2382          }
2383    
2384        /* If we are scanning a completed pattern, there are no forward references
2385        and all groups are complete. We need to detect whether this is a recursive
2386        call, as otherwise there will be an infinite loop. If it is a recursion,
2387        just skip over it. Simple recursions are easily detected. For mutual
2388        recursions we keep a chain on the stack. */
2389    
2390        else
2391          {
2392          recurse_check *r = recurses;
2393          const pcre_uchar *endgroup = scode;
2394    
2395          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2396          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2397    
2398      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        for (r = recurses; r != NULL; r = r->prev)
2399        if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;          if (r->group == scode) break;
2400          if (r != NULL) continue;   /* Mutual recursion */
2401          }
2402    
2403      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2404        stack chain to detect mutual recursions. */
2405    
2406      empty_branch = FALSE;      empty_branch = FALSE;
2407      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2408      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
2409    
2410      do      do
2411        {        {
2412        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2413          {          {
2414          empty_branch = TRUE;          empty_branch = TRUE;
2415          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2465  for (code = first_significant_code(code
2465        empty_branch = FALSE;        empty_branch = FALSE;
2466        do        do
2467          {          {
2468          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2469            empty_branch = TRUE;            empty_branch = TRUE;
2470          code += GET(code, 1);          code += GET(code, 1);
2471          }          }
# Line 2495  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2508        case OP_CRQUERY:        case OP_CRQUERY:
2509        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2510          case OP_CRPOSSTAR:
2511          case OP_CRPOSQUERY:
2512        break;        break;
2513    
2514        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2515        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2516        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2517          case OP_CRPOSPLUS:
2518        return FALSE;        return FALSE;
2519    
2520        case OP_CRRANGE:        case OP_CRRANGE:
2521        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2522          case OP_CRPOSRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2524        break;        break;
2525        }        }
# Line 2511  for (code = first_significant_code(code Line 2527  for (code = first_significant_code(code
2527    
2528      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2529    
2530        case OP_ANY:
2531        case OP_ALLANY:
2532        case OP_ANYBYTE:
2533    
2534      case OP_PROP:      case OP_PROP:
2535      case OP_NOTPROP:      case OP_NOTPROP:
2536        case OP_ANYNL:
2537    
2538        case OP_NOT_HSPACE:
2539        case OP_HSPACE:
2540        case OP_NOT_VSPACE:
2541        case OP_VSPACE:
2542      case OP_EXTUNI:      case OP_EXTUNI:
2543    
2544      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2545      case OP_DIGIT:      case OP_DIGIT:
2546      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2547      case OP_WHITESPACE:      case OP_WHITESPACE:
2548      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2549      case OP_WORDCHAR:      case OP_WORDCHAR:
2550      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2551      case OP_CHAR:      case OP_CHAR:
2552      case OP_CHARI:      case OP_CHARI:
2553      case OP_NOT:      case OP_NOT:
2554      case OP_NOTI:      case OP_NOTI:
2555    
2556      case OP_PLUS:      case OP_PLUS:
2557        case OP_PLUSI:
2558      case OP_MINPLUS:      case OP_MINPLUS:
2559      case OP_POSPLUS:      case OP_MINPLUSI:
2560      case OP_EXACT:  
2561      case OP_NOTPLUS:      case OP_NOTPLUS:
2562        case OP_NOTPLUSI:
2563      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2564        case OP_NOTMINPLUSI:
2565    
2566        case OP_POSPLUS:
2567        case OP_POSPLUSI:
2568      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2569        case OP_NOTPOSPLUSI:
2570    
2571        case OP_EXACT:
2572        case OP_EXACTI:
2573      case OP_NOTEXACT:      case OP_NOTEXACT:
2574        case OP_NOTEXACTI:
2575    
2576      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2577      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2578      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2579      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2580    
2581      return FALSE;      return FALSE;
2582    
2583      /* 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 2597  for (code = first_significant_code(code
2597      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2598      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2599      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2600      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)
2601        code += 2;        code += 2;
2602      break;      break;
2603    
# Line 2572  for (code = first_significant_code(code Line 2611  for (code = first_significant_code(code
2611      return TRUE;      return TRUE;
2612    
2613      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2614      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2615        followed by a multibyte character. */
2616    
2617  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2618      case OP_STAR:      case OP_STAR:
2619      case OP_STARI:      case OP_STARI:
2620        case OP_NOTSTAR:
2621        case OP_NOTSTARI:
2622    
2623      case OP_MINSTAR:      case OP_MINSTAR:
2624      case OP_MINSTARI:      case OP_MINSTARI:
2625        case OP_NOTMINSTAR:
2626        case OP_NOTMINSTARI:
2627    
2628      case OP_POSSTAR:      case OP_POSSTAR:
2629      case OP_POSSTARI:      case OP_POSSTARI:
2630        case OP_NOTPOSSTAR:
2631        case OP_NOTPOSSTARI:
2632    
2633      case OP_QUERY:      case OP_QUERY:
2634      case OP_QUERYI:      case OP_QUERYI:
2635        case OP_NOTQUERY:
2636        case OP_NOTQUERYI:
2637    
2638      case OP_MINQUERY:      case OP_MINQUERY:
2639      case OP_MINQUERYI:      case OP_MINQUERYI:
2640        case OP_NOTMINQUERY:
2641        case OP_NOTMINQUERYI:
2642    
2643      case OP_POSQUERY:      case OP_POSQUERY:
2644      case OP_POSQUERYI:      case OP_POSQUERYI:
2645        case OP_NOTPOSQUERY:
2646        case OP_NOTPOSQUERYI:
2647    
2648      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2649      break;      break;
2650    
2651      case OP_UPTO:      case OP_UPTO:
2652      case OP_UPTOI:      case OP_UPTOI:
2653        case OP_NOTUPTO:
2654        case OP_NOTUPTOI:
2655    
2656      case OP_MINUPTO:      case OP_MINUPTO:
2657      case OP_MINUPTOI:      case OP_MINUPTOI:
2658        case OP_NOTMINUPTO:
2659        case OP_NOTMINUPTOI:
2660    
2661      case OP_POSUPTO:      case OP_POSUPTO:
2662      case OP_POSUPTOI:      case OP_POSUPTOI:
2663        case OP_NOTPOSUPTO:
2664        case OP_NOTPOSUPTOI:
2665    
2666      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]);
2667      break;      break;
2668  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2673  for (code = first_significant_code(code
2673      case OP_MARK:      case OP_MARK:
2674      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2675      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2676      case OP_THEN_ARG:      case OP_THEN_ARG:
2677      code += code[1];      code += code[1];
2678      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2716  could_be_empty(const pcre_uchar *code, c
2716  {  {
2717  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2718    {    {
2719    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2720      return FALSE;      return FALSE;
2721    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2722    }    }
# Line 2662  return TRUE; Line 2726  return TRUE;
2726    
2727    
2728  /*************************************************  /*************************************************
2729  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2730  *************************************************/  *************************************************/
2731    
2732  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2733  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.  
2734    
2735  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2736  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2737  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:].  
2738    
2739  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2740  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2741  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2742  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2743  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2744  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2745  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2746  I think.         (c >= OP_STARI)?      OP_STARI :
2747                                 OP_STAR;
2748    }
2749    
 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.  
2750    
2751  In Perl, unescaped square brackets may also appear as part of class names. For  
2752  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  #ifdef SUPPORT_UCP
2753  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  /*************************************************
2754  seem right at all. PCRE does not allow closing square brackets in POSIX class  *        Check a character and a property        *
2755  names.  *************************************************/
2756    
2757    /* This function is called by check_auto_possessive() when a property item
2758    is adjacent to a fixed character.
2759    
2760  Arguments:  Arguments:
2761    ptr      pointer to the initial [    c            the character
2762    endptr   where to return the end pointer    ptype        the property type
2763      pdata        the data for the type
2764      negated      TRUE if it's a negated property (\P or \p{^)
2765    
2766  Returns:   TRUE or FALSE  Returns:       TRUE if auto-possessifying is OK
2767  */  */
2768    
2769  static BOOL  static BOOL
2770  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2771      BOOL negated)
2772  {  {
2773  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2774  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2775  for (++ptr; *ptr != 0; ptr++)  
2776    switch(ptype)
2777    {    {
2778    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    case PT_LAMP:
2779      ptr++;    return (prop->chartype == ucp_Lu ||
2780    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;            prop->chartype == ucp_Ll ||
2781    else            prop->chartype == ucp_Lt) == negated;
     {  
     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;  
 }  
2782    
2783      case PT_GC:
2784      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2785    
2786      case PT_PC:
2787      return (pdata == prop->chartype) == negated;
2788    
2789      case PT_SC:
2790      return (pdata == prop->script) == negated;
2791    
2792  /*************************************************    /* These are specials */
 *          Check POSIX class name                *  
 *************************************************/  
2793    
2794  /* This function is called to check the name given in a POSIX-style class entry    case PT_ALNUM:
2795  such as [:alnum:].    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2796              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2797    
2798  Arguments:    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2799    ptr        points to the first letter    means that Perl space and POSIX space are now identical. PCRE was changed
2800    len        the length of the name    at release 8.34. */
2801    
2802  Returns:     a value representing the name, or -1 if unknown    case PT_SPACE:    /* Perl space */
2803  */    case PT_PXSPACE:  /* POSIX space */
2804      switch(c)
2805        {
2806        HSPACE_CASES:
2807        VSPACE_CASES:
2808        return negated;
2809    
2810  static int      default:
2811  check_posix_name(const pcre_uchar *ptr, int len)      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2812  {      }
2813  const char *pn = posix_names;    break;  /* Control never reaches here */
2814  register int yield = 0;  
2815  while (posix_name_lengths[yield] != 0)    case PT_WORD:
2816    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2817    if (len == posix_name_lengths[yield] &&            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2818      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;            c == CHAR_UNDERSCORE) == negated;
2819    pn += posix_name_lengths[yield] + 1;  
2820    yield++;    case PT_CLIST:
2821      p = PRIV(ucd_caseless_sets) + prop->caseset;
2822      for (;;)
2823        {
2824        if (c < *p) return !negated;
2825        if (c == *p++) return negated;
2826        }
2827      break;  /* Control never reaches here */
2828    }    }
2829  return -1;  
2830    return FALSE;
2831  }  }
2832    #endif  /* SUPPORT_UCP */
2833    
2834    
2835    
2836  /*************************************************  /*************************************************
2837  *    Adjust OP_RECURSE items in repeated group   *  *        Fill the character property list        *
2838  *************************************************/  *************************************************/
2839    
2840  /* OP_RECURSE items contain an offset from the start of the regex to the group  /* Checks whether the code points to an opcode that can take part in auto-
2841  that is referenced. This means that groups can be replicated for fixed  possessification, and if so, fills a list with its properties.
 repetition simply by copying (because the recursion is allowed to refer to  
 earlier groups that are outside the current group). However, when a group is  
 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is  
 inserted before it, after it has been compiled. This means that any OP_RECURSE  
 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.  
   
 This function has been extended with the possibility of forward references for  
 recursions and subroutine calls. It must also check the list of such references  
 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).  
2842    
2843  Arguments:  Arguments:
2844    group      points to the start of the group    code        points to start of expression
2845    adjust     the amount by which the group is to be moved    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2846    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode    fcc         points to case-flipping table
2847    cd         contains pointers to tables etc.    list        points to output list
2848    save_hwm   the hwm forward reference pointer at the start of the group                list[0] will be filled with the opcode
2849                  list[1] will be non-zero if this opcode
2850                    can match an empty character string
2851                  list[2..7] depends on the opcode
2852    
2853  Returns:     nothing  Returns:      points to the start of the next opcode if *code is accepted
2854                  NULL if *code is not accepted
2855  */  */
2856    
2857  static void  static const pcre_uchar *
2858  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2859    pcre_uchar *save_hwm)    const pcre_uint8 *fcc, pcre_uint32 *list)
2860  {  {
2861  pcre_uchar *ptr = group;  pcre_uchar c = *code;
2862    pcre_uchar base;
2863    const pcre_uchar *end;
2864    pcre_uint32 chr;
2865    
2866  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  #ifdef SUPPORT_UCP
2867    pcre_uint32 *clist_dest;
2868    const pcre_uint32 *clist_src;
2869    #else
2870    utf = utf;  /* Suppress "unused parameter" compiler warning */
2871    #endif
2872    
2873    list[0] = c;
2874    list[1] = FALSE;
2875    code++;
2876    
2877    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2878    {    {
2879    int offset;    base = get_repeat_base(c);
2880    pcre_uchar *hc;    c -= (base - OP_STAR);
2881    
2882    /* See if this recursion is on the forward reference list. If so, adjust the    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2883    reference. */      code += IMM2_SIZE;
2884    
2885    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2886    
2887      switch(base)
2888      {      {
2889      offset = GET(hc, 0);      case OP_STAR:
2890      if (cd->start_code + offset == ptr + 1)      list[0] = OP_CHAR;
2891        {      break;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
     }  
   
   /* Otherwise, adjust the recursion offset if it's after the start of this  
   group. */  
   
   if (hc >= cd->hwm)  
     {  
     offset = GET(ptr, 1);  
     if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);  
     }  
   
   ptr += 1 + LINK_SIZE;  
   }  
 }  
   
   
   
 /*************************************************  
 *        Insert an automatic callout point       *  
 *************************************************/  
   
 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  
 callout points before each pattern item.  
   
 Arguments:  
   code           current code pointer  
   ptr            current pattern pointer  
   cd             pointers to tables etc  
   
 Returns:         new code pointer  
 */  
2892    
2893  static pcre_uchar *      case OP_STARI:
2894  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_CHARI;
2895  {      break;
 *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;  
 }  
2896    
2897        case OP_NOTSTAR:
2898        list[0] = OP_NOT;
2899        break;
2900    
2901        case OP_NOTSTARI:
2902        list[0] = OP_NOTI;
2903        break;
2904    
2905  /*************************************************      case OP_TYPESTAR:
2906  *         Complete a callout item                *      list[0] = *code;
2907  *************************************************/      code++;
2908        break;
2909        }
2910      c = list[0];
2911      }
2912    
2913  /* A callout item contains the length of the next item in the pattern, which  switch(c)
2914  we can't fill in till after we have reached the relevant point. This is used    {
2915  for both automatic and manual callouts.    case OP_NOT_DIGIT:
2916      case OP_DIGIT:
2917      case OP_NOT_WHITESPACE:
2918      case OP_WHITESPACE:
2919      case OP_NOT_WORDCHAR:
2920      case OP_WORDCHAR:
2921      case OP_ANY:
2922      case OP_ALLANY:
2923      case OP_ANYNL:
2924      case OP_NOT_HSPACE:
2925      case OP_HSPACE:
2926      case OP_NOT_VSPACE:
2927      case OP_VSPACE:
2928      case OP_EXTUNI:
2929      case OP_EODN:
2930      case OP_EOD:
2931      case OP_DOLL:
2932      case OP_DOLLM:
2933      return code;
2934    
2935  Arguments:    case OP_CHAR:
2936    previous_callout   points to previous callout item    case OP_NOT:
2937    ptr                current pattern pointer    GETCHARINCTEST(chr, code);
2938    cd                 pointers to tables etc    list[2] = chr;
2939      list[3] = NOTACHAR;
2940      return code;
2941    
2942  Returns:             nothing    case OP_CHARI:
2943  */    case OP_NOTI:
2944      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2945      GETCHARINCTEST(chr, code);
2946      list[2] = chr;
2947    
2948  static void  #ifdef SUPPORT_UCP
2949  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)    if (chr < 128 || (chr < 256 && !utf))
2950  {      list[3] = fcc[chr];
2951  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));    else
2952  PUT(previous_callout, 2 + LINK_SIZE, length);      list[3] = UCD_OTHERCASE(chr);
2953  }  #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2954      list[3] = (chr < 256) ? fcc[chr] : chr;
2955    #else
2956      list[3] = fcc[chr];
2957    #endif
2958    
2959      /* The othercase might be the same value. */
2960    
2961      if (chr == list[3])
2962        list[3] = NOTACHAR;
2963      else
2964        list[4] = NOTACHAR;
2965      return code;
2966    
2967  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2968  /*************************************************    case OP_PROP:
2969  *           Get othercase range                  *    case OP_NOTPROP:
2970  *************************************************/    if (code[0] != PT_CLIST)
2971        {
2972        list[2] = code[0];
2973        list[3] = code[1];
2974        return code + 2;
2975        }
2976    
2977  /* This function is passed the start and end of a class range, in UTF-8 mode    /* Convert only if we have enough space. */
 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.  
2978    
2979  Arguments:    clist_src = PRIV(ucd_caseless_sets) + code[1];
2980    cptr        points to starting character value; updated    clist_dest = list + 2;
2981    d           end value    code += 2;
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
2982    
2983  Yield:        -1 when no more    do {
2984                 0 when a range is returned       if (clist_dest >= list + 8)
2985                >0 the CASESET offset for char with multiple other cases         {
2986                  in this case, ocptr contains the original         /* Early return if there is not enough space. This should never
2987  */         happen, since all clists are shorter than 5 character now. */
2988           list[2] = code[0];
2989           list[3] = code[1];
2990           return code;
2991           }
2992         *clist_dest++ = *clist_src;
2993         }
2994      while(*clist_src++ != NOTACHAR);
2995    
2996  static int    /* All characters are stored. The terminating NOTACHAR
2997  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,    is copied form the clist itself. */
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2998    
2999  /* Find the first character that has an other case. If it has multiple other    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3000  cases, return its case offset value. */    return code;
3001    #endif
3002    
3003  for (c = *cptr; c <= d; c++)    case OP_NCLASS:
3004    {    case OP_CLASS:
3005    if ((co = UCD_CASESET(c)) != 0)  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3006      {    case OP_XCLASS:
3007      *ocptr = c++;   /* Character that has the set */    if (c == OP_XCLASS)
3008      *cptr = c;      /* Rest of input range */      end = code + GET(code, 0) - 1;
3009      return co;    else
3010      }  #endif
3011    if ((othercase = UCD_OTHERCASE(c)) != c) break;      end = code + 32 / sizeof(pcre_uchar);
   }  
3012    
3013  if (c > d) return -1;  /* Reached end of range */    switch(*end)
3014        {
3015        case OP_CRSTAR:
3016        case OP_CRMINSTAR:
3017        case OP_CRQUERY:
3018        case OP_CRMINQUERY:
3019        case OP_CRPOSSTAR:
3020        case OP_CRPOSQUERY:
3021        list[1] = TRUE;
3022        end++;
3023        break;
3024    
3025  *ocptr = othercase;      case OP_CRPLUS:
3026  next = othercase + 1;      case OP_CRMINPLUS:
3027        case OP_CRPOSPLUS:
3028        end++;
3029        break;
3030    
3031  for (++c; c <= d; c++)      case OP_CRRANGE:
3032    {      case OP_CRMINRANGE:
3033    if (UCD_OTHERCASE(c) != next) break;      case OP_CRPOSRANGE:
3034    next++;      list[1] = (GET2(end, 1) == 0);
3035        end += 1 + 2 * IMM2_SIZE;
3036        break;
3037        }
3038      list[2] = (pcre_uint32)(end - code);
3039      return end;
3040    }    }
3041    return NULL;    /* Opcode not accepted */
 *odptr = next - 1;     /* End of othercase range */  
 *cptr = c;             /* Rest of input range */  
 return 0;  
3042  }  }
3043    
3044    
3045    
3046  /*************************************************  /*************************************************
3047  *        Check a character and a property        *  *    Scan further character sets for match       *
3048  *************************************************/  *************************************************/
3049    
3050  /* This function is called by check_auto_possessive() when a property item  /* Checks whether the base and the current opcode have a common character, in
3051  is adjacent to a fixed character.  which case the base cannot be possessified.
3052    
3053  Arguments:  Arguments:
3054    c            the character    code        points to the byte code
3055    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3056    pdata        the data for the type    cd          static compile data
3057    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3058    
3059  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3060  */  */
3061    
3062  static BOOL  static BOOL
3063  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3064      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3065  {  {
3066  #ifdef SUPPORT_UCP  pcre_uchar c;
3067  const pcre_uint32 *p;  pcre_uint32 list[8];
3068    const pcre_uint32 *chr_ptr;
3069    const pcre_uint32 *ochr_ptr;
3070    const pcre_uint32 *list_ptr;
3071    const pcre_uchar *next_code;
3072    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3073    const pcre_uchar *xclass_flags;
3074  #endif  #endif
3075    const pcre_uint8 *class_bitset;
3076    const pcre_uint8 *set1, *set2, *set_end;
3077    pcre_uint32 chr;
3078    BOOL accepted, invert_bits;
3079    BOOL entered_a_group = FALSE;
3080    
3081    /* Note: the base_list[1] contains whether the current opcode has greedy
3082    (represented by a non-zero value) quantifier. This is a different from
3083    other character type lists, which stores here that the character iterator
3084    matches to an empty string (also represented by a non-zero value). */
3085    
3086  const ucd_record *prop = GET_UCD(c);  for(;;)
   
 switch(ptype)  
3087    {    {
3088    case PT_LAMP:    /* All operations move the code pointer forward.
3089    return (prop->chartype == ucp_Lu ||    Therefore infinite recursions are not possible. */
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3090    
3091    case PT_GC:    c = *code;
3092    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3093      /* Skip over callouts */
3094    
3095    case PT_PC:    if (c == OP_CALLOUT)
3096    return (pdata == prop->chartype) == negated;      {
3097        code += PRIV(OP_lengths)[c];
3098        continue;
3099        }
3100    
3101    case PT_SC:    if (c == OP_ALT)
3102    return (pdata == prop->script) == negated;      {
3103        do code += GET(code, 1); while (*code == OP_ALT);
3104        c = *code;
3105        }
3106    
3107    /* These are specials */    switch(c)
3108        {
3109        case OP_END:
3110        case OP_KETRPOS:
3111        /* TRUE only in greedy case. The non-greedy case could be replaced by
3112        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3113        uses more memory, which we cannot get at this stage.) */
3114    
3115    case PT_ALNUM:      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3116    
3117    case PT_SPACE:    /* Perl space */      case OP_KET:
3118    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3119            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3120            == negated;      cannot be converted to a possessive form. */
3121    
3122        if (base_list[1] == 0) return FALSE;
3123    
3124    case PT_PXSPACE:  /* POSIX space */      switch(*(code - GET(code, 1)))
3125    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        {
3126            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||        case OP_ASSERT:
3127            c == CHAR_FF || c == CHAR_CR)        case OP_ASSERT_NOT:
3128            == negated;        case OP_ASSERTBACK:
3129          case OP_ASSERTBACK_NOT:
3130          case OP_ONCE:
3131          case OP_ONCE_NC:
3132          /* Atomic sub-patterns and assertions can always auto-possessify their
3133          last iterator. However, if the group was entered as a result of checking
3134          a previous iterator, this is not possible. */
3135    
3136    case PT_WORD:        return !entered_a_group;
3137    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 (;;)  
     {  
     if ((unsigned int)c < *p) return !negated;  
     if ((unsigned int)c == *p++) return negated;  
     }  
   break;  /* Control never reaches here */  
 #endif  
   }  
3138    
3139  return FALSE;      code += PRIV(OP_lengths)[c];
3140  }      continue;
 #endif  /* SUPPORT_UCP */  
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          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3152          code = next_code + 1 + LINK_SIZE;
3153          next_code += GET(next_code, 1);
3154          }
3155    
3156  /*************************************************      entered_a_group = TRUE;
3157  *     Check if auto-possessifying is possible    *      continue;
 *************************************************/  
3158    
3159  /* This function is called for unlimited repeats of certain items, to see      case OP_BRAZERO:
3160  whether the next thing could possibly match the repeated item. If not, it makes      case OP_BRAMINZERO:
 sense to automatically possessify the repeated item.  
3161    
3162  Arguments:      next_code = code + 1;
3163    previous      pointer to the repeated opcode      if (*next_code != OP_BRA && *next_code != OP_CBRA
3164    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3165    ptr           next character in pattern  
3166    options       options bits      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3167    cd            contains pointers to tables etc.  
3168        /* The bracket content will be checked by the
3169        OP_BRA/OP_CBRA case above. */
3170        next_code += 1 + LINK_SIZE;
3171        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3172          return FALSE;
3173    
3174  Returns:        TRUE if possessifying is wanted      code += PRIV(OP_lengths)[c];
3175  */      continue;
3176    
3177  static BOOL      default:
3178  check_auto_possessive(const pcre_uchar *previous, BOOL utf,      break;
3179    const pcre_uchar *ptr, int options, compile_data *cd)      }
 {  
 pcre_uint32 c = NOTACHAR;  
 pcre_uint32 next;  
 int escape;  
 int op_code = *previous++;  
3180    
3181  /* Skip whitespace and comments in extended mode */    /* Check for a supported opcode, and load its properties. */
3182    
3183  if ((options & PCRE_EXTENDED) != 0)    code = get_chr_property_list(code, utf, cd->fcc, list);
3184    {    if (code == NULL) return FALSE;    /* Unsupported */
3185    for (;;)  
3186      /* If either opcode is a small character list, set pointers for comparing
3187      characters from that list with another list, or with a property. */
3188    
3189      if (base_list[0] == OP_CHAR)
3190      {      {
3191      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3192      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3193        {      }
3194        ptr++;    else if (list[0] == OP_CHAR)
3195        while (*ptr != 0)      {
3196          {      chr_ptr = list + 2;
3197          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }      list_ptr = base_list;
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3198      }      }
   }  
3199    
3200  /* If the next item is one that we can handle, get its value. A non-negative    /* Character bitsets can also be compared to certain opcodes. */
 value is a character, a negative value is an escape value. */  
3201    
3202  if (*ptr == CHAR_BACKSLASH)    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3203    {  #ifdef COMPILE_PCRE8
3204    int temperrorcode = 0;        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3205    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
3206  #endif  #endif
3207    next = *ptr++;        )
3208    }      {
3209  else return FALSE;  #ifdef COMPILE_PCRE8
3210        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3211    #else
3212        if (base_list[0] == OP_CLASS)
3213    #endif
3214          {
3215          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3216          list_ptr = list;
3217          }
3218        else
3219          {
3220          set1 = (pcre_uint8 *)(code - list[2]);
3221          list_ptr = base_list;
3222          }
3223    
3224  /* Skip whitespace and comments in extended mode */      invert_bits = FALSE;
3225        switch(list_ptr[0])
3226          {
3227          case OP_CLASS:
3228          case OP_NCLASS:
3229          set2 = (pcre_uint8 *)
3230            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3231          break;
3232    
3233  if ((options & PCRE_EXTENDED) != 0)  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3234    {        case OP_XCLASS:
3235    for (;;)        xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3236      {        if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3237      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;        if ((*xclass_flags & XCL_MAP) == 0)
3238      if (*ptr == CHAR_NUMBER_SIGN)          {
3239            /* No bits are set for characters < 256. */
3240            if (list[1] == 0) return TRUE;
3241            /* Might be an empty repeat. */
3242            continue;
3243            }
3244          set2 = (pcre_uint8 *)(xclass_flags + 1);
3245          break;
3246    #endif
3247    
3248          case OP_NOT_DIGIT:
3249          invert_bits = TRUE;
3250          /* Fall through */
3251          case OP_DIGIT:
3252          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3253          break;
3254    
3255          case OP_NOT_WHITESPACE:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_WHITESPACE:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3260          break;
3261    
3262          case OP_NOT_WORDCHAR:
3263          invert_bits = TRUE;
3264          /* Fall through */
3265          case OP_WORDCHAR:
3266          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3267          break;
3268    
3269          default:
3270          return FALSE;
3271          }
3272    
3273        /* Because the sets are unaligned, we need
3274        to perform byte comparison here. */
3275        set_end = set1 + 32;
3276        if (invert_bits)
3277        {        {
3278        ptr++;        do
       while (*ptr != 0)  
3279          {          {
3280          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3281          ptr++;          }
3282  #ifdef SUPPORT_UTF        while (set1 < set_end);
3283          if (utf) FORWARDCHAR(ptr);        }
3284  #endif      else
3285          {
3286          do
3287            {
3288            if ((*set1++ & *set2++) != 0) return FALSE;
3289          }          }
3290          while (set1 < set_end);
3291        }        }
3292      else break;  
3293        if (list[1] == 0) return TRUE;
3294        /* Might be an empty repeat. */
3295        continue;
3296      }      }
   }  
3297    
3298  /* If the next thing is itself optional, we have to give up. */    /* Some property combinations also acceptable. Unicode property opcodes are
3299      processed specially; the rest can be handled with a lookup table. */
3300    
3301  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||    else
3302    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      {
3303      return FALSE;      pcre_uint32 leftop, rightop;
3304    
3305  /* If the previous item is a character, get its value. */      leftop = base_list[0];
3306        rightop = list[0];
3307    
3308  if (op_code == OP_CHAR || op_code == OP_CHARI ||  #ifdef SUPPORT_UCP
3309      op_code == OP_NOT || op_code == OP_NOTI)      accepted = FALSE; /* Always set in non-unicode case. */
3310    //if (escape == 0) switch(op_code)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3311    {        {
3312  #ifdef SUPPORT_UTF        if (rightop == OP_EOD)
3313    GETCHARTEST(c, previous);          accepted = TRUE;
3314  #else        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3315    c = *previous;          {
3316  #endif          int n;
3317    }          const pcre_uint8 *p;
3318            BOOL same = leftop == rightop;
3319            BOOL lisprop = leftop == OP_PROP;
3320            BOOL risprop = rightop == OP_PROP;
3321            BOOL bothprop = lisprop && risprop;
3322    
3323            /* There's a table that specifies how each combination is to be
3324            processed:
3325              0   Always return FALSE (never auto-possessify)
3326              1   Character groups are distinct (possessify if both are OP_PROP)
3327              2   Check character categories in the same group (general or particular)
3328              3   Return TRUE if the two opcodes are not the same
3329              ... see comments below
3330            */
3331    
3332            n = propposstab[base_list[2]][list[2]];
3333            switch(n)
3334              {
3335              case 0: break;
3336              case 1: accepted = bothprop; break;
3337              case 2: accepted = (base_list[3] == list[3]) != same; break;
3338              case 3: accepted = !same; break;
3339    
3340  /* Now compare the next item with the previous opcode. First, handle cases when            case 4:  /* Left general category, right particular category */
3341  the next item is a character. */            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3342              break;
3343    
3344  if (escape == 0)            case 5:  /* Right general category, left particular category */
3345    {            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3346    /* For a caseless UTF match, the next character may have more than one other            break;
   case, which maps to the special PT_CLIST property. Check this first. */  
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
     {  
     int ocs = UCD_CASESET(next);  
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
3347    
3348    switch(op_code)            /* This code is logically tricky. Think hard before fiddling with it.
3349      {            The posspropstab table has four entries per row. Each row relates to
3350      case OP_CHAR:            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3351      return c != next;            Only WORD actually needs all four entries, but using repeats for the
3352              others means they can all use the same code below.
3353    
3354              The first two entries in each row are Unicode general categories, and
3355              apply always, because all the characters they include are part of the
3356              PCRE character set. The third and fourth entries are a general and a
3357              particular category, respectively, that include one or more relevant
3358              characters. One or the other is used, depending on whether the check
3359              is for a general or a particular category. However, in both cases the
3360              category contains more characters than the specials that are defined
3361              for the property being tested against. Therefore, it cannot be used
3362              in a NOTPROP case.
3363    
3364              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3365              Underscore is covered by ucp_P or ucp_Po. */
3366    
3367              case 6:  /* Left alphanum vs right general category */
3368              case 7:  /* Left space vs right general category */
3369              case 8:  /* Left word vs right general category */
3370              p = posspropstab[n-6];
3371              accepted = risprop && lisprop ==
3372                (list[3] != p[0] &&
3373                 list[3] != p[1] &&
3374                (list[3] != p[2] || !lisprop));
3375              break;
3376    
3377      /* For CHARI (caseless character) we must check the other case. If we have            case 9:   /* Right alphanum vs left general category */
3378      Unicode property support, we can use it to test the other case of            case 10:  /* Right space vs left general category */
3379      high-valued characters. We know that next can have only one other case,            case 11:  /* Right word vs left general category */
3380      because multi-other-case characters are dealt with above. */            p = posspropstab[n-9];
3381              accepted = lisprop && risprop ==
3382                (base_list[3] != p[0] &&
3383                 base_list[3] != p[1] &&
3384                (base_list[3] != p[2] || !risprop));
3385              break;
3386    
3387      case OP_CHARI:            case 12:  /* Left alphanum vs right particular category */
3388      if (c == next) return FALSE;            case 13:  /* Left space vs right particular category */
3389  #ifdef SUPPORT_UTF            case 14:  /* Left word vs right particular category */
3390      if (utf)            p = posspropstab[n-12];
3391        {            accepted = risprop && lisprop ==
3392        pcre_uint32 othercase;              (catposstab[p[0]][list[3]] &&
3393        if (next < 128) othercase = cd->fcc[next]; else               catposstab[p[1]][list[3]] &&
3394  #ifdef SUPPORT_UCP              (list[3] != p[3] || !lisprop));
3395        othercase = UCD_OTHERCASE(next);            break;
3396  #else  
3397        othercase = NOTACHAR;            case 15:  /* Right alphanum vs left particular category */
3398  #endif            case 16:  /* Right space vs left particular category */
3399        return c != othercase;            case 17:  /* Right word vs left particular category */
3400              p = posspropstab[n-15];
3401              accepted = lisprop && risprop ==
3402                (catposstab[p[0]][base_list[3]] &&
3403                 catposstab[p[1]][base_list[3]] &&
3404                (base_list[3] != p[3] || !risprop));
3405              break;
3406              }
3407            }
3408        }        }
3409    
3410      else      else
3411  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UCP */
3412      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3413        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3414      case OP_NOT:             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3415      return c == next;             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3416    
3417      case OP_NOTI:      if (!accepted) return FALSE;
3418      if (c == next) return TRUE;  
3419  #ifdef SUPPORT_UTF      if (list[1] == 0) return TRUE;
3420      if (utf)      /* Might be an empty repeat. */
3421        continue;
3422        }
3423    
3424      /* Control reaches here only if one of the items is a small character list.
3425      All characters are checked against the other side. */
3426    
3427      do
3428        {
3429        chr = *chr_ptr;
3430    
3431        switch(list_ptr[0])
3432        {        {
3433        pcre_uint32 othercase;        case OP_CHAR:
3434        if (next < 128) othercase = cd->fcc[next]; else        ochr_ptr = list_ptr + 2;
3435  #ifdef SUPPORT_UCP        do
3436        othercase = UCD_OTHERCASE(next);          {
3437  #else          if (chr == *ochr_ptr) return FALSE;
3438        othercase = NOTACHAR;          ochr_ptr++;
3439  #endif          }
3440        return c == othercase;        while(*ochr_ptr != NOTACHAR);
3441        }        break;
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3442    
3443      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.        case OP_NOT:
3444      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */        ochr_ptr = list_ptr + 2;
3445          do
3446            {
3447            if (chr == *ochr_ptr)
3448              break;
3449            ochr_ptr++;
3450            }
3451          while(*ochr_ptr != NOTACHAR);
3452          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3453          break;
3454    
3455      case OP_DIGIT:        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3456      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3457    
3458      case OP_NOT_DIGIT:        case OP_DIGIT:
3459      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3460          break;
3461    
3462      case OP_WHITESPACE:        case OP_NOT_DIGIT:
3463      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3464          break;
3465    
3466      case OP_NOT_WHITESPACE:        case OP_WHITESPACE:
3467      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3468          break;
3469    
3470      case OP_WORDCHAR:        case OP_NOT_WHITESPACE:
3471      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3472          break;
3473    
3474      case OP_NOT_WORDCHAR:        case OP_WORDCHAR:
3475      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3476          break;
3477    
3478      case OP_HSPACE:        case OP_NOT_WORDCHAR:
3479      case OP_NOT_HSPACE:        if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3480      switch(next)        break;
       {  
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3481    
3482        default:        case OP_HSPACE:
3483        return op_code != OP_NOT_HSPACE;        switch(chr)
3484        }          {
3485            HSPACE_CASES: return FALSE;
3486            default: break;
3487            }
3488          break;
3489    
3490      case OP_ANYNL:        case OP_NOT_HSPACE:
3491      case OP_VSPACE:        switch(chr)
3492      case OP_NOT_VSPACE:          {
3493      switch(next)          HSPACE_CASES: break;
3494        {          default: return FALSE;
3495        VSPACE_CASES:          }
3496        return op_code == OP_NOT_VSPACE;        break;
3497    
3498        default:        case OP_ANYNL:
3499        return op_code != OP_NOT_VSPACE;        case OP_VSPACE:
3500        }        switch(chr)
3501            {
3502            VSPACE_CASES: return FALSE;
3503            default: break;
3504            }
3505          break;
3506    
3507          case OP_NOT_VSPACE:
3508          switch(chr)
3509            {
3510            VSPACE_CASES: break;
3511            default: return FALSE;
3512            }
3513          break;
3514    
3515          case OP_DOLL:
3516          case OP_EODN:
3517          switch (chr)
3518            {
3519            case CHAR_CR:
3520            case CHAR_LF:
3521            case CHAR_VT:
3522            case CHAR_FF:
3523            case CHAR_NEL:
3524    #ifndef EBCDIC
3525            case 0x2028:
3526            case 0x2029:
3527    #endif  /* Not EBCDIC */
3528            return FALSE;
3529            }
3530          break;
3531    
3532          case OP_EOD:    /* Can always possessify before \z */
3533          break;
3534    
3535  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3536      case OP_PROP:        case OP_PROP:
3537      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3538          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3539                list_ptr[0] == OP_NOTPROP))
3540            return FALSE;
3541          break;
3542    #endif
3543    
3544      case OP_NOTPROP:        case OP_NCLASS:
3545      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3546          /* Fall through */
3547    
3548          case OP_CLASS:
3549          if (chr > 255) break;
3550          class_bitset = (pcre_uint8 *)
3551            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3552          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3553          break;
3554    
3555    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3556          case OP_XCLASS:
3557          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3558              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3559          break;
3560  #endif  #endif
3561    
3562      default:        default:
3563      return FALSE;        return FALSE;
3564          }
3565    
3566        chr_ptr++;
3567      }      }
3568      while(*chr_ptr != NOTACHAR);
3569    
3570      /* At least one character must be matched from this opcode. */
3571    
3572      if (list[1] == 0) return TRUE;
3573    }    }
3574    
3575  /* 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,
3576  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. */
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
3577    
3578  switch(op_code)  }
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3579    
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
3580    
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3581    
3582      case ESC_S:  /*************************************************
3583      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  *    Scan compiled regex for auto-possession     *
3584    *************************************************/
3585    
3586      case ESC_w:  /* Replaces single character iterations with their possessive alternatives
3587      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  if appropriate. This function modifies the compiled opcode!
3588    
3589      case ESC_W:  Arguments:
3590      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;    code        points to start of the byte code
3591      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3592      cd          static compile data
3593    
3594      case ESC_h:  Returns:      nothing
3595      case ESC_H:  */
     switch(c)  
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
3596    
3597      case ESC_v:  static void
3598      case ESC_V:  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3599      switch(c)  {
3600    register pcre_uchar c;
3601    const pcre_uchar *end;
3602    pcre_uchar *repeat_opcode;
3603    pcre_uint32 list[8];
3604    
3605    for (;;)
3606      {
3607      c = *code;
3608    
3609      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3610        {
3611        c -= get_repeat_base(c) - OP_STAR;
3612        end = (c <= OP_MINUPTO) ?
3613          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3614        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3615    
3616        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3617        {        {
3618        VSPACE_CASES:        switch(c)
3619        return escape != ESC_v;          {
3620            case OP_STAR:
3621            *code += OP_POSSTAR - OP_STAR;
3622            break;
3623    
3624        default:          case OP_MINSTAR:
3625        return escape == ESC_v;          *code += OP_POSSTAR - OP_MINSTAR;
3626        }          break;
3627    
3628            case OP_PLUS:
3629            *code += OP_POSPLUS - OP_PLUS;
3630            break;
3631    
3632      /* When PCRE_UCP is set, these values get generated for \d etc. Find          case OP_MINPLUS:
3633      their substitutions and process them. The result will always be either          *code += OP_POSPLUS - OP_MINPLUS;
3634      ESC_p or ESC_P. Then fall through to process those values. */          break;
3635    
3636  #ifdef SUPPORT_UCP          case OP_QUERY:
3637      case ESC_du:          *code += OP_POSQUERY - OP_QUERY;
3638      case ESC_DU:          break;
3639      case ESC_wu:  
3640      case ESC_WU:          case OP_MINQUERY:
3641      case ESC_su:          *code += OP_POSQUERY - OP_MINQUERY;
3642      case ESC_SU:          break;
3643        {  
3644        int temperrorcode = 0;          case OP_UPTO:
3645        ptr = substitutes[escape - ESC_DU];          *code += OP_POSUPTO - OP_UPTO;
3646        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);          break;
3647        if (temperrorcode != 0) return FALSE;  
3648        ptr++;    /* For compatibility */          case OP_MINUPTO:
3649            *code += OP_POSUPTO - OP_MINUPTO;
3650            break;
3651            }
3652        }        }
3653      /* Fall through */      c = *code;
3654        }
3655      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3656        {
3657    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3658        if (c == OP_XCLASS)
3659          repeat_opcode = code + GET(code, 1);
3660        else
3661    #endif
3662          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3663    
3664      case ESC_p:      c = *repeat_opcode;
3665      case ESC_P:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3666        {        {
3667        int ptype, pdata, errorcodeptr;        /* end must not be NULL. */
3668        BOOL negated;        end = get_chr_property_list(code, utf, cd->fcc, list);
3669    
3670        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 */  
3671    
3672        /* If the property item is optional, we have to give up. (When generated        if (compare_opcodes(end, utf, cd, list, end))
3673        from \d etc by PCRE_UCP, this test will have been applied much earlier,          {
3674        to the original \d etc. At this point, ptr will point to a zero byte. */          switch (c)
3675              {
3676              case OP_CRSTAR:
3677              case OP_CRMINSTAR:
3678              *repeat_opcode = OP_CRPOSSTAR;
3679              break;
3680    
3681        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||            case OP_CRPLUS:
3682          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)            case OP_CRMINPLUS:
3683            return FALSE;            *repeat_opcode = OP_CRPOSPLUS;
3684              break;
3685    
3686        /* Do the property check. */            case OP_CRQUERY:
3687              case OP_CRMINQUERY:
3688              *repeat_opcode = OP_CRPOSQUERY;
3689              break;
3690    
3691        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);            case OP_CRRANGE:
3692              case OP_CRMINRANGE:
3693              *repeat_opcode = OP_CRPOSRANGE;
3694              break;
3695              }
3696            }
3697        }        }
3698  #endif      c = *code;
   
     default:  
     return FALSE;  
3699      }      }
3700    
3701    /* In principle, support for Unicode properties should be integrated here as    switch(c)
3702    well. It means re-organizing the above code so as to get hold of the property      {
3703    values before switching on the op-code. However, I wonder how many patterns      case OP_END:
3704    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,      return;
   these op-codes are never generated.) */  
3705    
3706    case OP_DIGIT:      case OP_TYPESTAR:
3707    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||      case OP_TYPEMINSTAR:
3708           escape == ESC_h || escape == ESC_v || escape == ESC_R;      case OP_TYPEPLUS:
3709        case OP_TYPEMINPLUS:
3710        case OP_TYPEQUERY:
3711        case OP_TYPEMINQUERY:
3712        case OP_TYPEPOSSTAR:
3713        case OP_TYPEPOSPLUS:
3714        case OP_TYPEPOSQUERY:
3715        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3716        break;
3717    
3718    case OP_NOT_DIGIT:      case OP_TYPEUPTO:
3719    return escape == ESC_d;      case OP_TYPEMINUPTO:
3720        case OP_TYPEEXACT:
3721        case OP_TYPEPOSUPTO:
3722        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3723          code += 2;
3724        break;
3725    
3726    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3727        case OP_XCLASS:
3728        code += GET(code, 1);
3729        break;
3730    #endif
3731    
3732        case OP_MARK:
3733        case OP_PRUNE_ARG:
3734        case OP_SKIP_ARG:
3735        case OP_THEN_ARG:
3736        code += code[1];
3737        break;
3738        }
3739    
3740      /* Add in the fixed length from the table */
3741    
3742      code += PRIV(OP_lengths)[c];
3743    
3744      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3745      a multi-byte character. The length in the table is a minimum, so we have to
3746      arrange to skip the extra bytes. */
3747    
3748    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3749      if (utf) switch(c)
3750        {
3751        case OP_CHAR:
3752        case OP_CHARI:
3753        case OP_NOT:
3754        case OP_NOTI:
3755        case OP_STAR:
3756        case OP_MINSTAR:
3757        case OP_PLUS:
3758        case OP_MINPLUS:
3759        case OP_QUERY:
3760        case OP_MINQUERY:
3761        case OP_UPTO:
3762        case OP_MINUPTO:
3763        case OP_EXACT:
3764        case OP_POSSTAR:
3765        case OP_POSPLUS:
3766        case OP_POSQUERY:
3767        case OP_POSUPTO:
3768        case OP_STARI:
3769        case OP_MINSTARI:
3770        case OP_PLUSI:
3771        case OP_MINPLUSI:
3772        case OP_QUERYI:
3773        case OP_MINQUERYI:
3774        case OP_UPTOI:
3775        case OP_MINUPTOI:
3776        case OP_EXACTI:
3777        case OP_POSSTARI:
3778        case OP_POSPLUSI:
3779        case OP_POSQUERYI:
3780        case OP_POSUPTOI:
3781        case OP_NOTSTAR:
3782        case OP_NOTMINSTAR:
3783        case OP_NOTPLUS:
3784        case OP_NOTMINPLUS:
3785        case OP_NOTQUERY:
3786        case OP_NOTMINQUERY:
3787        case OP_NOTUPTO:
3788        case OP_NOTMINUPTO:
3789        case OP_NOTEXACT:
3790        case OP_NOTPOSSTAR:
3791        case OP_NOTPOSPLUS:
3792        case OP_NOTPOSQUERY:
3793        case OP_NOTPOSUPTO:
3794        case OP_NOTSTARI:
3795        case OP_NOTMINSTARI:
3796        case OP_NOTPLUSI:
3797        case OP_NOTMINPLUSI:
3798        case OP_NOTQUERYI:
3799        case OP_NOTMINQUERYI:
3800        case OP_NOTUPTOI:
3801        case OP_NOTMINUPTOI:
3802        case OP_NOTEXACTI:
3803        case OP_NOTPOSSTARI:
3804        case OP_NOTPOSPLUSI:
3805        case OP_NOTPOSQUERYI:
3806        case OP_NOTPOSUPTOI:
3807        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3808        break;
3809        }
3810    #else
3811      (void)(utf);  /* Keep compiler happy by referencing function argument */
3812    #endif
3813      }
3814    }
3815    
3816    
3817    
3818    /*************************************************
3819    *           Check for POSIX class syntax         *
3820    *************************************************/
3821    
3822    /* This function is called when the sequence "[:" or "[." or "[=" is
3823    encountered in a character class. It checks whether this is followed by a
3824    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3825    reach an unescaped ']' without the special preceding character, return FALSE.
3826    
3827    Originally, this function only recognized a sequence of letters between the
3828    terminators, but it seems that Perl recognizes any sequence of characters,
3829    though of course unknown POSIX names are subsequently rejected. Perl gives an
3830    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3831    didn't consider this to be a POSIX class. Likewise for [:1234:].
3832    
3833    The problem in trying to be exactly like Perl is in the handling of escapes. We
3834    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3835    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3836    below handles the special case of \], but does not try to do any other escape
3837    processing. This makes it different from Perl for cases such as [:l\ower:]
3838    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3839    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3840    I think.
3841    
3842    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3843    It seems that the appearance of a nested POSIX class supersedes an apparent
3844    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3845    a digit.
3846    
3847    In Perl, unescaped square brackets may also appear as part of class names. For
3848    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3849    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3850    seem right at all. PCRE does not allow closing square brackets in POSIX class
3851    names.
3852    
3853    Arguments:
3854      ptr      pointer to the initial [
3855      endptr   where to return the end pointer
3856    
3857    Returns:   TRUE or FALSE
3858    */
3859    
3860    static BOOL
3861    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3862    {
3863    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3864    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3865    for (++ptr; *ptr != CHAR_NULL; ptr++)
3866      {
3867      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3868        ptr++;
3869      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3870      else
3871        {
3872        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3873          {
3874          *endptr = ptr;
3875          return TRUE;
3876          }
3877        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3878             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3879              ptr[1] == CHAR_EQUALS_SIGN) &&
3880            check_posix_syntax(ptr, endptr))
3881          return FALSE;
3882        }
3883      }
3884    return FALSE;
3885    }
3886    
3887    
3888    
3889    
3890    /*************************************************
3891    *          Check POSIX class name                *
3892    *************************************************/
3893    
3894    /* This function is called to check the name given in a POSIX-style class entry
3895    such as [:alnum:].
3896    
3897    Arguments:
3898      ptr        points to the first letter
3899      len        the length of the name
3900    
3901    Returns:     a value representing the name, or -1 if unknown
3902    */
3903    
3904    static int
3905    check_posix_name(const pcre_uchar *ptr, int len)
3906    {
3907    const char *pn = posix_names;
3908    register int yield = 0;
3909    while (posix_name_lengths[yield] != 0)
3910      {
3911      if (len == posix_name_lengths[yield] &&
3912        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3913      pn += posix_name_lengths[yield] + 1;
3914      yield++;
3915      }
3916    return -1;
3917    }
3918    
3919    
3920    /*************************************************
3921    *    Adjust OP_RECURSE items in repeated group   *
3922    *************************************************/
3923    
3924    /* OP_RECURSE items contain an offset from the start of the regex to the group
3925    that is referenced. This means that groups can be replicated for fixed
3926    repetition simply by copying (because the recursion is allowed to refer to
3927    earlier groups that are outside the current group). However, when a group is
3928    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3929    inserted before it, after it has been compiled. This means that any OP_RECURSE
3930    items within it that refer to the group itself or any contained groups have to
3931    have their offsets adjusted. That one of the jobs of this function. Before it
3932    is called, the partially compiled regex must be temporarily terminated with
3933    OP_END.
3934    
3935    This function has been extended with the possibility of forward references for
3936    recursions and subroutine calls. It must also check the list of such references
3937    for the group we are dealing with. If it finds that one of the recursions in
3938    the current group is on this list, it adjusts the offset in the list, not the
3939    value in the reference (which is a group number).
3940    
3941    Arguments:
3942      group      points to the start of the group
3943      adjust     the amount by which the group is to be moved
3944      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3945      cd         contains pointers to tables etc.
3946      save_hwm   the hwm forward reference pointer at the start of the group
3947    
3948    Returns:     nothing
3949    */
3950    
3951    static void
3952    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3953      pcre_uchar *save_hwm)
3954    {
3955    pcre_uchar *ptr = group;
3956    
3957    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3958      {
3959      int offset;
3960      pcre_uchar *hc;
3961    
3962      /* See if this recursion is on the forward reference list. If so, adjust the
3963      reference. */
3964    
3965      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3966        {
3967        offset = (int)GET(hc, 0);
3968        if (cd->start_code + offset == ptr + 1)
3969          {
3970          PUT(hc, 0, offset + adjust);
3971          break;
3972          }
3973        }
3974    
3975      /* Otherwise, adjust the recursion offset if it's after the start of this
3976      group. */
3977    
3978      if (hc >= cd->hwm)
3979        {
3980        offset = (int)GET(ptr, 1);
3981        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3982        }
3983    
3984      ptr += 1 + LINK_SIZE;
3985      }
3986    }
3987    
3988    
3989    
3990    /*************************************************
3991    *        Insert an automatic callout point       *
3992    *************************************************/
3993    
3994    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3995    callout points before each pattern item.
3996    
3997    Arguments:
3998      code           current code pointer
3999      ptr            current pattern pointer
4000      cd             pointers to tables etc
4001    
4002    Returns:         new code pointer
4003    */
4004    
4005    static pcre_uchar *
4006    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4007    {
4008    *code++ = OP_CALLOUT;
4009    *code++ = 255;
4010    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4011    PUT(code, LINK_SIZE, 0);                       /* Default length */
4012    return code + 2 * LINK_SIZE;
4013    }
4014    
4015    
4016    
4017    /*************************************************
4018    *         Complete a callout item                *
4019    *************************************************/
4020    
4021    /* A callout item contains the length of the next item in the pattern, which
4022    we can't fill in till after we have reached the relevant point. This is used
4023    for both automatic and manual callouts.
4024    
4025    Arguments:
4026      previous_callout   points to previous callout item
4027      ptr                current pattern pointer
4028      cd                 pointers to tables etc
4029    
4030    Returns:             nothing
4031    */
4032    
4033    static void
4034    complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4035    {
4036    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4037    PUT(previous_callout, 2 + LINK_SIZE, length);
4038    }
4039    
4040    
4041    
4042    #ifdef SUPPORT_UCP
4043    /*************************************************
4044    *           Get othercase range                  *
4045    *************************************************/
4046    
4047    /* This function is passed the start and end of a class range, in UTF-8 mode
4048    with UCP support. It searches up the characters, looking for ranges of
4049    characters in the "other" case. Each call returns the next one, updating the
4050    start address. A character with multiple other cases is returned on its own
4051    with a special return value.
4052    
4053    case OP_WHITESPACE:  Arguments:
4054    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4055      d           end value
4056      ocptr       where to put start of othercase range
4057      odptr       where to put end of othercase range
4058    
4059    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4060    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4061                  >0 the CASESET offset for char with multiple other cases
4062                    in this case, ocptr contains the original
4063    */
4064    
4065    case OP_HSPACE:  static int
4066    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4067           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4068    {
4069    pcre_uint32 c, othercase, next;
4070    unsigned int co;
4071    
4072    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4073    return escape == ESC_h;  cases, return its case offset value. */
4074    
4075    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4076    case OP_ANYNL:    {
4077    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4078    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4079        *ocptr = c++;   /* Character that has the set */
4080        *cptr = c;      /* Rest of input range */
4081        return (int)co;
4082        }
4083      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4084      }
4085    
4086    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4087    
4088    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4089    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  range, which is either the end of the input range, or a character that has zero
4090           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4091    
4092    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4093    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4094    
4095    default:  for (++c; c <= d; c++)
4096    return FALSE;    {
4097      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4098      next++;
4099    }    }
4100    
4101  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4102    *cptr = c;             /* Rest of input range */
4103    return 0;
4104  }  }
4105    #endif  /* SUPPORT_UCP */
4106    
4107    
4108    
# Line 3418  switch(op_code) Line 4111  switch(op_code)
4111  *************************************************/  *************************************************/
4112    
4113  /* 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
4114  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
4115  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
4116  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4117    
4118  Arguments:  Arguments:
4119    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4120    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4121    options       the options word    options       the options word
4122    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4123    start         start of range character    start         start of range character
4124    end           end of range character    end           end of range character
4125    
4126  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4127                  the pointer to extra data is updated                  the pointer to extra data is updated
4128  */  */
# Line 3439  add_to_class(pcre_uint8 *classbits, pcre Line 4132  add_to_class(pcre_uint8 *classbits, pcre
4132    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4133  {  {
4134  pcre_uint32 c;  pcre_uint32 c;
4135    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4136  int n8 = 0;  int n8 = 0;
4137    
4138  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4139  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
4140  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
4141  range. */  range. */
4142    
4143  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4144    {    {
4145  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4146    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4147      {      {
4148      int rc;      int rc;
4149      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4150    
4151      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4152      c = start;      c = start;
4153    
4154      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4155        {        {
4156        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4157    
4158        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4159          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4160    
4161        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4162    
4163        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4164    
4165        /* 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
4166        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
4167        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4168    
4169        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4170        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4171        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 4175  if ((options & PCRE_CASELESS) != 0)
4175  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4176    
4177    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4178    
4179    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4180      {      {
4181      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4182      n8++;      n8++;
4183      }      }
4184    }    }
4185    
4186  /* 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
4187  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
4188  in all cases. */  in all cases. */
# Line 3507  in all cases. */ Line 4201  in all cases. */
4201    
4202  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4203    
4204  /* 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. */  
4205    
4206  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4207    {    {
4208    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4209      {    SETBIT(classbits, c);
4210      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4211    }    }
4212    
4213  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4214    {  if (start <= 0xff) start = 0xff + 1;
4215    
4216    if (end >= start)
4217      {
4218    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4219  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4220    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4221      {      {
4222      if (start < end)      if (start < end)
4223        {        {
4224        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4225        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4226        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4227        }        }
4228      else if (start == end)      else if (start == end)
4229        {        {
4230        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4231        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4232        }        }
4233      }      }
4234    else    else
4235  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4236    
4237    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4238    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4239    
4240  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4241      {}      {}
4242  #else  #else
4243    if (start < end)    if (start < end)
4244      {      {
4245      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4250  else
4250      {      {
4251      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4252      *uchardata++ = start;      *uchardata++ = start;
4253      }      }
4254  #endif  #endif
4255    
4256    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4257    }    }
4258    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4259    
4260  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4261  }  }
4262    
4263    
4264    
4265    
4266  /*************************************************  /*************************************************
4267  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4268  *************************************************/  *************************************************/
4269    
4270  /* 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
4271  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
4272  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
4273  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4277  Arguments:
4277    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4278    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4279    options       the options word    options       the options word
4280    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4281    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4282    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4283                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4284                    already know about                    already know about
4285    
4286  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4287                  the pointer to extra data is updated                  the pointer to extra data is updated
4288  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4296  while (p[0] < NOTACHAR)
4296    {    {
4297    int n = 0;    int n = 0;
4298    if (p[0] != except)    if (p[0] != except)
4299      {      {
4300      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4301      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4302      }      }
4303    p += n + 1;    p += n + 1;
4304    }    }
4305  return n8;  return n8;
4306  }  }
4307    
4308    
4309    
# Line 3624  Arguments: Line 4318  Arguments:
4318    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4319    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4320    options       the options word    options       the options word
4321    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4322    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4323    
4324  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4325                  the pointer to extra data is updated                  the pointer to extra data is updated
4326  */  */
4327    
4328  static int  static int
4329  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4330    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4331  {  {
4332  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4338  while (p[0] < NOTACHAR)
4338    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4339    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4340      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4341    p++;    p++;
4342    }    }
4343  return n8;  return n8;
4344  }  }
4345    
4346    
4347    
# Line 3662  to find out the amount of memory needed, Line 4356  to find out the amount of memory needed,
4356  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4357    
4358  Arguments:  Arguments:
4359    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4360    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4361    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4362    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4363    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4364    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4365    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4366    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4367    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4368    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4369                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4370      lengthptr         NULL during the real compile phase
4371                        points to length accumulator during pre-compile phase
4372    
4373  Returns:         TRUE on success  Returns:            TRUE on success
4374                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4375  */  */
4376    
4377  static BOOL  static BOOL
4378  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4379    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4380    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4381      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4382      branch_chain *bcptr, int cond_depth,
4383    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4384  {  {
4385  int repeat_type, op_type;  int repeat_type, op_type;
4386  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4387  int bravalue = 0;  int bravalue = 0;
4388  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4389  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4390  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4391    pcre_uint32 zeroreqchar, zerofirstchar;
4392    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4393  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4394  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4395  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4417  dynamically as we process the pattern. *
4417  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4418  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4419  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4420    #ifndef COMPILE_PCRE32
4421  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4422    #endif
4423  #else  #else
4424  BOOL utf = FALSE;  BOOL utf = FALSE;
4425  #endif  #endif
4426    
4427  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4428  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,
4429  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
4430  alternative calls for the different cases. */  alternative calls for the different cases. */
4431    
4432  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4454  to take the zero repeat into account. Th
4454  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4455  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4456    
4457  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4458    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4459    
4460  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4461  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 4477  for (;; ptr++)
4477    BOOL reset_bracount;    BOOL reset_bracount;
4478    int class_has_8bitchar;    int class_has_8bitchar;
4479    int class_one_char;    int class_one_char;
4480    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4481      BOOL xclass_has_prop;
4482    #endif
4483    int newoptions;    int newoptions;
4484    int recno;    int recno;
4485    int refsign;    int refsign;
4486    int skipbytes;    int skipbytes;
4487    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4488    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4489    int terminator;    int terminator;
4490    int mclength;    unsigned int mclength;
4491    int tempbracount;    unsigned int tempbracount;
4492    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4493    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4494    
4495    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4499  for (;; ptr++)
4499    /* 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
4500    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4501    
4502    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4503      {      {
4504      ptr = nestptr;      ptr = nestptr;
4505      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4574  for (;; ptr++)
4574    
4575    /* 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 */
4576    
4577    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4578      {      {
4579      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4580        {        {
# Line 3891  for (;; ptr++) Line 4597  for (;; ptr++)
4597          }          }
4598        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4599        }        }
4600        /* Control does not reach here. */
4601      }      }
4602    
4603    /* 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
4604    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. */  
4605    
4606    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4607      {      {
4608      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4609        {        {
4610          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4611          if (c != CHAR_NUMBER_SIGN) break;
4612        ptr++;        ptr++;
4613        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4614          {          {
4615          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4616              {                          /* IS_NEWLINE sets cd->nllen. */
4617              ptr += cd->nllen;
4618              break;
4619              }
4620          ptr++;          ptr++;
4621  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4622          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4623  #endif  #endif
4624          }          }
4625        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4626        }        }
4627      }      }
4628    
4629    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4630    
4631      is_quantifier =
4632        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4633        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4634    
4635      /* Fill in length of a previous callout, except when the next thing is a
4636      quantifier or when processing a property substitution string in UCP mode. */
4637    
4638      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4639           after_manual_callout-- <= 0)
4640        {
4641        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4642          complete_callout(previous_callout, ptr, cd);
4643        previous_callout = NULL;
4644        }
4645    
4646      /* Create auto callout, except for quantifiers, or while processing property
4647      strings that are substituted for \w etc in UCP mode. */
4648    
4649    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4650      {      {
4651      previous_callout = code;      previous_callout = code;
4652      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4653      }      }
4654    
4655      /* Process the next pattern item. */
4656    
4657    switch(c)    switch(c)
4658      {      {
4659      /* ===================================================================*/      /* ===================================================================*/
4660      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4661      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4662      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4663      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4664        *firstcharflagsptr = firstcharflags;
4665      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4666        *reqcharflagsptr = reqcharflags;
4667      *codeptr = code;      *codeptr = code;
4668      *ptrptr = ptr;