/[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 1518 by ph10, Fri Feb 6 16:10:27 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 47  supporting internal functions that are n Line 47  supporting internal functions that are n
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing pattern start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing pattern end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 250  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 265  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 292  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 319  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 372  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 387  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 452  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 518  static const char error_texts[] =
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 505  static const char error_texts[] = Line 539  static const char error_texts[] =
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553    ;    ;
554    
555  /* 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 689  static const pcre_uint8 ebcdic_chartab[]
689  #endif  #endif
690    
691    
692    /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730    /* This table is used to check whether auto-possessification is possible
731    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732    left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* 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 */
784      { 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 */
785      { 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 */
786      { 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 */
787      { 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 */
788      { 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 */
789      { 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 */
790      { 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 */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
870  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 886  find_error_text(int n)
886  const char *s = error_texts;  const char *s = error_texts;
887  for (; n > 0; n--)  for (; n > 0; n--)
888    {    {
889    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
890    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
891    }    }
892  return s;  return s;
893  }  }
894    
895    
896    
897  /*************************************************  /*************************************************
898  *           Expand the workspace                 *  *           Expand the workspace                 *
899  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 971  return (*p == CHAR_RIGHT_CURLY_BRACKET);
971  *************************************************/  *************************************************/
972    
973  /* 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
974  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
975  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.
976  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
977  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982    chptr          points to the data character    chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 992  Returns:         zero => a data characte
992  */  */
993    
994  static int  static int
995  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
997  {  {
998  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 1007  ptr--;                            /* Set
1007    
1008  /* 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. */
1009    
1010  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011    
1012  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013  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 1016  Otherwise further processing may be requ
1016  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017  /* Not alphanumeric */  /* Not alphanumeric */
1018  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1019  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)
1020      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021    
1022  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1023  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1067  else
1067            }            }
1068    
1069  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1070          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1071  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1072          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1073  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1074          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1075  #endif  #endif
1076            {            {
1077            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1118  else
1118      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119        {        {
1120        const pcre_uchar *p;        const pcre_uchar *p;
1121        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++)
1122          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124          {          {
1125          escape = ESC_k;          escape = ESC_k;
1126          break;          break;
# Line 961  else Line 1183  else
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      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
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      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
1191      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
1192      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
1193      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
1194      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
1195      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
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      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:
1202      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 1223  else
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          escape = -s;          escape = -s;
1229          break;          break;
# Line 1005  else Line 1231  else
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* 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
1235      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
1236      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
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \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
1244      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 1255  else
1255  #endif  #endif
1256      break;      break;
1257    
1258      /* \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
1259      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}. */
1260      If not, { is treated as a data character. */  
1261        case CHAR_o:
1262        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264          {
1265          ptr += 2;
1266          c = 0;
1267          overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269            {
1270            register pcre_uint32 cc = *ptr++;
1271            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272    #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274    #endif
1275            c = (c << 3) + cc - CHAR_0 ;
1276    #if defined COMPILE_PCRE8
1277            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE16
1279            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282    #endif
1283            }
1284          if (overflow)
1285            {
1286            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287            *errorcodeptr = ERR34;
1288            }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290            {
1291            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294          }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      case CHAR_x:      case CHAR_x:
1301      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1303        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305          {          {
# Line 1056  else Line 1316  else
1316  #endif  #endif
1317            }            }
1318          }          }
1319        break;        }    /* End JavaScript handling */
       }  
1320    
1321      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        c = 0;      else
1329        overflow = FALSE;        {
1330        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331          {          {
1332          register pcre_uint32 cc = *pt++;          ptr += 2;
1333          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1346          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1347  #endif  #endif
1348    
1349  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1353          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 */
1354          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355  #endif  #endif
1356    
1357  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1358          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1360          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1362          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363  #endif  #endif
1364          }            }
1365    
1366        if (overflow)          if (overflow)
1367          {            {
1368          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1370          }            }
1371    
1372        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373          {            {
1374          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375          ptr = pt;            }
         break;  
         }  
1376    
1377        /* 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.
1378        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1379        }          \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1387        {        else
1388        pcre_uint32 cc;                          /* Some compilers don't like */          {
1389        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1398        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400  #endif  #endif
1401        }            }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* 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 1410  else
1410    
1411      case CHAR_c:      case CHAR_c:
1412      c = *(++ptr);      c = *(++ptr);
1413      if (c == 0)      if (c == CHAR_NULL)
1414        {        {
1415        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1416        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1466  if ((options & PCRE_UCP) != 0 && escape
1466  return escape;  return escape;
1467  }  }
1468    
1469    
1470    
1471  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1472  /*************************************************  /*************************************************
1473  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1481  escape sequence.
1481  Argument:  Argument:
1482    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1483    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
1484    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
1485      pdataptr       points to an unsigned int that is set to the detailed property value
1486    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1487    
1488  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
1489  */  */
1490    
1491  static int  static BOOL
1492  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493      unsigned int *pdataptr, int *errorcodeptr)
1494  {  {
1495  pcre_uchar c;  pcre_uchar c;
1496  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1498  const pcre_uchar *ptr = *ptrptr;
1498  pcre_uchar name[32];  pcre_uchar name[32];
1499    
1500  c = *(++ptr);  c = *(++ptr);
1501  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1502    
1503  *negptr = FALSE;  *negptr = FALSE;
1504    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1515  if (c == CHAR_LEFT_CURLY_BRACKET)
1515    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516      {      {
1517      c = *(++ptr);      c = *(++ptr);
1518      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1519      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520      name[i] = c;      name[i] = c;
1521      }      }
# Line 1263  while (bot < top) Line 1545  while (bot < top)
1545    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);
1546    if (r == 0)    if (r == 0)
1547      {      {
1548      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1549      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1550        return TRUE;
1551      }      }
1552    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1553    }    }
1554    
1555  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1556  *ptrptr = ptr;  *ptrptr = ptr;
1557  return -1;  return FALSE;
1558    
1559  ERROR_RETURN:  ERROR_RETURN:
1560  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1561  *ptrptr = ptr;  *ptrptr = ptr;
1562  return -1;  return FALSE;
1563  }  }
1564  #endif  #endif
1565    
1566    
1567    
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 1308  read_repeat_counts(const pcre_uchar *p, Line 1590  read_repeat_counts(const pcre_uchar *p,
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597        {
1598        *errorcodeptr = ERR5;
1599        return p;
1600        }
1601    }    }
1602    
 /* Read the maximum value if there is one, and again do a paranoid on its size.  
 Also, max must not be less than min. */  
   
1603  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604    {    {
1605    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612            {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616        }        }
1617      if (max < min)      if (max < min)
1618        {        {
# Line 1340  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 1351  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1633  *      Find first significant op code            *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
# Line 1684  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671      case OP_NCREF:      case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673      case OP_NRREF:      case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
# Line 1700  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
# Line 1744  for (;;) Line 1727  for (;;)
1727    {    {
1728    int d;    int d;
1729    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1730    register int op = *cc;    register pcre_uchar op = *cc;
1731    
1732    switch (op)    switch (op)
1733      {      {
# Line 1824  for (;;) Line 1807  for (;;)
1807      case OP_COMMIT:      case OP_COMMIT:
1808      case OP_CREF:      case OP_CREF:
1809      case OP_DEF:      case OP_DEF:
1810        case OP_DNCREF:
1811        case OP_DNRREF:
1812      case OP_DOLL:      case OP_DOLL:
1813      case OP_DOLLM:      case OP_DOLLM:
1814      case OP_EOD:      case OP_EOD:
1815      case OP_EODN:      case OP_EODN:
1816      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1817      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1818      case OP_PRUNE:      case OP_PRUNE:
1819      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1835  for (;;)
1835      case OP_NOTI:      case OP_NOTI:
1836      branchlength++;      branchlength++;
1837      cc += 2;      cc += 2;
1838  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1839      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1840  #endif  #endif
1841      break;      break;
# Line 1864  for (;;) Line 1847  for (;;)
1847      case OP_EXACTI:      case OP_EXACTI:
1848      case OP_NOTEXACT:      case OP_NOTEXACT:
1849      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1850      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1851      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1852  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1853      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1854  #endif  #endif
1855      break;      break;
1856    
1857      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1858      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1859      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)
1860        cc += 2;        cc += 2;
1861      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1862      break;      break;
# Line 1909  for (;;) Line 1892  for (;;)
1892    
1893      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1894    
1895        case OP_CLASS:
1896        case OP_NCLASS:
1897  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1898      case OP_XCLASS:      case OP_XCLASS:
1899      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];      /* The original code caused an unsigned overflow in 64 bit systems,
1900      /* Fall through */      so now we use a conditional statement. */
1901        if (op == OP_XCLASS)
1902          cc += GET(cc, 1);
1903        else
1904          cc += PRIV(OP_lengths)[OP_CLASS];
1905    #else
1906        cc += PRIV(OP_lengths)[OP_CLASS];
1907  #endif  #endif
1908    
     case OP_CLASS:  
     case OP_NCLASS:  
     cc += PRIV(OP_lengths)[OP_CLASS];  
   
1909      switch (*cc)      switch (*cc)
1910        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1911        case OP_CRSTAR:        case OP_CRSTAR:
1912        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRQUERY:        case OP_CRQUERY:
1916        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1917          case OP_CRPOSSTAR:
1918          case OP_CRPOSPLUS:
1919          case OP_CRPOSQUERY:
1920        return -1;        return -1;
1921    
1922        case OP_CRRANGE:        case OP_CRRANGE:
1923        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1924          case OP_CRPOSRANGE:
1925        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1927        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1928        break;        break;
1929    
# Line 1999  for (;;) Line 1990  for (;;)
1990      case OP_QUERYI:      case OP_QUERYI:
1991      case OP_REF:      case OP_REF:
1992      case OP_REFI:      case OP_REFI:
1993        case OP_DNREF:
1994        case OP_DNREFI:
1995      case OP_SBRA:      case OP_SBRA:
1996      case OP_SBRAPOS:      case OP_SBRAPOS:
1997      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2028  for (;;)
2028    
2029    
2030    
   
2031  /*************************************************  /*************************************************
2032  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2033  *************************************************/  *************************************************/
# Line 2059  PRIV(find_bracket)(const pcre_uchar *cod Line 2051  PRIV(find_bracket)(const pcre_uchar *cod
2051  {  {
2052  for (;;)  for (;;)
2053    {    {
2054    register int c = *code;    register pcre_uchar c = *code;
2055    
2056    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2057    
# Line 2082  for (;;) Line 2074  for (;;)
2074    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2075             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2076      {      {
2077      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2078      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2079      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2080      }      }
# Line 2112  for (;;) Line 2104  for (;;)
2104        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2105        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2106        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2107        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)
2108          code += 2;          code += 2;
2109        break;        break;
2110    
2111        case OP_MARK:        case OP_MARK:
2112        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2113        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2114        case OP_THEN_ARG:        case OP_THEN_ARG:
2115        code += code[1];        code += code[1];
2116        break;        break;
# Line 2197  find_recurse(const pcre_uchar *code, BOO Line 2186  find_recurse(const pcre_uchar *code, BOO
2186  {  {
2187  for (;;)  for (;;)
2188    {    {
2189    register int c = *code;    register pcre_uchar c = *code;
2190    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2191    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2192    
# Line 2232  for (;;) Line 2221  for (;;)
2221        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2222        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2223        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2224        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)
2225          code += 2;          code += 2;
2226        break;        break;
2227    
2228        case OP_MARK:        case OP_MARK:
2229        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2230        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2231        case OP_THEN_ARG:        case OP_THEN_ARG:
2232        code += code[1];        code += code[1];
2233        break;        break;
# Line 2343  Arguments: Line 2329  Arguments:
2329    endcode     points to where to stop    endcode     points to where to stop
2330    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2331    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2332      recurses    chain of recurse_check to catch mutual recursion
2333    
2334  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2335  */  */
2336    
2337    typedef struct recurse_check {
2338      struct recurse_check *prev;
2339      const pcre_uchar *group;
2340    } recurse_check;
2341    
2342  static BOOL  static BOOL
2343  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2344    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2345  {  {
2346  register int c;  register pcre_uchar c;
2347    recurse_check this_recurse;
2348    
2349  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2350       code < endcode;       code < endcode;
2351       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 2373  for (code = first_significant_code(code
2373    
2374    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2375      {      {
2376      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2377        const pcre_uchar *endgroup = scode;
2378      BOOL empty_branch;      BOOL empty_branch;
2379    
2380      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2381        when called to scan a completed pattern by setting cd->start_workspace to
2382        NULL. */
2383    
2384        if (cd->start_workspace != NULL)
2385          {
2386          const pcre_uchar *tcode;
2387          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2388            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2389          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2390          }
2391    
2392      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      /* If the reference is to a completed group, we need to detect whether this
2393        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;      is a recursive call, as otherwise there will be an infinite loop. If it is
2394        a recursion, just skip over it. Simple recursions are easily detected. For
2395        mutual recursions we keep a chain on the stack. */
2396    
2397        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2398        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2399        else
2400          {
2401          recurse_check *r = recurses;
2402          for (r = recurses; r != NULL; r = r->prev)
2403            if (r->group == scode) break;
2404          if (r != NULL) continue;   /* Mutual recursion */
2405          }
2406    
2407      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2408        stack chain to detect mutual recursions. */
2409    
2410      empty_branch = FALSE;      empty_branch = FALSE;
2411      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2412      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
2413    
2414      do      do
2415        {        {
2416        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2417          {          {
2418          empty_branch = TRUE;          empty_branch = TRUE;
2419          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2469  for (code = first_significant_code(code
2469        empty_branch = FALSE;        empty_branch = FALSE;
2470        do        do
2471          {          {
2472          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2473            empty_branch = TRUE;            empty_branch = TRUE;
2474          code += GET(code, 1);          code += GET(code, 1);
2475          }          }
# Line 2495  for (code = first_significant_code(code Line 2511  for (code = first_significant_code(code
2511        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2512        case OP_CRQUERY:        case OP_CRQUERY:
2513        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2514          case OP_CRPOSSTAR:
2515          case OP_CRPOSQUERY:
2516        break;        break;
2517    
2518        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2519        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2520        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2521          case OP_CRPOSPLUS:
2522        return FALSE;        return FALSE;
2523    
2524        case OP_CRRANGE:        case OP_CRRANGE:
2525        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2526          case OP_CRPOSRANGE:
2527        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2528        break;        break;
2529        }        }
# Line 2511  for (code = first_significant_code(code Line 2531  for (code = first_significant_code(code
2531    
2532      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2533    
2534        case OP_ANY:
2535        case OP_ALLANY:
2536        case OP_ANYBYTE:
2537    
2538      case OP_PROP:      case OP_PROP:
2539      case OP_NOTPROP:      case OP_NOTPROP:
2540        case OP_ANYNL:
2541    
2542        case OP_NOT_HSPACE:
2543        case OP_HSPACE:
2544        case OP_NOT_VSPACE:
2545        case OP_VSPACE:
2546      case OP_EXTUNI:      case OP_EXTUNI:
2547    
2548      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2549      case OP_DIGIT:      case OP_DIGIT:
2550      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2551      case OP_WHITESPACE:      case OP_WHITESPACE:
2552      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2553      case OP_WORDCHAR:      case OP_WORDCHAR:
2554      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2555      case OP_CHAR:      case OP_CHAR:
2556      case OP_CHARI:      case OP_CHARI:
2557      case OP_NOT:      case OP_NOT:
2558      case OP_NOTI:      case OP_NOTI:
2559    
2560      case OP_PLUS:      case OP_PLUS:
2561        case OP_PLUSI:
2562      case OP_MINPLUS:      case OP_MINPLUS:
2563      case OP_POSPLUS:      case OP_MINPLUSI:
2564      case OP_EXACT:  
2565      case OP_NOTPLUS:      case OP_NOTPLUS:
2566        case OP_NOTPLUSI:
2567      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2568        case OP_NOTMINPLUSI:
2569    
2570        case OP_POSPLUS:
2571        case OP_POSPLUSI:
2572      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2573        case OP_NOTPOSPLUSI:
2574    
2575        case OP_EXACT:
2576        case OP_EXACTI:
2577      case OP_NOTEXACT:      case OP_NOTEXACT:
2578        case OP_NOTEXACTI:
2579    
2580      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2581      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2582      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2583      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2584    
2585      return FALSE;      return FALSE;
2586    
2587      /* 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 2601  for (code = first_significant_code(code
2601      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2602      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2603      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2604      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)
2605        code += 2;        code += 2;
2606      break;      break;
2607    
# Line 2572  for (code = first_significant_code(code Line 2615  for (code = first_significant_code(code
2615      return TRUE;      return TRUE;
2616    
2617      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2618      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2619        followed by a multibyte character. */
2620    
2621  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2622      case OP_STAR:      case OP_STAR:
2623      case OP_STARI:      case OP_STARI:
2624        case OP_NOTSTAR:
2625        case OP_NOTSTARI:
2626    
2627      case OP_MINSTAR:      case OP_MINSTAR:
2628      case OP_MINSTARI:      case OP_MINSTARI:
2629        case OP_NOTMINSTAR:
2630        case OP_NOTMINSTARI:
2631    
2632      case OP_POSSTAR:      case OP_POSSTAR:
2633      case OP_POSSTARI:      case OP_POSSTARI:
2634        case OP_NOTPOSSTAR:
2635        case OP_NOTPOSSTARI:
2636    
2637      case OP_QUERY:      case OP_QUERY:
2638      case OP_QUERYI:      case OP_QUERYI:
2639        case OP_NOTQUERY:
2640        case OP_NOTQUERYI:
2641    
2642      case OP_MINQUERY:      case OP_MINQUERY:
2643      case OP_MINQUERYI:      case OP_MINQUERYI:
2644        case OP_NOTMINQUERY:
2645        case OP_NOTMINQUERYI:
2646    
2647      case OP_POSQUERY:      case OP_POSQUERY:
2648      case OP_POSQUERYI:      case OP_POSQUERYI:
2649        case OP_NOTPOSQUERY:
2650        case OP_NOTPOSQUERYI:
2651    
2652      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2653      break;      break;
2654    
2655      case OP_UPTO:      case OP_UPTO:
2656      case OP_UPTOI:      case OP_UPTOI:
2657        case OP_NOTUPTO:
2658        case OP_NOTUPTOI:
2659    
2660      case OP_MINUPTO:      case OP_MINUPTO:
2661      case OP_MINUPTOI:      case OP_MINUPTOI:
2662        case OP_NOTMINUPTO:
2663        case OP_NOTMINUPTOI:
2664    
2665      case OP_POSUPTO:      case OP_POSUPTO:
2666      case OP_POSUPTOI:      case OP_POSUPTOI:
2667        case OP_NOTPOSUPTO:
2668        case OP_NOTPOSUPTOI:
2669    
2670      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]);
2671      break;      break;
2672  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2677  for (code = first_significant_code(code
2677      case OP_MARK:      case OP_MARK:
2678      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2679      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2680      case OP_THEN_ARG:      case OP_THEN_ARG:
2681      code += code[1];      code += code[1];
2682      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2720  could_be_empty(const pcre_uchar *code, c
2720  {  {
2721  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2722    {    {
2723    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2724      return FALSE;      return FALSE;
2725    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2726    }    }
# Line 2662  return TRUE; Line 2730  return TRUE;
2730    
2731    
2732  /*************************************************  /*************************************************
2733  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2734  *************************************************/  *************************************************/
2735    
2736  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2737  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.  
2738    
2739  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2740  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2741  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:].  
2742    
2743  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2744  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2745  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2746  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2747  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2748  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2749  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2750  I think.         (c >= OP_STARI)?      OP_STARI :
2751                                 OP_STAR;
2752    }
2753    
 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.  
2754    
2755  In Perl, unescaped square brackets may also appear as part of class names. For  
2756  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  #ifdef SUPPORT_UCP
2757  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  /*************************************************
2758  seem right at all. PCRE does not allow closing square brackets in POSIX class  *        Check a character and a property        *
2759  names.  *************************************************/
2760    
2761    /* This function is called by check_auto_possessive() when a property item
2762    is adjacent to a fixed character.
2763    
2764  Arguments:  Arguments:
2765    ptr      pointer to the initial [    c            the character
2766    endptr   where to return the end pointer    ptype        the property type
2767      pdata        the data for the type
2768      negated      TRUE if it's a negated property (\P or \p{^)
2769    
2770  Returns:   TRUE or FALSE  Returns:       TRUE if auto-possessifying is OK
2771  */  */
2772    
2773  static BOOL  static BOOL
2774  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2775      BOOL negated)
2776  {  {
2777  int terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2778  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2779  for (++ptr; *ptr != 0; ptr++)  
2780    switch(ptype)
2781    {    {
2782    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    case PT_LAMP:
2783      ptr++;    return (prop->chartype == ucp_Lu ||
2784    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;            prop->chartype == ucp_Ll ||
2785    else            prop->chartype == ucp_Lt) == negated;
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
 }  
2786    
2787      case PT_GC:
2788      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2789    
2790      case PT_PC:
2791      return (pdata == prop->chartype) == negated;
2792    
2793      case PT_SC:
2794      return (pdata == prop->script) == negated;
2795    
2796  /*************************************************    /* These are specials */
 *          Check POSIX class name                *  
 *************************************************/  
2797    
2798  /* This function is called to check the name given in a POSIX-style class entry    case PT_ALNUM:
2799  such as [:alnum:].    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2800              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2801    
2802  Arguments:    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2803    ptr        points to the first letter    means that Perl space and POSIX space are now identical. PCRE was changed
2804    len        the length of the name    at release 8.34. */
2805    
2806  Returns:     a value representing the name, or -1 if unknown    case PT_SPACE:    /* Perl space */
2807  */    case PT_PXSPACE:  /* POSIX space */
2808      switch(c)
2809        {
2810        HSPACE_CASES:
2811        VSPACE_CASES:
2812        return negated;
2813    
2814  static int      default:
2815  check_posix_name(const pcre_uchar *ptr, int len)      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2816  {      }
2817  const char *pn = posix_names;    break;  /* Control never reaches here */
2818  register int yield = 0;  
2819  while (posix_name_lengths[yield] != 0)    case PT_WORD:
2820    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2821    if (len == posix_name_lengths[yield] &&            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2822      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;            c == CHAR_UNDERSCORE) == negated;
2823    pn += posix_name_lengths[yield] + 1;  
2824    yield++;    case PT_CLIST:
2825      p = PRIV(ucd_caseless_sets) + prop->caseset;
2826      for (;;)
2827        {
2828        if (c < *p) return !negated;
2829        if (c == *p++) return negated;
2830        }
2831      break;  /* Control never reaches here */
2832    }    }
2833  return -1;  
2834    return FALSE;
2835  }  }
2836    #endif  /* SUPPORT_UCP */
2837    
2838    
2839    
2840  /*************************************************  /*************************************************
2841  *    Adjust OP_RECURSE items in repeated group   *  *        Fill the character property list        *
2842  *************************************************/  *************************************************/
2843    
2844  /* OP_RECURSE items contain an offset from the start of the regex to the group  /* Checks whether the code points to an opcode that can take part in auto-
2845  that is referenced. This means that groups can be replicated for fixed  possessification, and if so, fills a list with its properties.
 repetition simply by copying (because the recursion is allowed to refer to  
 earlier groups that are outside the current group). However, when a group is  
 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is  
 inserted before it, after it has been compiled. This means that any OP_RECURSE  
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
   
 This function has been extended with the possibility of forward references for  
 recursions and subroutine calls. It must also check the list of such references  
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2846    
2847  Arguments:  Arguments:
2848    group      points to the start of the group    code        points to start of expression
2849    adjust     the amount by which the group is to be moved    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2850    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode    fcc         points to case-flipping table
2851    cd         contains pointers to tables etc.    list        points to output list
2852    save_hwm   the hwm forward reference pointer at the start of the group                list[0] will be filled with the opcode
2853                  list[1] will be non-zero if this opcode
2854                    can match an empty character string
2855                  list[2..7] depends on the opcode
2856    
2857  Returns:     nothing  Returns:      points to the start of the next opcode if *code is accepted
2858                  NULL if *code is not accepted
2859  */  */
2860    
2861  static void  static const pcre_uchar *
2862  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2863    pcre_uchar *save_hwm)    const pcre_uint8 *fcc, pcre_uint32 *list)
2864  {  {
2865  pcre_uchar *ptr = group;  pcre_uchar c = *code;
2866    pcre_uchar base;
2867    const pcre_uchar *end;
2868    pcre_uint32 chr;
2869    
2870  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  #ifdef SUPPORT_UCP
2871    pcre_uint32 *clist_dest;
2872    const pcre_uint32 *clist_src;
2873    #else
2874    utf = utf;  /* Suppress "unused parameter" compiler warning */
2875    #endif
2876    
2877    list[0] = c;
2878    list[1] = FALSE;
2879    code++;
2880    
2881    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2882    {    {
2883    int offset;    base = get_repeat_base(c);
2884    pcre_uchar *hc;    c -= (base - OP_STAR);
2885    
2886    /* See if this recursion is on the forward reference list. If so, adjust the    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2887    reference. */      code += IMM2_SIZE;
2888    
2889    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2890    
2891      switch(base)
2892      {      {
2893      offset = GET(hc, 0);      case OP_STAR:
2894      if (cd->start_code + offset == ptr + 1)      list[0] = OP_CHAR;
2895        {      break;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
     }  
2896    
2897    /* Otherwise, adjust the recursion offset if it's after the start of this      case OP_STARI:
2898    group. */      list[0] = OP_CHARI;
2899        break;
2900    
2901    if (hc >= cd->hwm)      case OP_NOTSTAR:
2902      {      list[0] = OP_NOT;
2903      offset = GET(ptr, 1);      break;
     if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);  
     }  
2904    
2905    ptr += 1 + LINK_SIZE;      case OP_NOTSTARI:
2906        list[0] = OP_NOTI;
2907        break;
2908    
2909        case OP_TYPESTAR:
2910        list[0] = *code;
2911        code++;
2912        break;
2913        }
2914      c = list[0];
2915    }    }
 }  
2916    
2917    switch(c)
2918      {
2919      case OP_NOT_DIGIT:
2920      case OP_DIGIT:
2921      case OP_NOT_WHITESPACE:
2922      case OP_WHITESPACE:
2923      case OP_NOT_WORDCHAR:
2924      case OP_WORDCHAR:
2925      case OP_ANY:
2926      case OP_ALLANY:
2927      case OP_ANYNL:
2928      case OP_NOT_HSPACE:
2929      case OP_HSPACE:
2930      case OP_NOT_VSPACE:
2931      case OP_VSPACE:
2932      case OP_EXTUNI:
2933      case OP_EODN:
2934      case OP_EOD:
2935      case OP_DOLL:
2936      case OP_DOLLM:
2937      return code;
2938    
2939      case OP_CHAR:
2940      case OP_NOT:
2941      GETCHARINCTEST(chr, code);
2942      list[2] = chr;
2943      list[3] = NOTACHAR;
2944      return code;
2945    
2946  /*************************************************    case OP_CHARI:
2947  *        Insert an automatic callout point       *    case OP_NOTI:
2948  *************************************************/    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2949      GETCHARINCTEST(chr, code);
2950      list[2] = chr;
2951    
2952  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  #ifdef SUPPORT_UCP
2953  callout points before each pattern item.    if (chr < 128 || (chr < 256 && !utf))
2954        list[3] = fcc[chr];
2955      else
2956        list[3] = UCD_OTHERCASE(chr);
2957    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2958      list[3] = (chr < 256) ? fcc[chr] : chr;
2959    #else
2960      list[3] = fcc[chr];
2961    #endif
2962    
2963  Arguments:    /* The othercase might be the same value. */
   code           current code pointer  
   ptr            current pattern pointer  
   cd             pointers to tables etc  
2964    
2965  Returns:         new code pointer    if (chr == list[3])
2966  */      list[3] = NOTACHAR;
2967      else
2968        list[4] = NOTACHAR;
2969      return code;
2970    
2971  static pcre_uchar *  #ifdef SUPPORT_UCP
2972  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)    case OP_PROP:
2973  {    case OP_NOTPROP:
2974  *code++ = OP_CALLOUT;    if (code[0] != PT_CLIST)
2975  *code++ = 255;      {
2976  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */      list[2] = code[0];
2977  PUT(code, LINK_SIZE, 0);                       /* Default length */      list[3] = code[1];
2978  return code + 2 * LINK_SIZE;      return code + 2;
2979  }      }
2980    
2981      /* Convert only if we have enough space. */
2982    
2983      clist_src = PRIV(ucd_caseless_sets) + code[1];
2984      clist_dest = list + 2;
2985      code += 2;
2986    
2987  /*************************************************    do {
2988  *         Complete a callout item                *       if (clist_dest >= list + 8)
2989  *************************************************/         {
2990           /* Early return if there is not enough space. This should never
2991           happen, since all clists are shorter than 5 character now. */
2992           list[2] = code[0];
2993           list[3] = code[1];
2994           return code;
2995           }
2996         *clist_dest++ = *clist_src;
2997         }
2998      while(*clist_src++ != NOTACHAR);
2999    
3000  /* A callout item contains the length of the next item in the pattern, which    /* All characters are stored. The terminating NOTACHAR
3001  we can't fill in till after we have reached the relevant point. This is used    is copied form the clist itself. */
 for both automatic and manual callouts.  
3002    
3003  Arguments:    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3004    previous_callout   points to previous callout item    return code;
3005    ptr                current pattern pointer  #endif
   cd                 pointers to tables etc  
3006    
3007  Returns:             nothing    case OP_NCLASS:
3008  */    case OP_CLASS:
3009    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3010      case OP_XCLASS:
3011      if (c == OP_XCLASS)
3012        end = code + GET(code, 0) - 1;
3013      else
3014    #endif
3015        end = code + 32 / sizeof(pcre_uchar);
3016    
3017  static void    switch(*end)
3018  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      {
3019  {      case OP_CRSTAR:
3020  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));      case OP_CRMINSTAR:
3021  PUT(previous_callout, 2 + LINK_SIZE, length);      case OP_CRQUERY:
3022        case OP_CRMINQUERY:
3023        case OP_CRPOSSTAR:
3024        case OP_CRPOSQUERY:
3025        list[1] = TRUE;
3026        end++;
3027        break;
3028    
3029        case OP_CRPLUS:
3030        case OP_CRMINPLUS:
3031        case OP_CRPOSPLUS:
3032        end++;
3033        break;
3034    
3035        case OP_CRRANGE:
3036        case OP_CRMINRANGE:
3037        case OP_CRPOSRANGE:
3038        list[1] = (GET2(end, 1) == 0);
3039        end += 1 + 2 * IMM2_SIZE;
3040        break;
3041        }
3042      list[2] = (pcre_uint32)(end - code);
3043      return end;
3044      }
3045    return NULL;    /* Opcode not accepted */
3046  }  }
3047    
3048    
3049    
 #ifdef SUPPORT_UCP  
3050  /*************************************************  /*************************************************
3051  *           Get othercase range                  *  *    Scan further character sets for match       *
3052  *************************************************/  *************************************************/
3053    
3054  /* This function is passed the start and end of a class range, in UTF-8 mode  /* Checks whether the base and the current opcode have a common character, in
3055  with UCP support. It searches up the characters, looking for ranges of  which case the base cannot be possessified.
 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.  
3056    
3057  Arguments:  Arguments:
3058    cptr        points to starting character value; updated    code        points to the byte code
3059    d           end value    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3060    ocptr       where to put start of othercase range    cd          static compile data
3061    odptr       where to put end of othercase range    base_list   the data list of the base opcode
3062    
3063  Yield:        -1 when no more  Returns:      TRUE if the auto-possessification is possible
                0 when a range is returned  
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
3064  */  */
3065    
3066  static int  static BOOL
3067  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3068    pcre_uint32 *odptr)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3069  {  {
3070  pcre_uint32 c, othercase, next;  pcre_uchar c;
3071  int co;  pcre_uint32 list[8];
3072    const pcre_uint32 *chr_ptr;
3073    const pcre_uint32 *ochr_ptr;
3074    const pcre_uint32 *list_ptr;
3075    const pcre_uchar *next_code;
3076    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3077    const pcre_uchar *xclass_flags;
3078    #endif
3079    const pcre_uint8 *class_bitset;
3080    const pcre_uint8 *set1, *set2, *set_end;
3081    pcre_uint32 chr;
3082    BOOL accepted, invert_bits;
3083    BOOL entered_a_group = FALSE;
3084    
3085    /* Note: the base_list[1] contains whether the current opcode has greedy
3086    (represented by a non-zero value) quantifier. This is a different from
3087    other character type lists, which stores here that the character iterator
3088    matches to an empty string (also represented by a non-zero value). */
3089    
3090  /* Find the first character that has an other case. If it has multiple other  for(;;)
3091  cases, return its case offset value. */    {
3092      /* All operations move the code pointer forward.
3093      Therefore infinite recursions are not possible. */
3094    
3095  for (c = *cptr; c <= d; c++)    c = *code;
   {  
   if ((co = UCD_CASESET(c)) != 0)  
     {  
     *ocptr = c++;   /* Character that has the set */  
     *cptr = c;      /* Rest of input range */  
     return co;  
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
3096    
3097  if (c > d) return -1;  /* Reached end of range */    /* Skip over callouts */
3098    
3099  *ocptr = othercase;    if (c == OP_CALLOUT)
3100  next = othercase + 1;      {
3101        code += PRIV(OP_lengths)[c];
3102        continue;
3103        }
3104    
3105  for (++c; c <= d; c++)    if (c == OP_ALT)
3106    {      {
3107    if (UCD_OTHERCASE(c) != next) break;      do code += GET(code, 1); while (*code == OP_ALT);
3108    next++;      c = *code;
3109    }      }
3110    
3111  *odptr = next - 1;     /* End of othercase range */    switch(c)
3112  *cptr = c;             /* Rest of input range */      {
3113  return 0;      case OP_END:
3114  }      case OP_KETRPOS:
3115        /* TRUE only in greedy case. The non-greedy case could be replaced by
3116        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3117        uses more memory, which we cannot get at this stage.) */
3118    
3119        return base_list[1] != 0;
3120    
3121        case OP_KET:
3122        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3123        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3124        cannot be converted to a possessive form. */
3125    
3126        if (base_list[1] == 0) return FALSE;
3127    
3128        switch(*(code - GET(code, 1)))
3129          {
3130          case OP_ASSERT:
3131          case OP_ASSERT_NOT:
3132          case OP_ASSERTBACK:
3133          case OP_ASSERTBACK_NOT:
3134          case OP_ONCE:
3135          case OP_ONCE_NC:
3136          /* Atomic sub-patterns and assertions can always auto-possessify their
3137          last iterator. However, if the group was entered as a result of checking
3138          a previous iterator, this is not possible. */
3139    
3140  /*************************************************        return !entered_a_group;
3141  *        Check a character and a property        *        }
 *************************************************/  
3142    
3143  /* This function is called by check_auto_possessive() when a property item      code += PRIV(OP_lengths)[c];
3144  is adjacent to a fixed character.      continue;
3145    
3146  Arguments:      case OP_ONCE:
3147    c            the character      case OP_ONCE_NC:
3148    ptype        the property type      case OP_BRA:
3149    pdata        the data for the type      case OP_CBRA:
3150    negated      TRUE if it's a negated property (\P or \p{^)      next_code = code + GET(code, 1);
3151        code += PRIV(OP_lengths)[c];
3152    
3153  Returns:       TRUE if auto-possessifying is OK      while (*next_code == OP_ALT)
3154  */        {
3155          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3156          code = next_code + 1 + LINK_SIZE;
3157          next_code += GET(next_code, 1);
3158          }
3159    
3160  static BOOL      entered_a_group = TRUE;
3161  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)      continue;
 {  
 #ifdef SUPPORT_UCP  
 const pcre_uint32 *p;  
 #endif  
3162    
3163  const ucd_record *prop = GET_UCD(c);      case OP_BRAZERO:
3164        case OP_BRAMINZERO:
3165    
3166  switch(ptype)      next_code = code + 1;
3167    {      if (*next_code != OP_BRA && *next_code != OP_CBRA
3168    case PT_LAMP:          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3169    return (prop->chartype == ucp_Lu ||  
3170            prop->chartype == ucp_Ll ||      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3171            prop->chartype == ucp_Lt) == negated;  
3172        /* The bracket content will be checked by the
3173        OP_BRA/OP_CBRA case above. */
3174        next_code += 1 + LINK_SIZE;
3175        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3176          return FALSE;
3177    
3178    case PT_GC:      code += PRIV(OP_lengths)[c];
3179    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;      continue;
3180    
3181    case PT_PC:      default:
3182    return (pdata == prop->chartype) == negated;      break;
3183        }
3184    
3185    case PT_SC:    /* Check for a supported opcode, and load its properties. */
   return (pdata == prop->script) == negated;  
3186    
3187    /* These are specials */    code = get_chr_property_list(code, utf, cd->fcc, list);
3188      if (code == NULL) return FALSE;    /* Unsupported */
3189    
3190    case PT_ALNUM:    /* If either opcode is a small character list, set pointers for comparing
3191    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    characters from that list with another list, or with a property. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3192    
3193    case PT_SPACE:    /* Perl space */    if (base_list[0] == OP_CHAR)
3194    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      {
3195            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      chr_ptr = base_list + 2;
3196            == negated;      list_ptr = list;
3197        }
3198      else if (list[0] == OP_CHAR)
3199        {
3200        chr_ptr = list + 2;
3201        list_ptr = base_list;
3202        }
3203    
3204    case PT_PXSPACE:  /* POSIX space */    /* Character bitsets can also be compared to certain opcodes. */
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3205    
3206    case PT_WORD:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3207    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  #ifdef COMPILE_PCRE8
3208            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3209            c == CHAR_UNDERSCORE) == negated;        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3210    #endif
3211  #ifdef SUPPORT_UCP        )
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
3212      {      {
3213      if ((unsigned int)c < *p) return !negated;  #ifdef COMPILE_PCRE8
3214      if ((unsigned int)c == *p++) return negated;      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3215      }  #else
3216    break;  /* Control never reaches here */      if (base_list[0] == OP_CLASS)
3217  #endif  #endif
3218    }        {
3219          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3220          list_ptr = list;
3221          }
3222        else
3223          {
3224          set1 = (pcre_uint8 *)(code - list[2]);
3225          list_ptr = base_list;
3226          }
3227    
3228  return FALSE;      invert_bits = FALSE;
3229  }      switch(list_ptr[0])
3230  #endif  /* SUPPORT_UCP */        {
3231          case OP_CLASS:
3232          case OP_NCLASS:
3233          set2 = (pcre_uint8 *)
3234            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3235          break;
3236    
3237    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3238          case OP_XCLASS:
3239          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3240          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3241          if ((*xclass_flags & XCL_MAP) == 0)
3242            {
3243            /* No bits are set for characters < 256. */
3244            if (list[1] == 0) return TRUE;
3245            /* Might be an empty repeat. */
3246            continue;
3247            }
3248          set2 = (pcre_uint8 *)(xclass_flags + 1);
3249          break;
3250    #endif
3251    
3252          case OP_NOT_DIGIT:
3253          invert_bits = TRUE;
3254          /* Fall through */
3255          case OP_DIGIT:
3256          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3257          break;
3258    
3259  /*************************************************        case OP_NOT_WHITESPACE:
3260  *     Check if auto-possessifying is possible    *        invert_bits = TRUE;
3261  *************************************************/        /* Fall through */
3262          case OP_WHITESPACE:
3263          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3264          break;
3265    
3266  /* This function is called for unlimited repeats of certain items, to see        case OP_NOT_WORDCHAR:
3267  whether the next thing could possibly match the repeated item. If not, it makes        invert_bits = TRUE;
3268  sense to automatically possessify the repeated item.        /* Fall through */
3269          case OP_WORDCHAR:
3270          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3271          break;
3272    
3273  Arguments:        default:
3274    previous      pointer to the repeated opcode        return FALSE;
3275    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode        }
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3276    
3277  Returns:        TRUE if possessifying is wanted      /* Because the sets are unaligned, we need
3278  */      to perform byte comparison here. */
3279        set_end = set1 + 32;
3280  static BOOL      if (invert_bits)
 check_auto_possessive(const pcre_uchar *previous, BOOL utf,  
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_uint32 c = NOTACHAR;  
 pcre_uint32 next;  
 int escape;  
 int op_code = *previous++;  
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
3281        {        {
3282        ptr++;        do
       while (*ptr != 0)  
3283          {          {
3284          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3285          ptr++;          }
3286  #ifdef SUPPORT_UTF        while (set1 < set_end);
3287          if (utf) FORWARDCHAR(ptr);        }
3288  #endif      else
3289          {
3290          do
3291            {
3292            if ((*set1++ & *set2++) != 0) return FALSE;
3293          }          }
3294          while (set1 < set_end);
3295        }        }
3296      else break;  
3297        if (list[1] == 0) return TRUE;
3298        /* Might be an empty repeat. */
3299        continue;
3300      }      }
   }  
3301    
3302  /* If the next item is one that we can handle, get its value. A non-negative    /* Some property combinations also acceptable. Unicode property opcodes are
3303  value is a character, a negative value is an escape value. */    processed specially; the rest can be handled with a lookup table. */
3304    
3305  if (*ptr == CHAR_BACKSLASH)    else
3306    {      {
3307    int temperrorcode = 0;      pcre_uint32 leftop, rightop;
   escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3308    
3309  /* Skip whitespace and comments in extended mode */      leftop = base_list[0];
3310        rightop = list[0];
3311    
3312  if ((options & PCRE_EXTENDED) != 0)  #ifdef SUPPORT_UCP
3313    {      accepted = FALSE; /* Always set in non-unicode case. */
3314    for (;;)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
3315        {        {
3316        ptr++;        if (rightop == OP_EOD)
3317        while (*ptr != 0)          accepted = TRUE;
3318          {        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3319          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          {
3320          ptr++;          int n;
3321  #ifdef SUPPORT_UTF          const pcre_uint8 *p;
3322          if (utf) FORWARDCHAR(ptr);          BOOL same = leftop == rightop;
3323  #endif          BOOL lisprop = leftop == OP_PROP;
3324            BOOL risprop = rightop == OP_PROP;
3325            BOOL bothprop = lisprop && risprop;
3326    
3327            /* There's a table that specifies how each combination is to be
3328            processed:
3329              0   Always return FALSE (never auto-possessify)
3330              1   Character groups are distinct (possessify if both are OP_PROP)
3331              2   Check character categories in the same group (general or particular)
3332              3   Return TRUE if the two opcodes are not the same
3333              ... see comments below
3334            */
3335    
3336            n = propposstab[base_list[2]][list[2]];
3337            switch(n)
3338              {
3339              case 0: break;
3340              case 1: accepted = bothprop; break;
3341              case 2: accepted = (base_list[3] == list[3]) != same; break;
3342              case 3: accepted = !same; break;
3343    
3344              case 4:  /* Left general category, right particular category */
3345              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3346              break;
3347    
3348              case 5:  /* Right general category, left particular category */
3349              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3350              break;
3351    
3352              /* This code is logically tricky. Think hard before fiddling with it.
3353              The posspropstab table has four entries per row. Each row relates to
3354              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3355              Only WORD actually needs all four entries, but using repeats for the
3356              others means they can all use the same code below.
3357    
3358              The first two entries in each row are Unicode general categories, and
3359              apply always, because all the characters they include are part of the
3360              PCRE character set. The third and fourth entries are a general and a
3361              particular category, respectively, that include one or more relevant
3362              characters. One or the other is used, depending on whether the check
3363              is for a general or a particular category. However, in both cases the
3364              category contains more characters than the specials that are defined
3365              for the property being tested against. Therefore, it cannot be used
3366              in a NOTPROP case.
3367    
3368              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3369              Underscore is covered by ucp_P or ucp_Po. */
3370    
3371              case 6:  /* Left alphanum vs right general category */
3372              case 7:  /* Left space vs right general category */
3373              case 8:  /* Left word vs right general category */
3374              p = posspropstab[n-6];
3375              accepted = risprop && lisprop ==
3376                (list[3] != p[0] &&
3377                 list[3] != p[1] &&
3378                (list[3] != p[2] || !lisprop));
3379              break;
3380    
3381              case 9:   /* Right alphanum vs left general category */
3382              case 10:  /* Right space vs left general category */
3383              case 11:  /* Right word vs left general category */
3384              p = posspropstab[n-9];
3385              accepted = lisprop && risprop ==
3386                (base_list[3] != p[0] &&
3387                 base_list[3] != p[1] &&
3388                (base_list[3] != p[2] || !risprop));
3389              break;
3390    
3391              case 12:  /* Left alphanum vs right particular category */
3392              case 13:  /* Left space vs right particular category */
3393              case 14:  /* Left word vs right particular category */
3394              p = posspropstab[n-12];
3395              accepted = risprop && lisprop ==
3396                (catposstab[p[0]][list[3]] &&
3397                 catposstab[p[1]][list[3]] &&
3398                (list[3] != p[3] || !lisprop));
3399              break;
3400    
3401              case 15:  /* Right alphanum vs left particular category */
3402              case 16:  /* Right space vs left particular category */
3403              case 17:  /* Right word vs left particular category */
3404              p = posspropstab[n-15];
3405              accepted = lisprop && risprop ==
3406                (catposstab[p[0]][base_list[3]] &&
3407                 catposstab[p[1]][base_list[3]] &&
3408                (base_list[3] != p[3] || !risprop));
3409              break;
3410              }
3411          }          }
3412        }        }
     else break;  
     }  
   }  
3413    
3414  /* If the next thing is itself optional, we have to give up. */      else
3415    #endif  /* SUPPORT_UCP */
3416    
3417  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3418    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3419      return FALSE;             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3420    
3421  /* If the previous item is a character, get its value. */      if (!accepted) return FALSE;
3422    
3423  if (op_code == OP_CHAR || op_code == OP_CHARI ||      if (list[1] == 0) return TRUE;
3424      op_code == OP_NOT || op_code == OP_NOTI)      /* Might be an empty repeat. */
3425    //if (escape == 0) switch(op_code)      continue;
3426    {      }
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3427    
3428  /* Now compare the next item with the previous opcode. First, handle cases when    /* Control reaches here only if one of the items is a small character list.
3429  the next item is a character. */    All characters are checked against the other side. */
3430    
3431  if (escape == 0)    do
   {  
   /* For a caseless UTF match, the next character may have more than one other  
   case, which maps to the special PT_CLIST property. Check this first. */  
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3432      {      {
3433      int ocs = UCD_CASESET(next);      chr = *chr_ptr;
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
3434    
3435    switch(op_code)      switch(list_ptr[0])
3436      {        {
3437      case OP_CHAR:        case OP_CHAR:
3438      return c != next;        ochr_ptr = list_ptr + 2;
3439          do
3440            {
3441            if (chr == *ochr_ptr) return FALSE;
3442            ochr_ptr++;
3443            }
3444          while(*ochr_ptr != NOTACHAR);
3445          break;
3446    
3447      /* For CHARI (caseless character) we must check the other case. If we have        case OP_NOT:
3448      Unicode property support, we can use it to test the other case of        ochr_ptr = list_ptr + 2;
3449      high-valued characters. We know that next can have only one other case,        do
3450      because multi-other-case characters are dealt with above. */          {
3451            if (chr == *ochr_ptr)
3452              break;
3453            ochr_ptr++;
3454            }
3455          while(*ochr_ptr != NOTACHAR);
3456          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3457          break;
3458    
3459      case OP_CHARI:        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3460      if (c == next) return FALSE;        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
 #ifdef SUPPORT_UTF  
     if (utf)  
       {  
       pcre_uint32 othercase;  
       if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c != othercase;  
       }  
     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)  
       {  
       pcre_uint32 othercase;  
       if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
       }  
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3461    
3462      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.        case OP_DIGIT:
3463      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3464          break;
3465    
3466      case OP_DIGIT:        case OP_NOT_DIGIT:
3467      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3468          break;
3469    
3470      case OP_NOT_DIGIT:        case OP_WHITESPACE:
3471      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3472          break;
3473    
3474      case OP_WHITESPACE:        case OP_NOT_WHITESPACE:
3475      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3476          break;
3477    
3478      case OP_NOT_WHITESPACE:        case OP_WORDCHAR:
3479      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3480          break;
3481    
3482      case OP_WORDCHAR:        case OP_NOT_WORDCHAR:
3483      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3484          break;
3485    
3486      case OP_NOT_WORDCHAR:        case OP_HSPACE:
3487      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;        switch(chr)
3488            {
3489            HSPACE_CASES: return FALSE;
3490            default: break;
3491            }
3492          break;
3493    
3494      case OP_HSPACE:        case OP_NOT_HSPACE:
3495      case OP_NOT_HSPACE:        switch(chr)
3496      switch(next)          {
3497        {          HSPACE_CASES: break;
3498        HSPACE_CASES:          default: return FALSE;
3499        return op_code == OP_NOT_HSPACE;          }
3500          break;
3501    
3502        default:        case OP_ANYNL:
3503        return op_code != OP_NOT_HSPACE;        case OP_VSPACE:
3504        }        switch(chr)
3505            {
3506            VSPACE_CASES: return FALSE;
3507            default: break;
3508            }
3509          break;
3510    
3511      case OP_ANYNL:        case OP_NOT_VSPACE:
3512      case OP_VSPACE:        switch(chr)
3513      case OP_NOT_VSPACE:          {
3514      switch(next)          VSPACE_CASES: break;
3515        {          default: return FALSE;
3516        VSPACE_CASES:          }
3517        return op_code == OP_NOT_VSPACE;        break;
3518    
3519        default:        case OP_DOLL:
3520        return op_code != OP_NOT_VSPACE;        case OP_EODN:
3521        }        switch (chr)
3522            {
3523            case CHAR_CR:
3524            case CHAR_LF:
3525            case CHAR_VT:
3526            case CHAR_FF:
3527            case CHAR_NEL:
3528    #ifndef EBCDIC
3529            case 0x2028:
3530            case 0x2029:
3531    #endif  /* Not EBCDIC */
3532            return FALSE;
3533            }
3534          break;
3535    
3536          case OP_EOD:    /* Can always possessify before \z */
3537          break;
3538    
3539  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3540      case OP_PROP:        case OP_PROP:
3541      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3542          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3543                list_ptr[0] == OP_NOTPROP))
3544            return FALSE;
3545          break;
3546    #endif
3547    
3548      case OP_NOTPROP:        case OP_NCLASS:
3549      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3550          /* Fall through */
3551    
3552          case OP_CLASS:
3553          if (chr > 255) break;
3554          class_bitset = (pcre_uint8 *)
3555            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3556          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3557          break;
3558    
3559    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3560          case OP_XCLASS:
3561          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3562              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3563          break;
3564  #endif  #endif
3565    
3566      default:        default:
3567      return FALSE;        return FALSE;
3568          }
3569    
3570        chr_ptr++;
3571      }      }
3572      while(*chr_ptr != NOTACHAR);
3573    
3574      /* At least one character must be matched from this opcode. */
3575    
3576      if (list[1] == 0) return TRUE;
3577    }    }
3578    
3579  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3580  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  but some compilers complain about an unreachable statement. */
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
3581    
3582  switch(op_code)  }
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3583    
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
3584    
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3585    
3586      case ESC_S:  /*************************************************
3587      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  *    Scan compiled regex for auto-possession     *
3588    *************************************************/
3589    
3590      case ESC_w:  /* Replaces single character iterations with their possessive alternatives
3591      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  if appropriate. This function modifies the compiled opcode!
3592    
3593      case ESC_W:  Arguments:
3594      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;    code        points to start of the byte code
3595      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3596      cd          static compile data
3597    
3598      case ESC_h:  Returns:      nothing
3599      case ESC_H:  */
     switch(c)  
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
3600    
3601      case ESC_v:  static void
3602      case ESC_V:  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3603      switch(c)  {
3604        {  register pcre_uchar c;
3605        VSPACE_CASES:  const pcre_uchar *end;
3606        return escape != ESC_v;  pcre_uchar *repeat_opcode;
3607    pcre_uint32 list[8];
3608    
3609        default:  for (;;)
3610        return escape == ESC_v;    {
3611        }    c = *code;
3612    
3613      /* When PCRE_UCP is set, these values get generated for \d etc. Find    /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3614      their substitutions and process them. The result will always be either    it may compile without complaining, but may get into a loop here if the code
3615      ESC_p or ESC_P. Then fall through to process those values. */    pointer points to a bad value. This is, of course a documentated possibility,
3616      when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3617      just give up on this optimization. */
3618    
3619  #ifdef SUPPORT_UCP    if (c >= OP_TABLE_LENGTH) return;
3620      case ESC_du:  
3621      case ESC_DU:    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3622      case ESC_wu:      {
3623      case ESC_WU:      c -= get_repeat_base(c) - OP_STAR;
3624      case ESC_su:      end = (c <= OP_MINUPTO) ?
3625      case ESC_SU:        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3626        {      list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
       int temperrorcode = 0;  
       ptr = substitutes[escape - ESC_DU];  
       escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
3627    
3628      case ESC_p:      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
     case ESC_P:  
3629        {        {
3630        int ptype, pdata, errorcodeptr;        switch(c)
3631        BOOL negated;          {
3632            case OP_STAR:
3633            *code += OP_POSSTAR - OP_STAR;
3634            break;
3635    
3636            case OP_MINSTAR:
3637            *code += OP_POSSTAR - OP_MINSTAR;
3638            break;
3639    
3640            case OP_PLUS:
3641            *code += OP_POSPLUS - OP_PLUS;
3642            break;
3643    
3644        ptr--;      /* Make ptr point at the p or P */          case OP_MINPLUS:
3645        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);          *code += OP_POSPLUS - OP_MINPLUS;
3646        if (ptype < 0) return FALSE;          break;
       ptr++;      /* Point past the final curly ket */  
3647    
3648        /* If the property item is optional, we have to give up. (When generated          case OP_QUERY:
3649        from \d etc by PCRE_UCP, this test will have been applied much earlier,          *code += OP_POSQUERY - OP_QUERY;
3650        to the original \d etc. At this point, ptr will point to a zero byte. */          break;
3651    
3652        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||          case OP_MINQUERY:
3653          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)          *code += OP_POSQUERY - OP_MINQUERY;
3654            return FALSE;          break;
3655    
3656        /* Do the property check. */          case OP_UPTO:
3657            *code += OP_POSUPTO - OP_UPTO;
3658            break;
3659    
3660        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);          case OP_MINUPTO:
3661            *code += OP_POSUPTO - OP_MINUPTO;
3662            break;
3663            }
3664        }        }
3665        c = *code;
3666        }
3667      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3668        {
3669    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3670        if (c == OP_XCLASS)
3671          repeat_opcode = code + GET(code, 1);
3672        else
3673  #endif  #endif
3674          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3675    
3676      default:      c = *repeat_opcode;
3677      return FALSE;      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3678          {
3679          /* end must not be NULL. */
3680          end = get_chr_property_list(code, utf, cd->fcc, list);
3681    
3682          list[1] = (c & 1) == 0;
3683    
3684          if (compare_opcodes(end, utf, cd, list, end))
3685            {
3686            switch (c)
3687              {
3688              case OP_CRSTAR:
3689              case OP_CRMINSTAR:
3690              *repeat_opcode = OP_CRPOSSTAR;
3691              break;
3692    
3693              case OP_CRPLUS:
3694              case OP_CRMINPLUS:
3695              *repeat_opcode = OP_CRPOSPLUS;
3696              break;
3697    
3698              case OP_CRQUERY:
3699              case OP_CRMINQUERY:
3700              *repeat_opcode = OP_CRPOSQUERY;
3701              break;
3702    
3703              case OP_CRRANGE:
3704              case OP_CRMINRANGE:
3705              *repeat_opcode = OP_CRPOSRANGE;
3706              break;
3707              }
3708            }
3709          }
3710        c = *code;
3711      }      }
3712    
3713    /* In principle, support for Unicode properties should be integrated here as    switch(c)
3714    well. It means re-organizing the above code so as to get hold of the property      {
3715    values before switching on the op-code. However, I wonder how many patterns      case OP_END:
3716    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,      return;
   these op-codes are never generated.) */  
3717    
3718    case OP_DIGIT:      case OP_TYPESTAR:
3719    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||      case OP_TYPEMINSTAR:
3720           escape == ESC_h || escape == ESC_v || escape == ESC_R;      case OP_TYPEPLUS:
3721        case OP_TYPEMINPLUS:
3722        case OP_TYPEQUERY:
3723        case OP_TYPEMINQUERY:
3724        case OP_TYPEPOSSTAR:
3725        case OP_TYPEPOSPLUS:
3726        case OP_TYPEPOSQUERY:
3727        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3728        break;
3729    
3730    case OP_NOT_DIGIT:      case OP_TYPEUPTO:
3731    return escape == ESC_d;      case OP_TYPEMINUPTO:
3732        case OP_TYPEEXACT:
3733        case OP_TYPEPOSUPTO:
3734        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3735          code += 2;
3736        break;
3737    
3738    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3739        case OP_XCLASS:
3740        code += GET(code, 1);
3741        break;
3742    #endif
3743    
3744        case OP_MARK:
3745        case OP_PRUNE_ARG:
3746        case OP_SKIP_ARG:
3747        case OP_THEN_ARG:
3748        code += code[1];
3749        break;
3750        }
3751    
3752      /* Add in the fixed length from the table */
3753    
3754      code += PRIV(OP_lengths)[c];
3755    
3756      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3757      a multi-byte character. The length in the table is a minimum, so we have to
3758      arrange to skip the extra bytes. */
3759    
3760    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3761      if (utf) switch(c)
3762        {
3763        case OP_CHAR:
3764        case OP_CHARI:
3765        case OP_NOT:
3766        case OP_NOTI:
3767        case OP_STAR:
3768        case OP_MINSTAR:
3769        case OP_PLUS:
3770        case OP_MINPLUS:
3771        case OP_QUERY:
3772        case OP_MINQUERY:
3773        case OP_UPTO:
3774        case OP_MINUPTO:
3775        case OP_EXACT:
3776        case OP_POSSTAR:
3777        case OP_POSPLUS:
3778        case OP_POSQUERY:
3779        case OP_POSUPTO:
3780        case OP_STARI:
3781        case OP_MINSTARI:
3782        case OP_PLUSI:
3783        case OP_MINPLUSI:
3784        case OP_QUERYI:
3785        case OP_MINQUERYI:
3786        case OP_UPTOI:
3787        case OP_MINUPTOI:
3788        case OP_EXACTI:
3789        case OP_POSSTARI:
3790        case OP_POSPLUSI:
3791        case OP_POSQUERYI:
3792        case OP_POSUPTOI:
3793        case OP_NOTSTAR:
3794        case OP_NOTMINSTAR:
3795        case OP_NOTPLUS:
3796        case OP_NOTMINPLUS:
3797        case OP_NOTQUERY:
3798        case OP_NOTMINQUERY:
3799        case OP_NOTUPTO:
3800        case OP_NOTMINUPTO:
3801        case OP_NOTEXACT:
3802        case OP_NOTPOSSTAR:
3803        case OP_NOTPOSPLUS:
3804        case OP_NOTPOSQUERY:
3805        case OP_NOTPOSUPTO:
3806        case OP_NOTSTARI:
3807        case OP_NOTMINSTARI:
3808        case OP_NOTPLUSI:
3809        case OP_NOTMINPLUSI:
3810        case OP_NOTQUERYI:
3811        case OP_NOTMINQUERYI:
3812        case OP_NOTUPTOI:
3813        case OP_NOTMINUPTOI:
3814        case OP_NOTEXACTI:
3815        case OP_NOTPOSSTARI:
3816        case OP_NOTPOSPLUSI:
3817        case OP_NOTPOSQUERYI:
3818        case OP_NOTPOSUPTOI:
3819        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3820        break;
3821        }
3822    #else
3823      (void)(utf);  /* Keep compiler happy by referencing function argument */
3824    #endif
3825      }
3826    }
3827    
3828    
3829    
3830    /*************************************************
3831    *           Check for POSIX class syntax         *
3832    *************************************************/
3833    
3834    /* This function is called when the sequence "[:" or "[." or "[=" is
3835    encountered in a character class. It checks whether this is followed by a
3836    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3837    reach an unescaped ']' without the special preceding character, return FALSE.
3838    
3839    Originally, this function only recognized a sequence of letters between the
3840    terminators, but it seems that Perl recognizes any sequence of characters,
3841    though of course unknown POSIX names are subsequently rejected. Perl gives an
3842    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3843    didn't consider this to be a POSIX class. Likewise for [:1234:].
3844    
3845    The problem in trying to be exactly like Perl is in the handling of escapes. We
3846    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3847    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3848    below handles the special case of \], but does not try to do any other escape
3849    processing. This makes it different from Perl for cases such as [:l\ower:]
3850    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3851    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3852    I think.
3853    
3854    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3855    It seems that the appearance of a nested POSIX class supersedes an apparent
3856    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3857    a digit.
3858    
3859    In Perl, unescaped square brackets may also appear as part of class names. For
3860    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3861    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3862    seem right at all. PCRE does not allow closing square brackets in POSIX class
3863    names.
3864    
3865    Arguments:
3866      ptr      pointer to the initial [
3867      endptr   where to return the end pointer
3868    
3869    Returns:   TRUE or FALSE
3870    */
3871    
3872    static BOOL
3873    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3874    {
3875    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3876    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3877    for (++ptr; *ptr != CHAR_NULL; ptr++)
3878      {
3879      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3880        ptr++;
3881      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3882      else
3883        {
3884        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3885          {
3886          *endptr = ptr;
3887          return TRUE;
3888          }
3889        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3890             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3891              ptr[1] == CHAR_EQUALS_SIGN) &&
3892            check_posix_syntax(ptr, endptr))
3893          return FALSE;
3894        }
3895      }
3896    return FALSE;
3897    }
3898    
3899    
3900    
3901    
3902    /*************************************************
3903    *          Check POSIX class name                *
3904    *************************************************/
3905    
3906    /* This function is called to check the name given in a POSIX-style class entry
3907    such as [:alnum:].
3908    
3909    Arguments:
3910      ptr        points to the first letter
3911      len        the length of the name
3912    
3913    Returns:     a value representing the name, or -1 if unknown
3914    */
3915    
3916    static int
3917    check_posix_name(const pcre_uchar *ptr, int len)
3918    {
3919    const char *pn = posix_names;
3920    register int yield = 0;
3921    while (posix_name_lengths[yield] != 0)
3922      {
3923      if (len == posix_name_lengths[yield] &&
3924        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3925      pn += posix_name_lengths[yield] + 1;
3926      yield++;
3927      }
3928    return -1;
3929    }
3930    
3931    
3932    /*************************************************
3933    *    Adjust OP_RECURSE items in repeated group   *
3934    *************************************************/
3935    
3936    /* OP_RECURSE items contain an offset from the start of the regex to the group
3937    that is referenced. This means that groups can be replicated for fixed
3938    repetition simply by copying (because the recursion is allowed to refer to
3939    earlier groups that are outside the current group). However, when a group is
3940    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3941    inserted before it, after it has been compiled. This means that any OP_RECURSE
3942    items within it that refer to the group itself or any contained groups have to
3943    have their offsets adjusted. That one of the jobs of this function. Before it
3944    is called, the partially compiled regex must be temporarily terminated with
3945    OP_END.
3946    
3947    This function has been extended with the possibility of forward references for
3948    recursions and subroutine calls. It must also check the list of such references
3949    for the group we are dealing with. If it finds that one of the recursions in
3950    the current group is on this list, it adjusts the offset in the list, not the
3951    value in the reference (which is a group number).
3952    
3953    Arguments:
3954      group      points to the start of the group
3955      adjust     the amount by which the group is to be moved
3956      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3957      cd         contains pointers to tables etc.
3958      save_hwm   the hwm forward reference pointer at the start of the group
3959    
3960    Returns:     nothing
3961    */
3962    
3963    static void
3964    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3965      pcre_uchar *save_hwm)
3966    {
3967    pcre_uchar *ptr = group;
3968    
3969    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3970      {
3971      int offset;
3972      pcre_uchar *hc;
3973    
3974      /* See if this recursion is on the forward reference list. If so, adjust the
3975      reference. */
3976    
3977      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3978        {
3979        offset = (int)GET(hc, 0);
3980        if (cd->start_code + offset == ptr + 1)
3981          {
3982          PUT(hc, 0, offset + adjust);
3983          break;
3984          }
3985        }
3986    
3987      /* Otherwise, adjust the recursion offset if it's after the start of this
3988      group. */
3989    
3990      if (hc >= cd->hwm)
3991        {
3992        offset = (int)GET(ptr, 1);
3993        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3994        }
3995    
3996      ptr += 1 + LINK_SIZE;
3997      }
3998    }
3999    
4000    
4001    
4002    /*************************************************
4003    *        Insert an automatic callout point       *
4004    *************************************************/
4005    
4006    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4007    callout points before each pattern item.
4008    
4009    Arguments:
4010      code           current code pointer
4011      ptr            current pattern pointer
4012      cd             pointers to tables etc
4013    
4014    Returns:         new code pointer
4015    */
4016    
4017    static pcre_uchar *
4018    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4019    {
4020    *code++ = OP_CALLOUT;
4021    *code++ = 255;
4022    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4023    PUT(code, LINK_SIZE, 0);                       /* Default length */
4024    return code + 2 * LINK_SIZE;
4025    }
4026    
4027    
4028    
4029    /*************************************************
4030    *         Complete a callout item                *
4031    *************************************************/
4032    
4033    /* A callout item contains the length of the next item in the pattern, which
4034    we can't fill in till after we have reached the relevant point. This is used
4035    for both automatic and manual callouts.
4036    
4037    Arguments:
4038      previous_callout   points to previous callout item
4039      ptr                current pattern pointer
4040      cd                 pointers to tables etc
4041    
4042    Returns:             nothing
4043    */
4044    
4045    static void
4046    complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4047    {
4048    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4049    PUT(previous_callout, 2 + LINK_SIZE, length);
4050    }
4051    
4052    
4053    
4054    #ifdef SUPPORT_UCP
4055    /*************************************************
4056    *           Get othercase range                  *
4057    *************************************************/
4058    
4059    /* This function is passed the start and end of a class range, in UTF-8 mode
4060    with UCP support. It searches up the characters, looking for ranges of
4061    characters in the "other" case. Each call returns the next one, updating the
4062    start address. A character with multiple other cases is returned on its own
4063    with a special return value.
4064    
4065    case OP_WHITESPACE:  Arguments:
4066    return escape == ESC_S || escape == ESC_d || escape == ESC_w;    cptr        points to starting character value; updated
4067      d           end value
4068      ocptr       where to put start of othercase range
4069      odptr       where to put end of othercase range
4070    
4071    case OP_NOT_WHITESPACE:  Yield:        -1 when no more
4072    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;                 0 when a range is returned
4073                  >0 the CASESET offset for char with multiple other cases
4074                    in this case, ocptr contains the original
4075    */
4076    
4077    case OP_HSPACE:  static int
4078    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4079           escape == ESC_w || escape == ESC_v || escape == ESC_R;    pcre_uint32 *odptr)
4080    {
4081    pcre_uint32 c, othercase, next;
4082    unsigned int co;
4083    
4084    case OP_NOT_HSPACE:  /* Find the first character that has an other case. If it has multiple other
4085    return escape == ESC_h;  cases, return its case offset value. */
4086    
4087    /* Can't have \S in here because VT matches \S (Perl anomaly) */  for (c = *cptr; c <= d; c++)
4088    case OP_ANYNL:    {
4089    case OP_VSPACE:    if ((co = UCD_CASESET(c)) != 0)
4090    return escape == ESC_V || escape == ESC_d || escape == ESC_w;      {
4091        *ocptr = c++;   /* Character that has the set */
4092        *cptr = c;      /* Rest of input range */
4093        return (int)co;
4094        }
4095      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4096      }
4097    
4098    case OP_NOT_VSPACE:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_v || escape == ESC_R;  
4099    
4100    case OP_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4101    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  range, which is either the end of the input range, or a character that has zero
4102           escape == ESC_v || escape == ESC_R;  or more than one other cases. */
4103    
4104    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4105    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4106    
4107    default:  for (++c; c <= d; c++)
4108    return FALSE;    {
4109      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4110      next++;
4111    }    }
4112    
4113  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4114    *cptr = c;             /* Rest of input range */
4115    return 0;
4116  }  }
4117    #endif  /* SUPPORT_UCP */
4118    
4119    
4120    
# Line 3418  switch(op_code) Line 4123  switch(op_code)
4123  *************************************************/  *************************************************/
4124    
4125  /* 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
4126  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
4127  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
4128  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4129    
4130  Arguments:  Arguments:
4131    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4132    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4133    options       the options word    options       the options word
4134    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4135    start         start of range character    start         start of range character
4136    end           end of range character    end           end of range character
4137    
4138  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4139                  the pointer to extra data is updated                  the pointer to extra data is updated
4140  */  */
# Line 3439  add_to_class(pcre_uint8 *classbits, pcre Line 4144  add_to_class(pcre_uint8 *classbits, pcre
4144    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4145  {  {
4146  pcre_uint32 c;  pcre_uint32 c;
4147    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4148  int n8 = 0;  int n8 = 0;
4149    
4150  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4151  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
4152  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
4153  range. */  range. */
4154    
4155  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4156    {    {
4157  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4158    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4159      {      {
4160      int rc;      int rc;
4161      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4162    
4163      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4164      c = start;      c = start;
4165    
4166      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4167        {        {
4168        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4169    
4170        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4171          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4172    
4173        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4174    
4175        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4176    
4177        /* 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
4178        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
4179        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4180    
4181        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4182        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4183        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 4187  if ((options & PCRE_CASELESS) != 0)
4187  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4188    
4189    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4190    
4191    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4192      {      {
4193      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4194      n8++;      n8++;
4195      }      }
4196    }    }
4197    
4198  /* 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
4199  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
4200  in all cases. */  in all cases. */
# Line 3507  in all cases. */ Line 4213  in all cases. */
4213    
4214  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4215    
4216  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
 data. */  
4217    
4218  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4219    {    {
4220    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4221      {    SETBIT(classbits, c);
4222      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4223    }    }
4224    
4225  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4226    {  if (start <= 0xff) start = 0xff + 1;
4227    
4228    if (end >= start)
4229      {
4230    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4231  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4232    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4233      {      {
4234      if (start < end)      if (start < end)
4235        {        {
4236        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4237        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4238        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4239        }        }
4240      else if (start == end)      else if (start == end)
4241        {        {
4242        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4243        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4244        }        }
4245      }      }
4246    else    else
4247  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4248    
4249    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4250    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4251    
4252  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4253      {}      {}
4254  #else  #else
4255    if (start < end)    if (start < end)
4256      {      {
4257      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4262  else
4262      {      {
4263      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4264      *uchardata++ = start;      *uchardata++ = start;
4265      }      }
4266  #endif  #endif
4267    
4268    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4269    }    }
4270    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4271    
4272  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4273  }  }
4274    
4275    
4276    
4277    
4278  /*************************************************  /*************************************************
4279  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4280  *************************************************/  *************************************************/
4281    
4282  /* 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
4283  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
4284  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
4285  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4289  Arguments:
4289    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4290    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4291    options       the options word    options       the options word
4292    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4293    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4294    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4295                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4296                    already know about                    already know about
4297    
4298  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4299                  the pointer to extra data is updated                  the pointer to extra data is updated
4300  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4308  while (p[0] < NOTACHAR)
4308    {    {
4309    int n = 0;    int n = 0;
4310    if (p[0] != except)    if (p[0] != except)
4311      {      {
4312      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4313      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4314      }      }
4315    p += n + 1;    p += n + 1;
4316    }    }
4317  return n8;  return n8;
4318  }  }
4319    
4320    
4321    
# Line 3624  Arguments: Line 4330  Arguments:
4330    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4331    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4332    options       the options word    options       the options word
4333    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4334    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4335    
4336  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4337                  the pointer to extra data is updated                  the pointer to extra data is updated
4338  */  */
4339    
4340  static int  static int
4341  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4342    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4343  {  {
4344  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4350  while (p[0] < NOTACHAR)
4350    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4351    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4352      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4353    p++;    p++;
4354    }    }
4355  return n8;  return n8;
4356  }  }
4357    
4358    
4359    
# Line 3662  to find out the amount of memory needed, Line 4368  to find out the amount of memory needed,
4368  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4369    
4370  Arguments:  Arguments:
4371    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4372    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4373    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4374    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4375    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4376    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4377    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4378    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4379    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4380    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4381                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4382      lengthptr         NULL during the real compile phase
4383                        points to length accumulator during pre-compile phase
4384    
4385  Returns:         TRUE on success  Returns:            TRUE on success
4386                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4387  */  */
4388    
4389  static BOOL  static BOOL
4390  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4391    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4392    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4393      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4394      branch_chain *bcptr, int cond_depth,
4395    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4396  {  {
4397  int repeat_type, op_type;  int repeat_type, op_type;
4398  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4399  int bravalue = 0;  int bravalue = 0;
4400  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4401  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4402  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4403    pcre_uint32 zeroreqchar, zerofirstchar;
4404    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4405  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4406  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4407  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4429  dynamically as we process the pattern. *
4429  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4430  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4431  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4432    #ifndef COMPILE_PCRE32
4433  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4434    #endif
4435  #else  #else
4436  BOOL utf = FALSE;  BOOL utf = FALSE;
4437  #endif  #endif
4438    
4439  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4440  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,
4441  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
4442  alternative calls for the different cases. */  alternative calls for the different cases. */
4443    
4444  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4466  to take the zero repeat into account. Th
4466  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4467  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4468    
4469  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4470    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4471    
4472  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4473  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3774  for (;; ptr++) Line 4489  for (;; ptr++)
4489    BOOL reset_bracount;    BOOL reset_bracount;
4490    int class_has_8bitchar;    int class_has_8bitchar;
4491    int class_one_char;    int class_one_char;
4492    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4493      BOOL xclass_has_prop;
4494    #endif
4495    int newoptions;    int newoptions;
4496    int recno;    int recno;
4497    int refsign;    int refsign;
4498    int skipbytes;    int skipbytes;
4499    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4500    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4501    int terminator;    int terminator;
4502    int mclength;    unsigned int mclength;
4503    int tempbracount;    unsigned int tempbracount;
4504    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4505    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4506    
4507    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4511  for (;; ptr++)
4511    /* 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
4512    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4513    
4514    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4515      {      {
4516      ptr = nestptr;      ptr = nestptr;
4517      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4586  for (;; ptr++)
4586    
4587    /* 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 */
4588    
4589    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4590      {      {
4591      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4592        {        {
# Line 3891  for (;; ptr++) Line 4609  for (;; ptr++)
4609          }          }
4610        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4611        }        }
4612        /* Control does not reach here. */
4613      }      }
4614    
4615    /* 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
4616    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. */  
4617    
4618    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4619      {      {
4620      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4621        {        {
4622          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4623          if (c != CHAR_NUMBER_SIGN) break;
4624        ptr++;        ptr++;
4625        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4626          {          {
4627          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4628              {                          /* IS_NEWLINE sets cd->nllen. */
4629              ptr += cd->nllen;
4630              break;
4631              }
4632          ptr++;          ptr++;
4633  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4634          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4635  #endif  #endif
4636          }          }
4637        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4638        }        }
4639      }      }
4640    
4641    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4642    
4643      is_quantifier =
4644        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4645        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4646    
4647      /* Fill in length of a previous callout, except when the next thing is a
4648      quantifier or when processing a property substitution string in UCP mode. */
4649    
4650      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4651           after_manual_callout-- <= 0)
4652        {
4653        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4654          complete_callout(previous_callout, ptr, cd);
4655        previous_callout = NULL;
4656        }
4657    
4658      /* Create auto callout, except for quantifiers, or while processing property
4659      strings that are substituted for \w etc in UCP mode. */
4660    
4661    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4662      {      {
4663      previous_callout = code;      previous_callout = code;
4664      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4665      }      }
4666    
4667      /* Process the next pattern item. */
4668    
4669    switch(c)    switch(c)
4670      {      {
4671      /* ===================================================================*/      /* ===================================================================*/
4672      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4673      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4674      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4675      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4676        *firstcharflagsptr = firstcharflags;
4677      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4678        *reqcharflagsptr = reqcharflags;
4679      *codeptr = code;      *codeptr = code;
4680      *ptrptr = ptr;      *ptrptr = ptr;
4681      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4699  for (;; ptr++)
4699      previous = NULL;      previous = NULL;
4700      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4701        {        {
4702        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET)
4703            zerofirstcharflags = firstcharflags = REQ_NONE;
4704        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4705        }        }
4706      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4715  for (;; ptr++)
4715      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4716    
4717      case CHAR_DOT:      case CHAR_DOT:
4718      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4719      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4720        zerofirstcharflags = firstcharflags;
4721      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4722        zeroreqcharflags = reqcharflags;
4723      previous = code;      previous = code;
4724      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4725      break;      break;
# Line 4016  for (;; ptr++) Line 4748  for (;; ptr++)
4748        }        }
4749      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4750    
4751        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4752        used for "start of word" and "end of word". As these are otherwise illegal
4753        sequences, we don't break anything by recognizing them. They are replaced
4754        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4755        erroneous and are handled by the normal code below. */
4756    
4757      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4758        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4759          {
4760          nestptr = ptr + 7;
4761          ptr = sub_start_of_word - 1;
4762          continue;
4763          }
4764    
4765        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4766          {
4767          nestptr = ptr + 7;
4768          ptr = sub_end_of_word - 1;
4769          continue;
4770          }
4771    
4772        /* Handle a real character class. */
4773    
4774      previous = code;      previous = code;
4775    
4776      /* 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 4815  for (;; ptr++)
4815          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4816        {        {
4817        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4818        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4819        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4820          zerofirstcharflags = firstcharflags;
4821        break;        break;
4822        }        }
4823    
# Line 4072  for (;; ptr++) Line 4827  for (;; ptr++)
4827    
4828      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4829    
4830        /* Extended class (xclass) will be used when characters > 255
4831        might match. */
4832    
4833    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4834        xclass = FALSE;
4835        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4836        class_uchardata_base = class_uchardata;   /* Save the start */
4837    #endif
4838    
4839      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4840      class_has_8bitchar will be non-zero if the class contains at least one <      class_has_8bitchar will be non-zero if the class contains at least one <
4841      256 character; class_one_char will be 1 if the class contains just one      256 character; class_one_char will be 1 if the class contains just one
4842      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4843        are present in the class. */
4844    
4845      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4846      class_one_char = 0;      class_one_char = 0;
4847    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4848        xclass_has_prop = FALSE;
4849    #endif
4850    
4851      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a