/[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 334 by ph10, Fri Apr 11 15:48:14 2008 UTC revision 1392 by ph10, Wed Nov 6 18:00:09 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 87  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118  #define COMPILE_WORK_SIZE (4096)  /* 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
126    overrun before it actually does run off the end of the data block. */
127    
128    #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138    /* Repeated character flags. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
144  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
145  is invalid. */  is invalid. */
146    
147  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
148    
149    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150    in UTF-8 mode. */
151    
152  static const short int escapes[] = {  static const short int escapes[] = {
153       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
154       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
155     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
156  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
158  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
159     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
160  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
161  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
162       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
163         -ESC_D,                  -ESC_E,
164         0,                       -ESC_G,
165         -ESC_H,                  0,
166         0,                       -ESC_K,
167         0,                       0,
168         -ESC_N,                  0,
169         -ESC_P,                  -ESC_Q,
170         -ESC_R,                  -ESC_S,
171         0,                       0,
172         -ESC_V,                  -ESC_W,
173         -ESC_X,                  0,
174         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
175         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
176         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
177         CHAR_GRAVE_ACCENT,       7,
178         -ESC_b,                  0,
179         -ESC_d,                  ESC_e,
180         ESC_f,                   0,
181         -ESC_h,                  0,
182         0,                       -ESC_k,
183         0,                       0,
184         ESC_n,                   0,
185         -ESC_p,                  0,
186         ESC_r,                   -ESC_s,
187         ESC_tee,                 0,
188         -ESC_v,                  -ESC_w,
189         0,                       0,
190         -ESC_z
191  };  };
192    
193  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
194    
195    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196    
197  static const short int escapes[] = {  static const short int escapes[] = {
198  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
199  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 212  static const short int escapes[] = {
212  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
213  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
214  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
215  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
216  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
217  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
218  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 224  static const short int escapes[] = {
224    
225  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
227  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
228    string is built from string macros so that it works in UTF-8 mode on EBCDIC
229    platforms. */
230    
231  typedef struct verbitem {  typedef struct verbitem {
232    int   len;    int   len;                 /* Length of verb name */
233    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
234      int   op_arg;              /* Op when arg present, or -1 if not allowed */
235  } verbitem;  } verbitem;
236    
237  static const char verbnames[] =  static const char verbnames[] =
238    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
239    "COMMIT\0"    STRING_MARK0
240    "F\0"    STRING_ACCEPT0
241    "FAIL\0"    STRING_COMMIT0
242    "PRUNE\0"    STRING_F0
243    "SKIP\0"    STRING_FAIL0
244    "THEN";    STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248  static const verbitem verbs[] = {  static const verbitem verbs[] = {
249    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
250    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
251    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
252    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
253    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
254    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
255    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
256      { 4, OP_SKIP,   OP_SKIP_ARG  },
257      { 4, OP_THEN,   OP_THEN_ARG  }
258  };  };
259    
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 175  static const int verbcount = sizeof(verb Line 264  static const int verbcount = sizeof(verb
264  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
265  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
266  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
267  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270  static const char posix_names[] =  static const char posix_names[] =
271    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
272    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
273    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
274      STRING_word0  STRING_xdigit;
275    
276  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    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 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* 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
285  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
286  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 212  static const int posix_class_maps[] = { Line 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312    Unicode property escapes. */
313    
314    #ifdef SUPPORT_UCP
315    static const pcre_uchar string_PNd[]  = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pNd[]  = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    static const pcre_uchar string_PXsp[] = {
322      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
323      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324    static const pcre_uchar string_pXsp[] = {
325      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
326      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
327    static const pcre_uchar string_PXwd[] = {
328      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_pXwd[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    
334    static const pcre_uchar *substitutes[] = {
335      string_PNd,           /* \D */
336      string_pNd,           /* \d */
337      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339      string_PXwd,          /* \W */
340      string_pXwd           /* \w */
341    };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349    static const pcre_uchar string_pL[] =   {
350      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352    static const pcre_uchar string_pLl[] =  {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_pLu[] =  {
356      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_pXan[] = {
359      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_h[] =    {
362      CHAR_BACKSLASH, CHAR_h, '\0' };
363    static const pcre_uchar string_pXps[] = {
364      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
365      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
366    static const pcre_uchar string_PL[] =   {
367      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
368      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
369    static const pcre_uchar string_PLl[] =  {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    static const pcre_uchar string_PLu[] =  {
373      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
374      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
375    static const pcre_uchar string_PXan[] = {
376      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
377      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
378    static const pcre_uchar string_H[] =    {
379      CHAR_BACKSLASH, CHAR_H, '\0' };
380    static const pcre_uchar string_PXps[] = {
381      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
382      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
383    
384    static const pcre_uchar *posix_substitutes[] = {
385      string_pL,            /* alpha */
386      string_pLl,           /* lower */
387      string_pLu,           /* upper */
388      string_pXan,          /* alnum */
389      NULL,                 /* ascii */
390      string_h,             /* blank */
391      NULL,                 /* cntrl */
392      string_pNd,           /* digit */
393      NULL,                 /* graph */
394      NULL,                 /* print */
395      NULL,                 /* punct */
396      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398      NULL,                 /* xdigit */
399      /* Negated cases */
400      string_PL,            /* ^alpha */
401      string_PLl,           /* ^lower */
402      string_PLu,           /* ^upper */
403      string_PXan,          /* ^alnum */
404      NULL,                 /* ^ascii */
405      string_H,             /* ^blank */
406      NULL,                 /* ^cntrl */
407      string_PNd,           /* ^digit */
408      NULL,                 /* ^graph */
409      NULL,                 /* ^print */
410      NULL,                 /* ^punct */
411      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413      NULL                  /* ^xdigit */
414    };
415    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
416    #endif
417    
418  #define STRING(a)  # a  #define STRING(a)  # a
419  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 426  the number of relocations needed when a
426  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
427  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
428  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
429  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
430    
431    Each substring ends with \0 to insert a null character. This includes the final
432    substring, so that the whole string ends with \0\0, which can be detected when
433    counting through. */
434    
435  static const char error_texts[] =  static const char error_texts[] =
436    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 471  static const char error_texts[] =
471    /* 30 */    /* 30 */
472    "unknown POSIX class name\0"    "unknown POSIX class name\0"
473    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
474    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
475    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
480    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
481    "number after (?C is > 255\0"    "number after (?C is > 255\0"
482    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
483    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 494  static const char error_texts[] =
494    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
495    /* 50 */    /* 50 */
496    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
497    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
498    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
499    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
500    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
501    /* 55 */    /* 55 */
502    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
503    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
504    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
505    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
506    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507    /* 60 */    /* 60 */
508    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
509    "number is too big\0"    "number is too big\0"
510    "subpattern name expected\0"    "subpattern name expected\0"
511    "digit expected after (?+";    "digit expected after (?+\0"
512      "] is an invalid data character in JavaScript compatibility mode\0"
513      /* 65 */
514      "different names for subpatterns of the same number are not allowed\0"
515      "(*MARK) must have an argument\0"
516      "this version of PCRE is not compiled with Unicode property support\0"
517      "\\c must be followed by an ASCII character\0"
518      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
519      /* 70 */
520      "internal error: unknown opcode in find_fixedlength()\0"
521      "\\N is not supported in a class\0"
522      "too many forward references\0"
523      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524      "invalid UTF-16 string\0"
525      /* 75 */
526      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527      "character value in \\u.... sequence is too large\0"
528      "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      ;
537    
538  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
539  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 551  For convenience, we use the same bit def
551    
552  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
553    
554  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
555  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
556    into a subtraction and unsigned comparison). */
557    
558    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
559    
560    #ifndef EBCDIC
561    
562    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
563    UTF-8 mode. */
564    
565    static const pcre_uint8 digitab[] =
566    {    {
567    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
568    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 357  static const unsigned char digitab[] = Line 597  static const unsigned char digitab[] =
597    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
598    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
599    
600  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
601  static const unsigned char digitab[] =  
602    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
603    
604    static const pcre_uint8 digitab[] =
605    {    {
606    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
607    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 393  static const unsigned char digitab[] = Line 636  static const unsigned char digitab[] =
636    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
637    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
638    
639  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
640    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
641    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
642    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 429  static const unsigned char ebcdic_charta Line 672  static const unsigned char ebcdic_charta
672  #endif  #endif
673    
674    
675  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
676    between adjacent character-type opcodes. The left-hand (repeated) opcode is
677    used to select the row, and the right-hand opcode is use to select the column.
678    A value of 1 means that auto-possessification is OK. For example, the second
679    value in the first row means that \D+\d can be turned into \D++\d.
680    
681    The Unicode property types (\P and \p) have to be present to fill out the table
682    because of what their opcode values are, but the table values should always be
683    zero because property types are handled separately in the code. The last four
684    columns apply to items that cannot be repeated, so there is no need to have
685    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
686    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
687    
688    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
689    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
690    
691    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
692    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
693      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
694      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
695      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
696      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
697      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
698      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
699      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
704      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
705      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
706      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
707      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
708      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
709      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
710    };
711    
712  static BOOL  
713    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  /* This table is used to check whether auto-possessification is possible
714      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
715    left-hand (repeated) opcode is used to select the row, and the right-hand
716    opcode is used to select the column. The values are as follows:
717    
718      0   Always return FALSE (never auto-possessify)
719      1   Character groups are distinct (possessify if both are OP_PROP)
720      2   Check character categories in the same group (general or particular)
721      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
722    
723      4   Check left general category vs right particular category
724      5   Check right general category vs left particular category
725    
726      6   Left alphanum vs right general category
727      7   Left space vs right general category
728      8   Left word vs right general category
729    
730      9   Right alphanum vs left general category
731     10   Right space vs left general category
732     11   Right word vs left general category
733    
734     12   Left alphanum vs right particular category
735     13   Left space vs right particular category
736     14   Left word vs right particular category
737    
738     15   Right alphanum vs left particular category
739     16   Right space vs left particular category
740     17   Right word vs left particular category
741    */
742    
743    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
744    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
745      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
746      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
747      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
748      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
749      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
750      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
751      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
753      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
754      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
756    };
757    
758    /* This table is used to check whether auto-possessification is possible
759    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
760    specifies a general category and the other specifies a particular category. The
761    row is selected by the general category and the column by the particular
762    category. The value is 1 if the particular category is not part of the general
763    category. */
764    
765    static const pcre_uint8 catposstab[7][30] = {
766    /* 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 */
767      { 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772      { 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 */
773      { 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 */
774    };
775    
776    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
777    a general or particular category. The properties in each row are those
778    that apply to the character set in question. Duplication means that a little
779    unnecessary work is done when checking, but this keeps things much simpler
780    because they can all use the same code. For more details see the comment where
781    this table is used.
782    
783    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
784    "space", but from Perl 5.18 it's included, so both categories are treated the
785    same here. */
786    
787    static const pcre_uint8 posspropstab[3][4] = {
788      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
789      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
790      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
791    };
792    
793    /* This table is used when converting repeating opcodes into possessified
794    versions as a result of an explicit possessive quantifier such as ++. A zero
795    value means there is no possessified version - in those cases the item in
796    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
797    because all relevant opcodes are less than that. */
798    
799    static const pcre_uint8 opcode_possessify[] = {
800      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
802    
803      0,                       /* NOTI */
804      OP_POSSTAR, 0,           /* STAR, MINSTAR */
805      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
806      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
807      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
808      0,                       /* EXACT */
809      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
810    
811      OP_POSSTARI, 0,          /* STARI, MINSTARI */
812      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
813      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
814      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
815      0,                       /* EXACTI */
816      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
817    
818      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
819      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
820      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
821      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
822      0,                       /* NOTEXACT */
823      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
826      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
827      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
828      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
829      0,                       /* NOTEXACTI */
830      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
833      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
834      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
835      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
836      0,                       /* TYPEEXACT */
837      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
840      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
841      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
842      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
843      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
844    
845      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
846      0, 0,                    /* REF, REFI */
847      0, 0,                    /* DNREF, DNREFI */
848      0, 0                     /* RECURSE, CALLOUT */
849    };
850    
851    
852    
# Line 454  static const char * Line 867  static const char *
867  find_error_text(int n)  find_error_text(int n)
868  {  {
869  const char *s = error_texts;  const char *s = error_texts;
870  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
871      {
872      while (*s++ != CHAR_NULL) {};
873      if (*s == CHAR_NULL) return "Error text not found (please report)";
874      }
875  return s;  return s;
876  }  }
877    
878    
879    
880    /*************************************************
881    *           Expand the workspace                 *
882    *************************************************/
883    
884    /* This function is called during the second compiling phase, if the number of
885    forward references fills the existing workspace, which is originally a block on
886    the stack. A larger block is obtained from malloc() unless the ultimate limit
887    has been reached or the increase will be rather small.
888    
889    Argument: pointer to the compile data block
890    Returns:  0 if all went well, else an error number
891    */
892    
893    static int
894    expand_workspace(compile_data *cd)
895    {
896    pcre_uchar *newspace;
897    int newsize = cd->workspace_size * 2;
898    
899    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
900    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
901        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
902     return ERR72;
903    
904    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
905    if (newspace == NULL) return ERR21;
906    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
907    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
908    if (cd->workspace_size > COMPILE_WORK_SIZE)
909      (PUBL(free))((void *)cd->start_workspace);
910    cd->start_workspace = newspace;
911    cd->workspace_size = newsize;
912    return 0;
913    }
914    
915    
916    
917    /*************************************************
918    *            Check for counted repeat            *
919    *************************************************/
920    
921    /* This function is called when a '{' is encountered in a place where it might
922    start a quantifier. It looks ahead to see if it really is a quantifier or not.
923    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
924    where the ddds are digits.
925    
926    Arguments:
927      p         pointer to the first char after '{'
928    
929    Returns:    TRUE or FALSE
930    */
931    
932    static BOOL
933    is_counted_repeat(const pcre_uchar *p)
934    {
935    if (!IS_DIGIT(*p)) return FALSE;
936    p++;
937    while (IS_DIGIT(*p)) p++;
938    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
939    
940    if (*p++ != CHAR_COMMA) return FALSE;
941    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943    if (!IS_DIGIT(*p)) return FALSE;
944    p++;
945    while (IS_DIGIT(*p)) p++;
946    
947    return (*p == CHAR_RIGHT_CURLY_BRACKET);
948    }
949    
950    
951    
952  /*************************************************  /*************************************************
953  *            Handle escapes                      *  *            Handle escapes                      *
954  *************************************************/  *************************************************/
955    
956  /* 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
957  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
958  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
959  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
960  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
961  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
962    
963  Arguments:  Arguments:
964    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
965      chptr          points to a returned data character
966    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
967    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
968    options        the options bits    options        the options bits
969    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
970    
971  Returns:         zero or positive => a data character  Returns:         zero => a data character
972                   negative => a special escape sequence                   positive => a special escape sequence
973                     negative => a back reference
974                   on error, errorcodeptr is set                   on error, errorcodeptr is set
975  */  */
976    
977  static int  static int
978  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
979    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
980  {  {
981  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
982  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
983  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
984    pcre_uint32 c;
985    int escape = 0;
986    int i;
987    
988  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
989  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
990    
991  /* 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. */
992    
993  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
994    
995  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
996  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.
997  Otherwise further processing may be required. */  Otherwise further processing may be required. */
998    
999  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1000  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
1001  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1002    else if ((i = escapes[c - CHAR_0]) != 0)
1003      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1004    
1005  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1006  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
1007  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1008    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1009  #endif  #endif
1010    
1011  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1012    
1013  else  else
1014    {    {
1015    const uschar *oldptr;    const pcre_uchar *oldptr;
1016    BOOL braced, negated;    BOOL braced, negated, overflow;
1017      int s;
1018    
1019    switch (c)    switch (c)
1020      {      {
1021      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
1022      error. */      error. */
1023    
1024      case 'l':      case CHAR_l:
1025      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1026      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1027      break;      break;
1028    
1029      /* \g must be followed by one of a number of specific things:      case CHAR_u:
1030        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1031          {
1032          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1033          Otherwise it is a lowercase u letter. */
1034          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1035            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1037            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1038            {
1039            c = 0;
1040            for (i = 0; i < 4; ++i)
1041              {
1042              register pcre_uint32 cc = *(++ptr);
1043    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1044              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1045              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1046    #else           /* EBCDIC coding */
1047              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1048              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1049    #endif
1050              }
1051    
1052    #if defined COMPILE_PCRE8
1053            if (c > (utf ? 0x10ffffU : 0xffU))
1054    #elif defined COMPILE_PCRE16
1055            if (c > (utf ? 0x10ffffU : 0xffffU))
1056    #elif defined COMPILE_PCRE32
1057            if (utf && c > 0x10ffffU)
1058    #endif
1059              {
1060              *errorcodeptr = ERR76;
1061              }
1062            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1063            }
1064          }
1065        else
1066          *errorcodeptr = ERR37;
1067        break;
1068    
1069        case CHAR_U:
1070        /* In JavaScript, \U is an uppercase U letter. */
1071        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1072        break;
1073    
1074        /* In a character class, \g is just a literal "g". Outside a character
1075        class, \g must be followed by one of a number of specific things:
1076    
1077      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
1078      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
1079      5.10 feature.      5.10 feature.
1080    
1081      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1082      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
1083      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
1084      was \k.      was \k.
   
     (3) For Oniguruma compatibility we also support \g followed by a name or a  
     number either in angle brackets or in single quotes. However, these are  
     (possibly recursive) subroutine calls, _not_ backreferences. Just return  
     the -ESC_g code (cf \k). */  
1085    
1086      case 'g':      (3) For Oniguruma compatibility we also support \g followed by a name or a
1087      if (ptr[1] == '<' || ptr[1] == '\'')      number either in angle brackets or in single quotes. However, these are
1088        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1089        the ESC_g code (cf \k). */
1090    
1091        case CHAR_g:
1092        if (isclass) break;
1093        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1094        {        {
1095        c = -ESC_g;        escape = ESC_g;
1096        break;        break;
1097        }        }
1098    
1099      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
1100    
1101      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1102        {        {
1103        const uschar *p;        const pcre_uchar *p;
1104        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1105          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1106        if (*p != 0 && *p != '}')        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1107          {          {
1108          c = -ESC_k;          escape = ESC_k;
1109          break;          break;
1110          }          }
1111        braced = TRUE;        braced = TRUE;
# Line 571  else Line 1113  else
1113        }        }
1114      else braced = FALSE;      else braced = FALSE;
1115    
1116      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1117        {        {
1118        negated = TRUE;        negated = TRUE;
1119        ptr++;        ptr++;
1120        }        }
1121      else negated = FALSE;      else negated = FALSE;
1122    
1123      c = 0;      /* The integer range is limited by the machine's int representation. */
1124      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1125        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1126        while (IS_DIGIT(ptr[1]))
1127      if (c < 0)   /* Integer overflow */        {
1128          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1129            {
1130            overflow = TRUE;
1131            break;
1132            }
1133          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1134          }
1135        if (overflow) /* Integer overflow */
1136        {        {
1137          while (IS_DIGIT(ptr[1]))
1138            ptr++;
1139        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1140        break;        break;
1141        }        }
1142    
1143      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1144        {        {
1145        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1146        break;        break;
1147        }        }
1148    
1149      if (c == 0)      if (s == 0)
1150        {        {
1151        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1152        break;        break;
1153        }        }
1154    
1155      if (negated)      if (negated)
1156        {        {
1157        if (c > bracount)        if (s > bracount)
1158          {          {
1159          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1160          break;          break;
1161          }          }
1162        c = bracount - (c - 1);        s = bracount - (s - 1);
1163        }        }
1164    
1165      c = -(ESC_REF + c);      escape = -s;
1166      break;      break;
1167    
1168      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1169      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
1170      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1171        recommended to avoid the ambiguities in the old syntax.
1172    
1173      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
1174      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
1175      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
1176      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
1177      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
1178      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
1179      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1180    
1181      case '1': case '2': case '3': case '4': case '5':      Inside a character class, \ followed by a digit is always either a literal
1182      case '6': case '7': case '8': case '9':      8 or 9 or an octal number. */
1183    
1184        case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1185        case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1186    
1187      if (!isclass)      if (!isclass)
1188        {        {
1189        oldptr = ptr;        oldptr = ptr;
1190        c -= '0';        /* The integer range is limited by the machine's int representation. */
1191        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1192          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1193        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1194            {
1195            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1196              {
1197              overflow = TRUE;
1198              break;
1199              }
1200            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1201            }
1202          if (overflow) /* Integer overflow */
1203          {          {
1204            while (IS_DIGIT(ptr[1]))
1205              ptr++;
1206          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1207          break;          break;
1208          }          }
1209        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1210          {          {
1211          c = -(ESC_REF + c);          escape = -s;
1212          break;          break;
1213          }          }
1214        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1215        }        }
1216    
1217      /* 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
1218      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
1219      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
1220        changed so as not to insert the binary zero. */
1221    
1222      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1223        {  
1224        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1225    
1226      /* \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
1227      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
1228      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1229      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1230      than 3 octal digits. */      but no more than 3 octal digits. */
1231    
1232      case '0':      case CHAR_0:
1233      c -= '0';      c -= CHAR_0;
1234      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1235          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1236      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1237        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1238    #endif
1239      break;      break;
1240    
1241      /* \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
1242      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1243    
1244      case 'x':      case CHAR_o:
1245      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1246        {        {
1247        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1248        c = 0;        c = 0;
1249        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1250          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1251          {          {
1252          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1253          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1254          count++;  #ifdef COMPILE_PCRE32
1255            if (c >= 0x20000000l) { overflow = TRUE; break; }
1256  #ifndef EBCDIC  /* ASCII coding */  #endif
1257          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1258          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1259  #else           /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1260          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1261          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1262    #elif defined COMPILE_PCRE32
1263            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1264  #endif  #endif
1265          }          }
1266          if (overflow)
       if (*pt == '}')  
1267          {          {
1268          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1269          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1270          }          }
1271          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1272        /* If the sequence of hex digits does not end with '}', then we don't          {
1273        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1274            }
1275          else *errorcodeptr = ERR80;
1276        }        }
1277        break;
1278    
1279      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1280        numbers. Otherwise it is a lowercase x letter. */
1281    
1282      c = 0;      case CHAR_x:
1283      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1284        {        {
1285        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1286        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1287  #ifndef EBCDIC  /* ASCII coding */          {
1288        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1289        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1290              {
1291              register pcre_uint32 cc = *(++ptr);
1292    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1293              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1294              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1295  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1296        if (cc <= 'z') cc += 64;              /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1297        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1298  #endif  #endif
1299        }            }
1300            }
1301          }    /* End JavaScript handling */
1302    
1303        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1304        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1305        digits. If not, { used to be treated as a data character. However, Perl
1306        seems to read hex digits up to the first non-such, and ignore the rest, so
1307        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1308        now gives an error. */
1309    
1310        else
1311          {
1312          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1313            {
1314            ptr += 2;
1315            c = 0;
1316            overflow = FALSE;
1317            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1318              {
1319              register pcre_uint32 cc = *ptr++;
1320              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1321    
1322    #ifdef COMPILE_PCRE32
1323              if (c >= 0x10000000l) { overflow = TRUE; break; }
1324    #endif
1325    
1326    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1327              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1328              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1329    #else           /* EBCDIC coding */
1330              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1331              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1332    #endif
1333    
1334    #if defined COMPILE_PCRE8
1335              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1336    #elif defined COMPILE_PCRE16
1337              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1338    #elif defined COMPILE_PCRE32
1339              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1340    #endif
1341              }
1342    
1343            if (overflow)
1344              {
1345              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1346              *errorcodeptr = ERR34;
1347              }
1348    
1349            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1350              {
1351              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1352              }
1353    
1354            /* If the sequence of hex digits does not end with '}', give an error.
1355            We used just to recognize this construct and fall through to the normal
1356            \x handling, but nowadays Perl gives an error, which seems much more
1357            sensible, so we do too. */
1358    
1359            else *errorcodeptr = ERR79;
1360            }   /* End of \x{} processing */
1361    
1362          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1363    
1364          else
1365            {
1366            c = 0;
1367            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1368              {
1369              pcre_uint32 cc;                          /* Some compilers don't like */
1370              cc = *(++ptr);                           /* ++ in initializers */
1371    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1372              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1373              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1374    #else           /* EBCDIC coding */
1375              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1376              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1377    #endif
1378              }
1379            }     /* End of \xdd handling */
1380          }       /* End of Perl-style \x handling */
1381      break;      break;
1382    
1383      /* 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.
1384      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1385        coding is ASCII-specific, but then the whole concept of \cx is
1386      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1387    
1388      case 'c':      case CHAR_c:
1389      c = *(++ptr);      c = *(++ptr);
1390      if (c == 0)      if (c == CHAR_NULL)
1391        {        {
1392        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1393        break;        break;
1394        }        }
1395    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1396  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1397      if (c >= 'a' && c <= 'z') c -= 32;        {
1398          *errorcodeptr = ERR68;
1399          break;
1400          }
1401        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1402      c ^= 0x40;      c ^= 0x40;
1403  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1404      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1405      c ^= 0xC0;      c ^= 0xC0;
1406  #endif  #endif
1407      break;      break;
# Line 763  else Line 1423  else
1423      }      }
1424    }    }
1425    
1426    /* Perl supports \N{name} for character names, as well as plain \N for "not
1427    newline". PCRE does not support \N{name}. However, it does support
1428    quantification such as \N{2,3}. */
1429    
1430    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1431         !is_counted_repeat(ptr+2))
1432      *errorcodeptr = ERR37;
1433    
1434    /* If PCRE_UCP is set, we change the values for \d etc. */
1435    
1436    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1437      escape += (ESC_DU - ESC_D);
1438    
1439    /* Set the pointer to the final character before returning. */
1440    
1441  *ptrptr = ptr;  *ptrptr = ptr;
1442  return c;  *chptr = c;
1443    return escape;
1444  }  }
1445    
1446    
# Line 782  escape sequence. Line 1458  escape sequence.
1458  Argument:  Argument:
1459    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1460    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
1461    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
1462      pdataptr       points to an unsigned int that is set to the detailed property value
1463    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1464    
1465  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
1466  */  */
1467    
1468  static int  static BOOL
1469  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1470      unsigned int *pdataptr, int *errorcodeptr)
1471  {  {
1472  int c, i, bot, top;  pcre_uchar c;
1473  const uschar *ptr = *ptrptr;  int i, bot, top;
1474  char name[32];  const pcre_uchar *ptr = *ptrptr;
1475    pcre_uchar name[32];
1476    
1477  c = *(++ptr);  c = *(++ptr);
1478  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1479    
1480  *negptr = FALSE;  *negptr = FALSE;
1481    
1482  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1483  negation. */  negation. */
1484    
1485  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1486    {    {
1487    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1488      {      {
1489      *negptr = TRUE;      *negptr = TRUE;
1490      ptr++;      ptr++;
1491      }      }
1492    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1493      {      {
1494      c = *(++ptr);      c = *(++ptr);
1495      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1496      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1497      name[i] = c;      name[i] = c;
1498      }      }
1499    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1500    name[i] = 0;    name[i] = 0;
1501    }    }
1502    
# Line 834  else Line 1513  else
1513  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1514    
1515  bot = 0;  bot = 0;
1516  top = _pcre_utt_size;  top = PRIV(utt_size);
1517    
1518  while (bot < top)  while (bot < top)
1519    {    {
1520      int r;
1521    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1522    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1523    if (c == 0)    if (r == 0)
1524      {      {
1525      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1526      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1527        return TRUE;
1528      }      }
1529    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1530    }    }
1531    
1532  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1533  *ptrptr = ptr;  *ptrptr = ptr;
1534  return -1;  return FALSE;
1535    
1536  ERROR_RETURN:  ERROR_RETURN:
1537  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1538  *ptrptr = ptr;  *ptrptr = ptr;
1539  return -1;  return FALSE;
1540  }  }
1541  #endif  #endif
1542    
1543    
1544    
   
 /*************************************************  
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
1545  /*************************************************  /*************************************************
1546  *         Read repeat counts                     *  *         Read repeat counts                     *
1547  *************************************************/  *************************************************/
# Line 914  Returns:         pointer to '}' on succe Line 1561  Returns:         pointer to '}' on succe
1561                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1562  */  */
1563    
1564  static const uschar *  static const pcre_uchar *
1565  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1566  {  {
1567  int min = 0;  int min = 0;
1568  int max = -1;  int max = -1;
# Line 923  int max = -1; Line 1570  int max = -1;
1570  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1571  an integer overflow. */  an integer overflow. */
1572    
1573  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1574  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1575    {    {
1576    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1580  if (min < 0 || min > 65535)
1580  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1581  Also, max must not be less than min. */  Also, max must not be less than min. */
1582    
1583  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1584    {    {
1585    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1586      {      {
1587      max = 0;      max = 0;
1588      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1589      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1590        {        {
1591        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1610  return p;
1610    
1611    
1612  /*************************************************  /*************************************************
1613  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1614  *************************************************/  *************************************************/
1615    
1616  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1617  subpatterns, and counting them. If it finds a named pattern that matches the  for a fixed first character, or an anchoring op code etc. It skips over things
1618  name it is given, it returns its number. Alternatively, if the name is NULL, it  that do not influence this. For some calls, it makes sense to skip negative
1619  returns when it reaches a given numbered subpattern. This is used for forward  forward and all backward assertions, and also the \b assertion; for others it
1620  references to subpatterns. We know that if (?P< is encountered, the name will  does not.
 be terminated by '>' because that is checked in the first pass.  
1621    
1622  Arguments:  Arguments:
1623    ptr          current position in the pattern    code         pointer to the start of the group
1624    count        current count of capturing parens so far encountered    skipassert   TRUE if certain assertions are to be skipped
   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  
1625    
1626  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1627  */  */
1628    
1629  static int  static const pcre_uchar*
1630  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode)  
1631  {  {
1632  const uschar *thisname;  for (;;)
   
 for (; *ptr != 0; ptr++)  
1633    {    {
1634    int term;    switch ((int)*code)
1635        {
1636        case OP_ASSERT_NOT:
1637        case OP_ASSERTBACK:
1638        case OP_ASSERTBACK_NOT:
1639        if (!skipassert) return code;
1640        do code += GET(code, 1); while (*code == OP_ALT);
1641        code += PRIV(OP_lengths)[*code];
1642        break;
1643    
1644        case OP_WORD_BOUNDARY:
1645        case OP_NOT_WORD_BOUNDARY:
1646        if (!skipassert) return code;
1647        /* Fall through */
1648    
1649    /* Skip over backslashed characters and also entire \Q...\E */      case OP_CALLOUT:
1650        case OP_CREF:
1651        case OP_DNCREF:
1652        case OP_RREF:
1653        case OP_DNRREF:
1654        case OP_DEF:
1655        code += PRIV(OP_lengths)[*code];
1656        break;
1657    
1658    if (*ptr == '\\')      default:
1659      {      return code;
     if (*(++ptr) == 0) return -1;  
     if (*ptr == 'Q') for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != '\\');  
       if (*ptr == 0) return -1;  
       if (*(++ptr) == 'E') break;  
       }  
     continue;  
1660      }      }
1661      }
1662    /* Skip over character classes */  /* Control never reaches here */
1663    }
   if (*ptr == '[')  
     {  
     while (*(++ptr) != ']')  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == '\\')  
         {  
         if (*(++ptr) == 0) return -1;  
         if (*ptr == 'Q') for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != '\\');  
           if (*ptr == 0) return -1;  
           if (*(++ptr) == 'E') break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == '#')  
     {  
     while (*(++ptr) != 0 && *ptr != '\n');  
     if (*ptr == 0) return -1;  
     continue;  
     }  
   
   /* An opening parens must now be a real metacharacter */  
   
   if (*ptr != '(') continue;  
   if (ptr[1] != '?' && ptr[1] != '*')  
     {  
     count++;  
     if (name == NULL && count == lorn) return count;  
     continue;  
     }  
   
   ptr += 2;  
   if (*ptr == 'P') ptr++;                      /* Allow optional P */  
   
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
   
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
   
   if (name == NULL && count == lorn) return count;  
   term = *ptr++;  
   if (term == '<') term = '>';  
   thisname = ptr;  
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
   }  
   
 return -1;  
 }  
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, a change of option is important.  
 For some calls, it makes sense to skip negative forward and all backward  
 assertions, and also the \b assertion; for others it does not.  
   
 Arguments:  
   code         pointer to the start of the group  
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const uschar*  
 first_significant_code(const uschar *code, int *options, int optbit,  
   BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
     {  
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     case OP_WORD_BOUNDARY:  
     case OP_NOT_WORD_BOUNDARY:  
     if (!skipassert) return code;  
     /* Fall through */  
   
     case OP_CALLOUT:  
     case OP_CREF:  
     case OP_RREF:  
     case OP_DEF:  
     code += _pcre_OP_lengths[*code];  
     break;  
   
     default:  
     return code;  
     }  
   }  
 /* Control never reaches here */  
 }  
   
1664    
1665    
1666    
1667  /*************************************************  /*************************************************
1668  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1669  *************************************************/  *************************************************/
1670    
1671  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1672  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1673  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1674    temporarily terminated with OP_END when this function is called.
1675    
1676    This function is called when a backward assertion is encountered, so that if it
1677    fails, the error message can point to the correct place in the pattern.
1678    However, we cannot do this when the assertion contains subroutine calls,
1679    because they can be forward references. We solve this by remembering this case
1680    and doing the check at the end; a flag specifies which mode we are running in.
1681    
1682  Arguments:  Arguments:
1683    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1684    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1685      atend    TRUE if called when the pattern is complete
1686  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1687               or -2 if \C was encountered  
1688    Returns:   the fixed length,
1689                 or -1 if there is no fixed length,
1690                 or -2 if \C was encountered (in UTF-8 mode only)
1691                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1692                 or -4 if an unknown opcode was encountered (internal error)
1693  */  */
1694    
1695  static int  static int
1696  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1697  {  {
1698  int length = -1;  int length = -1;
1699    
1700  register int branchlength = 0;  register int branchlength = 0;
1701  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1702    
1703  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1704  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1168  branch, check the length against that of Line 1706  branch, check the length against that of
1706  for (;;)  for (;;)
1707    {    {
1708    int d;    int d;
1709    register int op = *cc;    pcre_uchar *ce, *cs;
1710      register pcre_uchar op = *cc;
1711    
1712    switch (op)    switch (op)
1713      {      {
1714        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1715        OP_BRA (normal non-capturing bracket) because the other variants of these
1716        opcodes are all concerned with unlimited repeated groups, which of course
1717        are not of fixed length. */
1718    
1719      case OP_CBRA:      case OP_CBRA:
1720      case OP_BRA:      case OP_BRA:
1721      case OP_ONCE:      case OP_ONCE:
1722        case OP_ONCE_NC:
1723      case OP_COND:      case OP_COND:
1724      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1725      if (d < 0) return d;      if (d < 0) return d;
1726      branchlength += d;      branchlength += d;
1727      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1728      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1729      break;      break;
1730    
1731      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1732      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1733      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1734        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1735        because they all imply an unlimited repeat. */
1736    
1737      case OP_ALT:      case OP_ALT:
1738      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1739      case OP_END:      case OP_END:
1740        case OP_ACCEPT:
1741        case OP_ASSERT_ACCEPT:
1742      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1743        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1744      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1746  for (;;)
1746      branchlength = 0;      branchlength = 0;
1747      break;      break;
1748    
1749        /* A true recursion implies not fixed length, but a subroutine call may
1750        be OK. If the subroutine is a forward reference, we can't deal with
1751        it until the end of the pattern, so return -3. */
1752    
1753        case OP_RECURSE:
1754        if (!atend) return -3;
1755        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1756        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1757        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1758        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1759        if (d < 0) return d;
1760        branchlength += d;
1761        cc += 1 + LINK_SIZE;
1762        break;
1763    
1764      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1765    
1766      case OP_ASSERT:      case OP_ASSERT:
# Line 1205  for (;;) Line 1768  for (;;)
1768      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1769      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1770      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1771      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1772        break;
1773    
1774      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1775    
1776      case OP_REVERSE:      case OP_MARK:
1777        case OP_PRUNE_ARG:
1778        case OP_SKIP_ARG:
1779        case OP_THEN_ARG:
1780        cc += cc[1] + PRIV(OP_lengths)[*cc];
1781        break;
1782    
1783        case OP_CALLOUT:
1784        case OP_CIRC:
1785        case OP_CIRCM:
1786        case OP_CLOSE:
1787        case OP_COMMIT:
1788      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1789      case OP_DEF:      case OP_DEF:
1790      case OP_OPT:      case OP_DNCREF:
1791      case OP_CALLOUT:      case OP_DNRREF:
1792      case OP_SOD:      case OP_DOLL:
1793      case OP_SOM:      case OP_DOLLM:
1794      case OP_EOD:      case OP_EOD:
1795      case OP_EODN:      case OP_EODN:
1796      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1797      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1798        case OP_PRUNE:
1799        case OP_REVERSE:
1800        case OP_RREF:
1801        case OP_SET_SOM:
1802        case OP_SKIP:
1803        case OP_SOD:
1804        case OP_SOM:
1805        case OP_THEN:
1806      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1807      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1808      break;      break;
1809    
1810      /* Handle literal characters */      /* Handle literal characters */
1811    
1812      case OP_CHAR:      case OP_CHAR:
1813      case OP_CHARNC:      case OP_CHARI:
1814      case OP_NOT:      case OP_NOT:
1815        case OP_NOTI:
1816      branchlength++;      branchlength++;
1817      cc += 2;      cc += 2;
1818  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1819      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1820  #endif  #endif
1821      break;      break;
1822    
# Line 1245  for (;;) Line 1824  for (;;)
1824      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1825    
1826      case OP_EXACT:      case OP_EXACT:
1827      branchlength += GET2(cc,1);      case OP_EXACTI:
1828      cc += 4;      case OP_NOTEXACT:
1829  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1830      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1831        {      cc += 2 + IMM2_SIZE;
1832        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1833        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834  #endif  #endif
1835      break;      break;
1836    
1837      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1838      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1839      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1840      cc += 4;        cc += 2;
1841        cc += 1 + IMM2_SIZE + 1;
1842      break;      break;
1843    
1844      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1268  for (;;) Line 1848  for (;;)
1848      cc += 2;      cc += 2;
1849      /* Fall through */      /* Fall through */
1850    
1851        case OP_HSPACE:
1852        case OP_VSPACE:
1853        case OP_NOT_HSPACE:
1854        case OP_NOT_VSPACE:
1855      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1856      case OP_DIGIT:      case OP_DIGIT:
1857      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1859  for (;;)
1859      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1860      case OP_WORDCHAR:      case OP_WORDCHAR:
1861      case OP_ANY:      case OP_ANY:
1862        case OP_ALLANY:
1863      branchlength++;      branchlength++;
1864      cc++;      cc++;
1865      break;      break;
1866    
1867      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1868        otherwise \C is coded as OP_ALLANY. */
1869    
1870      case OP_ANYBYTE:      case OP_ANYBYTE:
1871      return -2;      return -2;
1872    
1873      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1874    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1875      case OP_CLASS:      case OP_CLASS:
1876      case OP_NCLASS:      case OP_NCLASS:
1877      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1878        case OP_XCLASS:
1879        /* The original code caused an unsigned overflow in 64 bit systems,
1880        so now we use a conditional statement. */
1881        if (op == OP_XCLASS)
1882          cc += GET(cc, 1);
1883        else
1884          cc += PRIV(OP_lengths)[OP_CLASS];
1885    #else
1886        cc += PRIV(OP_lengths)[OP_CLASS];
1887    #endif
1888    
1889      switch (*cc)      switch (*cc)
1890        {        {
1891        case OP_CRSTAR:        case OP_CRSTAR:
1892        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1893          case OP_CRPLUS:
1894          case OP_CRMINPLUS:
1895        case OP_CRQUERY:        case OP_CRQUERY:
1896        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1897          case OP_CRPOSSTAR:
1898          case OP_CRPOSPLUS:
1899          case OP_CRPOSQUERY:
1900        return -1;        return -1;
1901    
1902        case OP_CRRANGE:        case OP_CRRANGE:
1903        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1904        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1905        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1906        cc += 5;        branchlength += (int)GET2(cc,1);
1907          cc += 1 + 2 * IMM2_SIZE;
1908        break;        break;
1909    
1910        default:        default:
# Line 1318  for (;;) Line 1914  for (;;)
1914    
1915      /* Anything else is variable length */      /* Anything else is variable length */
1916    
1917      default:      case OP_ANYNL:
1918        case OP_BRAMINZERO:
1919        case OP_BRAPOS:
1920        case OP_BRAPOSZERO:
1921        case OP_BRAZERO:
1922        case OP_CBRAPOS:
1923        case OP_EXTUNI:
1924        case OP_KETRMAX:
1925        case OP_KETRMIN:
1926        case OP_KETRPOS:
1927        case OP_MINPLUS:
1928        case OP_MINPLUSI:
1929        case OP_MINQUERY:
1930        case OP_MINQUERYI:
1931        case OP_MINSTAR:
1932        case OP_MINSTARI:
1933        case OP_MINUPTO:
1934        case OP_MINUPTOI:
1935        case OP_NOTMINPLUS:
1936        case OP_NOTMINPLUSI:
1937        case OP_NOTMINQUERY:
1938        case OP_NOTMINQUERYI:
1939        case OP_NOTMINSTAR:
1940        case OP_NOTMINSTARI:
1941        case OP_NOTMINUPTO:
1942        case OP_NOTMINUPTOI:
1943        case OP_NOTPLUS:
1944        case OP_NOTPLUSI:
1945        case OP_NOTPOSPLUS:
1946        case OP_NOTPOSPLUSI:
1947        case OP_NOTPOSQUERY:
1948        case OP_NOTPOSQUERYI:
1949        case OP_NOTPOSSTAR:
1950        case OP_NOTPOSSTARI:
1951        case OP_NOTPOSUPTO:
1952        case OP_NOTPOSUPTOI:
1953        case OP_NOTQUERY:
1954        case OP_NOTQUERYI:
1955        case OP_NOTSTAR:
1956        case OP_NOTSTARI:
1957        case OP_NOTUPTO:
1958        case OP_NOTUPTOI:
1959        case OP_PLUS:
1960        case OP_PLUSI:
1961        case OP_POSPLUS:
1962        case OP_POSPLUSI:
1963        case OP_POSQUERY:
1964        case OP_POSQUERYI:
1965        case OP_POSSTAR:
1966        case OP_POSSTARI:
1967        case OP_POSUPTO:
1968        case OP_POSUPTOI:
1969        case OP_QUERY:
1970        case OP_QUERYI:
1971        case OP_REF:
1972        case OP_REFI:
1973        case OP_DNREF:
1974        case OP_DNREFI:
1975        case OP_SBRA:
1976        case OP_SBRAPOS:
1977        case OP_SCBRA:
1978        case OP_SCBRAPOS:
1979        case OP_SCOND:
1980        case OP_SKIPZERO:
1981        case OP_STAR:
1982        case OP_STARI:
1983        case OP_TYPEMINPLUS:
1984        case OP_TYPEMINQUERY:
1985        case OP_TYPEMINSTAR:
1986        case OP_TYPEMINUPTO:
1987        case OP_TYPEPLUS:
1988        case OP_TYPEPOSPLUS:
1989        case OP_TYPEPOSQUERY:
1990        case OP_TYPEPOSSTAR:
1991        case OP_TYPEPOSUPTO:
1992        case OP_TYPEQUERY:
1993        case OP_TYPESTAR:
1994        case OP_TYPEUPTO:
1995        case OP_UPTO:
1996        case OP_UPTOI:
1997      return -1;      return -1;
1998    
1999        /* Catch unrecognized opcodes so that when new ones are added they
2000        are not forgotten, as has happened in the past. */
2001    
2002        default:
2003        return -4;
2004      }      }
2005    }    }
2006  /* Control never gets here */  /* Control never gets here */
# Line 1327  for (;;) Line 2008  for (;;)
2008    
2009    
2010    
   
2011  /*************************************************  /*************************************************
2012  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2013  *************************************************/  *************************************************/
2014    
2015  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2016  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2017    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2018    so that it can be called from pcre_study() when finding the minimum matching
2019    length.
2020    
2021  Arguments:  Arguments:
2022    code        points to start of expression    code        points to start of expression
2023    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2024    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2025    
2026  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
2027  */  */
2028    
2029  static const uschar *  const pcre_uchar *
2030  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2031  {  {
2032  for (;;)  for (;;)
2033    {    {
2034    register int c = *code;    register pcre_uchar c = *code;
2035    
2036    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2037    
2038    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1357  for (;;) Line 2041  for (;;)
2041    
2042    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2043    
2044      /* Handle recursion */
2045    
2046      else if (c == OP_REVERSE)
2047        {
2048        if (number < 0) return (pcre_uchar *)code;
2049        code += PRIV(OP_lengths)[c];
2050        }
2051    
2052    /* Handle capturing bracket */    /* Handle capturing bracket */
2053    
2054    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2055               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2056      {      {
2057      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2058      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2059      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2060      }      }
2061    
2062    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2063    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2064    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2065      must add in its length. */
2066    
2067    else    else
2068      {      {
# Line 1390  for (;;) Line 2084  for (;;)
2084        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2085        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2086        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2087        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2088            code += 2;
2089          break;
2090    
2091          case OP_MARK:
2092          case OP_PRUNE_ARG:
2093          case OP_SKIP_ARG:
2094          case OP_THEN_ARG:
2095          code += code[1];
2096        break;        break;
2097        }        }
2098    
2099      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2100    
2101      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2102    
2103    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2104    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2105    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2106    
2107  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2108      if (utf8) switch(c)      if (utf) switch(c)
2109        {        {
2110        case OP_CHAR:        case OP_CHAR:
2111        case OP_CHARNC:        case OP_CHARI:
2112        case OP_EXACT:        case OP_EXACT:
2113          case OP_EXACTI:
2114        case OP_UPTO:        case OP_UPTO:
2115          case OP_UPTOI:
2116        case OP_MINUPTO:        case OP_MINUPTO:
2117          case OP_MINUPTOI:
2118        case OP_POSUPTO:        case OP_POSUPTO:
2119          case OP_POSUPTOI:
2120        case OP_STAR:        case OP_STAR:
2121          case OP_STARI:
2122        case OP_MINSTAR:        case OP_MINSTAR:
2123          case OP_MINSTARI:
2124        case OP_POSSTAR:        case OP_POSSTAR:
2125          case OP_POSSTARI:
2126        case OP_PLUS:        case OP_PLUS:
2127          case OP_PLUSI:
2128        case OP_MINPLUS:        case OP_MINPLUS:
2129          case OP_MINPLUSI:
2130        case OP_POSPLUS:        case OP_POSPLUS:
2131          case OP_POSPLUSI:
2132        case OP_QUERY:        case OP_QUERY:
2133          case OP_QUERYI:
2134        case OP_MINQUERY:        case OP_MINQUERY:
2135          case OP_MINQUERYI:
2136        case OP_POSQUERY:        case OP_POSQUERY:
2137        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2138          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2139        break;        break;
2140        }        }
2141    #else
2142        (void)(utf);  /* Keep compiler happy by referencing function argument */
2143  #endif  #endif
2144      }      }
2145    }    }
# Line 1439  instance of OP_RECURSE. Line 2156  instance of OP_RECURSE.
2156    
2157  Arguments:  Arguments:
2158    code        points to start of expression    code        points to start of expression
2159    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2160    
2161  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2162  */  */
2163    
2164  static const uschar *  static const pcre_uchar *
2165  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2166  {  {
2167  for (;;)  for (;;)
2168    {    {
2169    register int c = *code;    register pcre_uchar c = *code;
2170    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2171    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2172    
# Line 1461  for (;;) Line 2178  for (;;)
2178    
2179    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2180    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2181    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2182      must add in its length. */
2183    
2184    else    else
2185      {      {
# Line 1483  for (;;) Line 2201  for (;;)
2201        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2202        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2203        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2204        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2205            code += 2;
2206          break;
2207    
2208          case OP_MARK:
2209          case OP_PRUNE_ARG:
2210          case OP_SKIP_ARG:
2211          case OP_THEN_ARG:
2212          code += code[1];
2213        break;        break;
2214        }        }
2215    
2216      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2217    
2218      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2219    
2220      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2221      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2222      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2223    
2224  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2225      if (utf8) switch(c)      if (utf) switch(c)
2226        {        {
2227        case OP_CHAR:        case OP_CHAR:
2228        case OP_CHARNC:        case OP_CHARI:
2229          case OP_NOT:
2230          case OP_NOTI:
2231        case OP_EXACT:        case OP_EXACT:
2232          case OP_EXACTI:
2233          case OP_NOTEXACT:
2234          case OP_NOTEXACTI:
2235        case OP_UPTO:        case OP_UPTO:
2236          case OP_UPTOI:
2237          case OP_NOTUPTO:
2238          case OP_NOTUPTOI:
2239        case OP_MINUPTO:        case OP_MINUPTO:
2240          case OP_MINUPTOI:
2241          case OP_NOTMINUPTO:
2242          case OP_NOTMINUPTOI:
2243        case OP_POSUPTO:        case OP_POSUPTO:
2244          case OP_POSUPTOI:
2245          case OP_NOTPOSUPTO:
2246          case OP_NOTPOSUPTOI:
2247        case OP_STAR:        case OP_STAR:
2248          case OP_STARI:
2249          case OP_NOTSTAR:
2250          case OP_NOTSTARI:
2251        case OP_MINSTAR:        case OP_MINSTAR:
2252          case OP_MINSTARI:
2253          case OP_NOTMINSTAR:
2254          case OP_NOTMINSTARI:
2255        case OP_POSSTAR:        case OP_POSSTAR:
2256          case OP_POSSTARI:
2257          case OP_NOTPOSSTAR:
2258          case OP_NOTPOSSTARI:
2259        case OP_PLUS:        case OP_PLUS:
2260          case OP_PLUSI:
2261          case OP_NOTPLUS:
2262          case OP_NOTPLUSI:
2263        case OP_MINPLUS:        case OP_MINPLUS:
2264          case OP_MINPLUSI:
2265          case OP_NOTMINPLUS:
2266          case OP_NOTMINPLUSI:
2267        case OP_POSPLUS:        case OP_POSPLUS:
2268          case OP_POSPLUSI:
2269          case OP_NOTPOSPLUS:
2270          case OP_NOTPOSPLUSI:
2271        case OP_QUERY:        case OP_QUERY:
2272          case OP_QUERYI:
2273          case OP_NOTQUERY:
2274          case OP_NOTQUERYI:
2275        case OP_MINQUERY:        case OP_MINQUERY:
2276          case OP_MINQUERYI:
2277          case OP_NOTMINQUERY:
2278          case OP_NOTMINQUERYI:
2279        case OP_POSQUERY:        case OP_POSQUERY:
2280        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2281          case OP_NOTPOSQUERY:
2282          case OP_NOTPOSQUERYI:
2283          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2284        break;        break;
2285        }        }
2286    #else
2287        (void)(utf);  /* Keep compiler happy by referencing function argument */
2288  #endif  #endif
2289      }      }
2290    }    }
# Line 1538  bracket whose current branch will alread Line 2307  bracket whose current branch will alread
2307  Arguments:  Arguments:
2308    code        points to start of search    code        points to start of search
2309    endcode     points to where to stop    endcode     points to where to stop
2310    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2311      cd          contains pointers to tables etc.
2312      recurses    chain of recurse_check to catch mutual recursion
2313    
2314  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2315  */  */
2316    
2317    typedef struct recurse_check {
2318      struct recurse_check *prev;
2319      const pcre_uchar *group;
2320    } recurse_check;
2321    
2322  static BOOL  static BOOL
2323  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2324      BOOL utf, compile_data *cd, recurse_check *recurses)
2325  {  {
2326  register int c;  register pcre_uchar c;
2327  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2328    
2329    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2330       code < endcode;       code < endcode;
2331       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2332    {    {
2333    const uschar *ccode;    const pcre_uchar *ccode;
2334    
2335    c = *code;    c = *code;
2336    
# Line 1565  for (code = first_significant_code(code Line 2344  for (code = first_significant_code(code
2344      continue;      continue;
2345      }      }
2346    
2347      /* For a recursion/subroutine call, if its end has been reached, which
2348      implies a backward reference subroutine call, we can scan it. If it's a
2349      forward reference subroutine call, we can't. To detect forward reference
2350      we have to scan up the list that is kept in the workspace. This function is
2351      called only when doing the real compile, not during the pre-compile that
2352      measures the size of the compiled pattern. */
2353    
2354      if (c == OP_RECURSE)
2355        {
2356        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2357        BOOL empty_branch;
2358    
2359        /* Test for forward reference or uncompleted reference. This is disabled
2360        when called to scan a completed pattern by setting cd->start_workspace to
2361        NULL. */
2362    
2363        if (cd->start_workspace != NULL)
2364          {
2365          const pcre_uchar *tcode;
2366          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2367            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2368          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2369          }
2370    
2371        /* If we are scanning a completed pattern, there are no forward references
2372        and all groups are complete. We need to detect whether this is a recursive
2373        call, as otherwise there will be an infinite loop. If it is a recursion,
2374        just skip over it. Simple recursions are easily detected. For mutual
2375        recursions we keep a chain on the stack. */
2376    
2377        else
2378          {
2379          recurse_check *r = recurses;
2380          const pcre_uchar *endgroup = scode;
2381    
2382          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2383          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2384    
2385          for (r = recurses; r != NULL; r = r->prev)
2386            if (r->group == scode) break;
2387          if (r != NULL) continue;   /* Mutual recursion */
2388          }
2389    
2390        /* Completed reference; scan the referenced group, remembering it on the
2391        stack chain to detect mutual recursions. */
2392    
2393        empty_branch = FALSE;
2394        this_recurse.prev = recurses;
2395        this_recurse.group = scode;
2396    
2397        do
2398          {
2399          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2400            {
2401            empty_branch = TRUE;
2402            break;
2403            }
2404          scode += GET(scode, 1);
2405          }
2406        while (*scode == OP_ALT);
2407    
2408        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2409        continue;
2410        }
2411    
2412    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2413    
2414    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2415          c == OP_BRAPOSZERO)
2416        {
2417        code += PRIV(OP_lengths)[c];
2418        do code += GET(code, 1); while (*code == OP_ALT);
2419        c = *code;
2420        continue;
2421        }
2422    
2423      /* A nested group that is already marked as "could be empty" can just be
2424      skipped. */
2425    
2426      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2427          c == OP_SCBRA || c == OP_SCBRAPOS)
2428      {      {
     code += _pcre_OP_lengths[c];  
2429      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2430      c = *code;      c = *code;
2431      continue;      continue;
# Line 1577  for (code = first_significant_code(code Line 2433  for (code = first_significant_code(code
2433    
2434    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2435    
2436    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2437          c == OP_CBRA || c == OP_CBRAPOS ||
2438          c == OP_ONCE || c == OP_ONCE_NC ||
2439          c == OP_COND)
2440      {      {
2441      BOOL empty_branch;      BOOL empty_branch;
2442      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2443    
2444      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2445        empty branch, so just skip over the conditional, because it could be empty.
2446        Otherwise, scan the individual branches of the group. */
2447    
2448      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
2449        code += GET(code, 1);        code += GET(code, 1);
2450        else
2451          {
2452          empty_branch = FALSE;
2453          do
2454            {
2455            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2456              empty_branch = TRUE;
2457            code += GET(code, 1);
2458            }
2459          while (*code == OP_ALT);
2460          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2461        }        }
2462      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2463      c = *code;      c = *code;
2464      continue;      continue;
2465      }      }
# Line 1603  for (code = first_significant_code(code Line 2470  for (code = first_significant_code(code
2470      {      {
2471      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2472      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2473      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2474      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2475      here. */      here. */
2476    
2477  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2478      case OP_XCLASS:      case OP_XCLASS:
2479      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2480      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1615  for (code = first_significant_code(code Line 2482  for (code = first_significant_code(code
2482    
2483      case OP_CLASS:      case OP_CLASS:
2484      case OP_NCLASS:      case OP_NCLASS:
2485      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2486    
2487  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2488      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2489  #endif  #endif
2490    
# Line 1627  for (code = first_significant_code(code Line 2494  for (code = first_significant_code(code
2494        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2495        case OP_CRQUERY:        case OP_CRQUERY:
2496        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2497          case OP_CRPOSSTAR:
2498          case OP_CRPOSQUERY:
2499        break;        break;
2500    
2501        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2502        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2503        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2504          case OP_CRPOSPLUS:
2505        return FALSE;        return FALSE;
2506    
2507        case OP_CRRANGE:        case OP_CRRANGE:
2508        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2509          case OP_CRPOSRANGE:
2510        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2511        break;        break;
2512        }        }
2513      break;      break;
2514    
2515      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2516    
2517        case OP_ANY:
2518        case OP_ALLANY:
2519        case OP_ANYBYTE:
2520    
2521        case OP_PROP:
2522        case OP_NOTPROP:
2523        case OP_ANYNL:
2524    
2525        case OP_NOT_HSPACE:
2526        case OP_HSPACE:
2527        case OP_NOT_VSPACE:
2528        case OP_VSPACE:
2529        case OP_EXTUNI:
2530    
2531        case OP_NOT_DIGIT:
2532        case OP_DIGIT:
2533        case OP_NOT_WHITESPACE:
2534        case OP_WHITESPACE:
2535        case OP_NOT_WORDCHAR:
2536        case OP_WORDCHAR:
2537    
2538        case OP_CHAR:
2539        case OP_CHARI:
2540        case OP_NOT:
2541        case OP_NOTI:
2542    
2543        case OP_PLUS:
2544        case OP_PLUSI:
2545        case OP_MINPLUS:
2546        case OP_MINPLUSI:
2547    
2548        case OP_NOTPLUS:
2549        case OP_NOTPLUSI:
2550        case OP_NOTMINPLUS:
2551        case OP_NOTMINPLUSI:
2552    
2553        case OP_POSPLUS:
2554        case OP_POSPLUSI:
2555        case OP_NOTPOSPLUS:
2556        case OP_NOTPOSPLUSI:
2557    
2558        case OP_EXACT:
2559        case OP_EXACTI:
2560        case OP_NOTEXACT:
2561        case OP_NOTEXACTI:
2562    
2563        case OP_TYPEPLUS:
2564        case OP_TYPEMINPLUS:
2565        case OP_TYPEPOSPLUS:
2566        case OP_TYPEEXACT:
2567    
2568        return FALSE;
2569    
2570        /* These are going to continue, as they may be empty, but we have to
2571        fudge the length for the \p and \P cases. */
2572    
2573        case OP_TYPESTAR:
2574        case OP_TYPEMINSTAR:
2575        case OP_TYPEPOSSTAR:
2576        case OP_TYPEQUERY:
2577        case OP_TYPEMINQUERY:
2578        case OP_TYPEPOSQUERY:
2579        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2580        break;
2581    
2582        /* Same for these */
2583    
2584        case OP_TYPEUPTO:
2585        case OP_TYPEMINUPTO:
2586        case OP_TYPEPOSUPTO:
2587        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2588          code += 2;
2589        break;
2590    
2591        /* End of branch */
2592    
2593        case OP_KET:
2594        case OP_KETRMAX:
2595        case OP_KETRMIN:
2596        case OP_KETRPOS:
2597        case OP_ALT:
2598        return TRUE;
2599    
2600        /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2601        MINUPTO, and POSUPTO and their caseless and negative versions may be
2602        followed by a multibyte character. */
2603    
2604    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2605        case OP_STAR:
2606        case OP_STARI:
2607        case OP_NOTSTAR:
2608        case OP_NOTSTARI:
2609    
2610        case OP_MINSTAR:
2611        case OP_MINSTARI:
2612        case OP_NOTMINSTAR:
2613        case OP_NOTMINSTARI:
2614    
2615        case OP_POSSTAR:
2616        case OP_POSSTARI:
2617        case OP_NOTPOSSTAR:
2618        case OP_NOTPOSSTARI:
2619    
2620        case OP_QUERY:
2621        case OP_QUERYI:
2622        case OP_NOTQUERY:
2623        case OP_NOTQUERYI:
2624    
2625        case OP_MINQUERY:
2626        case OP_MINQUERYI:
2627        case OP_NOTMINQUERY:
2628        case OP_NOTMINQUERYI:
2629    
2630        case OP_POSQUERY:
2631        case OP_POSQUERYI:
2632        case OP_NOTPOSQUERY:
2633        case OP_NOTPOSQUERYI:
2634    
2635        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2636        break;
2637    
2638        case OP_UPTO:
2639        case OP_UPTOI:
2640        case OP_NOTUPTO:
2641        case OP_NOTUPTOI:
2642    
2643        case OP_MINUPTO:
2644        case OP_MINUPTOI:
2645        case OP_NOTMINUPTO:
2646        case OP_NOTMINUPTOI:
2647    
2648        case OP_POSUPTO:
2649        case OP_POSUPTOI:
2650        case OP_NOTPOSUPTO:
2651        case OP_NOTPOSUPTOI:
2652    
2653        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2654        break;
2655    #endif
2656    
2657        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2658        string. */
2659    
2660        case OP_MARK:
2661        case OP_PRUNE_ARG:
2662        case OP_SKIP_ARG:
2663        case OP_THEN_ARG:
2664        code += code[1];
2665        break;
2666    
2667        /* None of the remaining opcodes are required to match a character. */
2668    
2669        default:
2670        break;
2671        }
2672      }
2673    
2674    return TRUE;
2675    }
2676    
2677    
2678    
2679    /*************************************************
2680    *    Scan compiled regex for non-emptiness       *
2681    *************************************************/
2682    
2683    /* This function is called to check for left recursive calls. We want to check
2684    the current branch of the current pattern to see if it could match the empty
2685    string. If it could, we must look outwards for branches at other levels,
2686    stopping when we pass beyond the bracket which is the subject of the recursion.
2687    This function is called only during the real compile, not during the
2688    pre-compile.
2689    
2690    Arguments:
2691      code        points to start of the recursion
2692      endcode     points to where to stop (current RECURSE item)
2693      bcptr       points to the chain of current (unclosed) branch starts
2694      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695      cd          pointers to tables etc
2696    
2697    Returns:      TRUE if what is matched could be empty
2698    */
2699    
2700    static BOOL
2701    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2702      branch_chain *bcptr, BOOL utf, compile_data *cd)
2703    {
2704    while (bcptr != NULL && bcptr->current_branch >= code)
2705      {
2706      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2707        return FALSE;
2708      bcptr = bcptr->outer;
2709      }
2710    return TRUE;
2711    }
2712    
2713    
2714    
2715    /*************************************************
2716    *        Base opcode of repeated opcodes         *
2717    *************************************************/
2718    
2719    /* Returns the base opcode for repeated single character type opcodes. If the
2720    opcode is not a repeated character type, it returns with the original value.
2721    
2722    Arguments:  c opcode
2723    Returns:    base opcode for the type
2724    */
2725    
2726    static pcre_uchar
2727    get_repeat_base(pcre_uchar c)
2728    {
2729    return (c > OP_TYPEPOSUPTO)? c :
2730           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2731           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2732           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2733           (c >= OP_STARI)?      OP_STARI :
2734                                 OP_STAR;
2735    }
2736    
2737    
2738    
2739    #ifdef SUPPORT_UCP
2740    /*************************************************
2741    *        Check a character and a property        *
2742    *************************************************/
2743    
2744    /* This function is called by check_auto_possessive() when a property item
2745    is adjacent to a fixed character.
2746    
2747    Arguments:
2748      c            the character
2749      ptype        the property type
2750      pdata        the data for the type
2751      negated      TRUE if it's a negated property (\P or \p{^)
2752    
2753    Returns:       TRUE if auto-possessifying is OK
2754    */
2755    
2756    static BOOL
2757    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2758      BOOL negated)
2759    {
2760    const pcre_uint32 *p;
2761    const ucd_record *prop = GET_UCD(c);
2762    
2763    switch(ptype)
2764      {
2765      case PT_LAMP:
2766      return (prop->chartype == ucp_Lu ||
2767              prop->chartype == ucp_Ll ||
2768              prop->chartype == ucp_Lt) == negated;
2769    
2770      case PT_GC:
2771      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2772    
2773      case PT_PC:
2774      return (pdata == prop->chartype) == negated;
2775    
2776      case PT_SC:
2777      return (pdata == prop->script) == negated;
2778    
2779      /* These are specials */
2780    
2781      case PT_ALNUM:
2782      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2783              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2784    
2785      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2786      means that Perl space and POSIX space are now identical. PCRE was changed
2787      at release 8.34. */
2788    
2789      case PT_SPACE:    /* Perl space */
2790      case PT_PXSPACE:  /* POSIX space */
2791      switch(c)
2792        {
2793        HSPACE_CASES:
2794        VSPACE_CASES:
2795        return negated;
2796    
2797        default:
2798        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2799        }
2800      break;  /* Control never reaches here */
2801    
2802      case PT_WORD:
2803      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2804              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2805              c == CHAR_UNDERSCORE) == negated;
2806    
2807      case PT_CLIST:
2808      p = PRIV(ucd_caseless_sets) + prop->caseset;
2809      for (;;)
2810        {
2811        if (c < *p) return !negated;
2812        if (c == *p++) return negated;
2813        }
2814      break;  /* Control never reaches here */
2815      }
2816    
2817    return FALSE;
2818    }
2819    #endif  /* SUPPORT_UCP */
2820    
2821    
2822    
2823    /*************************************************
2824    *        Fill the character property list        *
2825    *************************************************/
2826    
2827    /* Checks whether the code points to an opcode that can take part in auto-
2828    possessification, and if so, fills a list with its properties.
2829    
2830    Arguments:
2831      code        points to start of expression
2832      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2833      fcc         points to case-flipping table
2834      list        points to output list
2835                  list[0] will be filled with the opcode
2836                  list[1] will be non-zero if this opcode
2837                    can match an empty character string
2838                  list[2..7] depends on the opcode
2839    
2840    Returns:      points to the start of the next opcode if *code is accepted
2841                  NULL if *code is not accepted
2842    */
2843    
2844    static const pcre_uchar *
2845    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2846      const pcre_uint8 *fcc, pcre_uint32 *list)
2847    {
2848    pcre_uchar c = *code;
2849    const pcre_uchar *end;
2850    const pcre_uint32 *clist_src;
2851    pcre_uint32 *clist_dest;
2852    pcre_uint32 chr;
2853    pcre_uchar base;
2854    
2855    list[0] = c;
2856    list[1] = FALSE;
2857    code++;
2858    
2859    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2860      {
2861      base = get_repeat_base(c);
2862      c -= (base - OP_STAR);
2863    
2864      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2865        code += IMM2_SIZE;
2866    
2867      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2868    
2869      switch(base)
2870        {
2871        case OP_STAR:
2872        list[0] = OP_CHAR;
2873        break;
2874    
2875        case OP_STARI:
2876        list[0] = OP_CHARI;
2877        break;
2878    
2879        case OP_NOTSTAR:
2880        list[0] = OP_NOT;
2881        break;
2882    
2883        case OP_NOTSTARI:
2884        list[0] = OP_NOTI;
2885        break;
2886    
2887        case OP_TYPESTAR:
2888        list[0] = *code;
2889        code++;
2890        break;
2891        }
2892      c = list[0];
2893      }
2894    
2895    switch(c)
2896      {
2897      case OP_NOT_DIGIT:
2898      case OP_DIGIT:
2899      case OP_NOT_WHITESPACE:
2900      case OP_WHITESPACE:
2901      case OP_NOT_WORDCHAR:
2902      case OP_WORDCHAR:
2903      case OP_ANY:
2904      case OP_ALLANY:
2905      case OP_ANYNL:
2906      case OP_NOT_HSPACE:
2907      case OP_HSPACE:
2908      case OP_NOT_VSPACE:
2909      case OP_VSPACE:
2910      case OP_EXTUNI:
2911      case OP_EODN:
2912      case OP_EOD:
2913      case OP_DOLL:
2914      case OP_DOLLM:
2915      return code;
2916    
2917      case OP_CHAR:
2918      case OP_NOT:
2919      GETCHARINCTEST(chr, code);
2920      list[2] = chr;
2921      list[3] = NOTACHAR;
2922      return code;
2923    
2924      case OP_CHARI:
2925      case OP_NOTI:
2926      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2927      GETCHARINCTEST(chr, code);
2928      list[2] = chr;
2929    
2930    #ifdef SUPPORT_UCP
2931      if (chr < 128 || (chr < 256 && !utf))
2932        list[3] = fcc[chr];
2933      else
2934        list[3] = UCD_OTHERCASE(chr);
2935    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2936      list[3] = (chr < 256) ? fcc[chr] : chr;
2937    #else
2938      list[3] = fcc[chr];
2939    #endif
2940    
2941      /* The othercase might be the same value. */
2942    
2943      if (chr == list[3])
2944        list[3] = NOTACHAR;
2945      else
2946        list[4] = NOTACHAR;
2947      return code;
2948    
2949    #ifdef SUPPORT_UCP
2950      case OP_PROP:
2951      case OP_NOTPROP:
2952      if (code[0] != PT_CLIST)
2953        {
2954        list[2] = code[0];
2955        list[3] = code[1];
2956        return code + 2;
2957        }
2958    
2959      /* Convert only if we have enough space. */
2960    
2961      clist_src = PRIV(ucd_caseless_sets) + code[1];
2962      clist_dest = list + 2;
2963      code += 2;
2964    
2965      do {
2966         if (clist_dest >= list + 8)
2967           {
2968           /* Early return if there is not enough space. This should never
2969           happen, since all clists are shorter than 5 character now. */
2970           list[2] = code[0];
2971           list[3] = code[1];
2972           return code;
2973           }
2974         *clist_dest++ = *clist_src;
2975         }
2976      while(*clist_src++ != NOTACHAR);
2977    
2978      /* All characters are stored. The terminating NOTACHAR
2979      is copied form the clist itself. */
2980    
2981      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2982      return code;
2983    #endif
2984    
2985      case OP_NCLASS:
2986      case OP_CLASS:
2987    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2988      case OP_XCLASS:
2989      if (c == OP_XCLASS)
2990        end = code + GET(code, 0) - 1;
2991      else
2992    #endif
2993        end = code + 32 / sizeof(pcre_uchar);
2994    
2995      switch(*end)
2996        {
2997        case OP_CRSTAR:
2998        case OP_CRMINSTAR:
2999        case OP_CRQUERY:
3000        case OP_CRMINQUERY:
3001        case OP_CRPOSSTAR:
3002        case OP_CRPOSQUERY:
3003        list[1] = TRUE;
3004        end++;
3005        break;
3006    
3007        case OP_CRPLUS:
3008        case OP_CRMINPLUS:
3009        case OP_CRPOSPLUS:
3010        end++;
3011        break;
3012    
3013        case OP_CRRANGE:
3014        case OP_CRMINRANGE:
3015        case OP_CRPOSRANGE:
3016        list[1] = (GET2(end, 1) == 0);
3017        end += 1 + 2 * IMM2_SIZE;
3018        break;
3019        }
3020      list[2] = end - code;
3021      return end;
3022      }
3023    return NULL;    /* Opcode not accepted */
3024    }
3025    
3026    
3027    
3028    /*************************************************
3029    *    Scan further character sets for match       *
3030    *************************************************/
3031    
3032    /* Checks whether the base and the current opcode have a common character, in
3033    which case the base cannot be possessified.
3034    
3035    Arguments:
3036      code        points to the byte code
3037      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3038      cd          static compile data
3039      base_list   the data list of the base opcode
3040    
3041    Returns:      TRUE if the auto-possessification is possible
3042    */
3043    
3044    static BOOL
3045    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3046      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3047    {
3048    pcre_uchar c;
3049    pcre_uint32 list[8];
3050    const pcre_uint32 *chr_ptr;
3051    const pcre_uint32 *ochr_ptr;
3052    const pcre_uint32 *list_ptr;
3053    const pcre_uchar *next_code;
3054    const pcre_uint8 *class_bitset;
3055    const pcre_uint32 *set1, *set2, *set_end;
3056    pcre_uint32 chr;
3057    BOOL accepted, invert_bits;
3058    
3059    /* Note: the base_list[1] contains whether the current opcode has greedy
3060    (represented by a non-zero value) quantifier. This is a different from
3061    other character type lists, which stores here that the character iterator
3062    matches to an empty string (also represented by a non-zero value). */
3063    
3064    for(;;)
3065      {
3066      /* All operations move the code pointer forward.
3067      Therefore infinite recursions are not possible. */
3068    
3069      c = *code;
3070    
3071      /* Skip over callouts */
3072    
3073      if (c == OP_CALLOUT)
3074        {
3075        code += PRIV(OP_lengths)[c];
3076        continue;
3077        }
3078    
3079      if (c == OP_ALT)
3080        {
3081        do code += GET(code, 1); while (*code == OP_ALT);
3082        c = *code;
3083        }
3084    
3085      switch(c)
3086        {
3087        case OP_END:
3088        case OP_KETRPOS:
3089        /* TRUE only in greedy case. The non-greedy case could be replaced by
3090        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3091        uses more memory, which we cannot get at this stage.) */
3092    
3093        return base_list[1] != 0;
3094    
3095        case OP_KET:
3096        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3097        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3098        cannot be converted to a possessive form. */
3099    
3100        if (base_list[1] == 0) return FALSE;
3101    
3102        switch(*(code - GET(code, 1)))
3103          {
3104          case OP_ASSERT:
3105          case OP_ASSERT_NOT:
3106          case OP_ASSERTBACK:
3107          case OP_ASSERTBACK_NOT:
3108          case OP_ONCE:
3109          case OP_ONCE_NC:
3110          /* Atomic sub-patterns and assertions can always auto-possessify their
3111          last iterator. */
3112          return TRUE;
3113          }
3114    
3115        code += PRIV(OP_lengths)[c];
3116        continue;
3117    
3118        case OP_ONCE:
3119        case OP_ONCE_NC:
3120        case OP_BRA:
3121        case OP_CBRA:
3122        next_code = code + GET(code, 1);
3123        code += PRIV(OP_lengths)[c];
3124    
3125        while (*next_code == OP_ALT)
3126          {
3127          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3128          code = next_code + 1 + LINK_SIZE;
3129          next_code += GET(next_code, 1);
3130          }
3131        continue;
3132    
3133        case OP_BRAZERO:
3134        case OP_BRAMINZERO:
3135    
3136        next_code = code + 1;
3137        if (*next_code != OP_BRA && *next_code != OP_CBRA
3138            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3139    
3140        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3141    
3142        /* The bracket content will be checked by the
3143        OP_BRA/OP_CBRA case above. */
3144        next_code += 1 + LINK_SIZE;
3145        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3146          return FALSE;
3147    
3148        code += PRIV(OP_lengths)[c];
3149        continue;
3150        }
3151    
3152      /* Check for a supported opcode, and load its properties. */
3153    
3154      code = get_chr_property_list(code, utf, cd->fcc, list);
3155      if (code == NULL) return FALSE;    /* Unsupported */
3156    
3157      /* If either opcode is a small character list, set pointers for comparing
3158      characters from that list with another list, or with a property. */
3159    
3160      if (base_list[0] == OP_CHAR)
3161        {
3162        chr_ptr = base_list + 2;
3163        list_ptr = list;
3164        }
3165      else if (list[0] == OP_CHAR)
3166        {
3167        chr_ptr = list + 2;
3168        list_ptr = base_list;
3169        }
3170    
3171      /* Character bitsets can also be compared to certain opcodes. */
3172    
3173      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3174    #ifdef COMPILE_PCRE8
3175          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3176          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3177    #endif
3178          )
3179        {
3180    #ifdef COMPILE_PCRE8
3181        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3182    #else
3183        if (base_list[0] == OP_CLASS)
3184    #endif
3185          {
3186          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3187          list_ptr = list;
3188          }
3189        else
3190          {
3191          set1 = (pcre_uint32 *)(code - list[2]);
3192          list_ptr = base_list;
3193          }
3194    
3195        invert_bits = FALSE;
3196        switch(list_ptr[0])
3197          {
3198          case OP_CLASS:
3199          case OP_NCLASS:
3200          set2 = (pcre_uint32 *)
3201            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3202          break;
3203    
3204          /* OP_XCLASS cannot be supported here, because its bitset
3205          is not necessarily complete. E.g: [a-\0x{200}] is stored
3206          as a character range, and the appropriate bits are not set. */
3207    
3208          case OP_NOT_DIGIT:
3209            invert_bits = TRUE;
3210            /* Fall through */
3211          case OP_DIGIT:
3212            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3213            break;
3214    
3215          case OP_NOT_WHITESPACE:
3216            invert_bits = TRUE;
3217            /* Fall through */
3218          case OP_WHITESPACE:
3219            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3220            break;
3221    
3222          case OP_NOT_WORDCHAR:
3223            invert_bits = TRUE;
3224            /* Fall through */
3225          case OP_WORDCHAR:
3226            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3227            break;
3228    
3229          default:
3230          return FALSE;
3231          }
3232    
3233        /* Compare 4 bytes to improve speed. */
3234        set_end = set1 + (32 / 4);
3235        if (invert_bits)
3236          {
3237          do
3238            {
3239            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3240            }
3241          while (set1 < set_end);
3242          }
3243        else
3244          {
3245          do
3246            {
3247            if ((*set1++ & *set2++) != 0) return FALSE;
3248            }
3249          while (set1 < set_end);
3250          }
3251    
3252        if (list[1] == 0) return TRUE;
3253        /* Might be an empty repeat. */
3254        continue;
3255        }
3256    
3257      /* Some property combinations also acceptable. Unicode property opcodes are
3258      processed specially; the rest can be handled with a lookup table. */
3259    
3260      else
3261        {
3262        pcre_uint32 leftop, rightop;
3263    
3264        leftop = base_list[0];
3265        rightop = list[0];
3266    
3267    #ifdef SUPPORT_UCP
3268        accepted = FALSE; /* Always set in non-unicode case. */
3269        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3270          {
3271          if (rightop == OP_EOD)
3272            accepted = TRUE;
3273          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3274            {
3275            int n;
3276            const pcre_uint8 *p;
3277            BOOL same = leftop == rightop;
3278            BOOL lisprop = leftop == OP_PROP;
3279            BOOL risprop = rightop == OP_PROP;
3280            BOOL bothprop = lisprop && risprop;
3281    
3282            /* There's a table that specifies how each combination is to be
3283            processed:
3284              0   Always return FALSE (never auto-possessify)
3285              1   Character groups are distinct (possessify if both are OP_PROP)
3286              2   Check character categories in the same group (general or particular)
3287              3   Return TRUE if the two opcodes are not the same
3288              ... see comments below
3289            */
3290    
3291            n = propposstab[base_list[2]][list[2]];
3292            switch(n)
3293              {
3294              case 0: break;
3295              case 1: accepted = bothprop; break;
3296              case 2: accepted = (base_list[3] == list[3]) != same; break;
3297              case 3: accepted = !same; break;
3298    
3299              case 4:  /* Left general category, right particular category */
3300              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3301              break;
3302    
3303              case 5:  /* Right general category, left particular category */
3304              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3305              break;
3306    
3307              /* This code is logically tricky. Think hard before fiddling with it.
3308              The posspropstab table has four entries per row. Each row relates to
3309              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3310              Only WORD actually needs all four entries, but using repeats for the
3311              others means they can all use the same code below.
3312    
3313              The first two entries in each row are Unicode general categories, and
3314              apply always, because all the characters they include are part of the
3315              PCRE character set. The third and fourth entries are a general and a
3316              particular category, respectively, that include one or more relevant
3317              characters. One or the other is used, depending on whether the check
3318              is for a general or a particular category. However, in both cases the
3319              category contains more characters than the specials that are defined
3320              for the property being tested against. Therefore, it cannot be used
3321              in a NOTPROP case.
3322    
3323              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3324              Underscore is covered by ucp_P or ucp_Po. */
3325    
3326              case 6:  /* Left alphanum vs right general category */
3327              case 7:  /* Left space vs right general category */
3328              case 8:  /* Left word vs right general category */
3329              p = posspropstab[n-6];
3330              accepted = risprop && lisprop ==
3331                (list[3] != p[0] &&
3332                 list[3] != p[1] &&
3333                (list[3] != p[2] || !lisprop));
3334              break;
3335    
3336              case 9:   /* Right alphanum vs left general category */
3337              case 10:  /* Right space vs left general category */
3338              case 11:  /* Right word vs left general category */
3339              p = posspropstab[n-9];
3340              accepted = lisprop && risprop ==
3341                (base_list[3] != p[0] &&
3342                 base_list[3] != p[1] &&
3343                (base_list[3] != p[2] || !risprop));
3344              break;
3345    
3346              case 12:  /* Left alphanum vs right particular category */
3347              case 13:  /* Left space vs right particular category */
3348              case 14:  /* Left word vs right particular category */
3349              p = posspropstab[n-12];
3350              accepted = risprop && lisprop ==
3351                (catposstab[p[0]][list[3]] &&
3352                 catposstab[p[1]][list[3]] &&
3353                (list[3] != p[3] || !lisprop));
3354              break;
3355    
3356              case 15:  /* Right alphanum vs left particular category */
3357              case 16:  /* Right space vs left particular category */
3358              case 17:  /* Right word vs left particular category */
3359              p = posspropstab[n-15];
3360              accepted = lisprop && risprop ==
3361                (catposstab[p[0]][base_list[3]] &&
3362                 catposstab[p[1]][base_list[3]] &&
3363                (base_list[3] != p[3] || !risprop));
3364              break;
3365              }
3366            }
3367          }
3368    
3369        else
3370    #endif  /* SUPPORT_UCP */
3371    
3372        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3373               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3374               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3375    
3376        if (!accepted)
3377          return FALSE;
3378    
3379        if (list[1] == 0) return TRUE;
3380        /* Might be an empty repeat. */
3381        continue;
3382        }
3383    
3384      /* Control reaches here only if one of the items is a small character list.
3385      All characters are checked against the other side. */
3386    
3387      do
3388        {
3389        chr = *chr_ptr;
3390    
3391        switch(list_ptr[0])
3392          {
3393          case OP_CHAR:
3394          ochr_ptr = list_ptr + 2;
3395          do
3396            {
3397            if (chr == *ochr_ptr) return FALSE;
3398            ochr_ptr++;
3399            }
3400          while(*ochr_ptr != NOTACHAR);
3401          break;
3402    
3403          case OP_NOT:
3404          ochr_ptr = list_ptr + 2;
3405          do
3406            {
3407            if (chr == *ochr_ptr)
3408              break;
3409            ochr_ptr++;
3410            }
3411          while(*ochr_ptr != NOTACHAR);
3412          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3413          break;
3414    
3415          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3416          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3417    
3418          case OP_DIGIT:
3419          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3420          break;
3421    
3422          case OP_NOT_DIGIT:
3423          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3424          break;
3425    
3426          case OP_WHITESPACE:
3427          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3428          break;
3429    
3430          case OP_NOT_WHITESPACE:
3431          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3432          break;
3433    
3434          case OP_WORDCHAR:
3435          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3436          break;
3437    
3438          case OP_NOT_WORDCHAR:
3439          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3440          break;
3441    
3442          case OP_HSPACE:
3443          switch(chr)
3444            {
3445            HSPACE_CASES: return FALSE;
3446            default: break;
3447            }
3448          break;
3449    
3450          case OP_NOT_HSPACE:
3451          switch(chr)
3452            {
3453            HSPACE_CASES: break;
3454            default: return FALSE;
3455            }
3456          break;
3457    
3458          case OP_ANYNL:
3459          case OP_VSPACE:
3460          switch(chr)
3461            {
3462            VSPACE_CASES: return FALSE;
3463            default: break;
3464            }
3465          break;
3466    
3467          case OP_NOT_VSPACE:
3468          switch(chr)
3469            {
3470            VSPACE_CASES: break;
3471            default: return FALSE;
3472            }
3473          break;
3474    
3475          case OP_DOLL:
3476          case OP_EODN:
3477          switch (chr)
3478            {
3479            case CHAR_CR:
3480            case CHAR_LF:
3481            case CHAR_VT:
3482            case CHAR_FF:
3483            case CHAR_NEL:
3484    #ifndef EBCDIC
3485            case 0x2028:
3486            case 0x2029:
3487    #endif  /* Not EBCDIC */
3488            return FALSE;
3489            }
3490          break;
3491    
3492          case OP_EOD:    /* Can always possessify before \z */
3493          break;
3494    
3495    #ifdef SUPPORT_UCP
3496          case OP_PROP:
3497          case OP_NOTPROP:
3498          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3499                list_ptr[0] == OP_NOTPROP))
3500            return FALSE;
3501          break;
3502    #endif
3503    
3504          case OP_NCLASS:
3505          if (chr > 255) return FALSE;
3506          /* Fall through */
3507    
3508          case OP_CLASS:
3509          if (chr > 255) break;
3510          class_bitset = (pcre_uint8 *)
3511            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3512          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3513          break;
3514    
3515    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3516          case OP_XCLASS:
3517          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3518              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3519          break;
3520    #endif
3521    
3522          default:
3523          return FALSE;
3524          }
3525    
3526        chr_ptr++;
3527        }
3528      while(*chr_ptr != NOTACHAR);
3529    
3530      /* At least one character must be matched from this opcode. */
3531    
3532      if (list[1] == 0) return TRUE;
3533      }
3534    
3535    return FALSE;
3536    }
3537    
3538    
3539    
3540    /*************************************************
3541    *    Scan compiled regex for auto-possession     *
3542    *************************************************/
3543    
3544    /* Replaces single character iterations with their possessive alternatives
3545    if appropriate. This function modifies the compiled opcode!
3546    
3547    Arguments:
3548      code        points to start of the byte code
3549      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3550      cd          static compile data
3551    
3552    Returns:      nothing
3553    */
3554    
3555    static void
3556    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3557    {
3558    register pcre_uchar c;
3559    const pcre_uchar *end;
3560    pcre_uchar *repeat_opcode;
3561    pcre_uint32 list[8];
3562    
3563    for (;;)
3564      {
3565      c = *code;
3566    
3567      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3568        {
3569        c -= get_repeat_base(c) - OP_STAR;
3570        end = (c <= OP_MINUPTO) ?
3571          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3572        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3573    
3574        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3575          {
3576          switch(c)
3577            {
3578            case OP_STAR:
3579            *code += OP_POSSTAR - OP_STAR;
3580            break;
3581    
3582            case OP_MINSTAR:
3583            *code += OP_POSSTAR - OP_MINSTAR;
3584            break;
3585    
3586            case OP_PLUS:
3587            *code += OP_POSPLUS - OP_PLUS;
3588            break;
3589    
3590            case OP_MINPLUS:
3591            *code += OP_POSPLUS - OP_MINPLUS;
3592            break;
3593    
3594            case OP_QUERY:
3595            *code += OP_POSQUERY - OP_QUERY;
3596            break;
3597    
3598            case OP_MINQUERY:
3599            *code += OP_POSQUERY - OP_MINQUERY;
3600            break;
3601    
3602            case OP_UPTO:
3603            *code += OP_POSUPTO - OP_UPTO;
3604            break;
3605    
3606            case OP_MINUPTO:
3607            *code += OP_MINUPTO - OP_UPTO;
3608            break;
3609            }
3610          }
3611        c = *code;
3612        }
3613      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3614        {
3615    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3616        if (c == OP_XCLASS)
3617          repeat_opcode = code + GET(code, 1);
3618        else
3619    #endif
3620          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3621    
3622        c = *repeat_opcode;
3623        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3624          {