/[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 1446 by ph10, Mon Jan 13 17:00:49 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287      STRING_word0  STRING_xdigit;
288    
289  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 212  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325    Unicode property escapes. */
326    
327    #ifdef SUPPORT_UCP
328    static const pcre_uchar string_PNd[]  = {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pNd[]  = {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_PXsp[] = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXsp[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXwd[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXwd[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *substitutes[] = {
348      string_PNd,           /* \D */
349      string_pNd,           /* \d */
350      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352      string_PXwd,          /* \W */
353      string_pXwd           /* \w */
354    };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362    static const pcre_uchar string_pL[] =   {
363      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    static const pcre_uchar string_pLl[] =  {
366      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368    static const pcre_uchar string_pLu[] =  {
369      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pXan[] = {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_h[] =    {
375      CHAR_BACKSLASH, CHAR_h, '\0' };
376    static const pcre_uchar string_pXps[] = {
377      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379    static const pcre_uchar string_PL[] =   {
380      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382    static const pcre_uchar string_PLl[] =  {
383      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PLu[] =  {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PXan[] = {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_H[] =    {
392      CHAR_BACKSLASH, CHAR_H, '\0' };
393    static const pcre_uchar string_PXps[] = {
394      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396    
397    static const pcre_uchar *posix_substitutes[] = {
398      string_pL,            /* alpha */
399      string_pLl,           /* lower */
400      string_pLu,           /* upper */
401      string_pXan,          /* alnum */
402      NULL,                 /* ascii */
403      string_h,             /* blank */
404      NULL,                 /* cntrl */
405      string_pNd,           /* digit */
406      NULL,                 /* graph */
407      NULL,                 /* print */
408      NULL,                 /* punct */
409      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411      NULL,                 /* xdigit */
412      /* Negated cases */
413      string_PL,            /* ^alpha */
414      string_PLl,           /* ^lower */
415      string_PLu,           /* ^upper */
416      string_PXan,          /* ^alnum */
417      NULL,                 /* ^ascii */
418      string_H,             /* ^blank */
419      NULL,                 /* ^cntrl */
420      string_PNd,           /* ^digit */
421      NULL,                 /* ^graph */
422      NULL,                 /* ^print */
423      NULL,                 /* ^punct */
424      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426      NULL                  /* ^xdigit */
427    };
428    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429    #endif
430    
431  #define STRING(a)  # a  #define STRING(a)  # a
432  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 439  the number of relocations needed when a
439  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
440  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
441  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
442  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
443    
444    Each substring ends with \0 to insert a null character. This includes the final
445    substring, so that the whole string ends with \0\0, which can be detected when
446    counting through. */
447    
448  static const char error_texts[] =  static const char error_texts[] =
449    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 484  static const char error_texts[] =
484    /* 30 */    /* 30 */
485    "unknown POSIX class name\0"    "unknown POSIX class name\0"
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
493    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494    "number after (?C is > 255\0"    "number after (?C is > 255\0"
495    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
496    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 507  static const char error_texts[] =
507    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508    /* 50 */    /* 50 */
509    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
510    "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"
511    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
512    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
513    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
514    /* 55 */    /* 55 */
515    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
516    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
517    "\\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"
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+";    "digit expected after (?+\0"
525      "] is an invalid data character in JavaScript compatibility mode\0"
526      /* 65 */
527      "different names for subpatterns of the same number are not allowed\0"
528      "(*MARK) must have an argument\0"
529      "this version of PCRE is not compiled with Unicode property support\0"
530      "\\c must be followed by an ASCII character\0"
531      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532      /* 70 */
533      "internal error: unknown opcode in find_fixedlength()\0"
534      "\\N is not supported in a class\0"
535      "too many forward references\0"
536      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537      "invalid UTF-16 string\0"
538      /* 75 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      ;
551    
552  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
553  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 565  For convenience, we use the same bit def
565    
566  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
567    
568  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
569  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
570    into a subtraction and unsigned comparison). */
571    
572    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
573    
574    #ifndef EBCDIC
575    
576    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
577    UTF-8 mode. */
578    
579    static const pcre_uint8 digitab[] =
580    {    {
581    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
582    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 611  static const unsigned char digitab[] =
611    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
612    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
613    
614  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
615  static const unsigned char digitab[] =  
616    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
617    
618    static const pcre_uint8 digitab[] =
619    {    {
620    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
621    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 650  static const unsigned char digitab[] =
650    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
651    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
652    
653  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
654    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
655    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
656    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 686  static const unsigned char ebcdic_charta
686  #endif  #endif
687    
688    
689  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
690    between adjacent character-type opcodes. The left-hand (repeated) opcode is
691    used to select the row, and the right-hand opcode is use to select the column.
692    A value of 1 means that auto-possessification is OK. For example, the second
693    value in the first row means that \D+\d can be turned into \D++\d.
694    
695    The Unicode property types (\P and \p) have to be present to fill out the table
696    because of what their opcode values are, but the table values should always be
697    zero because property types are handled separately in the code. The last four
698    columns apply to items that cannot be repeated, so there is no need to have
699    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701    
702    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704    
705    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
707      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
708      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
709      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
710      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
711      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
712      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
713      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
714      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
718      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
720      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
724    };
725    
726    
727  static BOOL  /* This table is used to check whether auto-possessification is possible
728    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729      int *, int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
730    opcode is used to select the column. The values are as follows:
731    
732      0   Always return FALSE (never auto-possessify)
733      1   Character groups are distinct (possessify if both are OP_PROP)
734      2   Check character categories in the same group (general or particular)
735      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736    
737      4   Check left general category vs right particular category
738      5   Check right general category vs left particular category
739    
740      6   Left alphanum vs right general category
741      7   Left space vs right general category
742      8   Left word vs right general category
743    
744      9   Right alphanum vs left general category
745     10   Right space vs left general category
746     11   Right word vs left general category
747    
748     12   Left alphanum vs right particular category
749     13   Left space vs right particular category
750     14   Left word vs right particular category
751    
752     15   Right alphanum vs left particular category
753     16   Right space vs left particular category
754     17   Right word vs left particular category
755    */
756    
757    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
760      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
761      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
762      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
763      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
764      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
765      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
766      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
767      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
768      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
769      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
770    };
771    
772    /* This table is used to check whether auto-possessification is possible
773    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774    specifies a general category and the other specifies a particular category. The
775    row is selected by the general category and the column by the particular
776    category. The value is 1 if the particular category is not part of the general
777    category. */
778    
779    static const pcre_uint8 catposstab[7][30] = {
780    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
782      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
783      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
784      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
785      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
788    };
789    
790    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791    a general or particular category. The properties in each row are those
792    that apply to the character set in question. Duplication means that a little
793    unnecessary work is done when checking, but this keeps things much simpler
794    because they can all use the same code. For more details see the comment where
795    this table is used.
796    
797    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798    "space", but from Perl 5.18 it's included, so both categories are treated the
799    same here. */
800    
801    static const pcre_uint8 posspropstab[3][4] = {
802      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
803      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
804      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
805    };
806    
807    /* This table is used when converting repeating opcodes into possessified
808    versions as a result of an explicit possessive quantifier such as ++. A zero
809    value means there is no possessified version - in those cases the item in
810    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811    because all relevant opcodes are less than that. */
812    
813    static const pcre_uint8 opcode_possessify[] = {
814      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
815      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
816    
817      0,                       /* NOTI */
818      OP_POSSTAR, 0,           /* STAR, MINSTAR */
819      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
820      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
821      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
822      0,                       /* EXACT */
823      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_POSSTARI, 0,          /* STARI, MINSTARI */
826      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
827      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
828      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
829      0,                       /* EXACTI */
830      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
833      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
834      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
835      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
836      0,                       /* NOTEXACT */
837      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
840      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
841      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
842      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
843      0,                       /* NOTEXACTI */
844      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845    
846      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
847      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
848      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
849      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
850      0,                       /* TYPEEXACT */
851      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852    
853      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
854      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
855      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
856      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
857      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858    
859      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
860      0, 0,                    /* REF, REFI */
861      0, 0,                    /* DNREF, DNREFI */
862      0, 0                     /* RECURSE, CALLOUT */
863    };
864    
865    
866    
# Line 454  static const char * Line 881  static const char *
881  find_error_text(int n)  find_error_text(int n)
882  {  {
883  const char *s = error_texts;  const char *s = error_texts;
884  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
885      {
886      while (*s++ != CHAR_NULL) {};
887      if (*s == CHAR_NULL) return "Error text not found (please report)";
888      }
889  return s;  return s;
890  }  }
891    
892    
893    
894    /*************************************************
895    *           Expand the workspace                 *
896    *************************************************/
897    
898    /* This function is called during the second compiling phase, if the number of
899    forward references fills the existing workspace, which is originally a block on
900    the stack. A larger block is obtained from malloc() unless the ultimate limit
901    has been reached or the increase will be rather small.
902    
903    Argument: pointer to the compile data block
904    Returns:  0 if all went well, else an error number
905    */
906    
907    static int
908    expand_workspace(compile_data *cd)
909    {
910    pcre_uchar *newspace;
911    int newsize = cd->workspace_size * 2;
912    
913    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
914    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
915        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
916     return ERR72;
917    
918    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
919    if (newspace == NULL) return ERR21;
920    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
921    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
922    if (cd->workspace_size > COMPILE_WORK_SIZE)
923      (PUBL(free))((void *)cd->start_workspace);
924    cd->start_workspace = newspace;
925    cd->workspace_size = newsize;
926    return 0;
927    }
928    
929    
930    
931    /*************************************************
932    *            Check for counted repeat            *
933    *************************************************/
934    
935    /* This function is called when a '{' is encountered in a place where it might
936    start a quantifier. It looks ahead to see if it really is a quantifier or not.
937    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
938    where the ddds are digits.
939    
940    Arguments:
941      p         pointer to the first char after '{'
942    
943    Returns:    TRUE or FALSE
944    */
945    
946    static BOOL
947    is_counted_repeat(const pcre_uchar *p)
948    {
949    if (!IS_DIGIT(*p)) return FALSE;
950    p++;
951    while (IS_DIGIT(*p)) p++;
952    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953    
954    if (*p++ != CHAR_COMMA) return FALSE;
955    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956    
957    if (!IS_DIGIT(*p)) return FALSE;
958    p++;
959    while (IS_DIGIT(*p)) p++;
960    
961    return (*p == CHAR_RIGHT_CURLY_BRACKET);
962    }
963    
964    
965    
966  /*************************************************  /*************************************************
967  *            Handle escapes                      *  *            Handle escapes                      *
968  *************************************************/  *************************************************/
969    
970  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
971  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
972  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.
973  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
974  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
975  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
976    
977  Arguments:  Arguments:
978    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
979      chptr          points to a returned data character
980    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
981    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
982    options        the options bits    options        the options bits
983    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
984    
985  Returns:         zero or positive => a data character  Returns:         zero => a data character
986                   negative => a special escape sequence                   positive => a special escape sequence
987                     negative => a back reference
988                   on error, errorcodeptr is set                   on error, errorcodeptr is set
989  */  */
990    
991  static int  static int
992  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
994  {  {
995  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
996  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
997  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
998    pcre_uint32 c;
999    int escape = 0;
1000    int i;
1001    
1002  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1003  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1004    
1005  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1006    
1007  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008    
1009  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
1011  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1012    
1013  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1014  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
1015  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1016    else if ((i = escapes[c - CHAR_0]) != 0)
1017      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018    
1019  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1020  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
1021  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1022    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1023  #endif  #endif
1024    
1025  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1026    
1027  else  else
1028    {    {
1029    const uschar *oldptr;    const pcre_uchar *oldptr;
1030    BOOL braced, negated;    BOOL braced, negated, overflow;
1031      int s;
1032    
1033    switch (c)    switch (c)
1034      {      {
1035      /* 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
1036      error. */      error. */
1037    
1038      case 'l':      case CHAR_l:
1039      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1040      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1041      break;      break;
1042    
1043      /* \g must be followed by one of a number of specific things:      case CHAR_u:
1044        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1045          {
1046          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047          Otherwise it is a lowercase u letter. */
1048          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1050            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1051            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1052            {
1053            c = 0;
1054            for (i = 0; i < 4; ++i)
1055              {
1056              register pcre_uint32 cc = *(++ptr);
1057    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1058              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1059              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1060    #else           /* EBCDIC coding */
1061              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1062              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063    #endif
1064              }
1065    
1066    #if defined COMPILE_PCRE8
1067            if (c > (utf ? 0x10ffffU : 0xffU))
1068    #elif defined COMPILE_PCRE16
1069            if (c > (utf ? 0x10ffffU : 0xffffU))
1070    #elif defined COMPILE_PCRE32
1071            if (utf && c > 0x10ffffU)
1072    #endif
1073              {
1074              *errorcodeptr = ERR76;
1075              }
1076            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1077            }
1078          }
1079        else
1080          *errorcodeptr = ERR37;
1081        break;
1082    
1083        case CHAR_U:
1084        /* In JavaScript, \U is an uppercase U letter. */
1085        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1086        break;
1087    
1088        /* In a character class, \g is just a literal "g". Outside a character
1089        class, \g must be followed by one of a number of specific things:
1090    
1091      (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
1092      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
1093      5.10 feature.      5.10 feature.
1094    
1095      (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
1096      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
1097      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
1098      was \k.      was \k.
1099    
1100      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1101      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1102      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1103      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1104    
1105      case 'g':      case CHAR_g:
1106      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
1107        {      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1108        c = -ESC_g;        {
1109        break;        escape = ESC_g;
1110        }        break;
1111          }
1112    
1113      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
1114    
1115      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116        {        {
1117        const uschar *p;        const pcre_uchar *p;
1118        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120        if (*p != 0 && *p != '}')        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121          {          {
1122          c = -ESC_k;          escape = ESC_k;
1123          break;          break;
1124          }          }
1125        braced = TRUE;        braced = TRUE;
# Line 571  else Line 1127  else
1127        }        }
1128      else braced = FALSE;      else braced = FALSE;
1129    
1130      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1131        {        {
1132        negated = TRUE;        negated = TRUE;
1133        ptr++;        ptr++;
1134        }        }
1135      else negated = FALSE;      else negated = FALSE;
1136    
1137      c = 0;      /* The integer range is limited by the machine's int representation. */
1138      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1139        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1140        while (IS_DIGIT(ptr[1]))
1141      if (c < 0)   /* Integer overflow */        {
1142          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1143            {
1144            overflow = TRUE;
1145            break;
1146            }
1147          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1148          }
1149        if (overflow) /* Integer overflow */
1150        {        {
1151          while (IS_DIGIT(ptr[1]))
1152            ptr++;
1153        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1154        break;        break;
1155        }        }
1156    
1157      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1158        {        {
1159        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1160        break;        break;
1161        }        }
1162    
1163      if (c == 0)      if (s == 0)
1164        {        {
1165        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1166        break;        break;
1167        }        }
1168    
1169      if (negated)      if (negated)
1170        {        {
1171        if (c > bracount)        if (s > bracount)
1172          {          {
1173          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1174          break;          break;
1175          }          }
1176        c = bracount - (c - 1);        s = bracount - (s - 1);
1177        }        }
1178    
1179      c = -(ESC_REF + c);      escape = -s;
1180      break;      break;
1181    
1182      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1183      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1184      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185        recommended to avoid the ambiguities in the old syntax.
1186    
1187      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1188      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1189      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1190      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1191      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1193      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1194    
1195      case '1': case '2': case '3': case '4': case '5':      Inside a character class, \ followed by a digit is always either a literal
1196      case '6': case '7': case '8': case '9':      8 or 9 or an octal number. */
1197    
1198        case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199        case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1200    
1201      if (!isclass)      if (!isclass)
1202        {        {
1203        oldptr = ptr;        oldptr = ptr;
1204        c -= '0';        /* The integer range is limited by the machine's int representation. */
1205        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1206          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1207        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1208            {
1209            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1210              {
1211              overflow = TRUE;
1212              break;
1213              }
1214            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1215            }
1216          if (overflow) /* Integer overflow */
1217          {          {
1218            while (IS_DIGIT(ptr[1]))
1219              ptr++;
1220          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1221          break;          break;
1222          }          }
1223        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1224          {          {
1225          c = -(ESC_REF + c);          escape = -s;
1226          break;          break;
1227          }          }
1228        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1229        }        }
1230    
1231      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1232      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1234        changed so as not to insert the binary zero. */
1235    
1236      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1237        {  
1238        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1239    
1240      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1241      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1242      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
1243      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,
1244      than 3 octal digits. */      but no more than 3 octal digits. */
1245    
1246      case '0':      case CHAR_0:
1247      c -= '0';      c -= CHAR_0;
1248      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1249          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1250      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1251        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1252    #endif
1253      break;      break;
1254    
1255      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1256      than 0xff in 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. */  
1257    
1258      case 'x':      case CHAR_o:
1259      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260        {        {
1261        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1262        c = 0;        c = 0;
1263        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1264          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265          {          {
1266          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1267          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1268          count++;  #ifdef COMPILE_PCRE32
1269            if (c >= 0x20000000l) { overflow = TRUE; break; }
1270  #ifndef EBCDIC  /* ASCII coding */  #endif
1271          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1272          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1273  #else           /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1275          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE32
1277            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278  #endif  #endif
1279          }          }
1280          if (overflow)
       if (*pt == '}')  
1281          {          {
1282          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1284          }          }
1285          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286        /* If the sequence of hex digits does not end with '}', then we don't          {
1287        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288            }
1289          else *errorcodeptr = ERR80;
1290        }        }
1291        break;
1292    
1293      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294        numbers. Otherwise it is a lowercase x letter. */
1295    
1296      c = 0;      case CHAR_x:
1297      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298        {        {
1299        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301  #ifndef EBCDIC  /* ASCII coding */          {
1302        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1303        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1304              {
1305              register pcre_uint32 cc = *(++ptr);
1306    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1307              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1308              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1309  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1310        if (cc <= 'z') cc += 64;              /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1311        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1312  #endif  #endif
1313        }            }
1314            }
1315          }    /* End JavaScript handling */
1316    
1317        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319        digits. If not, { used to be treated as a data character. However, Perl
1320        seems to read hex digits up to the first non-such, and ignore the rest, so
1321        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322        now gives an error. */
1323    
1324        else
1325          {
1326          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327            {
1328            ptr += 2;
1329            c = 0;
1330            overflow = FALSE;
1331            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332              {
1333              register pcre_uint32 cc = *ptr++;
1334              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1335    
1336    #ifdef COMPILE_PCRE32
1337              if (c >= 0x10000000l) { overflow = TRUE; break; }
1338    #endif
1339    
1340    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1341              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1342              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343    #else           /* EBCDIC coding */
1344              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1345              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346    #endif
1347    
1348    #if defined COMPILE_PCRE8
1349              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350    #elif defined COMPILE_PCRE16
1351              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352    #elif defined COMPILE_PCRE32
1353              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354    #endif
1355              }
1356    
1357            if (overflow)
1358              {
1359              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360              *errorcodeptr = ERR34;
1361              }
1362    
1363            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364              {
1365              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366              }
1367    
1368            /* If the sequence of hex digits does not end with '}', give an error.
1369            We used just to recognize this construct and fall through to the normal
1370            \x handling, but nowadays Perl gives an error, which seems much more
1371            sensible, so we do too. */
1372    
1373            else *errorcodeptr = ERR79;
1374            }   /* End of \x{} processing */
1375    
1376          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377    
1378          else
1379            {
1380            c = 0;
1381            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382              {
1383              pcre_uint32 cc;                          /* Some compilers don't like */
1384              cc = *(++ptr);                           /* ++ in initializers */
1385    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1386              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1387              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388    #else           /* EBCDIC coding */
1389              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1390              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391    #endif
1392              }
1393            }     /* End of \xdd handling */
1394          }       /* End of Perl-style \x handling */
1395      break;      break;
1396    
1397      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1398      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
1399        coding is ASCII-specific, but then the whole concept of \cx is
1400      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1401    
1402      case 'c':      case CHAR_c:
1403      c = *(++ptr);      c = *(++ptr);
1404      if (c == 0)      if (c == CHAR_NULL)
1405        {        {
1406        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1407        break;        break;
1408        }        }
1409    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1410  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1411      if (c >= 'a' && c <= 'z') c -= 32;        {
1412          *errorcodeptr = ERR68;
1413          break;
1414          }
1415        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1416      c ^= 0x40;      c ^= 0x40;
1417  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1418      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1419      c ^= 0xC0;      c ^= 0xC0;
1420  #endif  #endif
1421      break;      break;
# Line 763  else Line 1437  else
1437      }      }
1438    }    }
1439    
1440    /* Perl supports \N{name} for character names, as well as plain \N for "not
1441    newline". PCRE does not support \N{name}. However, it does support
1442    quantification such as \N{2,3}. */
1443    
1444    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1445         !is_counted_repeat(ptr+2))
1446      *errorcodeptr = ERR37;
1447    
1448    /* If PCRE_UCP is set, we change the values for \d etc. */
1449    
1450    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1451      escape += (ESC_DU - ESC_D);
1452    
1453    /* Set the pointer to the final character before returning. */
1454    
1455  *ptrptr = ptr;  *ptrptr = ptr;
1456  return c;  *chptr = c;
1457    return escape;
1458  }  }
1459    
1460    
# Line 782  escape sequence. Line 1472  escape sequence.
1472  Argument:  Argument:
1473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1474    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1475    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1476      pdataptr       points to an unsigned int that is set to the detailed property value
1477    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1478    
1479  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1480  */  */
1481    
1482  static int  static BOOL
1483  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484      unsigned int *pdataptr, int *errorcodeptr)
1485  {  {
1486  int c, i, bot, top;  pcre_uchar c;
1487  const uschar *ptr = *ptrptr;  int i, bot, top;
1488  char name[32];  const pcre_uchar *ptr = *ptrptr;
1489    pcre_uchar name[32];
1490    
1491  c = *(++ptr);  c = *(++ptr);
1492  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1493    
1494  *negptr = FALSE;  *negptr = FALSE;
1495    
1496  /* \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
1497  negation. */  negation. */
1498    
1499  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1500    {    {
1501    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1502      {      {
1503      *negptr = TRUE;      *negptr = TRUE;
1504      ptr++;      ptr++;
1505      }      }
1506    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507      {      {
1508      c = *(++ptr);      c = *(++ptr);
1509      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1510      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511      name[i] = c;      name[i] = c;
1512      }      }
1513    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1514    name[i] = 0;    name[i] = 0;
1515    }    }
1516    
# Line 834  else Line 1527  else
1527  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1528    
1529  bot = 0;  bot = 0;
1530  top = _pcre_utt_size;  top = PRIV(utt_size);
1531    
1532  while (bot < top)  while (bot < top)
1533    {    {
1534      int r;
1535    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1536    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537    if (c == 0)    if (r == 0)
1538      {      {
1539      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1540      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1541        return TRUE;
1542      }      }
1543    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1544    }    }
1545    
1546  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1547  *ptrptr = ptr;  *ptrptr = ptr;
1548  return -1;  return FALSE;
1549    
1550  ERROR_RETURN:  ERROR_RETURN:
1551  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1552  *ptrptr = ptr;  *ptrptr = ptr;
1553  return -1;  return FALSE;
1554  }  }
1555  #endif  #endif
1556    
1557    
1558    
   
 /*************************************************  
 *            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 == '}');  
 }  
   
   
   
1559  /*************************************************  /*************************************************
1560  *         Read repeat counts                     *  *         Read repeat counts                     *
1561  *************************************************/  *************************************************/
# Line 914  Returns:         pointer to '}' on succe Line 1575  Returns:         pointer to '}' on succe
1575                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1576  */  */
1577    
1578  static const uschar *  static const pcre_uchar *
1579  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)
1580  {  {
1581  int min = 0;  int min = 0;
1582  int max = -1;  int max = -1;
# Line 923  int max = -1; Line 1584  int max = -1;
1584  /* 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
1585  an integer overflow. */  an integer overflow. */
1586    
1587  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1588  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1589    {    {
1590    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1594  if (min < 0 || min > 65535)
1594  /* 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.
1595  Also, max must not be less than min. */  Also, max must not be less than min. */
1596    
1597  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1598    {    {
1599    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1600      {      {
1601      max = 0;      max = 0;
1602      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1603      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1604        {        {
1605        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1624  return p;
1624    
1625    
1626  /*************************************************  /*************************************************
1627  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1628  *************************************************/  *************************************************/
1629    
1630  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1631  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
1632  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
1633  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
1634  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.  
1635    
1636  Arguments:  Arguments:
1637    ptr          current position in the pattern    code         pointer to the start of the group
1638    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  
1639    
1640  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1641  */  */
1642    
1643  static int  static const pcre_uchar*
1644  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode)  
1645  {  {
1646  const uschar *thisname;  for (;;)
   
 for (; *ptr != 0; ptr++)  
1647    {    {
1648    int term;    switch ((int)*code)
   
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == '\\')  
1649      {      {
1650      if (*(++ptr) == 0) return -1;      case OP_ASSERT_NOT:
1651      if (*ptr == 'Q') for (;;)      case OP_ASSERTBACK:
1652        {      case OP_ASSERTBACK_NOT:
1653        while (*(++ptr) != 0 && *ptr != '\\');      if (!skipassert) return code;
1654        if (*ptr == 0) return -1;      do code += GET(code, 1); while (*code == OP_ALT);
1655        if (*(++ptr) == 'E') break;      code += PRIV(OP_lengths)[*code];
1656        }      break;
     continue;  
     }  
1657    
1658    /* Skip over character classes */      case OP_WORD_BOUNDARY:
1659        case OP_NOT_WORD_BOUNDARY:
1660    if (*ptr == '[')      if (!skipassert) return code;
1661      {      /* Fall through */
     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 */  
1662    
1663      case OP_CALLOUT:      case OP_CALLOUT:
1664      case OP_CREF:      case OP_CREF:
1665        case OP_DNCREF:
1666      case OP_RREF:      case OP_RREF:
1667        case OP_DNRREF:
1668      case OP_DEF:      case OP_DEF:
1669      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1670      break;      break;
1671    
1672      default:      default:
# Line 1137  for (;;) Line 1678  for (;;)
1678    
1679    
1680    
   
1681  /*************************************************  /*************************************************
1682  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1683  *************************************************/  *************************************************/
1684    
1685  /* 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,
1686  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.
1687  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
1688    temporarily terminated with OP_END when this function is called.
1689    
1690    This function is called when a backward assertion is encountered, so that if it
1691    fails, the error message can point to the correct place in the pattern.
1692    However, we cannot do this when the assertion contains subroutine calls,
1693    because they can be forward references. We solve this by remembering this case
1694    and doing the check at the end; a flag specifies which mode we are running in.
1695    
1696  Arguments:  Arguments:
1697    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1698    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699      atend    TRUE if called when the pattern is complete
1700  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1701               or -2 if \C was encountered  
1702    Returns:   the fixed length,
1703                 or -1 if there is no fixed length,
1704                 or -2 if \C was encountered (in UTF-8 mode only)
1705                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706                 or -4 if an unknown opcode was encountered (internal error)
1707  */  */
1708    
1709  static int  static int
1710  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711  {  {
1712  int length = -1;  int length = -1;
1713    
1714  register int branchlength = 0;  register int branchlength = 0;
1715  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1716    
1717  /* 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
1718  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 1720  branch, check the length against that of
1720  for (;;)  for (;;)
1721    {    {
1722    int d;    int d;
1723    register int op = *cc;    pcre_uchar *ce, *cs;
1724      register pcre_uchar op = *cc;
1725    
1726    switch (op)    switch (op)
1727      {      {
1728        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729        OP_BRA (normal non-capturing bracket) because the other variants of these
1730        opcodes are all concerned with unlimited repeated groups, which of course
1731        are not of fixed length. */
1732    
1733      case OP_CBRA:      case OP_CBRA:
1734      case OP_BRA:      case OP_BRA:
1735      case OP_ONCE:      case OP_ONCE:
1736        case OP_ONCE_NC:
1737      case OP_COND:      case OP_COND:
1738      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739      if (d < 0) return d;      if (d < 0) return d;
1740      branchlength += d;      branchlength += d;
1741      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1742      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1743      break;      break;
1744    
1745      /* 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.
1746      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
1747      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
1748        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749        because they all imply an unlimited repeat. */
1750    
1751      case OP_ALT:      case OP_ALT:
1752      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1753      case OP_END:      case OP_END:
1754        case OP_ACCEPT:
1755        case OP_ASSERT_ACCEPT:
1756      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1757        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1758      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1760  for (;;)
1760      branchlength = 0;      branchlength = 0;
1761      break;      break;
1762    
1763        /* A true recursion implies not fixed length, but a subroutine call may
1764        be OK. If the subroutine is a forward reference, we can't deal with
1765        it until the end of the pattern, so return -3. */
1766    
1767        case OP_RECURSE:
1768        if (!atend) return -3;
1769        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1770        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1771        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1772        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1773        if (d < 0) return d;
1774        branchlength += d;
1775        cc += 1 + LINK_SIZE;
1776        break;
1777    
1778      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1779    
1780      case OP_ASSERT:      case OP_ASSERT:
# Line 1205  for (;;) Line 1782  for (;;)
1782      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1783      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1784      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1785      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1786        break;
1787    
1788      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1789    
1790      case OP_REVERSE:      case OP_MARK:
1791        case OP_PRUNE_ARG:
1792        case OP_SKIP_ARG:
1793        case OP_THEN_ARG:
1794        cc += cc[1] + PRIV(OP_lengths)[*cc];
1795        break;
1796    
1797        case OP_CALLOUT:
1798        case OP_CIRC:
1799        case OP_CIRCM:
1800        case OP_CLOSE:
1801        case OP_COMMIT:
1802      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1803      case OP_DEF:      case OP_DEF:
1804      case OP_OPT:      case OP_DNCREF:
1805      case OP_CALLOUT:      case OP_DNRREF:
1806      case OP_SOD:      case OP_DOLL:
1807      case OP_SOM:      case OP_DOLLM:
1808      case OP_EOD:      case OP_EOD:
1809      case OP_EODN:      case OP_EODN:
1810      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1811      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1812        case OP_PRUNE:
1813        case OP_REVERSE:
1814        case OP_RREF:
1815        case OP_SET_SOM:
1816        case OP_SKIP:
1817        case OP_SOD:
1818        case OP_SOM:
1819        case OP_THEN:
1820      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1821      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1822      break;      break;
1823    
1824      /* Handle literal characters */      /* Handle literal characters */
1825    
1826      case OP_CHAR:      case OP_CHAR:
1827      case OP_CHARNC:      case OP_CHARI:
1828      case OP_NOT:      case OP_NOT:
1829        case OP_NOTI:
1830      branchlength++;      branchlength++;
1831      cc += 2;      cc += 2;
1832  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1833      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1834  #endif  #endif
1835      break;      break;
1836    
# Line 1245  for (;;) Line 1838  for (;;)
1838      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1839    
1840      case OP_EXACT:      case OP_EXACT:
1841      branchlength += GET2(cc,1);      case OP_EXACTI:
1842      cc += 4;      case OP_NOTEXACT:
1843  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1844      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1845        {      cc += 2 + IMM2_SIZE;
1846        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1847        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848  #endif  #endif
1849      break;      break;
1850    
1851      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1852      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1853      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854      cc += 4;        cc += 2;
1855        cc += 1 + IMM2_SIZE + 1;
1856      break;      break;
1857    
1858      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1268  for (;;) Line 1862  for (;;)
1862      cc += 2;      cc += 2;
1863      /* Fall through */      /* Fall through */
1864    
1865        case OP_HSPACE:
1866        case OP_VSPACE:
1867        case OP_NOT_HSPACE:
1868        case OP_NOT_VSPACE:
1869      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1870      case OP_DIGIT:      case OP_DIGIT:
1871      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1873  for (;;)
1873      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1874      case OP_WORDCHAR:      case OP_WORDCHAR:
1875      case OP_ANY:      case OP_ANY:
1876        case OP_ALLANY:
1877      branchlength++;      branchlength++;
1878      cc++;      cc++;
1879      break;      break;
1880    
1881      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1882        otherwise \C is coded as OP_ALLANY. */
1883    
1884      case OP_ANYBYTE:      case OP_ANYBYTE:
1885      return -2;      return -2;
1886    
1887      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1888    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1889      case OP_CLASS:      case OP_CLASS:
1890      case OP_NCLASS:      case OP_NCLASS:
1891      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892        case OP_XCLASS:
1893        /* The original code caused an unsigned overflow in 64 bit systems,
1894        so now we use a conditional statement. */
1895        if (op == OP_XCLASS)
1896          cc += GET(cc, 1);
1897        else
1898          cc += PRIV(OP_lengths)[OP_CLASS];
1899    #else
1900        cc += PRIV(OP_lengths)[OP_CLASS];
1901    #endif
1902    
1903      switch (*cc)      switch (*cc)
1904        {        {
1905        case OP_CRSTAR:        case OP_CRSTAR:
1906        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1907          case OP_CRPLUS:
1908          case OP_CRMINPLUS:
1909        case OP_CRQUERY:        case OP_CRQUERY:
1910        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1911          case OP_CRPOSSTAR:
1912          case OP_CRPOSPLUS:
1913          case OP_CRPOSQUERY:
1914        return -1;        return -1;
1915    
1916        case OP_CRRANGE:        case OP_CRRANGE:
1917        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1918        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1919        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920        cc += 5;        branchlength += (int)GET2(cc,1);
1921          cc += 1 + 2 * IMM2_SIZE;
1922        break;        break;
1923    
1924        default:        default:
# Line 1318  for (;;) Line 1928  for (;;)
1928    
1929      /* Anything else is variable length */      /* Anything else is variable length */
1930    
1931      default:      case OP_ANYNL:
1932        case OP_BRAMINZERO:
1933        case OP_BRAPOS:
1934        case OP_BRAPOSZERO:
1935        case OP_BRAZERO:
1936        case OP_CBRAPOS:
1937        case OP_EXTUNI:
1938        case OP_KETRMAX:
1939        case OP_KETRMIN:
1940        case OP_KETRPOS:
1941        case OP_MINPLUS:
1942        case OP_MINPLUSI:
1943        case OP_MINQUERY:
1944        case OP_MINQUERYI:
1945        case OP_MINSTAR:
1946        case OP_MINSTARI:
1947        case OP_MINUPTO:
1948        case OP_MINUPTOI:
1949        case OP_NOTMINPLUS:
1950        case OP_NOTMINPLUSI:
1951        case OP_NOTMINQUERY:
1952        case OP_NOTMINQUERYI:
1953        case OP_NOTMINSTAR:
1954        case OP_NOTMINSTARI:
1955        case OP_NOTMINUPTO:
1956        case OP_NOTMINUPTOI:
1957        case OP_NOTPLUS:
1958        case OP_NOTPLUSI:
1959        case OP_NOTPOSPLUS:
1960        case OP_NOTPOSPLUSI:
1961        case OP_NOTPOSQUERY:
1962        case OP_NOTPOSQUERYI:
1963        case OP_NOTPOSSTAR:
1964        case OP_NOTPOSSTARI:
1965        case OP_NOTPOSUPTO:
1966        case OP_NOTPOSUPTOI:
1967        case OP_NOTQUERY:
1968        case OP_NOTQUERYI:
1969        case OP_NOTSTAR:
1970        case OP_NOTSTARI:
1971        case OP_NOTUPTO:
1972        case OP_NOTUPTOI:
1973        case OP_PLUS:
1974        case OP_PLUSI:
1975        case OP_POSPLUS:
1976        case OP_POSPLUSI:
1977        case OP_POSQUERY:
1978        case OP_POSQUERYI:
1979        case OP_POSSTAR:
1980        case OP_POSSTARI:
1981        case OP_POSUPTO:
1982        case OP_POSUPTOI:
1983        case OP_QUERY:
1984        case OP_QUERYI:
1985        case OP_REF:
1986        case OP_REFI:
1987        case OP_DNREF:
1988        case OP_DNREFI:
1989        case OP_SBRA:
1990        case OP_SBRAPOS:
1991        case OP_SCBRA:
1992        case OP_SCBRAPOS:
1993        case OP_SCOND:
1994        case OP_SKIPZERO:
1995        case OP_STAR:
1996        case OP_STARI:
1997        case OP_TYPEMINPLUS:
1998        case OP_TYPEMINQUERY:
1999        case OP_TYPEMINSTAR:
2000        case OP_TYPEMINUPTO:
2001        case OP_TYPEPLUS:
2002        case OP_TYPEPOSPLUS:
2003        case OP_TYPEPOSQUERY:
2004        case OP_TYPEPOSSTAR:
2005        case OP_TYPEPOSUPTO:
2006        case OP_TYPEQUERY:
2007        case OP_TYPESTAR:
2008        case OP_TYPEUPTO:
2009        case OP_UPTO:
2010        case OP_UPTOI:
2011      return -1;      return -1;
2012    
2013        /* Catch unrecognized opcodes so that when new ones are added they
2014        are not forgotten, as has happened in the past. */
2015    
2016        default:
2017        return -4;
2018      }      }
2019    }    }
2020  /* Control never gets here */  /* Control never gets here */
# Line 1327  for (;;) Line 2022  for (;;)
2022    
2023    
2024    
   
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
2028    
2029  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2030  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2031    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032    so that it can be called from pcre_study() when finding the minimum matching
2033    length.
2034    
2035  Arguments:  Arguments:
2036    code        points to start of expression    code        points to start of expression
2037    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2038    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2039    
2040  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
2041  */  */
2042    
2043  static const uschar *  const pcre_uchar *
2044  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045  {  {
2046  for (;;)  for (;;)
2047    {    {
2048    register int c = *code;    register pcre_uchar c = *code;
2049    
2050    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2051    
2052    /* 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 2055  for (;;)
2055    
2056    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2057    
2058      /* Handle recursion */
2059    
2060      else if (c == OP_REVERSE)
2061        {
2062        if (number < 0) return (pcre_uchar *)code;
2063        code += PRIV(OP_lengths)[c];
2064        }
2065    
2066    /* Handle capturing bracket */    /* Handle capturing bracket */
2067    
2068    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2069               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070      {      {
2071      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2072      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2073      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2074      }      }
2075    
2076    /* 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
2077    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
2078    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079      must add in its length. */
2080    
2081    else    else
2082      {      {
# Line 1390  for (;;) Line 2098  for (;;)
2098        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2099        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2100        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2101        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102            code += 2;
2103          break;
2104    
2105          case OP_MARK:
2106          case OP_PRUNE_ARG:
2107          case OP_SKIP_ARG:
2108          case OP_THEN_ARG:
2109          code += code[1];
2110        break;        break;
2111        }        }
2112    
2113      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2114    
2115      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2116    
2117    /* 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
2118    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
2119    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2120    
2121  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2122      if (utf8) switch(c)      if (utf) switch(c)
2123        {        {
2124        case OP_CHAR:        case OP_CHAR:
2125        case OP_CHARNC:        case OP_CHARI:
2126        case OP_EXACT:        case OP_EXACT:
2127          case OP_EXACTI:
2128        case OP_UPTO:        case OP_UPTO:
2129          case OP_UPTOI:
2130        case OP_MINUPTO:        case OP_MINUPTO:
2131          case OP_MINUPTOI:
2132        case OP_POSUPTO:        case OP_POSUPTO:
2133          case OP_POSUPTOI:
2134        case OP_STAR:        case OP_STAR:
2135          case OP_STARI:
2136        case OP_MINSTAR:        case OP_MINSTAR:
2137          case OP_MINSTARI:
2138        case OP_POSSTAR:        case OP_POSSTAR:
2139          case OP_POSSTARI:
2140        case OP_PLUS:        case OP_PLUS:
2141          case OP_PLUSI:
2142        case OP_MINPLUS:        case OP_MINPLUS:
2143          case OP_MINPLUSI:
2144        case OP_POSPLUS:        case OP_POSPLUS:
2145          case OP_POSPLUSI:
2146        case OP_QUERY:        case OP_QUERY:
2147          case OP_QUERYI:
2148        case OP_MINQUERY:        case OP_MINQUERY:
2149          case OP_MINQUERYI:
2150        case OP_POSQUERY:        case OP_POSQUERY:
2151        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2152          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2153        break;        break;
2154        }        }
2155    #else
2156        (void)(utf);  /* Keep compiler happy by referencing function argument */
2157  #endif  #endif
2158      }      }
2159    }    }
# Line 1439  instance of OP_RECURSE. Line 2170  instance of OP_RECURSE.
2170    
2171  Arguments:  Arguments:
2172    code        points to start of expression    code        points to start of expression
2173    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2174    
2175  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
2176  */  */
2177    
2178  static const uschar *  static const pcre_uchar *
2179  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2180  {  {
2181  for (;;)  for (;;)
2182    {    {
2183    register int c = *code;    register pcre_uchar c = *code;
2184    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2185    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2186    
# Line 1461  for (;;) Line 2192  for (;;)
2192    
2193    /* 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
2194    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
2195    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2196      must add in its length. */
2197    
2198    else    else
2199      {      {
# Line 1483  for (;;) Line 2215  for (;;)
2215        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2216        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2217        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2218        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219            code += 2;
2220          break;
2221    
2222          case OP_MARK:
2223          case OP_PRUNE_ARG:
2224          case OP_SKIP_ARG:
2225          case OP_THEN_ARG:
2226          code += code[1];
2227        break;        break;
2228        }        }
2229    
2230      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2231    
2232      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2233    
2234      /* 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
2235      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
2236      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2237    
2238  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2239      if (utf8) switch(c)      if (utf) switch(c)
2240        {        {
2241        case OP_CHAR:        case OP_CHAR:
2242        case OP_CHARNC:        case OP_CHARI:
2243          case OP_NOT:
2244          case OP_NOTI:
2245        case OP_EXACT:        case OP_EXACT:
2246          case OP_EXACTI:
2247          case OP_NOTEXACT:
2248          case OP_NOTEXACTI:
2249        case OP_UPTO:        case OP_UPTO:
2250          case OP_UPTOI:
2251          case OP_NOTUPTO:
2252          case OP_NOTUPTOI:
2253        case OP_MINUPTO:        case OP_MINUPTO:
2254          case OP_MINUPTOI:
2255          case OP_NOTMINUPTO:
2256          case OP_NOTMINUPTOI:
2257        case OP_POSUPTO:        case OP_POSUPTO:
2258          case OP_POSUPTOI:
2259          case OP_NOTPOSUPTO:
2260          case OP_NOTPOSUPTOI:
2261        case OP_STAR:        case OP_STAR:
2262          case OP_STARI:
2263          case OP_NOTSTAR:
2264          case OP_NOTSTARI:
2265        case OP_MINSTAR:        case OP_MINSTAR:
2266          case OP_MINSTARI:
2267          case OP_NOTMINSTAR:
2268          case OP_NOTMINSTARI:
2269        case OP_POSSTAR:        case OP_POSSTAR:
2270          case OP_POSSTARI:
2271          case OP_NOTPOSSTAR:
2272          case OP_NOTPOSSTARI:
2273        case OP_PLUS:        case OP_PLUS:
2274          case OP_PLUSI:
2275          case OP_NOTPLUS:
2276          case OP_NOTPLUSI:
2277        case OP_MINPLUS:        case OP_MINPLUS:
2278          case OP_MINPLUSI:
2279          case OP_NOTMINPLUS:
2280          case OP_NOTMINPLUSI:
2281        case OP_POSPLUS:        case OP_POSPLUS:
2282          case OP_POSPLUSI:
2283          case OP_NOTPOSPLUS:
2284          case OP_NOTPOSPLUSI:
2285        case OP_QUERY:        case OP_QUERY:
2286          case OP_QUERYI:
2287          case OP_NOTQUERY:
2288          case OP_NOTQUERYI:
2289        case OP_MINQUERY:        case OP_MINQUERY:
2290          case OP_MINQUERYI:
2291          case OP_NOTMINQUERY:
2292          case OP_NOTMINQUERYI:
2293        case OP_POSQUERY:        case OP_POSQUERY:
2294        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2295          case OP_NOTPOSQUERY:
2296          case OP_NOTPOSQUERYI:
2297          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2298        break;        break;
2299        }        }
2300    #else
2301        (void)(utf);  /* Keep compiler happy by referencing function argument */
2302  #endif  #endif
2303      }      }
2304    }    }
# Line 1538  bracket whose current branch will alread Line 2321  bracket whose current branch will alread
2321  Arguments:  Arguments:
2322    code        points to start of search    code        points to start of search
2323    endcode     points to where to stop    endcode     points to where to stop
2324    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325      cd          contains pointers to tables etc.
2326      recurses    chain of recurse_check to catch mutual recursion
2327    
2328  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2329  */  */
2330    
2331    typedef struct recurse_check {
2332      struct recurse_check *prev;
2333      const pcre_uchar *group;
2334    } recurse_check;
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338      BOOL utf, compile_data *cd, recurse_check *recurses)
2339  {  {
2340  register int c;  register pcre_uchar c;
2341  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2342    
2343    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346    {    {
2347    const uschar *ccode;    const pcre_uchar *ccode;
2348    
2349    c = *code;    c = *code;
2350    
# Line 1565  for (code = first_significant_code(code Line 2358  for (code = first_significant_code(code
2358      continue;      continue;
2359      }      }
2360    
2361      /* For a recursion/subroutine call, if its end has been reached, which
2362      implies a backward reference subroutine call, we can scan it. If it's a
2363      forward reference subroutine call, we can't. To detect forward reference
2364      we have to scan up the list that is kept in the workspace. This function is
2365      called only when doing the real compile, not during the pre-compile that
2366      measures the size of the compiled pattern. */
2367    
2368      if (c == OP_RECURSE)
2369        {
2370        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371        BOOL empty_branch;
2372    
2373        /* Test for forward reference or uncompleted reference. This is disabled
2374        when called to scan a completed pattern by setting cd->start_workspace to
2375        NULL. */
2376    
2377        if (cd->start_workspace != NULL)
2378          {
2379          const pcre_uchar *tcode;
2380          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383          }
2384    
2385        /* If we are scanning a completed pattern, there are no forward references
2386        and all groups are complete. We need to detect whether this is a recursive
2387        call, as otherwise there will be an infinite loop. If it is a recursion,
2388        just skip over it. Simple recursions are easily detected. For mutual
2389        recursions we keep a chain on the stack. */
2390    
2391        else
2392          {
2393          recurse_check *r = recurses;
2394          const pcre_uchar *endgroup = scode;
2395    
2396          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2398    
2399          for (r = recurses; r != NULL; r = r->prev)
2400            if (r->group == scode) break;
2401          if (r != NULL) continue;   /* Mutual recursion */
2402          }
2403    
2404        /* Completed reference; scan the referenced group, remembering it on the
2405        stack chain to detect mutual recursions. */
2406    
2407        empty_branch = FALSE;
2408        this_recurse.prev = recurses;
2409        this_recurse.group = scode;
2410    
2411        do
2412          {
2413          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414            {
2415            empty_branch = TRUE;
2416            break;
2417            }
2418          scode += GET(scode, 1);
2419          }
2420        while (*scode == OP_ALT);
2421    
2422        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2423        continue;
2424        }
2425    
2426    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2427    
2428    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2429          c == OP_BRAPOSZERO)
2430        {
2431        code += PRIV(OP_lengths)[c];
2432        do code += GET(code, 1); while (*code == OP_ALT);
2433        c = *code;
2434        continue;
2435        }
2436    
2437      /* A nested group that is already marked as "could be empty" can just be
2438      skipped. */
2439    
2440      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2441          c == OP_SCBRA || c == OP_SCBRAPOS)
2442      {      {
     code += _pcre_OP_lengths[c];  
2443      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2444      c = *code;      c = *code;
2445      continue;      continue;
# Line 1577  for (code = first_significant_code(code Line 2447  for (code = first_significant_code(code
2447    
2448    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2449    
2450    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2451          c == OP_CBRA || c == OP_CBRAPOS ||
2452          c == OP_ONCE || c == OP_ONCE_NC ||
2453          c == OP_COND)
2454      {      {
2455      BOOL empty_branch;      BOOL empty_branch;
2456      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2457    
2458      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2459        empty branch, so just skip over the conditional, because it could be empty.
2460        Otherwise, scan the individual branches of the group. */
2461    
2462      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;  
2463        code += GET(code, 1);        code += GET(code, 1);
2464        else
2465          {
2466          empty_branch = FALSE;
2467          do
2468            {
2469            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470              empty_branch = TRUE;
2471            code += GET(code, 1);
2472            }
2473          while (*code == OP_ALT);
2474          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2475        }        }
2476      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2477      c = *code;      c = *code;
2478      continue;      continue;
2479      }      }
# Line 1603  for (code = first_significant_code(code Line 2484  for (code = first_significant_code(code
2484      {      {
2485      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2486      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2487      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2488      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"
2489      here. */      here. */
2490    
2491  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2492      case OP_XCLASS:      case OP_XCLASS:
2493      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2494      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1615  for (code = first_significant_code(code Line 2496  for (code = first_significant_code(code
2496    
2497      case OP_CLASS:      case OP_CLASS:
2498      case OP_NCLASS:      case OP_NCLASS:
2499      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2500    
2501  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2502      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2503  #endif  #endif
2504    
# Line 1627  for (code = first_significant_code(code Line 2508  for (code = first_significant_code(code
2508        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2509        case OP_CRQUERY:        case OP_CRQUERY:
2510        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2511          case OP_CRPOSSTAR:
2512          case OP_CRPOSQUERY:
2513        break;        break;
2514    
2515        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2516        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2517        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2518          case OP_CRPOSPLUS:
2519        return FALSE;        return FALSE;
2520    
2521        case OP_CRRANGE:        case OP_CRRANGE:
2522        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2523          case OP_CRPOSRANGE:
2524        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2525        break;        break;
2526        }        }
# Line 1643  for (code = first_significant_code(code Line 2528  for (code = first_significant_code(code
2528    
2529      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2530    
2531        case OP_ANY:
2532        case OP_ALLANY:
2533        case OP_ANYBYTE:
2534    
2535      case OP_PROP:      case OP_PROP:
2536      case OP_NOTPROP:      case OP_NOTPROP:
2537        case OP_ANYNL:
2538    
2539        case OP_NOT_HSPACE:
2540        case OP_HSPACE:
2541        case OP_NOT_VSPACE:
2542        case OP_VSPACE:
2543      case OP_EXTUNI:      case OP_EXTUNI:
2544    
2545      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2546      case OP_DIGIT:      case OP_DIGIT:
2547      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2548      case OP_WHITESPACE:      case OP_WHITESPACE:
2549      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2550      case OP_WORDCHAR:      case OP_WORDCHAR:
2551      case OP_ANY:  
     case OP_ANYBYTE:  
2552      case OP_CHAR:      case OP_CHAR:
2553      case OP_CHARNC:      case OP_CHARI:
2554      case OP_NOT:      case OP_NOT:
2555        case OP_NOTI:
2556    
2557      case OP_PLUS:      case OP_PLUS:
2558        case OP_PLUSI:
2559      case OP_MINPLUS:      case OP_MINPLUS:
2560      case OP_POSPLUS:      case OP_MINPLUSI:
2561      case OP_EXACT:  
2562      case OP_NOTPLUS:      case OP_NOTPLUS:
2563        case OP_NOTPLUSI:
2564      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2565        case OP_NOTMINPLUSI:
2566    
2567        case OP_POSPLUS:
2568        case OP_POSPLUSI:
2569      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2570        case OP_NOTPOSPLUSI:
2571    
2572        case OP_EXACT:
2573        case OP_EXACTI:
2574      case OP_NOTEXACT:      case OP_NOTEXACT:
2575        case OP_NOTEXACTI:
2576    
2577      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2578      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2579      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2580      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2581    
2582      return FALSE;      return FALSE;
2583    
2584      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 1688  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2599      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2600      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2601      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602          code += 2;
2603        break;
2604    
2605        /* End of branch */
2606    
2607        case OP_KET:
2608        case OP_KETRMAX:
2609        case OP_KETRMIN:
2610        case OP_KETRPOS:
2611        case OP_ALT:
2612        return TRUE;
2613    
2614        /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615        MINUPTO, and POSUPTO and their caseless and negative versions may be
2616        followed by a multibyte character. */
2617    
2618    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619        case OP_STAR:
2620        case OP_STARI:
2621        case OP_NOTSTAR:
2622        case OP_NOTSTARI:
2623    
2624        case OP_MINSTAR:
2625        case OP_MINSTARI:
2626        case OP_NOTMINSTAR:
2627        case OP_NOTMINSTARI:
2628    
2629        case OP_POSSTAR:
2630        case OP_POSSTARI:
2631        case OP_NOTPOSSTAR:
2632        case OP_NOTPOSSTARI:
2633    
2634        case OP_QUERY:
2635        case OP_QUERYI:
2636        case OP_NOTQUERY:
2637        case OP_NOTQUERYI:
2638    
2639        case OP_MINQUERY:
2640        case OP_MINQUERYI:
2641        case OP_NOTMINQUERY:
2642        case OP_NOTMINQUERYI:
2643    
2644        case OP_POSQUERY:
2645        case OP_POSQUERYI:
2646        case OP_NOTPOSQUERY:
2647        case OP_NOTPOSQUERYI:
2648    
2649        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650        break;
2651    
2652        case OP_UPTO:
2653        case OP_UPTOI:
2654        case OP_NOTUPTO:
2655        case OP_NOTUPTOI:
2656    
2657        case OP_MINUPTO:
2658        case OP_MINUPTOI:
2659        case OP_NOTMINUPTO:
2660        case OP_NOTMINUPTOI:
2661    
2662        case OP_POSUPTO:
2663        case OP_POSUPTOI:
2664        case OP_NOTPOSUPTO:
2665        case OP_NOTPOSUPTOI:
2666    
2667        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668        break;
2669    #endif
2670    
2671        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2672        string. */
2673    
2674        case OP_MARK:
2675        case OP_PRUNE_ARG:
2676        case OP_SKIP_ARG:
2677        case OP_THEN_ARG:
2678        code += code[1];
2679        break;
2680    
2681        /* None of the remaining opcodes are required to match a character. */
2682    
2683        default:
2684        break;
2685        }
2686      }
2687    
2688    return TRUE;
2689    }
2690    
2691    
2692    
2693    /*************************************************
2694    *    Scan compiled regex for non-emptiness       *
2695    *************************************************/
2696    
2697    /* This function is called to check for left recursive calls. We want to check
2698    the current branch of the current pattern to see if it could match the empty
2699    string. If it could, we must look outwards for branches at other levels,
2700    stopping when we pass beyond the bracket which is the subject of the recursion.
2701    This function is called only during the real compile, not during the
2702    pre-compile.
2703    
2704    Arguments:
2705      code        points to start of the recursion
2706      endcode     points to where to stop (current RECURSE item)
2707      bcptr       points to the chain of current (unclosed) branch starts
2708      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709      cd          pointers to tables etc
2710    
2711    Returns:      TRUE if what is matched could be empty
2712    */
2713    
2714    static BOOL
2715    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716      branch_chain *bcptr, BOOL utf, compile_data *cd)
2717    {
2718    while (bcptr != NULL && bcptr->current_branch >= code)
2719      {
2720      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721        return FALSE;
2722      bcptr = bcptr->outer;
2723      }
2724    return TRUE;
2725    }
2726    
2727    
2728    
2729    /*************************************************
2730    *        Base opcode of repeated opcodes         *
2731    *************************************************/
2732    
2733    /* Returns the base opcode for repeated single character type opcodes. If the
2734    opcode is not a repeated character type, it returns with the original value.
2735    
2736    Arguments:  c opcode
2737    Returns:    base opcode for the type
2738    */
2739    
2740    static pcre_uchar
2741    get_repeat_base(pcre_uchar c)
2742    {
2743    return (c > OP_TYPEPOSUPTO)? c :
2744           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2745           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2746           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2747           (c >= OP_STARI)?      OP_STARI :
2748                                 OP_STAR;
2749    }
2750    
2751    
2752    
2753    #ifdef SUPPORT_UCP
2754    /*************************************************
2755    *        Check a character and a property        *
2756    *************************************************/
2757    
2758    /* This function is called by check_auto_possessive() when a property item
2759    is adjacent to a fixed character.
2760    
2761    Arguments:
2762      c            the character
2763      ptype        the property type
2764      pdata        the data for the type
2765      negated      TRUE if it's a negated property (\P or \p{^)
2766    
2767    Returns:       TRUE if auto-possessifying is OK
2768    */
2769    
2770    static BOOL
2771    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772      BOOL negated)
2773    {
2774    const pcre_uint32 *p;
2775    const ucd_record *prop = GET_UCD(c);
2776    
2777    switch(ptype)
2778      {
2779      case PT_LAMP:
2780      return (prop->chartype == ucp_Lu ||
2781              prop->chartype == ucp_Ll ||
2782              prop->chartype == ucp_Lt) == negated;
2783    
2784      case PT_GC:
2785      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2786    
2787      case PT_PC:
2788      return (pdata == prop->chartype) == negated;
2789    
2790      case PT_SC:
2791      return (pdata == prop->script) == negated;
2792    
2793      /* These are specials */
2794    
2795      case PT_ALNUM:
2796      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2798    
2799      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800      means that Perl space and POSIX space are now identical. PCRE was changed
2801      at release 8.34. */
2802    
2803      case PT_SPACE:    /* Perl space */
2804      case PT_PXSPACE:  /* POSIX space */
2805      switch(c)
2806        {
2807        HSPACE_CASES:
2808        VSPACE_CASES:
2809        return negated;
2810    
2811        default:
2812        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813        }
2814      break;  /* Control never reaches here */
2815    
2816      case PT_WORD:
2817      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819              c == CHAR_UNDERSCORE) == negated;
2820    
2821      case PT_CLIST:
2822      p = PRIV(ucd_caseless_sets) + prop->caseset;
2823      for (;;)
2824        {
2825        if (c < *p) return !negated;
2826        if (c == *p++) return negated;
2827        }
2828      break;  /* Control never reaches here */
2829      }
2830    
2831    return FALSE;
2832    }
2833    #endif  /* SUPPORT_UCP */
2834    
2835    
2836    
2837    /*************************************************
2838    *        Fill the character property list        *
2839    *************************************************/
2840    
2841    /* Checks whether the code points to an opcode that can take part in auto-
2842    possessification, and if so, fills a list with its properties.
2843    
2844    Arguments:
2845      code        points to start of expression
2846      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847      fcc         points to case-flipping table
2848      list        points to output list
2849                  list[0] will be filled with the opcode
2850                  list[1] will be non-zero if this opcode
2851                    can match an empty character string
2852                  list[2..7] depends on the opcode
2853    
2854    Returns:      points to the start of the next opcode if *code is accepted
2855                  NULL if *code is not accepted
2856    */
2857    
2858    static const pcre_uchar *
2859    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860      const pcre_uint8 *fcc, pcre_uint32 *list)
2861    {
2862    pcre_uchar c = *code;
2863    pcre_uchar base;
2864    const pcre_uchar *end;
2865    pcre_uint32 chr;
2866    
2867    #ifdef SUPPORT_UCP
2868    pcre_uint32 *clist_dest;
2869    const pcre_uint32 *clist_src;
2870    #else
2871    utf = utf;  /* Suppress "unused parameter" compiler warning */
2872    #endif
2873    
2874    list[0] = c;
2875    list[1] = FALSE;
2876    code++;
2877    
2878    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879      {
2880      base = get_repeat_base(c);
2881      c -= (base - OP_STAR);
2882    
2883      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884        code += IMM2_SIZE;
2885    
2886      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2887    
2888      switch(base)
2889        {
2890        case OP_STAR:
2891        list[0] = OP_CHAR;
2892        break;
2893    
2894        case OP_STARI:
2895        list[0] = OP_CHARI;
2896        break;
2897    
2898        case OP_NOTSTAR:
2899        list[0] = OP_NOT;
2900        break;
2901    
2902        case OP_NOTSTARI:
2903        list[0] = OP_NOTI;
2904        break;
2905    
2906        case OP_TYPESTAR:
2907        list[0] = *code;
2908        code++;
2909        break;
2910        }
2911      c = list[0];
2912      }
2913    
2914    switch(c)
2915      {
2916      case OP_NOT_DIGIT:
2917      case OP_DIGIT:
2918      case OP_NOT_WHITESPACE:
2919      case OP_WHITESPACE:
2920      case OP_NOT_WORDCHAR:
2921      case OP_WORDCHAR:
2922      case OP_ANY:
2923      case OP_ALLANY:
2924      case OP_ANYNL:
2925      case OP_NOT_HSPACE:
2926      case OP_HSPACE:
2927      case OP_NOT_VSPACE:
2928      case OP_VSPACE:
2929      case OP_EXTUNI:
2930      case OP_EODN:
2931      case OP_EOD:
2932      case OP_DOLL:
2933      case OP_DOLLM:
2934      return code;
2935    
2936      case OP_CHAR:
2937      case OP_NOT:
2938      GETCHARINCTEST(chr, code);
2939      list[2] = chr;
2940      list[3] = NOTACHAR;
2941      return code;
2942    
2943      case OP_CHARI:
2944      case OP_NOTI:
2945      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946      GETCHARINCTEST(chr, code);
2947      list[2] = chr;
2948    
2949    #ifdef SUPPORT_UCP
2950      if (chr < 128 || (chr < 256 && !utf))
2951        list[3] = fcc[chr];
2952      else
2953        list[3] = UCD_OTHERCASE(chr);
2954    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955      list[3] = (chr < 256) ? fcc[chr] : chr;
2956    #else
2957      list[3] = fcc[chr];
2958    #endif
2959    
2960      /* The othercase might be the same value. */
2961    
2962      if (chr == list[3])
2963        list[3] = NOTACHAR;
2964      else
2965        list[4] = NOTACHAR;
2966      return code;
2967    
2968    #ifdef SUPPORT_UCP
2969      case OP_PROP:
2970      case OP_NOTPROP:
2971      if (code[0] != PT_CLIST)
2972        {
2973        list[2] = code[0];
2974        list[3] = code[1];
2975        return code + 2;
2976        }
2977    
2978      /* Convert only if we have enough space. */
2979    
2980      clist_src = PRIV(ucd_caseless_sets) + code[1];
2981      clist_dest = list + 2;
2982      code += 2;
2983    
2984      do {
2985         if (clist_dest >= list + 8)
2986           {
2987           /* Early return if there is not enough space. This should never
2988           happen, since all clists are shorter than 5 character now. */
2989           list[2] = code[0];
2990           list[3] = code[1];
2991           return code;
2992           }
2993         *clist_dest++ = *clist_src;
2994         }
2995      while(*clist_src++ != NOTACHAR);
2996    
2997      /* All characters are stored. The terminating NOTACHAR
2998      is copied form the clist itself. */
2999    
3000      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001      return code;
3002    #endif
3003    
3004      case OP_NCLASS:
3005      case OP_CLASS:
3006    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007      case OP_XCLASS:
3008      if (c == OP_XCLASS)
3009        end = code + GET(code, 0) - 1;
3010      else
3011    #endif
3012        end = code + 32 / sizeof(pcre_uchar);
3013    
3014      switch(*end)
3015        {
3016        case OP_CRSTAR:
3017        case OP_CRMINSTAR:
3018        case OP_CRQUERY:
3019        case OP_CRMINQUERY:
3020        case OP_CRPOSSTAR:
3021        case OP_CRPOSQUERY:
3022        list[1] = TRUE;
3023        end++;
3024        break;
3025    
3026        case OP_CRPLUS:
3027        case OP_CRMINPLUS:
3028        case OP_CRPOSPLUS:
3029        end++;
3030        break;
3031    
3032        case OP_CRRANGE:
3033        case OP_CRMINRANGE:
3034        case OP_CRPOSRANGE:
3035        list[1] = (GET2(end, 1) == 0);
3036        end += 1 + 2 * IMM2_SIZE;
3037        break;
3038        }
3039      list[2] = end - code;
3040      return end;
3041      }
3042    return NULL;    /* Opcode not accepted */
3043    }
3044    
3045    
3046    
3047    /*************************************************
3048    *    Scan further character sets for match       *
3049    *************************************************/
3050    
3051    /* Checks whether the base and the current opcode have a common character, in
3052    which case the base cannot be possessified.
3053    
3054    Arguments:
3055      code        points to the byte code
3056      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057      cd          static compile data
3058      base_list   the data list of the base opcode
3059    
3060    Returns:      TRUE if the auto-possessification is possible
3061    */
3062    
3063    static BOOL
3064    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066    {
3067    pcre_uchar c;
3068    pcre_uint32 list[8];
3069    const pcre_uint32 *chr_ptr;
3070    const pcre_uint32 *ochr_ptr;
3071    const pcre_uint32 *list_ptr;
3072    const pcre_uchar *next_code;
3073    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3074    const pcre_uchar *xclass_flags;
3075    #endif
3076    const pcre_uint8 *class_bitset;
3077    const pcre_uint8 *set1, *set2, *set_end;
3078    pcre_uint32 chr;
3079    BOOL accepted, invert_bits;
3080    
3081    /* Note: the base_list[1] contains whether the current opcode has greedy
3082    (represented by a non-zero value) quantifier. This is a different from
3083    other character type lists, which stores here that the character iterator
3084    matches to an empty string (also represented by a non-zero value). */
3085    
3086    for(;;)
3087      {
3088      /* All operations move the code pointer forward.
3089      Therefore infinite recursions are not possible. */
3090    
3091      c = *code;
3092    
3093      /* Skip over callouts */
3094    
3095      if (c == OP_CALLOUT)
3096        {
3097        code += PRIV(OP_lengths)[c];
3098        continue;
3099        }
3100    
3101      if (c == OP_ALT)
3102        {
3103        do code += GET(code, 1); while (*code == OP_ALT);
3104        c = *code;
3105        }
3106    
3107      switch(c)
3108        {
3109        case OP_END:
3110        case OP_KETRPOS:
3111        /* TRUE only in greedy case. The non-greedy case could be replaced by
3112        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3113        uses more memory, which we cannot get at this stage.) */
3114    
3115        return base_list[1] != 0;
3116    
3117        case OP_KET:
3118        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3119        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3120        cannot be converted to a possessive form. */
3121    
3122        if (base_list[1] == 0) return FALSE;
3123    
3124        switch(*(code - GET(code, 1)))
3125          {
3126          case OP_ASSERT:
3127          case OP_ASSERT_NOT:
3128          case OP_ASSERTBACK:
3129          case OP_ASSERTBACK_NOT:
3130          case OP_ONCE:
3131          case OP_ONCE_NC:
3132          /* Atomic sub-patterns and assertions can always auto-possessify their
3133          last iterator. */
3134          return TRUE;
3135          }
3136    
3137        code += PRIV(OP_lengths)[c];
3138        continue;
3139    
3140        case OP_ONCE:
3141        case OP_ONCE_NC:
3142        case OP_BRA:
3143        case OP_CBRA:
3144        next_code = code + GET(code, 1);
3145        code += PRIV(OP_lengths)[c];
3146    
3147        while (*next_code == OP_ALT)
3148          {
3149          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3150          code = next_code + 1 + LINK_SIZE;
3151          next_code += GET(next_code, 1);
3152          }
3153        continue;
3154    
3155        case OP_BRAZERO:
3156        case OP_BRAMINZERO:
3157    
3158        next_code = code + 1;
3159        if (*next_code != OP_BRA && *next_code != OP_CBRA
3160            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3161    
3162        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3163    
3164        /* The bracket content will be checked by the
3165        OP_BRA/OP_CBRA case above. */
3166        next_code += 1 + LINK_SIZE;
3167        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3168          return FALSE;
3169    
3170        code += PRIV(OP_lengths)[c];
3171        continue;
3172        }
3173    
3174      /* Check for a supported opcode, and load its properties. */
3175    
3176      code = get_chr_property_list(code, utf, cd->fcc, list);
3177      if (code == NULL) return FALSE;    /* Unsupported */
3178    
3179      /* If either opcode is a small character list, set pointers for comparing
3180      characters from that list with another list, or with a property. */
3181    
3182      if (base_list[0] == OP_CHAR)
3183        {
3184        chr_ptr = base_list + 2;
3185        list_ptr = list;
3186        }
3187      else if (list[0] == OP_CHAR)
3188        {
3189        chr_ptr = list + 2;
3190        list_ptr = base_list;
3191        }
3192    
3193      /* Character bitsets can also be compared to certain opcodes. */
3194    
3195      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3196    #ifdef COMPILE_PCRE8
3197          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3198          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3199    #endif
3200          )
3201        {
3202    #ifdef COMPILE_PCRE8
3203        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3204    #else
3205        if (base_list[0] == OP_CLASS)
3206    #endif
3207          {
3208          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3209          list_ptr = list;
3210          }
3211        else
3212          {
3213          set1 = (pcre_uint8 *)(code - list[2]);
3214          list_ptr = base_list;
3215          }
3216    
3217        invert_bits = FALSE;
3218        switch(list_ptr[0])
3219          {
3220          case OP_CLASS:
3221          case OP_NCLASS:
3222          set2 = (pcre_uint8 *)
3223            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3224          break;
3225    
3226    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3227          case OP_XCLASS:
3228          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3229          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3230          if ((*xclass_flags & XCL_MAP) == 0)
3231            {
3232            /* No bits are set for characters < 256. */
3233            if (list[1] == 0) return TRUE;
3234            /* Might be an empty repeat. */
3235            continue;
3236            }
3237          set2 = (pcre_uint8 *)(xclass_flags + 1);
3238          break;
3239    #endif
3240    
3241          case OP_NOT_DIGIT:
3242          invert_bits = TRUE;
3243          /* Fall through */
3244          case OP_DIGIT:
3245          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3246          break;
3247    
3248          case OP_NOT_WHITESPACE:
3249          invert_bits = TRUE;
3250          /* Fall through */
3251          case OP_WHITESPACE:
3252          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3253          break;
3254    
3255          case OP_NOT_WORDCHAR:
3256          invert_bits = TRUE;
3257          /* Fall through */
3258          case OP_WORDCHAR:
3259          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3260          break;
3261    
3262          default:
3263          return FALSE;
3264          }
3265    
3266        /* Because the sets are unaligned, we need
3267        to perform byte comparison here. */
3268        set_end = set1 + 32;
3269        if (invert_bits)
3270          {
3271          do
3272            {
3273            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3274            }
3275          while (set1 < set_end);
3276          }
3277        else
3278          {
3279          do
3280            {
3281            if ((*set1++ & *set2++) != 0) return FALSE;
3282            }
3283          while (set1 < set_end);
3284          }
3285    
3286        if (list[1] == 0) return TRUE;
3287        /* Might be an empty repeat. */
3288        continue;
3289        }
3290    
3291      /* Some property combinations also acceptable. Unicode property opcodes are
3292      processed specially; the rest can be handled with a lookup table. */
3293    
3294      else
3295        {
3296        pcre_uint32 leftop, rightop;
3297    
3298        leftop = base_list[0];
3299        rightop = list[0];
3300    
3301    #ifdef SUPPORT_UCP
3302        accepted = FALSE; /* Always set in non-unicode case. */
3303        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3304          {
3305          if (rightop == OP_EOD)
3306            accepted = TRUE;
3307          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3308            {
3309            int n;
3310            const pcre_uint8 *p;
3311            BOOL same = leftop == rightop;
3312            BOOL lisprop = leftop == OP_PROP;
3313            BOOL risprop = rightop == OP_PROP;
3314            BOOL bothprop = lisprop && risprop;
3315    
3316            /* There's a table that specifies how each combination is to be
3317            processed:
3318              0   Always return FALSE (never auto-possessify)
3319              1   Character groups are distinct (possessify if both are OP_PROP)
3320              2   Check character categories in the same group (general or particular)
3321              3   Return TRUE if the two opcodes are not the same
3322              ... see comments below
3323            */
3324    
3325            n = propposstab[base_list[2]][list[2]];
3326            switch(n)
3327              {
3328              case 0: break;
3329              case 1: accepted = bothprop; break;
3330              case 2: accepted = (base_list[3] == list[3]) != same; break;
3331              case 3: accepted = !same; break;
3332    
3333              case 4:  /* Left general category, right particular category */
3334              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3335              break;
3336    
3337              case 5:  /* Right general category, left particular category */
3338              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3339              break;
3340    
3341              /* This code is logically tricky. Think hard before fiddling with it.
3342              The posspropstab table has four entries per row. Each row relates to
3343              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3344              Only WORD actually needs all four entries, but using repeats for the
3345              others means they can all use the same code below.
3346    
3347              The first two entries in each row are Unicode general categories, and
3348              apply always, because all the characters they include are part of the
3349              PCRE character set. The third and fourth entries are a general and a
3350              particular category, respectively, that include one or more relevant
3351              characters. One or the other is used, depending on whether the check
3352              is for a general or a particular category. However, in both cases the
3353              category contains more characters than the specials that are defined
3354              for the property being tested against. Therefore, it cannot be used
3355              in a NOTPROP case.
3356    
3357              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3358              Underscore is covered by ucp_P or ucp_Po. */
3359    
3360              case 6:  /* Left alphanum vs right general category */
3361              case 7:  /* Left space vs right general category */
3362              case 8:  /* Left word vs right general category */
3363              p = posspropstab[n-6];
3364              accepted = risprop && lisprop ==
3365                (list[3] != p[0] &&
3366                 list[3] != p[1] &&
3367                (list[3] != p[2] || !lisprop));
3368              break;
3369    
3370              case 9:   /* Right alphanum vs left general category */
3371              case 10:  /* Right space vs left general category */
3372              case 11:  /* Right word vs left general category */
3373              p = posspropstab[n-9];
3374              accepted = lisprop && risprop ==
3375                (base_list[3] != p[0] &&
3376                 base_list[3] != p[1] &&
3377                (base_list[3] != p[2] || !risprop));
3378              break;
3379    
3380              case 12:  /* Left alphanum vs right particular category */
3381              case 13:  /* Left space vs right particular category */
3382              case 14:  /* Left word vs right particular category */
3383              p = posspropstab[n-12];
3384              accepted = risprop && lisprop ==
3385                (catposstab[p[0]][list[3]] &&
3386                 catposstab[p[1]][list[3]] &&
3387                (list[3] != p[3] || !lisprop));
3388              break;
3389    
3390              case 15:  /* Right alphanum vs left particular category */
3391              case 16:  /* Right space vs left particular category */
3392              case 17:  /* Right word vs left particular category */
3393              p = posspropstab[n-15];
3394              accepted = lisprop && risprop ==
3395                (catposstab[p[0]][base_list[3]] &&
3396                 catposstab[p[1]][base_list[3]] &&
3397                (base_list[3] != p[3] || !risprop));
3398              break;
3399              }
3400            }
3401          }
3402    
3403        else
3404    #endif  /* SUPPORT_UCP */
3405    
3406        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3407               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3408               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3409    
3410        if (!accepted)
3411          return FALSE;
3412    
3413        if (list[1] == 0) return TRUE;
3414        /* Might be an empty repeat. */
3415        continue;
3416        }
3417    
3418      /* Control reaches here only if one of the items is a small character list.
3419      All characters are checked against the other side. */
3420    
3421      do
3422        {
3423        chr = *chr_ptr;
3424    
3425        switch(list_ptr[0])
3426          {
3427          case OP_CHAR:
3428          ochr_ptr = list_ptr + 2;
3429          do
3430            {
3431            if (chr == *ochr_ptr) return FALSE;
3432            ochr_ptr++;
3433            }
3434          while(*ochr_ptr != NOTACHAR);
3435          break;
3436    
3437          case OP_NOT:
3438          ochr_ptr = list_ptr + 2;
3439          do
3440            {
3441            if (chr == *ochr_ptr)
3442              break;
3443            ochr_ptr++;
3444            }
3445          while(*ochr_ptr != NOTACHAR);
3446          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3447          break;
3448    
3449          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3450          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3451    
3452          case OP_DIGIT:
3453          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3454          break;
3455    
3456          case OP_NOT_DIGIT:
3457          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3458          break;
3459    
3460          case OP_WHITESPACE:
3461          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3462          break;
3463    
3464          case OP_NOT_WHITESPACE:
3465          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3466          break;
3467    
3468          case OP_WORDCHAR:
3469          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3470          break;
3471    
3472          case OP_NOT_WORDCHAR:
3473          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3474          break;
3475    
3476          case OP_HSPACE:
3477          switch(chr)
3478            {
3479            HSPACE_CASES: return FALSE;
3480            default: break;
3481            }
3482          break;
3483    
3484          case OP_NOT_HSPACE:
3485          switch(chr)
3486            {
3487            HSPACE_CASES: break;
3488            default: return FALSE;
3489            }
3490          break;
3491    
3492          case OP_ANYNL:
3493          case OP_VSPACE:
3494          switch(chr)
3495            {
3496            VSPACE_CASES: return FALSE;
3497            default: break;
3498            }
3499          break;
3500    
3501          case OP_NOT_VSPACE:
3502          switch(chr)
3503            {
3504            VSPACE_CASES: break;
3505            default: return FALSE;
3506            }
3507          break;
3508    
3509          case OP_DOLL:
3510          case OP_EODN:
3511          switch (chr)
3512            {
3513            case CHAR_CR:
3514            case CHAR_LF:
3515            case CHAR_VT:
3516            case CHAR_FF:
3517            case CHAR_NEL:
3518    #ifndef EBCDIC
3519            case 0x2028:
3520            case 0x2029:
3521    #endif  /* Not EBCDIC */
3522            return FALSE;
3523            }
3524          break;
3525    
3526          case OP_EOD:    /* Can always possessify before \z */
3527          break;
3528    
3529    #ifdef SUPPORT_UCP
3530          case OP_PROP:
3531          case OP_NOTPROP:
3532          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3533                list_ptr[0] == OP_NOTPROP))
3534            return FALSE;
3535          break;
3536    #endif
3537    
3538          case OP_NCLASS:
3539          if (chr > 255) return FALSE;
3540          /* Fall through */
3541    
3542          case OP_CLASS:
3543          if (chr > 255) break;
3544          class_bitset = (pcre_uint8 *)
3545            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3546          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3547          break;
3548    
3549    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3550          case OP_XCLASS:
3551          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3552              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3553          break;
3554    #endif
3555    
3556          default:
3557          return FALSE;
3558          }
3559    
3560        chr_ptr++;
3561        }
3562      while(*chr_ptr != NOTACHAR);
3563    
3564      /* At least one character must be matched from this opcode. */
3565    
3566      if (list[1] == 0) return TRUE;
3567      }
3568    
3569    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3570    but some compilers complain about an unreachable statement. */
3571    
3572    }
3573    
3574    
3575    
3576    /*************************************************
3577    *    Scan compiled regex for auto-possession     *
3578    *************************************************/
3579    
3580    /* Replaces single character iterations with their possessive alternatives
3581    if appropriate. This function modifies the compiled opcode!
3582    
3583    Arguments:
3584      code        points to start of the byte code
3585      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3586      cd          static compile data
3587    
3588    Returns:      nothing
3589    */
3590    
3591    static void
3592    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3593    {
3594    register pcre_uchar c;
3595    const pcre_uchar *end;
3596    pcre_uchar *repeat_opcode;
3597    pcre_uint32 list[8];
3598    
3599    for (;;)
3600      {
3601      c = *code;
3602    
3603      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3604        {
3605        c -= get_repeat_base(c) - OP_STAR;
3606        end = (c <= OP_MINUPTO) ?
3607          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3608        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3609    
3610        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3611          {
3612          switch(c)
3613            {
3614            case OP_STAR:
3615            *code += OP_POSSTAR - OP_STAR;
3616            break;
3617    
3618            case OP_MINSTAR:
3619            *code += OP_POSSTAR - OP_MINSTAR;
3620            break;
3621    
3622            case OP_PLUS:
3623            *code += OP_POSPLUS - OP_PLUS;
3624            break;
3625    
3626            case OP_MINPLUS:
3627            *code += OP_POSPLUS - OP_MINPLUS;
3628            break;
3629    
3630            case OP_QUERY:
3631            *code += OP_POSQUERY - OP_QUERY;
3632            break;
3633    
3634            case OP_MINQUERY:
3635            *code += OP_POSQUERY - OP_MINQUERY;
3636            break;
3637    
3638            case OP_UPTO:
3639            *code += OP_POSUPTO - OP_UPTO;
3640            break;
3641    
3642            case OP_MINUPTO:
3643            *code += OP_MINUPTO - OP_UPTO;
3644            break;
3645            }