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

Diff of /code/trunk/pcre_compile.c

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

revision 1067 by chpe, Tue Oct 16 15:54:22 2012 UTC revision 1411 by ph10, Fri Dec 6 17:11:44 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 250  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 265  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 292  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 319  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 372  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 387  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 452  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 518  static const char error_texts[] =
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 505  static const char error_texts[] = Line 539  static const char error_texts[] =
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550    ;    ;
551    
552  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 686  static const pcre_uint8 ebcdic_chartab[]
686  #endif  #endif
687    
688    
689    /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726    
727    /* This table is used to check whether auto-possessification is possible
728    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729    left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
782      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
783      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
784      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
867  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 883  find_error_text(int n)
883  const char *s = error_texts;  const char *s = error_texts;
884  for (; n > 0; n--)  for (; n > 0; n--)
885    {    {
886    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
887    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
888    }    }
889  return s;  return s;
890  }  }
891    
892    
893    
894  /*************************************************  /*************************************************
895  *           Expand the workspace                 *  *           Expand the workspace                 *
896  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 968  return (*p == CHAR_RIGHT_CURLY_BRACKET);
968  *************************************************/  *************************************************/
969    
970  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
971  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
972  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
973  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
974  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979    chptr          points to the data character    chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 989  Returns:         zero => a data characte
989  */  */
990    
991  static int  static int
992  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
994  {  {
995  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 1004  ptr--;                            /* Set
1004    
1005  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1006    
1007  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008    
1009  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 795  Otherwise further processing may be requ Line 1013  Otherwise further processing may be requ
1013  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1014  /* Not alphanumeric */  /* Not alphanumeric */
1015  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1016  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
1017      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018    
1019  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1020  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1064  else
1064            }            }
1065    
1066  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1067          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1068  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1069          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1070  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1071          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1072  #endif  #endif
1073            {            {
1074            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1115  else
1115      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116        {        {
1117        const pcre_uchar *p;        const pcre_uchar *p;
1118        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121          {          {
1122          escape = ESC_k;          escape = ESC_k;
1123          break;          break;
# Line 961  else Line 1180  else
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1188      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1189      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1190      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1191      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195        Inside a character class, \ followed by a digit is always either a literal
1196        8 or 9 or an octal number. */
1197    
1198      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1220  else
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          escape = -s;          escape = -s;
1226          break;          break;
# Line 1005  else Line 1228  else
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1232      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1241      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1252  else
1252  #endif  #endif
1253      break;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1257      If not, { is treated as a data character. */  
1258        case CHAR_o:
1259        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260          {
1261          ptr += 2;
1262          c = 0;
1263          overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265            {
1266            register pcre_uint32 cc = *ptr++;
1267            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268    #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270    #endif
1271            c = (c << 3) + cc - CHAR_0 ;
1272    #if defined COMPILE_PCRE8
1273            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274    #elif defined COMPILE_PCRE16
1275            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278    #endif
1279            }
1280          if (overflow)
1281            {
1282            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283            *errorcodeptr = ERR34;
1284            }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286            {
1287            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290          }
1291        break;
1292    
1293        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296      case CHAR_x:      case CHAR_x:
1297      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1299        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301          {          {
# Line 1056  else Line 1312  else
1312  #endif  #endif
1313            }            }
1314          }          }
1315        break;        }    /* End JavaScript handling */
       }  
1316    
1317      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1320        seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        c = 0;      else
1325        overflow = FALSE;        {
1326        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327          {          {
1328          register pcre_uint32 cc = *pt++;          ptr += 2;
1329          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1330            overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1337          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1338  #endif  #endif
1339    
1340  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1344          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1345          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346  #endif  #endif
1347    
1348  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1349          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1351          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1353          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354  #endif  #endif
1355          }            }
1356    
1357        if (overflow)          if (overflow)
1358          {            {
1359          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1361          }            }
1362    
1363        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364          {            {
1365          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366          ptr = pt;            }
         break;  
         }  
1367    
1368        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1369        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1370        }          \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1378        {        else
1379        pcre_uint32 cc;                          /* Some compilers don't like */          {
1380        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1389        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391  #endif  #endif
1392        }            }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1401  else
1401    
1402      case CHAR_c:      case CHAR_c:
1403      c = *(++ptr);      c = *(++ptr);
1404      if (c == 0)      if (c == CHAR_NULL)
1405        {        {
1406        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1407        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1457  if ((options & PCRE_UCP) != 0 && escape
1457  return escape;  return escape;
1458  }  }
1459    
1460    
1461    
1462  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1463  /*************************************************  /*************************************************
1464  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1472  escape sequence.
1472  Argument:  Argument:
1473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1474    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1475    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1476      pdataptr       points to an unsigned int that is set to the detailed property value
1477    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1478    
1479  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1480  */  */
1481    
1482  static int  static BOOL
1483  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484      unsigned int *pdataptr, int *errorcodeptr)
1485  {  {
1486  pcre_uchar c;  pcre_uchar c;
1487  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1489  const pcre_uchar *ptr = *ptrptr;
1489  pcre_uchar name[32];  pcre_uchar name[32];
1490    
1491  c = *(++ptr);  c = *(++ptr);
1492  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1493    
1494  *negptr = FALSE;  *negptr = FALSE;
1495    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1506  if (c == CHAR_LEFT_CURLY_BRACKET)
1506    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507      {      {
1508      c = *(++ptr);      c = *(++ptr);
1509      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1510      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511      name[i] = c;      name[i] = c;
1512      }      }
# Line 1263  while (bot < top) Line 1536  while (bot < top)
1536    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537    if (r == 0)    if (r == 0)
1538      {      {
1539      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1540      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1541        return TRUE;
1542      }      }
1543    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1544    }    }
1545    
1546  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1547  *ptrptr = ptr;  *ptrptr = ptr;
1548  return -1;  return FALSE;
1549    
1550  ERROR_RETURN:  ERROR_RETURN:
1551  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1552  *ptrptr = ptr;  *ptrptr = ptr;
1553  return -1;  return FALSE;
1554  }  }
1555  #endif  #endif
1556    
1557    
1558    
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 1351  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       pcre_uchar term;  
       const pcre_uchar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == (int)(ptr - thisname) &&  
           STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1627  *      Find first significant op code            *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
# Line 1684  for (;;) Line 1662  for (;;)
1662    
1663      case OP_CALLOUT:      case OP_CALLOUT:
1664      case OP_CREF:      case OP_CREF:
1665      case OP_NCREF:      case OP_DNCREF:
1666      case OP_RREF:      case OP_RREF:
1667      case OP_NRREF:      case OP_DNRREF:
1668      case OP_DEF:      case OP_DEF:
1669      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1670      break;      break;
# Line 1700  for (;;) Line 1678  for (;;)
1678    
1679    
1680    
   
1681  /*************************************************  /*************************************************
1682  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1683  *************************************************/  *************************************************/
# Line 1744  for (;;) Line 1721  for (;;)
1721    {    {
1722    int d;    int d;
1723    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1724    register int op = *cc;    register pcre_uchar op = *cc;
1725    
1726    switch (op)    switch (op)
1727      {      {
# Line 1824  for (;;) Line 1801  for (;;)
1801      case OP_COMMIT:      case OP_COMMIT:
1802      case OP_CREF:      case OP_CREF:
1803      case OP_DEF:      case OP_DEF:
1804        case OP_DNCREF:
1805        case OP_DNRREF:
1806      case OP_DOLL:      case OP_DOLL:
1807      case OP_DOLLM:      case OP_DOLLM:
1808      case OP_EOD:      case OP_EOD:
1809      case OP_EODN:      case OP_EODN:
1810      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812      case OP_PRUNE:      case OP_PRUNE:
1813      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1829  for (;;)
1829      case OP_NOTI:      case OP_NOTI:
1830      branchlength++;      branchlength++;
1831      cc += 2;      cc += 2;
1832  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1833      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834  #endif  #endif
1835      break;      break;
# Line 1864  for (;;) Line 1841  for (;;)
1841      case OP_EXACTI:      case OP_EXACTI:
1842      case OP_NOTEXACT:      case OP_NOTEXACT:
1843      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1844      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1845      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1846  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1847      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848  #endif  #endif
1849      break;      break;
1850    
1851      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1852      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1853      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854        cc += 2;        cc += 2;
1855      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1856      break;      break;
# Line 1909  for (;;) Line 1886  for (;;)
1886    
1887      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1888    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1889      case OP_CLASS:      case OP_CLASS:
1890      case OP_NCLASS:      case OP_NCLASS:
1891    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892        case OP_XCLASS:
1893        /* The original code caused an unsigned overflow in 64 bit systems,
1894        so now we use a conditional statement. */
1895        if (op == OP_XCLASS)
1896          cc += GET(cc, 1);
1897        else
1898          cc += PRIV(OP_lengths)[OP_CLASS];
1899    #else
1900      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1901    #endif
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918          case OP_CRPOSRANGE:
1919        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1921        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1922        break;        break;
1923    
# Line 1999  for (;;) Line 1984  for (;;)
1984      case OP_QUERYI:      case OP_QUERYI:
1985      case OP_REF:      case OP_REF:
1986      case OP_REFI:      case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989      case OP_SBRA:      case OP_SBRA:
1990      case OP_SBRAPOS:      case OP_SBRAPOS:
1991      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
# Line 2059  PRIV(find_bracket)(const pcre_uchar *cod Line 2045  PRIV(find_bracket)(const pcre_uchar *cod
2045  {  {
2046  for (;;)  for (;;)
2047    {    {
2048    register int c = *code;    register pcre_uchar c = *code;
2049    
2050    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2051    
# Line 2082  for (;;) Line 2068  for (;;)
2068    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2069             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070      {      {
2071      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2072      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2073      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2074      }      }
# Line 2112  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102          code += 2;          code += 2;
2103        break;        break;
2104    
2105        case OP_MARK:        case OP_MARK:
2106        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2108        case OP_THEN_ARG:        case OP_THEN_ARG:
2109        code += code[1];        code += code[1];
2110        break;        break;
# Line 2197  find_recurse(const pcre_uchar *code, BOO Line 2180  find_recurse(const pcre_uchar *code, BOO
2180  {  {
2181  for (;;)  for (;;)
2182    {    {
2183    register int c = *code;    register pcre_uchar c = *code;
2184    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2185    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2186    
# Line 2232  for (;;) Line 2215  for (;;)
2215        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2216        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2217        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2218        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219          code += 2;          code += 2;
2220        break;        break;
2221    
2222        case OP_MARK:        case OP_MARK:
2223        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2224        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2225        code += code[1];        case OP_THEN_ARG:
       break;  
   
       case OP_THEN_ARG:  
2226        code += code[1];        code += code[1];
2227        break;        break;
2228        }        }
# Line 2343  Arguments: Line 2323  Arguments:
2323    endcode     points to where to stop    endcode     points to where to stop
2324    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2326      recurses    chain of recurse_check to catch mutual recursion
2327    
2328  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2329  */  */
2330    
2331    typedef struct recurse_check {
2332      struct recurse_check *prev;
2333      const pcre_uchar *group;
2334    } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2339  {  {
2340  register int c;  register pcre_uchar c;
2341    recurse_check this_recurse;
2342    
2343  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2379  for (code = first_significant_code(code Line 2367  for (code = first_significant_code(code
2367    
2368    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2369      {      {
2370      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371      BOOL empty_branch;      BOOL empty_branch;
2372    
2373      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2374        when called to scan a completed pattern by setting cd->start_workspace to
2375        NULL. */
2376    
2377        if (cd->start_workspace != NULL)
2378          {
2379          const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383          }
2384    
2385        /* If we are scanning a completed pattern, there are no forward references
2386        and all groups are complete. We need to detect whether this is a recursive
2387        call, as otherwise there will be an infinite loop. If it is a recursion,
2388        just skip over it. Simple recursions are easily detected. For mutual
2389        recursions we keep a chain on the stack. */
2390    
2391        else
2392          {
2393          recurse_check *r = recurses;
2394          const pcre_uchar *endgroup = scode;
2395    
2396          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        for (r = recurses; r != NULL; r = r->prev)
2400        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;          if (r->group == scode) break;
2401          if (r != NULL) continue;   /* Mutual recursion */
2402          }
2403    
2404      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2405        stack chain to detect mutual recursions. */
2406    
2407      empty_branch = FALSE;      empty_branch = FALSE;
2408      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2409      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
2410    
2411      do      do
2412        {        {
2413        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414          {          {
2415          empty_branch = TRUE;          empty_branch = TRUE;
2416          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2466  for (code = first_significant_code(code
2466        empty_branch = FALSE;        empty_branch = FALSE;
2467        do        do
2468          {          {
2469          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470            empty_branch = TRUE;            empty_branch = TRUE;
2471          code += GET(code, 1);          code += GET(code, 1);
2472          }          }
# Line 2495  for (code = first_significant_code(code Line 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 2511  for (code = first_significant_code(code Line 2528  for (code = first_significant_code(code
2528    
2529      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      case OP_NOTPROP:
2537        case OP_ANYNL:
2538    
2539        case OP_NOT_HSPACE:
2540        case OP_HSPACE:
2541        case OP_NOT_VSPACE:
2542        case OP_VSPACE:
2543      case OP_EXTUNI:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARI:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555      case OP_NOTI:      case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558        case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
2560      case OP_POSPLUS:      case OP_MINPLUSI:
2561      case OP_EXACT:  
2562      case OP_NOTPLUS:      case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2558  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2599      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2600      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2601      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602        code += 2;        code += 2;
2603      break;      break;
2604    
# Line 2572  for (code = first_significant_code(code Line 2612  for (code = first_significant_code(code
2612      return TRUE;      return TRUE;
2613    
2614      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619      case OP_STAR:      case OP_STAR:
2620      case OP_STARI:      case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624      case OP_MINSTAR:      case OP_MINSTAR:
2625      case OP_MINSTARI:      case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629      case OP_POSSTAR:      case OP_POSSTAR:
2630      case OP_POSSTARI:      case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634      case OP_QUERY:      case OP_QUERY:
2635      case OP_QUERYI:      case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639      case OP_MINQUERY:      case OP_MINQUERY:
2640      case OP_MINQUERYI:      case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644      case OP_POSQUERY:      case OP_POSQUERY:
2645      case OP_POSQUERYI:      case OP_POSQUERYI:
2646        case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650      break;      break;
2651    
2652      case OP_UPTO:      case OP_UPTO:
2653      case OP_UPTOI:      case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657      case OP_MINUPTO:      case OP_MINUPTO:
2658      case OP_MINUPTOI:      case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662      case OP_POSUPTO:      case OP_POSUPTO:
2663      case OP_POSUPTOI:      case OP_POSUPTOI:
2664        case OP_NOTPOSUPTO:
2665        case OP_NOTPOSUPTOI:
2666    
2667      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668      break;      break;
2669  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2674  for (code = first_significant_code(code
2674      case OP_MARK:      case OP_MARK:
2675      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2676      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2677      case OP_THEN_ARG:      case OP_THEN_ARG:
2678      code += code[1];      code += code[1];
2679      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2717  could_be_empty(const pcre_uchar *code, c
2717  {  {
2718  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2719    {    {
2720    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721      return FALSE;      return FALSE;
2722    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2723    }    }
# Line 2662  return TRUE; Line 2727  return TRUE;
2727    
2728    
2729  /*************************************************  /*************************************************
2730  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2731  *************************************************/  *************************************************/
2732    
2733  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2734  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2735    
2736  Returns:   TRUE or FALSE  Arguments:  c opcode
2737    Returns:    base opcode for the type
2738  */  */
2739    
2740  static BOOL  static pcre_uchar
2741  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2742  {  {
2743  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2744  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2748      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2749  }  }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754  /*************************************************  /*************************************************
2755  *          Check POSIX class name                *  *        Check a character and a property        *
2756  *************************************************/  *************************************************/
2757    
2758  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2759  such as [:alnum:].  is adjacent to a fixed character.
2760    
2761  Arguments:  Arguments:
2762    ptr        points to the first letter    c            the character
2763    len        the length of the name    ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2768  */  */
2769    
2770  static int  static BOOL
2771  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773  {  {
2774  const char *pn = posix_names;  const pcre_uint32 *p;
2775  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2776    
2777  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2778  that is referenced. This means that groups can be replicated for fixed    {
2779  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2780  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2781  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2782  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2783    
2784  This function has been extended with the possibility of forward references for    case PT_GC:
2785  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2786    
2787  Arguments:    case PT_PC:
2788    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2789    
2790  Returns:     nothing    case PT_SC:
2791  */    return (pdata == prop->script) == negated;
2792    
2793  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2794    
2795  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2796    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2798    
2799    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806      {      {
2807      offset = GET(hc, 0);      HSPACE_CASES:
2808      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2809        {      return negated;
2810        PUT(hc, 0, offset + adjust);  
2811        break;      default:
2812        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813      }      }
2814      break;  /* Control never reaches here */
2815    
2816    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2817    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821    if (hc >= cd->hwm)    case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824      {      {
2825      offset = GET(ptr, 1);      if (c < *p) return !negated;
2826      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2827      }      }
2828      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2829    }    }
2830    
2831    return FALSE;
2832  }  }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837  /*************************************************  /*************************************************
2838  *        Insert an automatic callout point       *  *        Fill the character property list        *
2839  *************************************************/  *************************************************/
2840    
2841  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2842  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2843    
2844  Arguments:  Arguments:
2845    code           current code pointer    code        points to start of expression
2846    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847    cd             pointers to tables etc    fcc         points to case-flipping table
2848      list        points to output list
2849                  list[0] will be filled with the opcode
2850                  list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852                  list[2..7] depends on the opcode
2853    
2854  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856  */  */
2857    
2858  static pcre_uchar *  static const pcre_uchar *
2859  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860  {    const pcre_uint8 *fcc, pcre_uint32 *list)
 *code++ = OP_CALLOUT;  
 *code++ = 255;  
 PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  
 PUT(code, LINK_SIZE, 0);                       /* Default length */  
 return code + 2 * LINK_SIZE;  
 }  
   
   
   
 /*************************************************  
 *         Complete a callout item                *  
 *************************************************/  
   
 /* A callout item contains the length of the next item in the pattern, which  
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
   
 Arguments:  
   previous_callout   points to previous callout item  
   ptr                current pattern pointer  
   cd                 pointers to tables etc  
   
 Returns:             nothing  
 */  
   
 static void  
 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  
2861  {  {
2862  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  pcre_uchar c = *code;
2863  PUT(previous_callout, 2 + LINK_SIZE, length);  pcre_uchar base;
2864  }  const pcre_uchar *end;
2865    pcre_uint32 chr;
   
2866    
2867  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2868  /*************************************************  pcre_uint32 *clist_dest;
2869  *           Get othercase range                  *  const pcre_uint32 *clist_src;
2870  *************************************************/  #else
2871    utf = utf;  /* Suppress "unused parameter" compiler warning */
2872  /* This function is passed the start and end of a class range, in UTF-8 mode  #endif
 with UCP support. It searches up the characters, looking for ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
2873    
2874  Arguments:  list[0] = c;
2875    cptr        points to starting character value; updated  list[1] = FALSE;
2876    d           end value  code++;
   ocptr       where to put start of othercase range  
   odptr       where to put end of othercase range  
2877    
2878  Yield:        -1 when no more  if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879                 0 when a range is returned    {
2880                >0 the CASESET offset for char with multiple other cases    base = get_repeat_base(c);
2881                  in this case, ocptr contains the original    c -= (base - OP_STAR);
 */  
2882    
2883  static int    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      code += IMM2_SIZE;
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2885    
2886  /* Find the first character that has an other case. If it has multiple other    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 cases, return its case offset value. */  
2887    
2888  for (c = *cptr; c <= d; c++)    switch(base)
   {  
   if ((co = UCD_CASESET(c)) != 0)  
2889      {      {
2890      *ocptr = c++;   /* Character that has the set */      case OP_STAR:
2891      *cptr = c;      /* Rest of input range */      list[0] = OP_CHAR;
2892      return co;      break;
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
   
 if (c > d) return -1;  /* Reached end of range */  
   
 *ocptr = othercase;  
 next = othercase + 1;  
   
 for (++c; c <= d; c++)  
   {  
   if (UCD_OTHERCASE(c) != next) break;  
   next++;  
   }  
2893    
2894  *odptr = next - 1;     /* End of othercase range */      case OP_STARI:
2895  *cptr = c;             /* Rest of input range */      list[0] = OP_CHARI;
2896  return 0;      break;
 }  
2897    
2898        case OP_NOTSTAR:
2899        list[0] = OP_NOT;
2900        break;
2901    
2902        case OP_NOTSTARI:
2903        list[0] = OP_NOTI;
2904        break;
2905    
2906  /*************************************************      case OP_TYPESTAR:
2907  *        Check a character and a property        *      list[0] = *code;
2908  *************************************************/      code++;
2909        break;
2910        }
2911      c = list[0];
2912      }
2913    
2914  /* This function is called by check_auto_possessive() when a property item  switch(c)
2915  is adjacent to a fixed character.    {
2916      case OP_NOT_DIGIT:
2917      case OP_DIGIT:
2918      case OP_NOT_WHITESPACE:
2919      case OP_WHITESPACE:
2920      case OP_NOT_WORDCHAR:
2921      case OP_WORDCHAR:
2922      case OP_ANY:
2923      case OP_ALLANY:
2924      case OP_ANYNL:
2925      case OP_NOT_HSPACE:
2926      case OP_HSPACE:
2927      case OP_NOT_VSPACE:
2928      case OP_VSPACE:
2929      case OP_EXTUNI:
2930      case OP_EODN:
2931      case OP_EOD:
2932      case OP_DOLL:
2933      case OP_DOLLM:
2934      return code;
2935    
2936  Arguments:    case OP_CHAR:
2937    c            the character    case OP_NOT:
2938    ptype        the property type    GETCHARINCTEST(chr, code);
2939    pdata        the data for the type    list[2] = chr;
2940    negated      TRUE if it's a negated property (\P or \p{^)    list[3] = NOTACHAR;
2941      return code;
2942    
2943  Returns:       TRUE if auto-possessifying is OK    case OP_CHARI:
2944  */    case OP_NOTI:
2945      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946      GETCHARINCTEST(chr, code);
2947      list[2] = chr;
2948    
 static BOOL  
 check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  
 {  
2949  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2950  const pcre_uint32 *p;    if (chr < 128 || (chr < 256 && !utf))
2951        list[3] = fcc[chr];
2952      else
2953        list[3] = UCD_OTHERCASE(chr);
2954    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955      list[3] = (chr < 256) ? fcc[chr] : chr;
2956    #else
2957      list[3] = fcc[chr];
2958  #endif  #endif
2959    
2960  const ucd_record *prop = GET_UCD(c);    /* The othercase might be the same value. */
2961    
2962  switch(ptype)    if (chr == list[3])
2963    {      list[3] = NOTACHAR;
2964    case PT_LAMP:    else
2965    return (prop->chartype == ucp_Lu ||      list[4] = NOTACHAR;
2966            prop->chartype == ucp_Ll ||    return code;
           prop->chartype == ucp_Lt) == negated;  
2967    
2968    case PT_GC:  #ifdef SUPPORT_UCP
2969    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;    case OP_PROP:
2970      case OP_NOTPROP:
2971      if (code[0] != PT_CLIST)
2972        {
2973        list[2] = code[0];
2974        list[3] = code[1];
2975        return code + 2;
2976        }
2977    
2978    case PT_PC:    /* Convert only if we have enough space. */
   return (pdata == prop->chartype) == negated;  
2979    
2980    case PT_SC:    clist_src = PRIV(ucd_caseless_sets) + code[1];
2981    return (pdata == prop->script) == negated;    clist_dest = list + 2;
2982      code += 2;
2983    
2984    /* These are specials */    do {
2985         if (clist_dest >= list + 8)
2986           {
2987           /* Early return if there is not enough space. This should never
2988           happen, since all clists are shorter than 5 character now. */
2989           list[2] = code[0];
2990           list[3] = code[1];
2991           return code;
2992           }
2993         *clist_dest++ = *clist_src;
2994         }
2995      while(*clist_src++ != NOTACHAR);
2996    
2997    case PT_ALNUM:    /* All characters are stored. The terminating NOTACHAR
2998    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    is copied form the clist itself. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
2999    
3000    case PT_SPACE:    /* Perl space */    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    return code;
3002            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  #endif
           == negated;  
3003    
3004    case PT_PXSPACE:  /* POSIX space */    case OP_NCLASS:
3005    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    case OP_CLASS:
3006            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007            c == CHAR_FF || c == CHAR_CR)    case OP_XCLASS:
3008            == negated;    if (c == OP_XCLASS)
3009        end = code + GET(code, 0) - 1;
3010      else
3011    #endif
3012        end = code + 32 / sizeof(pcre_uchar);
3013    
3014    case PT_WORD:    switch(*end)
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   
 #ifdef SUPPORT_UCP  
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3015      {      {
3016      if ((unsigned int)c < *p) return !negated;      case OP_CRSTAR:
3017      if ((unsigned int)c == *p++) return negated;      case OP_CRMINSTAR:
3018        case OP_CRQUERY:
3019        case OP_CRMINQUERY:
3020        case OP_CRPOSSTAR:
3021        case OP_CRPOSQUERY:
3022        list[1] = TRUE;
3023        end++;
3024        break;
3025    
3026        case OP_CRPLUS:
3027        case OP_CRMINPLUS:
3028        case OP_CRPOSPLUS:
3029        end++;
3030        break;
3031    
3032        case OP_CRRANGE:
3033        case OP_CRMINRANGE:
3034        case OP_CRPOSRANGE:
3035        list[1] = (GET2(end, 1) == 0);
3036        end += 1 + 2 * IMM2_SIZE;
3037        break;
3038      }      }
3039    break;  /* Control never reaches here */    list[2] = end - code;
3040  #endif    return end;
3041    }    }
3042    return NULL;    /* Opcode not accepted */
 return FALSE;  
3043  }  }
 #endif  /* SUPPORT_UCP */  
3044    
3045    
3046    
3047  /*************************************************  /*************************************************
3048  *     Check if auto-possessifying is possible    *  *    Scan further character sets for match       *
3049  *************************************************/  *************************************************/
3050    
3051  /* This function is called for unlimited repeats of certain items, to see  /* Checks whether the base and the current opcode have a common character, in
3052  whether the next thing could possibly match the repeated item. If not, it makes  which case the base cannot be possessified.
 sense to automatically possessify the repeated item.  
3053    
3054  Arguments:  Arguments:
3055    previous      pointer to the repeated opcode    code        points to the byte code
3056    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057    ptr           next character in pattern    cd          static compile data
3058    options       options bits    base_list   the data list of the base opcode
   cd            contains pointers to tables etc.  
3059    
3060  Returns:        TRUE if possessifying is wanted  Returns:      TRUE if the auto-possessification is possible
3061  */  */
3062    
3063  static BOOL  static BOOL
3064  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065    const pcre_uchar *ptr, int options, compile_data *cd)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066  {  {
3067  pcre_uint32 c = NOTACHAR;  pcre_uchar c;
3068  pcre_uint32 next;  pcre_uint32 list[8];
3069  int escape;  const pcre_uint32 *chr_ptr;
3070  int op_code = *previous++;  const pcre_uint32 *ochr_ptr;
3071    const pcre_uint32 *list_ptr;
3072  /* Skip whitespace and comments in extended mode */  const pcre_uchar *next_code;
3073    const pcre_uint8 *class_bitset;
3074    const pcre_uint32 *set1, *set2, *set_end;
3075    pcre_uint32 chr;
3076    BOOL accepted, invert_bits;
3077    
3078    /* Note: the base_list[1] contains whether the current opcode has greedy
3079    (represented by a non-zero value) quantifier. This is a different from
3080    other character type lists, which stores here that the character iterator
3081    matches to an empty string (also represented by a non-zero value). */
3082    
3083  if ((options & PCRE_EXTENDED) != 0)  for(;;)
3084    {    {
3085    for (;;)    /* All operations move the code pointer forward.
3086      Therefore infinite recursions are not possible. */
3087    
3088      c = *code;
3089    
3090      /* Skip over callouts */
3091    
3092      if (c == OP_CALLOUT)
3093      {      {
3094      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += PRIV(OP_lengths)[c];
3095      if (*ptr == CHAR_NUMBER_SIGN)      continue;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3096      }      }
   }  
3097    
3098  /* If the next item is one that we can handle, get its value. A non-negative    if (c == OP_ALT)
3099  value is a character, a negative value is an escape value. */      {
3100        do code += GET(code, 1); while (*code == OP_ALT);
3101        c = *code;
3102        }
3103    
3104  if (*ptr == CHAR_BACKSLASH)    switch(c)
3105    {      {
3106    int temperrorcode = 0;      case OP_END:
3107    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);      case OP_KETRPOS:
3108    if (temperrorcode != 0) return FALSE;      /* TRUE only in greedy case. The non-greedy case could be replaced by
3109    ptr++;    /* Point after the escape sequence */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3110    }      uses more memory, which we cannot get at this stage.) */
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3111    
3112  /* Skip whitespace and comments in extended mode */      return base_list[1] != 0;
3113    
3114  if ((options & PCRE_EXTENDED) != 0)      case OP_KET:
3115    {      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3116    for (;;)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3117      {      cannot be converted to a possessive form. */
3118      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
3119      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[1] == 0) return FALSE;
3120    
3121        switch(*(code - GET(code, 1)))
3122          {
3123          case OP_ASSERT:
3124          case OP_ASSERT_NOT:
3125          case OP_ASSERTBACK:
3126          case OP_ASSERTBACK_NOT:
3127          case OP_ONCE:
3128          case OP_ONCE_NC:
3129          /* Atomic sub-patterns and assertions can always auto-possessify their
3130          last iterator. */
3131          return TRUE;
3132          }
3133    
3134        code += PRIV(OP_lengths)[c];
3135        continue;
3136    
3137        case OP_ONCE:
3138        case OP_ONCE_NC:
3139        case OP_BRA:
3140        case OP_CBRA:
3141        next_code = code + GET(code, 1);
3142        code += PRIV(OP_lengths)[c];
3143    
3144        while (*next_code == OP_ALT)
3145        {        {
3146        ptr++;        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3147        while (*ptr != 0)        code = next_code + 1 + LINK_SIZE;
3148          {        next_code += GET(next_code, 1);
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
3149        }        }
3150      else break;      continue;
     }  
   }  
3151    
3152  /* If the next thing is itself optional, we have to give up. */      case OP_BRAZERO:
3153        case OP_BRAMINZERO:
3154    
3155  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      next_code = code + 1;
3156    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      if (*next_code != OP_BRA && *next_code != OP_CBRA
3157      return FALSE;          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3158    
3159        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3160    
3161        /* The bracket content will be checked by the
3162        OP_BRA/OP_CBRA case above. */
3163        next_code += 1 + LINK_SIZE;
3164        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3165          return FALSE;
3166    
3167  /* If the previous item is a character, get its value. */      code += PRIV(OP_lengths)[c];
3168        continue;
3169        }
3170    
3171  if (op_code == OP_CHAR || op_code == OP_CHARI ||    /* Check for a supported opcode, and load its properties. */
     op_code == OP_NOT || op_code == OP_NOTI)  
   //if (escape == 0) switch(op_code)  
   {  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3172    
3173  /* Now compare the next item with the previous opcode. First, handle cases when    code = get_chr_property_list(code, utf, cd->fcc, list);
3174  the next item is a character. */    if (code == NULL) return FALSE;    /* Unsupported */
3175    
3176  if (escape == 0)    /* If either opcode is a small character list, set pointers for comparing
3177    {    characters from that list with another list, or with a property. */
3178    /* For a caseless UTF match, the next character may have more than one other  
3179    case, which maps to the special PT_CLIST property. Check this first. */    if (base_list[0] == OP_CHAR)
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3180      {      {
3181      int ocs = UCD_CASESET(next);      chr_ptr = base_list + 2;
3182      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      list_ptr = list;
3183      }      }
3184  #endif    else if (list[0] == OP_CHAR)
   
   switch(op_code)  
3185      {      {
3186      case OP_CHAR:      chr_ptr = list + 2;
3187      return c != next;      list_ptr = base_list;
3188        }
3189    
3190      /* For CHARI (caseless character) we must check the other case. If we have    /* Character bitsets can also be compared to certain opcodes. */
     Unicode property support, we can use it to test the other case of  
     high-valued characters. We know that next can have only one other case,  
     because multi-other-case characters are dealt with above. */  
3191    
3192      case OP_CHARI:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3193      if (c == next) return FALSE;  #ifdef COMPILE_PCRE8
3194  #ifdef SUPPORT_UTF        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3195      if (utf)        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3196        {  #endif
3197        pcre_uint32 othercase;        )
3198        if (next < 128) othercase = cd->fcc[next]; else      {
3199  #ifdef SUPPORT_UCP  #ifdef COMPILE_PCRE8
3200        othercase = UCD_OTHERCASE(next);      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3201  #else  #else
3202        othercase = NOTACHAR;      if (base_list[0] == OP_CLASS)
3203  #endif  #endif
3204        return c != othercase;        {
3205          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3206          list_ptr = list;
3207        }        }
3208      else      else
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
   
     case OP_NOT:  
     return c == next;  
   
     case OP_NOTI:  
     if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
3209        {        {
3210        pcre_uint32 othercase;        set1 = (pcre_uint32 *)(code - list[2]);
3211        if (next < 128) othercase = cd->fcc[next]; else        list_ptr = base_list;
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
3212        }        }
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3213    
3214      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      invert_bits = FALSE;
3215      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      switch(list_ptr[0])
3216          {
3217          case OP_CLASS:
3218          case OP_NCLASS:
3219          set2 = (pcre_uint32 *)
3220            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3221          break;
3222    
3223      case OP_DIGIT:        /* OP_XCLASS cannot be supported here, because its bitset
3224      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        is not necessarily complete. E.g: [a-\0x{200}] is stored
3225          as a character range, and the appropriate bits are not set. */
3226    
3227      case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
3228      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;          invert_bits = TRUE;
3229            /* Fall through */
3230          case OP_DIGIT:
3231            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3232            break;
3233    
3234      case OP_WHITESPACE:        case OP_NOT_WHITESPACE:
3235      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;          invert_bits = TRUE;
3236            /* Fall through */
3237          case OP_WHITESPACE:
3238            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3239            break;
3240    
3241      case OP_NOT_WHITESPACE:        case OP_NOT_WORDCHAR:
3242      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;          invert_bits = TRUE;
3243            /* Fall through */
3244          case OP_WORDCHAR:
3245            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3246            break;
3247    
3248      case OP_WORDCHAR:        default:
3249      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        return FALSE;
3250          }
3251    
3252      case OP_NOT_WORDCHAR:      /* Compare 4 bytes to improve speed. */
3253      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;      set_end = set1 + (32 / 4);
3254        if (invert_bits)
3255          {
3256          do
3257            {
3258            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3259            }
3260          while (set1 < set_end);
3261          }
3262        else
3263          {
3264          do
3265            {
3266            if ((*set1++ & *set2++) != 0) return FALSE;
3267            }
3268          while (set1 < set_end);
3269          }
3270    
3271      case OP_HSPACE:      if (list[1] == 0) return TRUE;
3272      case OP_NOT_HSPACE:      /* Might be an empty repeat. */
3273      switch(next)      continue;
3274        }
3275    
3276      /* Some property combinations also acceptable. Unicode property opcodes are
3277      processed specially; the rest can be handled with a lookup table. */
3278    
3279      else
3280        {
3281        pcre_uint32 leftop, rightop;
3282    
3283        leftop = base_list[0];
3284        rightop = list[0];
3285    
3286    #ifdef SUPPORT_UCP
3287        accepted = FALSE; /* Always set in non-unicode case. */
3288        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3289        {        {
3290        HSPACE_CASES:        if (rightop == OP_EOD)
3291        return op_code == OP_NOT_HSPACE;          accepted = TRUE;
3292          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3293            {
3294            int n;
3295            const pcre_uint8 *p;
3296            BOOL same = leftop == rightop;
3297            BOOL lisprop = leftop == OP_PROP;
3298            BOOL risprop = rightop == OP_PROP;
3299            BOOL bothprop = lisprop && risprop;
3300    
3301            /* There's a table that specifies how each combination is to be
3302            processed:
3303              0   Always return FALSE (never auto-possessify)
3304              1   Character groups are distinct (possessify if both are OP_PROP)
3305              2   Check character categories in the same group (general or particular)
3306              3   Return TRUE if the two opcodes are not the same
3307              ... see comments below
3308            */
3309    
3310            n = propposstab[base_list[2]][list[2]];
3311            switch(n)
3312              {
3313              case 0: break;
3314              case 1: accepted = bothprop; break;
3315              case 2: accepted = (base_list[3] == list[3]) != same; break;
3316              case 3: accepted = !same; break;
3317    
3318        default:            case 4:  /* Left general category, right particular category */
3319        return op_code != OP_NOT_HSPACE;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3320              break;
3321    
3322              case 5:  /* Right general category, left particular category */
3323              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3324              break;
3325    
3326              /* This code is logically tricky. Think hard before fiddling with it.
3327              The posspropstab table has four entries per row. Each row relates to
3328              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3329              Only WORD actually needs all four entries, but using repeats for the
3330              others means they can all use the same code below.
3331    
3332              The first two entries in each row are Unicode general categories, and
3333              apply always, because all the characters they include are part of the
3334              PCRE character set. The third and fourth entries are a general and a
3335              particular category, respectively, that include one or more relevant
3336              characters. One or the other is used, depending on whether the check
3337              is for a general or a particular category. However, in both cases the
3338              category contains more characters than the specials that are defined
3339              for the property being tested against. Therefore, it cannot be used
3340              in a NOTPROP case.
3341    
3342              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3343              Underscore is covered by ucp_P or ucp_Po. */
3344    
3345              case 6:  /* Left alphanum vs right general category */
3346              case 7:  /* Left space vs right general category */
3347              case 8:  /* Left word vs right general category */
3348              p = posspropstab[n-6];
3349              accepted = risprop && lisprop ==
3350                (list[3] != p[0] &&
3351                 list[3] != p[1] &&
3352                (list[3] != p[2] || !lisprop));
3353              break;
3354    
3355              case 9:   /* Right alphanum vs left general category */
3356              case 10:  /* Right space vs left general category */
3357              case 11:  /* Right word vs left general category */
3358              p = posspropstab[n-9];
3359              accepted = lisprop && risprop ==
3360                (base_list[3] != p[0] &&
3361                 base_list[3] != p[1] &&
3362                (base_list[3] != p[2] || !risprop));
3363              break;
3364    
3365              case 12:  /* Left alphanum vs right particular category */
3366              case 13:  /* Left space vs right particular category */
3367              case 14:  /* Left word vs right particular category */
3368              p = posspropstab[n-12];
3369              accepted = risprop && lisprop ==
3370                (catposstab[p[0]][list[3]] &&
3371                 catposstab[p[1]][list[3]] &&
3372                (list[3] != p[3] || !lisprop));
3373              break;
3374    
3375              case 15:  /* Right alphanum vs left particular category */
3376              case 16:  /* Right space vs left particular category */
3377              case 17:  /* Right word vs left particular category */
3378              p = posspropstab[n-15];
3379              accepted = lisprop && risprop ==
3380                (catposstab[p[0]][base_list[3]] &&
3381                 catposstab[p[1]][base_list[3]] &&
3382                (base_list[3] != p[3] || !risprop));
3383              break;
3384              }
3385            }
3386        }        }
3387    
3388      case OP_ANYNL:      else
3389      case OP_VSPACE:  #endif  /* SUPPORT_UCP */
3390      case OP_NOT_VSPACE:  
3391      switch(next)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3392               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3393               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3394    
3395        if (!accepted)
3396          return FALSE;
3397    
3398        if (list[1] == 0) return TRUE;
3399        /* Might be an empty repeat. */
3400        continue;
3401        }
3402    
3403      /* Control reaches here only if one of the items is a small character list.
3404      All characters are checked against the other side. */
3405    
3406      do
3407        {
3408        chr = *chr_ptr;
3409    
3410        switch(list_ptr[0])
3411        {        {
3412        VSPACE_CASES:        case OP_CHAR:
3413        return op_code == OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3414          do
3415            {
3416            if (chr == *ochr_ptr) return FALSE;
3417            ochr_ptr++;
3418            }
3419          while(*ochr_ptr != NOTACHAR);
3420          break;
3421    
3422        default:        case OP_NOT:
3423        return op_code != OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3424        }        do
3425            {
3426            if (chr == *ochr_ptr)
3427              break;
3428            ochr_ptr++;
3429            }
3430          while(*ochr_ptr != NOTACHAR);
3431          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3432          break;
3433    
3434          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3435          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3436    
3437          case OP_DIGIT:
3438          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3439          break;
3440    
3441          case OP_NOT_DIGIT:
3442          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3443          break;
3444    
3445          case OP_WHITESPACE:
3446          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3447          break;
3448    
3449          case OP_NOT_WHITESPACE:
3450          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3451          break;
3452    
3453          case OP_WORDCHAR:
3454          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3455          break;
3456    
3457          case OP_NOT_WORDCHAR:
3458          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3459          break;
3460    
3461          case OP_HSPACE:
3462          switch(chr)
3463            {
3464            HSPACE_CASES: return FALSE;
3465            default: break;
3466            }
3467          break;
3468    
3469          case OP_NOT_HSPACE:
3470          switch(chr)
3471            {
3472            HSPACE_CASES: break;
3473            default: return FALSE;
3474            }
3475          break;
3476    
3477          case OP_ANYNL:
3478          case OP_VSPACE:
3479          switch(chr)
3480            {
3481            VSPACE_CASES: return FALSE;
3482            default: break;
3483            }
3484          break;
3485    
3486          case OP_NOT_VSPACE:
3487          switch(chr)
3488            {
3489            VSPACE_CASES: break;
3490            default: return FALSE;
3491            }
3492          break;
3493    
3494          case OP_DOLL:
3495          case OP_EODN:
3496          switch (chr)
3497            {
3498            case CHAR_CR:
3499            case CHAR_LF:
3500            case CHAR_VT:
3501            case CHAR_FF:
3502            case CHAR_NEL:
3503    #ifndef EBCDIC
3504            case 0x2028:
3505            case 0x2029:
3506    #endif  /* Not EBCDIC */
3507            return FALSE;
3508            }
3509          break;
3510    
3511          case OP_EOD:    /* Can always possessify before \z */
3512          break;
3513    
3514  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3515      case OP_PROP:        case OP_PROP:
3516      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3517          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3518                list_ptr[0] == OP_NOTPROP))
3519            return FALSE;
3520          break;
3521    #endif
3522    
3523      case OP_NOTPROP:        case OP_NCLASS:
3524      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3525          /* Fall through */
3526    
3527          case OP_CLASS:
3528          if (chr > 255) break;
3529          class_bitset = (pcre_uint8 *)
3530            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3531          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3532          break;
3533    
3534    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3535          case OP_XCLASS:
3536          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3537              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3538          break;
3539  #endif  #endif
3540    
3541      default:        default:
3542      return FALSE;        return FALSE;
3543          }
3544    
3545        chr_ptr++;
3546      }      }
3547      while(*chr_ptr != NOTACHAR);
3548    
3549      /* At least one character must be matched from this opcode. */
3550    
3551      if (list[1] == 0) return TRUE;
3552    }    }
3553    
3554  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  return FALSE;
3555  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  }
3556  generated only when PCRE_UCP is *not* set, that is, when only ASCII  
3557  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
3558  replaced by OP_PROP codes when PCRE_UCP is set. */  
3559    /*************************************************
3560    *    Scan compiled regex for auto-possession     *
3561    *************************************************/
3562    
3563    /* Replaces single character iterations with their possessive alternatives
3564    if appropriate. This function modifies the compiled opcode!
3565    
3566    Arguments:
3567      code        points to start of the byte code
3568      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3569      cd          static compile data
3570    
3571    Returns:      nothing
3572    */
3573    
3574    static void
3575    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3576    {
3577    register pcre_uchar c;
3578    const pcre_uchar *end;
3579    pcre_uchar *repeat_opcode;
3580    pcre_uint32 list[8];
3581    
3582  switch(op_code)  for (;;)
3583    {    {
3584    case OP_CHAR:    c = *code;
3585    case OP_CHARI:  
3586    switch(escape)    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3587      {      {
3588      case ESC_d:      c -= get_repeat_base(c) - OP_STAR;
3589      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      end = (c <= OP_MINUPTO) ?
3590          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3591        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3592    
3593        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3594          {
3595          switch(c)
3596            {
3597            case OP_STAR:
3598            *code += OP_POSSTAR - OP_STAR;
3599            break;
3600    
3601            case OP_MINSTAR:
3602            *code += OP_POSSTAR - OP_MINSTAR;
3603            break;
3604    
3605      case ESC_D:          case OP_PLUS:
3606      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;          *code += OP_POSPLUS - OP_PLUS;
3607            break;
3608    
3609      case ESC_s:          case OP_MINPLUS:
3610      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;          *code += OP_POSPLUS - OP_MINPLUS;
3611            break;
3612    
3613      case ESC_S:          case OP_QUERY:
3614      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;          *code += OP_POSQUERY - OP_QUERY;
3615            break;
3616    
3617      case ESC_w:          case OP_MINQUERY:
3618      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;          *code += OP_POSQUERY - OP_MINQUERY;
3619            break;
3620    
3621      case ESC_W:          case OP_UPTO:
3622      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;          *code += OP_POSUPTO - OP_UPTO;
3623            break;
3624    
3625      case ESC_h:          case OP_MINUPTO:
3626      case ESC_H:          *code += OP_MINUPTO - OP_UPTO;
3627      switch(c)          break;
3628        {          }
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
3629        }        }
3630        c = *code;
3631        }
3632      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3633        {
3634    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3635        if (c == OP_XCLASS)
3636          repeat_opcode = code + GET(code, 1);
3637        else
3638    #endif
3639          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3640    
3641      case ESC_v:      c = *repeat_opcode;
3642      case ESC_V:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
     switch(c)  
3643        {        {
3644        VSPACE_CASES:        /* end must not be NULL. */
3645        return escape != ESC_v;        end = get_chr_property_list(code, utf, cd->fcc, list);
3646    
3647        default:        list[1] = (c & 1) == 0;
       return escape == ESC_v;  
       }  
3648    
3649      /* When PCRE_UCP is set, these values get generated for \d etc. Find        if (compare_opcodes(end, utf, cd, list, end))
3650      their substitutions and process them. The result will always be either          {
3651      ESC_p or ESC_P. Then fall through to process those values. */          switch (c)
3652              {
3653              case OP_CRSTAR:
3654              case OP_CRMINSTAR:
3655              *repeat_opcode = OP_CRPOSSTAR;
3656              break;
3657    
3658  #ifdef SUPPORT_UCP            case OP_CRPLUS:
3659      case ESC_du:            case OP_CRMINPLUS:
3660      case ESC_DU:            *repeat_opcode = OP_CRPOSPLUS;
3661      case ESC_wu:            break;
3662      case ESC_WU:  
3663      case ESC_su:            case OP_CRQUERY:
3664      case ESC_SU:            case OP_CRMINQUERY:
3665        {            *repeat_opcode = OP_CRPOSQUERY;
3666        int temperrorcode = 0;            break;
3667        ptr = substitutes[escape - ESC_DU];  
3668        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);            case OP_CRRANGE:
3669        if (temperrorcode != 0) return FALSE;            case OP_CRMINRANGE:
3670        ptr++;    /* For compatibility */            *repeat_opcode = OP_CRPOSRANGE;
3671              break;
3672              }
3673            }
3674        }        }
3675      /* Fall through */      c = *code;
3676        }
3677    
3678      case ESC_p:    switch(c)
3679      case ESC_P:      {
3680        {      case OP_END:
3681        int ptype, pdata, errorcodeptr;      return;
3682        BOOL negated;  
3683        case OP_TYPESTAR:
3684        case OP_TYPEMINSTAR:
3685        case OP_TYPEPLUS:
3686        case OP_TYPEMINPLUS:
3687        case OP_TYPEQUERY:
3688        case OP_TYPEMINQUERY:
3689        case OP_TYPEPOSSTAR:
3690        case OP_TYPEPOSPLUS:
3691        case OP_TYPEPOSQUERY:
3692        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3693        break;
3694    
3695        case OP_TYPEUPTO:
3696        case OP_TYPEMINUPTO:
3697        case OP_TYPEEXACT:
3698        case OP_TYPEPOSUPTO:
3699        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3700          code += 2;
3701        break;
3702    
3703    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3704        case OP_XCLASS:
3705        code += GET(code, 1);
3706        break;
3707    #endif
3708    
3709        case OP_MARK:
3710        case OP_PRUNE_ARG:
3711        case OP_SKIP_ARG:
3712        case OP_THEN_ARG:
3713        code += code[1];
3714        break;
3715        }
3716    
3717        ptr--;      /* Make ptr point at the p or P */    /* Add in the fixed length from the table */
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3718    
3719        /* If the property item is optional, we have to give up. (When generated    code += PRIV(OP_lengths)[c];
3720        from \d etc by PCRE_UCP, this test will have been applied much earlier,  
3721        to the original \d etc. At this point, ptr will point to a zero byte. */    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3722      a multi-byte character. The length in the table is a minimum, so we have to
3723      arrange to skip the extra bytes. */
3724    
3725        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3726          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    if (utf) switch(c)
3727            return FALSE;      {
3728        case OP_CHAR:
3729        case OP_CHARI:
3730        case OP_NOT:
3731        case OP_NOTI:
3732        case OP_STAR:
3733        case OP_MINSTAR:
3734        case OP_PLUS:
3735        case OP_MINPLUS:
3736        case OP_QUERY:
3737        case OP_MINQUERY:
3738        case OP_UPTO:
3739        case OP_MINUPTO:
3740        case OP_EXACT:
3741        case OP_POSSTAR:
3742        case OP_POSPLUS:
3743        case OP_POSQUERY:
3744        case OP_POSUPTO:
3745        case OP_STARI:
3746        case OP_MINSTARI:
3747        case OP_PLUSI:
3748        case OP_MINPLUSI:
3749        case OP_QUERYI:
3750        case OP_MINQUERYI:
3751        case OP_UPTOI:
3752        case OP_MINUPTOI:
3753        case OP_EXACTI:
3754        case OP_POSSTARI:
3755        case OP_POSPLUSI:
3756        case OP_POSQUERYI:
3757        case OP_POSUPTOI:
3758        case OP_NOTSTAR:
3759        case OP_NOTMINSTAR:
3760        case OP_NOTPLUS:
3761        case OP_NOTMINPLUS:
3762        case OP_NOTQUERY:
3763        case OP_NOTMINQUERY:
3764        case OP_NOTUPTO:
3765        case OP_NOTMINUPTO:
3766        case OP_NOTEXACT:
3767        case OP_NOTPOSSTAR:
3768        case OP_NOTPOSPLUS:
3769        case OP_NOTPOSQUERY:
3770        case OP_NOTPOSUPTO:
3771        case OP_NOTSTARI:
3772        case OP_NOTMINSTARI:
3773        case OP_NOTPLUSI:
3774        case OP_NOTMINPLUSI:
3775        case OP_NOTQUERYI:
3776        case OP_NOTMINQUERYI:
3777        case OP_NOTUPTOI:
3778        case OP_NOTMINUPTOI:
3779        case OP_NOTEXACTI:
3780        case OP_NOTPOSSTARI:
3781        case OP_NOTPOSPLUSI:
3782        case OP_NOTPOSQUERYI:
3783        case OP_NOTPOSUPTOI:
3784        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3785        break;
3786        }
3787    #else
3788      (void)(utf);  /* Keep compiler happy by referencing function argument */
3789    #endif
3790      }
3791    }
3792    
3793    
3794    
3795    /*************************************************
3796    *           Check for POSIX class syntax         *
3797    *************************************************/
3798    
3799    /* This function is called when the sequence "[:" or "[." or "[=" is
3800    encountered in a character class. It checks whether this is followed by a
3801    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3802    reach an unescaped ']' without the special preceding character, return FALSE.
3803    
3804    Originally, this function only recognized a sequence of letters between the
3805    terminators, but it seems that Perl recognizes any sequence of characters,
3806    though of course unknown POSIX names are subsequently rejected. Perl gives an
3807    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3808    didn't consider this to be a POSIX class. Likewise for [:1234:].
3809    
3810    The problem in trying to be exactly like Perl is in the handling of escapes. We
3811    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3812    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3813    below handles the special case of \], but does not try to do any other escape
3814    processing. This makes it different from Perl for cases such as [:l\ower:]
3815    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3816    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3817    I think.
3818    
3819    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3820    It seems that the appearance of a nested POSIX class supersedes an apparent
3821    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3822    a digit.
3823    
3824    In Perl, unescaped square brackets may also appear as part of class names. For
3825    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3826    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3827    seem right at all. PCRE does not allow closing square brackets in POSIX class
3828    names.
3829    
3830    Arguments:
3831      ptr      pointer to the initial [
3832      endptr   where to return the end pointer
3833    
3834    Returns:   TRUE or FALSE
3835    */
3836    
3837    static BOOL
3838    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3839    {
3840    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3841    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3842    for (++ptr; *ptr != CHAR_NULL; ptr++)
3843      {
3844      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3845        ptr++;
3846      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3847      else
3848        {
3849        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3850          {
3851          *endptr = ptr;
3852          return TRUE;
3853          }
3854        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3855             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3856              ptr[1] == CHAR_EQUALS_SIGN) &&
3857            check_posix_syntax(ptr, endptr))
3858          return FALSE;
3859        }
3860      }
3861    return FALSE;
3862    }
3863    
3864    
3865    
3866    
3867    /*************************************************
3868    *          Check POSIX class name                *
3869    *************************************************/
3870    
3871    /* This function is called to check the name given in a POSIX-style class entry
3872    such as [:alnum:].
3873    
3874    Arguments:
3875      ptr        points to the first letter
3876      len        the length of the name
3877    
3878    Returns:     a value representing the name, or -1 if unknown
3879    */
3880    
3881    static int
3882    check_posix_name(const pcre_uchar *ptr, int len)
3883    {
3884    const char *pn = posix_names;
3885    register int yield = 0;
3886    while (posix_name_lengths[yield] != 0)
3887      {
3888      if (len == posix_name_lengths[yield] &&
3889        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3890      pn += posix_name_lengths[yield] + 1;
3891      yield++;
3892      }
3893    return -1;
3894    }
3895    
3896    
3897    /*************************************************
3898    *    Adjust OP_RECURSE items in repeated group   *
3899    *************************************************/
3900    
3901    /* OP_RECURSE items contain an offset from the start of the regex to the group
3902    that is referenced. This means that groups can be replicated for fixed
3903    repetition simply by copying (because the recursion is allowed to refer to
3904    earlier groups that are outside the current group). However, when a group is
3905    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3906    inserted before it, after it has been compiled. This means that any OP_RECURSE
3907    items within it that refer to the group itself or any contained groups have to
3908    have their offsets adjusted. That one of the jobs of this function. Before it
3909    is called, the partially compiled regex must be temporarily terminated with
3910    OP_END.
3911    
3912    This function has been extended with the possibility of forward references for
3913    recursions and subroutine calls. It must also check the list of such references
3914    for the group we are dealing with. If it finds that one of the recursions in
3915    the current group is on this list, it adjusts the offset in the list, not the
3916    value in the reference (which is a group number).
3917    
3918    Arguments:
3919      group      points to the start of the group
3920      adjust     the amount by which the group is to be moved
3921      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3922      cd         contains pointers to tables etc.
3923      save_hwm   the hwm forward reference pointer at the start of the group
3924    
3925    Returns:     nothing
3926    */
3927    
3928    static void
3929    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3930      pcre_uchar *save_hwm)
3931    {
3932    pcre_uchar *ptr = group;
3933    
3934    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3935      {
3936      int offset;
3937      pcre_uchar *hc;
3938    
3939      /* See if this recursion is on the forward reference list. If so, adjust the
3940      reference. */
3941    
3942      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3943        {
3944        offset = (int)GET(hc, 0);
3945        if (cd->start_code + offset == ptr + 1)
3946          {
3947          PUT(hc, 0, offset + adjust);
3948          break;
3949          }
3950        }
3951    
3952      /* Otherwise, adjust the recursion offset if it's after the start of this
3953      group. */
3954    
3955      if (hc >= cd->hwm)
3956        {
3957        offset = (int)GET(ptr, 1);
3958        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3959        }
3960    
3961      ptr += 1 + LINK_SIZE;
3962      }
3963    }
3964    
3965    
3966    
3967    /*************************************************
3968    *        Insert an automatic callout point       *
3969    *************************************************/
3970    
3971    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3972    callout points before each pattern item.
3973    
3974    Arguments:
3975      code           current code pointer
3976      ptr            current pattern pointer
3977      cd             pointers to tables etc
3978    
3979    Returns:         new code pointer
3980    */
3981    
3982    static pcre_uchar *
3983    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3984    {
3985    *code++ = OP_CALLOUT;
3986    *code++ = 255;
3987    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3988    PUT(code, LINK_SIZE, 0);                       /* Default length */
3989    return code + 2 * LINK_SIZE;
3990    }
3991    
3992    
3993    
3994    /*************************************************
3995    *         Complete a callout item                *
3996    *************************************************/
3997    
3998    /* A callout item contains the length of the next item in the pattern, which
3999    we can't fill in till after we have reached the relevant point. This is used
4000    for both automatic and manual callouts.
4001    
4002        /* Do the property check. */  Arguments:
4003      previous_callout   points to previous callout item
4004      ptr                current pattern pointer
4005      cd                 pointers to tables etc
4006    
4007        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
4008        }  */
 #endif  
4009    
4010      default:  static void
4011      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4012      }  {
4013    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4014    PUT(previous_callout, 2 + LINK_SIZE, length);
4015    }
4016    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
4017    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
4018    
4019    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4020    return escape == ESC_d;  /*************************************************
4021    *           Get othercase range                  *
4022    *************************************************/
4023    
4024    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4025    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4026    characters in the "other" case. Each call returns the next one, updating the
4027    start address. A character with multiple other cases is returned on its own
4028    with a special return value.
4029    
4030    case OP_NOT_WHITESPACE:  Arguments:
4031    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
4032      d           end value
4033      ocptr       where to put start of othercase range
4034      odptr       where to put end of othercase range
4035    
4036    case OP_HSPACE:  Yield:        -1 when no more
4037    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
4038           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
4039                    in this case, ocptr contains the original
4040    */
4041    
4042    case OP_NOT_HSPACE:  static int
4043    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4044      pcre_uint32 *odptr)
4045    {
4046    pcre_uint32 c, othercase, next;
4047    unsigned int co;
4048    
4049    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
4050    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
4051    
4052    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4053    return escape == ESC_v || escape == ESC_R;    {
4054      if ((co = UCD_CASESET(c)) != 0)
4055        {
4056        *ocptr = c++;   /* Character that has the set */
4057        *cptr = c;      /* Rest of input range */
4058        return (int)co;
4059        }
4060      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4061      }
4062    
4063    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  
          escape == ESC_v || escape == ESC_R;  
4064    
4065    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4066    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4067    
4068    default:  for (++c; c <= d; c++)
4069    return FALSE;    {
4070      if (UCD_OTHERCASE(c) != next) break;
4071      next++;
4072    }    }
4073    
4074  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4075    *cptr = c;             /* Rest of input range */
4076    return 0;
4077  }  }
4078    #endif  /* SUPPORT_UCP */
4079    
4080    
4081    
# Line 3418  switch(op_code) Line 4084  switch(op_code)
4084  *************************************************/  *************************************************/
4085    
4086  /* 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
4087  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
4088  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
4089  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4090    
4091  Arguments:  Arguments:
4092    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4093    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4094    options       the options word    options       the options word
4095    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4096    start         start of range character    start         start of range character
4097    end           end of range character    end           end of range character
4098    
4099  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4100                  the pointer to extra data is updated                  the pointer to extra data is updated
4101  */  */
# Line 3441  add_to_class(pcre_uint8 *classbits, pcre Line 4107  add_to_class(pcre_uint8 *classbits, pcre
4107  pcre_uint32 c;  pcre_uint32 c;
4108  int n8 = 0;  int n8 = 0;
4109    
4110  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4111  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
4112  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
4113  range. */  range. */
4114    
4115  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4116    {    {
4117  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4118    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4119      {      {
4120      int rc;      int rc;
4121      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4122    
4123      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4124      c = start;      c = start;
4125    
4126      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4127        {        {
4128        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4129    
4130        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4131          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4132    
4133        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4134    
4135        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4136    
4137        /* 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
4138        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
4139        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4140    
4141        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4142        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4143        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 4147  if ((options & PCRE_CASELESS) != 0)
4147  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4148    
4149    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4150    
4151    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4152      {      {
4153      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4154      n8++;      n8++;
4155      }      }
4156    }    }
4157    
4158  /* 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
4159  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
4160  in all cases. */  in all cases. */
# Line 3514  if (end < 0x100) Line 4180  if (end < 0x100)
4180    {    {
4181    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4182      {      {
4183      n8++;      n8++;
4184      SETBIT(classbits, c);      SETBIT(classbits, c);
4185      }      }
4186    }    }
4187    
4188  else  else
4189    {    {
4190    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4191    
4192  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4193    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4194      {      {
4195      if (start < end)      if (start < end)
4196        {        {
4197        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4198        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4199        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4200        }        }
4201      else if (start == end)      else if (start == end)
4202        {        {
4203        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4204        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4205        }        }
4206      }      }
4207    else    else
4208  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4209    
4210    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4211    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4212    
4213  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4214      {}      {}
4215  #else  #else
4216    if (start < end)    if (start < end)
4217      {      {
4218      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4223  else
4223      {      {
4224      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4225      *uchardata++ = start;      *uchardata++ = start;
4226      }      }
4227  #endif  #endif
4228    
4229    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4230    }    }
4231    
4232  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4233  }  }
4234    
4235    
4236    
4237    
4238  /*************************************************  /*************************************************
4239  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4240  *************************************************/  *************************************************/
4241    
4242  /* 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
4243  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
4244  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
4245  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4249  Arguments:
4249    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4250    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4251    options       the options word    options       the options word
4252    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4253    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4254    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4255                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4256                    already know about                    already know about
4257    
4258  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4259                  the pointer to extra data is updated                  the pointer to extra data is updated
4260  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4268  while (p[0] < NOTACHAR)
4268    {    {
4269    int n = 0;    int n = 0;
4270    if (p[0] != except)    if (p[0] != except)
4271      {      {
4272      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4273      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4274      }      }
4275    p += n + 1;    p += n + 1;
4276    }    }
4277  return n8;  return n8;
4278  }  }
4279    
4280    
4281    
# Line 3624  Arguments: Line 4290  Arguments:
4290    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4291    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4292    options       the options word    options       the options word
4293    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4294    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4295    
4296  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4297                  the pointer to extra data is updated                  the pointer to extra data is updated
4298  */  */
4299    
4300  static int  static int
4301  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4302    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4303  {  {
4304  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4310  while (p[0] < NOTACHAR)
4310    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4311    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4312      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4313    p++;    p++;
4314    }    }
4315  return n8;  return n8;
4316  }  }
4317    
4318    
4319    
# Line 3662  to find out the amount of memory needed, Line 4328  to find out the amount of memory needed,
4328  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4329    
4330  Arguments:  Arguments:
4331    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4332    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4333    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4334    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4335    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4336    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4337    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4338    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4339    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4340    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4341                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4342      lengthptr         NULL during the real compile phase
4343                        points to length accumulator during pre-compile phase
4344    
4345  Returns:         TRUE on success  Returns:            TRUE on success
4346                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4347  */  */
4348    
4349  static BOOL  static BOOL
4350  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4351    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4352    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4353      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4354      branch_chain *bcptr, int cond_depth,
4355    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4356  {  {
4357  int repeat_type, op_type;  int repeat_type, op_type;
4358  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4359  int bravalue = 0;  int bravalue = 0;
4360  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4361  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4362  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4363    pcre_uint32 zeroreqchar, zerofirstchar;
4364    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4365  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4366  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4367  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4389  dynamically as we process the pattern. *
4389  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4390  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4391  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4392    #ifndef COMPILE_PCRE32
4393  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4394    #endif
4395  #else  #else
4396  BOOL utf = FALSE;  BOOL utf = FALSE;
4397  #endif  #endif
4398    
4399  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4400  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,
4401  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
4402  alternative calls for the different cases. */  alternative calls for the different cases. */
4403    
4404  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4426  to take the zero repeat into account. Th
4426  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4427  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4428    
4429  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4430    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4431    
4432  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4433  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3778  for (;; ptr++) Line 4453  for (;; ptr++)
4453    int recno;    int recno;
4454    int refsign;    int refsign;
4455    int skipbytes;    int skipbytes;
4456    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4457    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4458    int terminator;    int terminator;
4459    int mclength;    unsigned int mclength;
4460    int tempbracount;    unsigned int tempbracount;
4461    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4462    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4463    
4464    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4468  for (;; ptr++)
4468    /* 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
4469    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4470    
4471    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4472      {      {
4473      ptr = nestptr;      ptr = nestptr;
4474      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4543  for (;; ptr++)
4543    
4544    /* 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 */
4545    
4546    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4547      {      {
4548      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4549        {        {
# Line 3891  for (;; ptr++) Line 4566  for (;; ptr++)
4566          }          }
4567        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4568        }        }
4569        /* Control does not reach here. */
4570      }      }
4571    
4572    /* 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
4573    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. */  
4574    
4575    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4576      {      {
4577      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4578        {        {
4579          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4580          if (c != CHAR_NUMBER_SIGN) break;
4581        ptr++;        ptr++;
4582        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4583          {          {
4584          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4585              {                          /* IS_NEWLINE sets cd->nllen. */
4586              ptr += cd->nllen;
4587              break;
4588              }
4589          ptr++;          ptr++;
4590  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4591          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4592  #endif  #endif
4593          }          }
4594        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4595        }        }
4596      }      }
4597    
4598    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4599    
4600      is_quantifier =
4601        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4602        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4603    
4604      /* Fill in length of a previous callout, except when the next thing is a
4605      quantifier or when processing a property substitution string in UCP mode. */
4606    
4607      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4608           after_manual_callout-- <= 0)
4609        {
4610        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4611          complete_callout(previous_callout, ptr, cd);
4612        previous_callout = NULL;
4613        }
4614    
4615      /* Create auto callout, except for quantifiers, or while processing property
4616      strings that are substituted for \w etc in UCP mode. */
4617    
4618    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4619      {      {
4620      previous_callout = code;      previous_callout = code;
4621      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4622      }      }
4623    
4624      /* Process the next pattern item. */
4625    
4626    switch(c)    switch(c)
4627      {      {
4628      /* ===================================================================*/      /* ===================================================================*/
4629      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4630      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4631      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4632      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4633        *firstcharflagsptr = firstcharflags;
4634      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4635        *reqcharflagsptr = reqcharflags;
4636      *codeptr = code;      *codeptr = code;
4637      *ptrptr = ptr;      *ptrptr = ptr;
4638      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4656  for (;; ptr++)
4656      previous = NULL;      previous = NULL;
4657      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4658        {        {
4659        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4660        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4661        }        }
4662      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4671  for (;; ptr++)
4671      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4672    
4673      case CHAR_DOT:      case CHAR_DOT:
4674      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4675      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4676        zerofirstcharflags = firstcharflags;
4677      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4678        zeroreqcharflags = reqcharflags;
4679      previous = code;      previous = code;
4680      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4681      break;      break;
# Line 4015  for (;; ptr++) Line 4703  for (;; ptr++)
4703        goto FAILED;        goto FAILED;
4704        }        }
4705      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4706    
4707        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4708        used for "start of word" and "end of word". As these are otherwise illegal
4709        sequences, we don't break anything by recognizing them. They are replaced
4710        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4711        erroneous and are handled by the normal code below. */
4712    
4713      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4714        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4715          {
4716          nestptr = ptr + 7;
4717          ptr = sub_start_of_word - 1;
4718          continue;
4719          }
4720    
4721        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4722          {
4723          nestptr = ptr + 7;
4724          ptr = sub_end_of_word - 1;
4725          continue;
4726          }
4727    
4728        /* Handle a real character class. */
4729    
4730      previous = code;      previous = code;
4731    
4732      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
# Line 4061  for (;; ptr++) Line 4771  for (;; ptr++)
4771          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4772        {        {
4773        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4774        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4775        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4776          zerofirstcharflags = firstcharflags;
4777        break;        break;
4778        }        }
4779    
# Line 4097  for (;; ptr++) Line 4808  for (;; ptr++)
4808      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4809      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4810    
4811      if (c != 0) do      if (c != CHAR_NULL) do
4812        {        {
4813        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4814    
4815  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
4816        if (utf && HAS_EXTRALEN(c))        if (utf && HAS_EXTRALEN(c))
4817          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4818          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
# Line 4112  for (;; ptr++) Line 4823  for (;; ptr++)
4823        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4824        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
4825        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4826        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4827        however. */        however. */
4828    
4829        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4176  for (;; ptr++) Line 4887  for (;; ptr++)
4887          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4888          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4889    
4890          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4891            posix_class = 0;            posix_class = 0;
4892    
4893          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
4894          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4895            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4896            directly. */
4897    
4898  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4899          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4900            {            {
4901              unsigned int ptype = 0;
4902            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4903    
4904              /* The posix_substitutes table specifies which POSIX classes can be
4905              converted to \p or \P items. */
4906    
4907            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4908              {              {
4909              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4910              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4911              continue;              continue;
4912              }              }
4913    
4914              /* There are three other classes that generate special property calls
4915              that are recognized only in an XCLASS. */
4916    
4917              else switch(posix_class)
4918                {
4919                case PC_GRAPH:
4920                ptype = PT_PXGRAPH;
4921                /* Fall through */
4922                case PC_PRINT:
4923                if (ptype == 0) ptype = PT_PXPRINT;
4924                /* Fall through */
4925                case PC_PUNCT:
4926                if (ptype == 0) ptype = PT_PXPUNCT;
4927                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4928                *class_uchardata++ = ptype;
4929                *class_uchardata++ = 0;
4930                ptr = tempptr + 1;
4931                continue;
4932    
4933                /* For all other POSIX classes, no special action is taken in UCP
4934                mode. Fall through to the non_UCP case. */
4935    
4936                default:
4937                break;
4938                }
4939            }            }
4940  #endif  #endif
4941          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, or when UCP makes no difference, we build the
4942          chunk of local store because we may be adding and subtracting from it,          bit map for the POSIX class in a chunk of local store because we may be
4943          and we don't want to subtract bits that may be in the main map already.          adding and subtracting from it, and we don't want to subtract bits that
4944          At the end we or the result into the bit map that is being built. */          may be in the main map already. At the end we or the result into the
4945            bit map that is being built. */
4946    
4947          posix_class *= 3;          posix_class *= 3;
4948    
# Line 4253  for (;; ptr++) Line 4998  for (;; ptr++)
4998    
4999        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
5000          {          {
5001          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5002              TRUE);
5003          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
5004            if (escape == 0) c = ec;
         if (escape == 0)  
           c = ec;  
5005          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5006          else if (escape == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
5007            {            {
5008            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
5009            goto FAILED;            goto FAILED;
# Line 4316  for (;; ptr++) Line 5059  for (;; ptr++)
5059              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5060              continue;              continue;
5061    
5062              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5063              if it was previously set by something earlier in the character              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was