/[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 96 by nigel, Fri Mar 2 13:10:43 2007 UTC revision 1389 by ph10, Tue Nov 5 18:05:29 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
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    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* 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. */
70    
71    #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
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #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         *
95  *************************************************/  *************************************************/
# Line 72  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    /* 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 COMPILE_WORK_SIZE (4096)  #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  #if !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       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -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       0,      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,      0, -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 106  static const short int escapes[] = { Line 203  static const short int escapes[] = {
203  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
206  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
207  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
208  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
209  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
210  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
211  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
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 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
215  /*  D0 */   '}',     0,      0,       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,     0, -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,
219  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
220  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 222  static const short int escapes[] = {
222  #endif  #endif
223    
224    
225  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
227  as this is assumed for handling case independence. */  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  static const char *const posix_names[] = {  platforms. */
230    "alpha", "lower", "upper",  
231    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
232    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
233      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;
236    
237    static const char verbnames[] =
238      "\0"                       /* Empty name is a shorthand for MARK */
239      STRING_MARK0
240      STRING_ACCEPT0
241      STRING_COMMIT0
242      STRING_F0
243      STRING_FAIL0
244      STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248    static const verbitem verbs[] = {
249      { 0, -1,        OP_MARK },
250      { 4, -1,        OP_MARK },
251      { 6, OP_ACCEPT, -1 },
252      { 6, OP_COMMIT, -1 },
253      { 1, OP_FAIL,   -1 },
254      { 4, OP_FAIL,   -1 },
255      { 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);
261    
262    
263    /* Tables of names of POSIX character classes and their lengths. The names are
264    now all in a single string, to reduce the number of relocations when a shared
265    library is dynamically loaded. The list of lengths is terminated by a zero
266    length entry. The first three must be alpha, lower, upper, as this is assumed
267    for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270    static const char posix_names[] =
271      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
272      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
273      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
274      STRING_word0  STRING_xdigit;
275    
276  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
285  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
286  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 164  static const int posix_class_maps[] = { Line 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312    Unicode property escapes. */
313    
314    #ifdef SUPPORT_UCP
315    static const pcre_uchar string_PNd[]  = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pNd[]  = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    static const pcre_uchar string_PXsp[] = {
322      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
323      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324    static const pcre_uchar string_pXsp[] = {
325      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
326      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
327    static const pcre_uchar string_PXwd[] = {
328      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_pXwd[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    
334    static const pcre_uchar *substitutes[] = {
335      string_PNd,           /* \D */
336      string_pNd,           /* \d */
337      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339      string_PXwd,          /* \W */
340      string_pXwd           /* \w */
341    };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349    static const pcre_uchar string_pL[] =   {
350      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352    static const pcre_uchar string_pLl[] =  {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_pLu[] =  {
356      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_pXan[] = {
359      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_h[] =    {
362      CHAR_BACKSLASH, CHAR_h, '\0' };
363    static const pcre_uchar string_pXps[] = {
364      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
365      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
366    static const pcre_uchar string_PL[] =   {
367      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
368      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
369    static const pcre_uchar string_PLl[] =  {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    static const pcre_uchar string_PLu[] =  {
373      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
374      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
375    static const pcre_uchar string_PXan[] = {
376      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
377      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
378    static const pcre_uchar string_H[] =    {
379      CHAR_BACKSLASH, CHAR_H, '\0' };
380    static const pcre_uchar string_PXps[] = {
381      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
382      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
383    
384    static const pcre_uchar *posix_substitutes[] = {
385      string_pL,            /* alpha */
386      string_pLl,           /* lower */
387      string_pLu,           /* upper */
388      string_pXan,          /* alnum */
389      NULL,                 /* ascii */
390      string_h,             /* blank */
391      NULL,                 /* cntrl */
392      string_pNd,           /* digit */
393      NULL,                 /* graph */
394      NULL,                 /* print */
395      NULL,                 /* punct */
396      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398      NULL,                 /* xdigit */
399      /* Negated cases */
400      string_PL,            /* ^alpha */
401      string_PLl,           /* ^lower */
402      string_PLu,           /* ^upper */
403      string_PXan,          /* ^alnum */
404      NULL,                 /* ^ascii */
405      string_H,             /* ^blank */
406      NULL,                 /* ^cntrl */
407      string_PNd,           /* ^digit */
408      NULL,                 /* ^graph */
409      NULL,                 /* ^print */
410      NULL,                 /* ^punct */
411      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413      NULL                  /* ^xdigit */
414    };
415    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
416    #endif
417    
418  #define STRING(a)  # a  #define STRING(a)  # a
419  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 421  static const int posix_class_maps[] = {
421  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
422  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
423  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
424  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
425    the number of relocations needed when a shared library is loaded dynamically,
426  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
427    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
428    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
429    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
430    "unrecognized character follows \\",  
431    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
432    substring, so that the whole string ends with \0\0, which can be detected when
433    counting through. */
434    
435    static const char error_texts[] =
436      "no error\0"
437      "\\ at end of pattern\0"
438      "\\c at end of pattern\0"
439      "unrecognized character follows \\\0"
440      "numbers out of order in {} quantifier\0"
441    /* 5 */    /* 5 */
442    "number too big in {} quantifier",    "number too big in {} quantifier\0"
443    "missing terminating ] for character class",    "missing terminating ] for character class\0"
444    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
445    "range out of order in character class",    "range out of order in character class\0"
446    "nothing to repeat",    "nothing to repeat\0"
447    /* 10 */    /* 10 */
448    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
449    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
450    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
451    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
452    "missing )",    "missing )\0"
453    /* 15 */    /* 15 */
454    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
455    "erroffset passed as NULL",    "erroffset passed as NULL\0"
456    "unknown option bit(s) set",    "unknown option bit(s) set\0"
457    "missing ) after comment",    "missing ) after comment\0"
458    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
459    /* 20 */    /* 20 */
460    "regular expression too large",    "regular expression is too large\0"
461    "failed to get memory",    "failed to get memory\0"
462    "unmatched parentheses",    "unmatched parentheses\0"
463    "internal error: code overflow",    "internal error: code overflow\0"
464    "unrecognized character after (?<",    "unrecognized character after (?<\0"
465    /* 25 */    /* 25 */
466    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
467    "malformed number or name after (?(",    "malformed number or name after (?(\0"
468    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
469    "assertion expected after (?(",    "assertion expected after (?(\0"
470    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
471    /* 30 */    /* 30 */
472    "unknown POSIX class name",    "unknown POSIX class name\0"
473    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
474    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
475    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large",    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)",    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
480    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
481    "number after (?C is > 255",    "number after (?C is > 255\0"
482    "closing ) for (?C expected",    "closing ) for (?C expected\0"
483    /* 40 */    /* 40 */
484    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
485    "unrecognized character after (?P",    "unrecognized character after (?P\0"
486    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
487    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
488    "invalid UTF-8 string",    "invalid UTF-8 string\0"
489    /* 45 */    /* 45 */
490    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
491    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
492    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
493    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
494    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
495    /* 50 */    /* 50 */
496    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
497    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
498    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
499    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
500    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
501    /* 55 */    /* 55 */
502    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
503    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
504    "\\g is not followed by an (optionally braced) non-zero number"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
505  };    "a numbered reference must not be zero\0"
506      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507      /* 60 */
508      "(*VERB) not recognized or malformed\0"
509      "number is too big\0"
510      "subpattern name expected\0"
511      "digit expected after (?+\0"
512      "] is an invalid data character in JavaScript compatibility mode\0"
513      /* 65 */
514      "different names for subpatterns of the same number are not allowed\0"
515      "(*MARK) must have an argument\0"
516      "this version of PCRE is not compiled with Unicode property support\0"
517      "\\c must be followed by an ASCII character\0"
518      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
519      /* 70 */
520      "internal error: unknown opcode in find_fixedlength()\0"
521      "\\N is not supported in a class\0"
522      "too many forward references\0"
523      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524      "invalid UTF-16 string\0"
525      /* 75 */
526      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527      "character value in \\u.... sequence is too large\0"
528      "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      ;
536    
537  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
538  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 262  For convenience, we use the same bit def Line 550  For convenience, we use the same bit def
550    
551  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
552    
553  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
554  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
555    into a subtraction and unsigned comparison). */
556    
557    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
558    
559    #ifndef EBCDIC
560    
561    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
562    UTF-8 mode. */
563    
564    static const pcre_uint8 digitab[] =
565    {    {
566    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
567    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 298  static const unsigned char digitab[] = Line 596  static const unsigned char digitab[] =
596    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
597    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
598    
599  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
600  static const unsigned char digitab[] =  
601    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
602    
603    static const pcre_uint8 digitab[] =
604    {    {
605    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
606    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 312  static const unsigned char digitab[] = Line 613  static const unsigned char digitab[] =
613    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
614    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
615    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
616    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
617    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
618    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
619    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 334  static const unsigned char digitab[] = Line 635  static const unsigned char digitab[] =
635    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
636    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
637    
638  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
639    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
640    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
641    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 346  static const unsigned char ebcdic_charta Line 647  static const unsigned char ebcdic_charta
647    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
648    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
649    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
650    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
651    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
652    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
653    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 370  static const unsigned char ebcdic_charta Line 671  static const unsigned char ebcdic_charta
671  #endif  #endif
672    
673    
674  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
675    between adjacent character-type opcodes. The left-hand (repeated) opcode is
676    used to select the row, and the right-hand opcode is use to select the column.
677    A value of 1 means that auto-possessification is OK. For example, the second
678    value in the first row means that \D+\d can be turned into \D++\d.
679    
680    The Unicode property types (\P and \p) have to be present to fill out the table
681    because of what their opcode values are, but the table values should always be
682    zero because property types are handled separately in the code. The last four
683    columns apply to items that cannot be repeated, so there is no need to have
684    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
685    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
686    
687    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
688    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
689    
690    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
691    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
692      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
693      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
694      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
696      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
697      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
698      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
699      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
703      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
704      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
705      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
706      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
707      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
708      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
709    };
710    
711    
712    /* This table is used to check whether auto-possessification is possible
713    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
714    left-hand (repeated) opcode is used to select the row, and the right-hand
715    opcode is used to select the column. The values are as follows:
716    
717      0   Always return FALSE (never auto-possessify)
718      1   Character groups are distinct (possessify if both are OP_PROP)
719      2   Check character categories in the same group (general or particular)
720      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
721    
722      4   Check left general category vs right particular category
723      5   Check right general category vs left particular category
724    
725      6   Left alphanum vs right general category
726      7   Left space vs right general category
727      8   Left word vs right general category
728    
729      9   Right alphanum vs left general category
730     10   Right space vs left general category
731     11   Right word vs left general category
732    
733     12   Left alphanum vs right particular category
734     13   Left space vs right particular category
735     14   Left word vs right particular category
736    
737     15   Right alphanum vs left particular category
738     16   Right space vs left particular category
739     17   Right word vs left particular category
740    */
741    
742    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
743    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
744      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
745      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
746      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
747      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
748      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
749      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
750      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
751      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
752      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
753      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
754      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
755    };
756    
757    /* This table is used to check whether auto-possessification is possible
758    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
759    specifies a general category and the other specifies a particular category. The
760    row is selected by the general category and the column by the particular
761    category. The value is 1 if the particular category is not part of the general
762    category. */
763    
764    static const pcre_uint8 catposstab[7][30] = {
765    /* 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 */
766      { 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 */
767      { 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772      { 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 */
773    };
774    
775    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
776    a general or particular category. The properties in each row are those
777    that apply to the character set in question. Duplication means that a little
778    unnecessary work is done when checking, but this keeps things much simpler
779    because they can all use the same code. For more details see the comment where
780    this table is used.
781    
782    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
783    "space", but from Perl 5.18 it's included, so both categories are treated the
784    same here. */
785    
786    static const pcre_uint8 posspropstab[3][4] = {
787      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
788      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
789      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
790    };
791    
792    /* This table is used when converting repeating opcodes into possessified
793    versions as a result of an explicit possessive quantifier such as ++. A zero
794    value means there is no possessified version - in those cases the item in
795    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
796    because all relevant opcodes are less than that. */
797    
798    static const pcre_uint8 opcode_possessify[] = {
799      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
800      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
801    
802      0,                       /* NOTI */
803      OP_POSSTAR, 0,           /* STAR, MINSTAR */
804      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
805      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
806      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
807      0,                       /* EXACT */
808      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
809    
810      OP_POSSTARI, 0,          /* STARI, MINSTARI */
811      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
812      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
813      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
814      0,                       /* EXACTI */
815      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
816    
817      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
818      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
819      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
820      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
821      0,                       /* NOTEXACT */
822      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
823    
824      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
825      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
826      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
827      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
828      0,                       /* NOTEXACTI */
829      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
830    
831      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
832      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
833      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
834      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
835      0,                       /* TYPEEXACT */
836      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
837    
838      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
839      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
840      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
841      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
842      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
843    
844      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
845      0, 0,                    /* REF, REFI */
846      0, 0,                    /* DNREF, DNREFI */
847      0, 0                     /* RECURSE, CALLOUT */
848    };
849    
850    
851    
852    /*************************************************
853    *            Find an error text                  *
854    *************************************************/
855    
856    /* The error texts are now all in one long string, to save on relocations. As
857    some of the text is of unknown length, we can't use a table of offsets.
858    Instead, just count through the strings. This is not a performance issue
859    because it happens only when there has been a compilation error.
860    
861    Argument:   the error number
862    Returns:    pointer to the error string
863    */
864    
865    static const char *
866    find_error_text(int n)
867    {
868    const char *s = error_texts;
869    for (; n > 0; n--)
870      {
871      while (*s++ != CHAR_NULL) {};
872      if (*s == CHAR_NULL) return "Error text not found (please report)";
873      }
874    return s;
875    }
876    
877    
878    
879    /*************************************************
880    *           Expand the workspace                 *
881    *************************************************/
882    
883    /* This function is called during the second compiling phase, if the number of
884    forward references fills the existing workspace, which is originally a block on
885    the stack. A larger block is obtained from malloc() unless the ultimate limit
886    has been reached or the increase will be rather small.
887    
888    Argument: pointer to the compile data block
889    Returns:  0 if all went well, else an error number
890    */
891    
892    static int
893    expand_workspace(compile_data *cd)
894    {
895    pcre_uchar *newspace;
896    int newsize = cd->workspace_size * 2;
897    
898    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
899    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
900        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
901     return ERR72;
902    
903    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
904    if (newspace == NULL) return ERR21;
905    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
906    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
907    if (cd->workspace_size > COMPILE_WORK_SIZE)
908      (PUBL(free))((void *)cd->start_workspace);
909    cd->start_workspace = newspace;
910    cd->workspace_size = newsize;
911    return 0;
912    }
913    
914    
915    
916    /*************************************************
917    *            Check for counted repeat            *
918    *************************************************/
919    
920    /* This function is called when a '{' is encountered in a place where it might
921    start a quantifier. It looks ahead to see if it really is a quantifier or not.
922    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
923    where the ddds are digits.
924    
925    Arguments:
926      p         pointer to the first char after '{'
927    
928    Returns:    TRUE or FALSE
929    */
930    
931  static BOOL  static BOOL
932    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
933      int *, branch_chain *, compile_data *, int *);  {
934    if (!IS_DIGIT(*p)) return FALSE;
935    p++;
936    while (IS_DIGIT(*p)) p++;
937    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
938    
939    if (*p++ != CHAR_COMMA) return FALSE;
940    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
941    
942    if (!IS_DIGIT(*p)) return FALSE;
943    p++;
944    while (IS_DIGIT(*p)) p++;
945    
946    return (*p == CHAR_RIGHT_CURLY_BRACKET);
947    }
948    
949    
950    
# Line 383  static BOOL Line 953  static BOOL
953  *************************************************/  *************************************************/
954    
955  /* 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
956  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
957  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.
958  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
959  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
960  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
961    
962  Arguments:  Arguments:
963    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
964      chptr          points to a returned data character
965    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
966    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
967    options        the options bits    options        the options bits
968    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
969    
970  Returns:         zero or positive => a data character  Returns:         zero => a data character
971                   negative => a special escape sequence                   positive => a special escape sequence
972                   on error, errorptr is set                   negative => a back reference
973                     on error, errorcodeptr is set
974  */  */
975    
976  static int  static int
977  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
978    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
979  {  {
980  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
981  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
982  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
983    pcre_uint32 c;
984    int escape = 0;
985    int i;
986    
987  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
988  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
989    
990  /* 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. */
991    
992  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
993    
994  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
995  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.
996  Otherwise further processing may be required. */  Otherwise further processing may be required. */
997    
998  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
999  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1000  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1001    else if ((i = escapes[c - CHAR_0]) != 0)
1002  #else          /* EBCDIC coding */    { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1003  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  
1004  else if ((i = escapes[c - 0x48]) != 0)  c = i;  #else           /* EBCDIC coding */
1005    /* Not alphanumeric */
1006    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1007    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1008  #endif  #endif
1009    
1010  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1011    
1012  else  else
1013    {    {
1014    const uschar *oldptr;    const pcre_uchar *oldptr;
1015    BOOL braced, negated;    BOOL braced, negated, overflow;
1016      int s;
1017    
1018    switch (c)    switch (c)
1019      {      {
1020      /* 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
1021      error. */      error. */
1022    
1023      case 'l':      case CHAR_l:
1024      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1025      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1026      break;      break;
1027    
1028      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
1029      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1030      This is a Perl 5.10 feature. */        {
1031          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1032          Otherwise it is a lowercase u letter. */
1033          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1034            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1035            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1037            {
1038            c = 0;
1039            for (i = 0; i < 4; ++i)
1040              {
1041              register pcre_uint32 cc = *(++ptr);
1042    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1044              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1045    #else           /* EBCDIC coding */
1046              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1047              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048    #endif
1049              }
1050    
1051    #if defined COMPILE_PCRE8
1052            if (c > (utf ? 0x10ffffU : 0xffU))
1053    #elif defined COMPILE_PCRE16
1054            if (c > (utf ? 0x10ffffU : 0xffffU))
1055    #elif defined COMPILE_PCRE32
1056            if (utf && c > 0x10ffffU)
1057    #endif
1058              {
1059              *errorcodeptr = ERR76;
1060              }
1061            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1062            }
1063          }
1064        else
1065          *errorcodeptr = ERR37;
1066        break;
1067    
1068        case CHAR_U:
1069        /* In JavaScript, \U is an uppercase U letter. */
1070        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1071        break;
1072    
1073        /* In a character class, \g is just a literal "g". Outside a character
1074        class, \g must be followed by one of a number of specific things:
1075    
1076        (1) A number, either plain or braced. If positive, it is an absolute
1077        backreference. If negative, it is a relative backreference. This is a Perl
1078        5.10 feature.
1079    
1080        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1081        is part of Perl's movement towards a unified syntax for back references. As
1082        this is synonymous with \k{name}, we fudge it up by pretending it really
1083        was \k.
1084    
1085        (3) For Oniguruma compatibility we also support \g followed by a name or a
1086        number either in angle brackets or in single quotes. However, these are
1087        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1088        the ESC_g code (cf \k). */
1089    
1090        case CHAR_g:
1091        if (isclass) break;
1092        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1093          {
1094          escape = ESC_g;
1095          break;
1096          }
1097    
1098      case 'g':      /* Handle the Perl-compatible cases */
1099      if (ptr[1] == '{')  
1100        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1101        {        {
1102          const pcre_uchar *p;
1103          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1104            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1105          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1106            {
1107            escape = ESC_k;
1108            break;
1109            }
1110        braced = TRUE;        braced = TRUE;
1111        ptr++;        ptr++;
1112        }        }
1113      else braced = FALSE;      else braced = FALSE;
1114    
1115      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1116        {        {
1117        negated = TRUE;        negated = TRUE;
1118        ptr++;        ptr++;
1119        }        }
1120      else negated = FALSE;      else negated = FALSE;
1121    
1122      c = 0;      /* The integer range is limited by the machine's int representation. */
1123      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1124        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1125        while (IS_DIGIT(ptr[1]))
1126          {
1127          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1128            {
1129            overflow = TRUE;
1130            break;
1131            }
1132          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1133          }
1134        if (overflow) /* Integer overflow */
1135          {
1136          while (IS_DIGIT(ptr[1]))
1137            ptr++;
1138          *errorcodeptr = ERR61;
1139          break;
1140          }
1141    
1142      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1143        {        {
1144        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1145        return 0;        break;
1146          }
1147    
1148        if (s == 0)
1149          {
1150          *errorcodeptr = ERR58;
1151          break;
1152        }        }
1153    
1154      if (negated)      if (negated)
1155        {        {
1156        if (c > bracount)        if (s > bracount)
1157          {          {
1158          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1159          return 0;          break;
1160          }          }
1161        c = bracount - (c - 1);        s = bracount - (s - 1);
1162        }        }
1163    
1164      c = -(ESC_REF + c);      escape = -s;
1165      break;      break;
1166    
1167      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1168      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
1169      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1170        recommended to avoid the ambiguities in the old syntax.
1171    
1172      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
1173      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
1174      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
1175      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
1176      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
1177      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
1178      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1179    
1180        Inside a character class, \ followed by a digit is always either a literal
1181        8 or 9 or an octal number. */
1182    
1183      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1184      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1185    
1186      if (!isclass)      if (!isclass)
1187        {        {
1188        oldptr = ptr;        oldptr = ptr;
1189        c -= '0';        /* The integer range is limited by the machine's int representation. */
1190        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1191          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1192        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1193            {
1194            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1195              {
1196              overflow = TRUE;
1197              break;
1198              }
1199            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1200            }
1201          if (overflow) /* Integer overflow */
1202            {
1203            while (IS_DIGIT(ptr[1]))
1204              ptr++;
1205            *errorcodeptr = ERR61;
1206            break;
1207            }
1208          if (s < 8 || s <= bracount)  /* Check for back reference */
1209          {          {
1210          c = -(ESC_REF + c);          escape = -s;
1211          break;          break;
1212          }          }
1213        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1214        }        }
1215    
1216      /* 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
1217      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
1218      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
1219        changed so as not to insert the binary zero. */
1220    
1221      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1222        {  
1223        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1224    
1225      /* \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
1226      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
1227      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
1228      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,
1229      than 3 octal digits. */      but no more than 3 octal digits. */
1230    
1231      case '0':      case CHAR_0:
1232      c -= '0';      c -= CHAR_0;
1233      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1234          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1235      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1236        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1237    #endif
1238      break;      break;
1239    
1240      /* \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
1241      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. */  
1242    
1243      case 'x':      case CHAR_o:
1244      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1245        {        {
1246        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1247        c = 0;        c = 0;
1248        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1249          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1250          {          {
1251          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1252          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1253          count++;  #ifdef COMPILE_PCRE32
1254            if (c >= 0x20000000l) { overflow = TRUE; break; }
1255  #if !EBCDIC    /* ASCII coding */  #endif
1256          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1257          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1258  #else          /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1259          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1260          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1261    #elif defined COMPILE_PCRE32
1262            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1263  #endif  #endif
1264          }          }
1265          if (overflow)
       if (*pt == '}')  
1266          {          {
1267          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1268          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1269          }          }
1270          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1271        /* If the sequence of hex digits does not end with '}', then we don't          {
1272        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1273            }
1274          else *errorcodeptr = ERR80;
1275        }        }
1276        break;
1277    
1278      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1279        numbers. Otherwise it is a lowercase x letter. */
1280    
1281      c = 0;      case CHAR_x:
1282      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1283        {        {
1284        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1285        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1286  #if !EBCDIC    /* ASCII coding */          {
1287        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1288        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1289  #else          /* EBCDIC coding */            {
1290        if (cc <= 'z') cc += 64;              /* Convert to upper case */            register pcre_uint32 cc = *(++ptr);
1291        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1292              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1293              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1294    #else           /* EBCDIC coding */
1295              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1296              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1297  #endif  #endif
1298        }            }
1299            }
1300          }    /* End JavaScript handling */
1301    
1302        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1303        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1304        digits. If not, { used to be treated as a data character. However, Perl
1305        seems to read hex digits up to the first non-such, and ignore the rest, so
1306        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1307        now gives an error. */
1308    
1309        else
1310          {
1311          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1312            {
1313            ptr += 2;
1314            c = 0;
1315            overflow = FALSE;
1316            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1317              {
1318              register pcre_uint32 cc = *ptr++;
1319              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1320    
1321    #ifdef COMPILE_PCRE32
1322              if (c >= 0x10000000l) { overflow = TRUE; break; }
1323    #endif
1324    
1325    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1326              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1327              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1328    #else           /* EBCDIC coding */
1329              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1330              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1331    #endif
1332    
1333    #if defined COMPILE_PCRE8
1334              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1335    #elif defined COMPILE_PCRE16
1336              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1337    #elif defined COMPILE_PCRE32
1338              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1339    #endif
1340              }
1341    
1342            if (overflow)
1343              {
1344              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1345              *errorcodeptr = ERR34;
1346              }
1347    
1348            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1349              {
1350              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1351              }
1352    
1353            /* If the sequence of hex digits does not end with '}', give an error.
1354            We used just to recognize this construct and fall through to the normal
1355            \x handling, but nowadays Perl gives an error, which seems much more
1356            sensible, so we do too. */
1357    
1358            else *errorcodeptr = ERR79;
1359            }   /* End of \x{} processing */
1360    
1361          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1362    
1363          else
1364            {
1365            c = 0;
1366            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1367              {
1368              pcre_uint32 cc;                          /* Some compilers don't like */
1369              cc = *(++ptr);                           /* ++ in initializers */
1370    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1371              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1372              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1373    #else           /* EBCDIC coding */
1374              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1375              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1376    #endif
1377              }
1378            }     /* End of \xdd handling */
1379          }       /* End of Perl-style \x handling */
1380      break;      break;
1381    
1382      /* 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.
1383      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
1384        coding is ASCII-specific, but then the whole concept of \cx is
1385      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1386    
1387      case 'c':      case CHAR_c:
1388      c = *(++ptr);      c = *(++ptr);
1389      if (c == 0)      if (c == CHAR_NULL)
1390        {        {
1391        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1392        return 0;        break;
1393        }        }
1394    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1395  #if !EBCDIC    /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1396      if (c >= 'a' && c <= 'z') c -= 32;        {
1397          *errorcodeptr = ERR68;
1398          break;
1399          }
1400        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1401      c ^= 0x40;      c ^= 0x40;
1402  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1403      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1404      c ^= 0xC0;      c ^= 0xC0;
1405  #endif  #endif
1406      break;      break;
1407    
1408      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1409      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1410      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
1411      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
1412      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1413    
1414      default:      default:
1415      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1422  else
1422      }      }
1423    }    }
1424    
1425    /* Perl supports \N{name} for character names, as well as plain \N for "not
1426    newline". PCRE does not support \N{name}. However, it does support
1427    quantification such as \N{2,3}. */
1428    
1429    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1430         !is_counted_repeat(ptr+2))
1431      *errorcodeptr = ERR37;
1432    
1433    /* If PCRE_UCP is set, we change the values for \d etc. */
1434    
1435    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1436      escape += (ESC_DU - ESC_D);
1437    
1438    /* Set the pointer to the final character before returning. */
1439    
1440  *ptrptr = ptr;  *ptrptr = ptr;
1441  return c;  *chptr = c;
1442    return escape;
1443  }  }
1444    
1445    
# Line 656  escape sequence. Line 1457  escape sequence.
1457  Argument:  Argument:
1458    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1459    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
1460    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
1461      pdataptr       points to an unsigned int that is set to the detailed property value
1462    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1463    
1464  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
1465  */  */
1466    
1467  static int  static BOOL
1468  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1469      unsigned int *pdataptr, int *errorcodeptr)
1470  {  {
1471  int c, i, bot, top;  pcre_uchar c;
1472  const uschar *ptr = *ptrptr;  int i, bot, top;
1473  char name[32];  const pcre_uchar *ptr = *ptrptr;
1474    pcre_uchar name[32];
1475    
1476  c = *(++ptr);  c = *(++ptr);
1477  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1478    
1479  *negptr = FALSE;  *negptr = FALSE;
1480    
1481  /* \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
1482  negation. */  negation. */
1483    
1484  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1485    {    {
1486    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1487      {      {
1488      *negptr = TRUE;      *negptr = TRUE;
1489      ptr++;      ptr++;
1490      }      }
1491    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1492      {      {
1493      c = *(++ptr);      c = *(++ptr);
1494      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1495      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1496      name[i] = c;      name[i] = c;
1497      }      }
1498    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1499    name[i] = 0;    name[i] = 0;
1500    }    }
1501    
# Line 708  else Line 1512  else
1512  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1513    
1514  bot = 0;  bot = 0;
1515  top = _pcre_utt_size;  top = PRIV(utt_size);
1516    
1517  while (bot < top)  while (bot < top)
1518    {    {
1519      int r;
1520    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1521    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1522    if (c == 0)    if (r == 0)
1523      {      {
1524      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1525      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1526        return TRUE;
1527      }      }
1528    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1529    }    }
1530    
1531  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1532  *ptrptr = ptr;  *ptrptr = ptr;
1533  return -1;  return FALSE;
1534    
1535  ERROR_RETURN:  ERROR_RETURN:
1536  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1537  *ptrptr = ptr;  *ptrptr = ptr;
1538  return -1;  return FALSE;
1539  }  }
1540  #endif  #endif
1541    
1542    
1543    
   
 /*************************************************  
 *            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 == '}');  
 }  
   
   
   
1544  /*************************************************  /*************************************************
1545  *         Read repeat counts                     *  *         Read repeat counts                     *
1546  *************************************************/  *************************************************/
# Line 788  Returns:         pointer to '}' on succe Line 1560  Returns:         pointer to '}' on succe
1560                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1561  */  */
1562    
1563  static const uschar *  static const pcre_uchar *
1564  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)
1565  {  {
1566  int min = 0;  int min = 0;
1567  int max = -1;  int max = -1;
# Line 797  int max = -1; Line 1569  int max = -1;
1569  /* 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
1570  an integer overflow. */  an integer overflow. */
1571    
1572  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1573  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1574    {    {
1575    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1579  if (min < 0 || min > 65535)
1579  /* 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.
1580  Also, max must not be less than min. */  Also, max must not be less than min. */
1581    
1582  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1583    {    {
1584    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1585      {      {
1586      max = 0;      max = 0;
1587      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1588      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1589        {        {
1590        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1609  return p;
1609    
1610    
1611  /*************************************************  /*************************************************
1612  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1613  *************************************************/  *************************************************/
1614    
1615  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1616  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
1617  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
1618  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
1619  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.  
1620    
1621  Arguments:  Arguments:
1622    ptr          current position in the pattern    code         pointer to the start of the group
1623    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  
1624    
1625  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1626  */  */
1627    
1628  static int  static const pcre_uchar*
1629  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode)  
1630  {  {
1631  const uschar *thisname;  for (;;)
   
 for (; *ptr != 0; ptr++)  
   {  
   int term;  
   
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   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;  
     }  
   
   /* Skip over character classes */  
   
   if (*ptr == '[')  
     {  
     while (*(++ptr) != ']')  
       {  
       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] != '?')  
     {  
     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 (;;)  
1632    {    {
1633    switch ((int)*code)    switch ((int)*code)
1634      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1635      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1636      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1637      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1638      if (!skipassert) return code;      if (!skipassert) return code;
1639      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1640      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1641      break;      break;
1642    
1643      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1647  for (;;)
1647    
1648      case OP_CALLOUT:      case OP_CALLOUT:
1649      case OP_CREF:      case OP_CREF:
1650        case OP_DNCREF:
1651      case OP_RREF:      case OP_RREF:
1652        case OP_DNRREF:
1653      case OP_DEF:      case OP_DEF:
1654      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1655      break;      break;
1656    
1657      default:      default:
# Line 1010  for (;;) Line 1663  for (;;)
1663    
1664    
1665    
   
1666  /*************************************************  /*************************************************
1667  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1668  *************************************************/  *************************************************/
1669    
1670  /* 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,
1671  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.
1672  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
1673    temporarily terminated with OP_END when this function is called.
1674    
1675    This function is called when a backward assertion is encountered, so that if it
1676    fails, the error message can point to the correct place in the pattern.
1677    However, we cannot do this when the assertion contains subroutine calls,
1678    because they can be forward references. We solve this by remembering this case
1679    and doing the check at the end; a flag specifies which mode we are running in.
1680    
1681  Arguments:  Arguments:
1682    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1683    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1684      atend    TRUE if called when the pattern is complete
1685  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1686               or -2 if \C was encountered  
1687    Returns:   the fixed length,
1688                 or -1 if there is no fixed length,
1689                 or -2 if \C was encountered (in UTF-8 mode only)
1690                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1691                 or -4 if an unknown opcode was encountered (internal error)
1692  */  */
1693    
1694  static int  static int
1695  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1696  {  {
1697  int length = -1;  int length = -1;
1698    
1699  register int branchlength = 0;  register int branchlength = 0;
1700  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1701    
1702  /* 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
1703  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1041  branch, check the length against that of Line 1705  branch, check the length against that of
1705  for (;;)  for (;;)
1706    {    {
1707    int d;    int d;
1708    register int op = *cc;    pcre_uchar *ce, *cs;
1709      register pcre_uchar op = *cc;
1710    
1711    switch (op)    switch (op)
1712      {      {
1713        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1714        OP_BRA (normal non-capturing bracket) because the other variants of these
1715        opcodes are all concerned with unlimited repeated groups, which of course
1716        are not of fixed length. */
1717    
1718      case OP_CBRA:      case OP_CBRA:
1719      case OP_BRA:      case OP_BRA:
1720      case OP_ONCE:      case OP_ONCE:
1721        case OP_ONCE_NC:
1722      case OP_COND:      case OP_COND:
1723      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1724      if (d < 0) return d;      if (d < 0) return d;
1725      branchlength += d;      branchlength += d;
1726      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1727      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1728      break;      break;
1729    
1730      /* 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.
1731      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
1732      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
1733        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1734        because they all imply an unlimited repeat. */
1735    
1736      case OP_ALT:      case OP_ALT:
1737      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1738      case OP_END:      case OP_END:
1739        case OP_ACCEPT:
1740        case OP_ASSERT_ACCEPT:
1741      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1742        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1743      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1745  for (;;)
1745      branchlength = 0;      branchlength = 0;
1746      break;      break;
1747    
1748        /* A true recursion implies not fixed length, but a subroutine call may
1749        be OK. If the subroutine is a forward reference, we can't deal with
1750        it until the end of the pattern, so return -3. */
1751    
1752        case OP_RECURSE:
1753        if (!atend) return -3;
1754        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1755        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1756        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1757        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1758        if (d < 0) return d;
1759        branchlength += d;
1760        cc += 1 + LINK_SIZE;
1761        break;
1762    
1763      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1764    
1765      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1767  for (;;)
1767      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1768      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1769      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1770      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1771        break;
1772    
1773      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1774    
1775      case OP_REVERSE:      case OP_MARK:
1776        case OP_PRUNE_ARG:
1777        case OP_SKIP_ARG:
1778        case OP_THEN_ARG:
1779        cc += cc[1] + PRIV(OP_lengths)[*cc];
1780        break;
1781    
1782        case OP_CALLOUT:
1783        case OP_CIRC:
1784        case OP_CIRCM:
1785        case OP_CLOSE:
1786        case OP_COMMIT:
1787      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1788      case OP_DEF:      case OP_DEF:
1789      case OP_OPT:      case OP_DNCREF:
1790      case OP_CALLOUT:      case OP_DNRREF:
1791      case OP_SOD:      case OP_DOLL:
1792      case OP_SOM:      case OP_DOLLM:
1793      case OP_EOD:      case OP_EOD:
1794      case OP_EODN:      case OP_EODN:
1795      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1796      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1797        case OP_PRUNE:
1798        case OP_REVERSE:
1799        case OP_RREF:
1800        case OP_SET_SOM:
1801        case OP_SKIP:
1802        case OP_SOD:
1803        case OP_SOM:
1804        case OP_THEN:
1805      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1806      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1807      break;      break;
1808    
1809      /* Handle literal characters */      /* Handle literal characters */
1810    
1811      case OP_CHAR:      case OP_CHAR:
1812      case OP_CHARNC:      case OP_CHARI:
1813      case OP_NOT:      case OP_NOT:
1814        case OP_NOTI:
1815      branchlength++;      branchlength++;
1816      cc += 2;      cc += 2;
1817  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1818      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1819  #endif  #endif
1820      break;      break;
1821    
# Line 1119  for (;;) Line 1823  for (;;)
1823      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1824    
1825      case OP_EXACT:      case OP_EXACT:
1826      branchlength += GET2(cc,1);      case OP_EXACTI:
1827      cc += 4;      case OP_NOTEXACT:
1828  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1829      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1830        {      cc += 2 + IMM2_SIZE;
1831        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1832        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1833  #endif  #endif
1834      break;      break;
1835    
1836      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1837      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1838      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1839          cc += 2;
1840        cc += 1 + IMM2_SIZE + 1;
1841      break;      break;
1842    
1843      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1847  for (;;)
1847      cc += 2;      cc += 2;
1848      /* Fall through */      /* Fall through */
1849    
1850        case OP_HSPACE:
1851        case OP_VSPACE:
1852        case OP_NOT_HSPACE:
1853        case OP_NOT_VSPACE:
1854      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1855      case OP_DIGIT:      case OP_DIGIT:
1856      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1858  for (;;)
1858      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1859      case OP_WORDCHAR:      case OP_WORDCHAR:
1860      case OP_ANY:      case OP_ANY:
1861        case OP_ALLANY:
1862      branchlength++;      branchlength++;
1863      cc++;      cc++;
1864      break;      break;
1865    
1866      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1867        otherwise \C is coded as OP_ALLANY. */
1868    
1869      case OP_ANYBYTE:      case OP_ANYBYTE:
1870      return -2;      return -2;
1871    
1872      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1873    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1874      case OP_CLASS:      case OP_CLASS:
1875      case OP_NCLASS:      case OP_NCLASS:
1876      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1877        case OP_XCLASS:
1878        /* The original code caused an unsigned overflow in 64 bit systems,
1879        so now we use a conditional statement. */
1880        if (op == OP_XCLASS)
1881          cc += GET(cc, 1);
1882        else
1883          cc += PRIV(OP_lengths)[OP_CLASS];
1884    #else
1885        cc += PRIV(OP_lengths)[OP_CLASS];
1886    #endif
1887    
1888      switch (*cc)      switch (*cc)
1889        {        {
1890        case OP_CRSTAR:        case OP_CRSTAR:
1891        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1892          case OP_CRPLUS:
1893          case OP_CRMINPLUS:
1894        case OP_CRQUERY:        case OP_CRQUERY:
1895        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1896          case OP_CRPOSSTAR:
1897          case OP_CRPOSPLUS:
1898          case OP_CRPOSQUERY:
1899        return -1;        return -1;
1900    
1901        case OP_CRRANGE:        case OP_CRRANGE:
1902        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1903        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1904        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1905        cc += 5;        branchlength += (int)GET2(cc,1);
1906          cc += 1 + 2 * IMM2_SIZE;
1907        break;        break;
1908    
1909        default:        default:
# Line 1191  for (;;) Line 1913  for (;;)
1913    
1914      /* Anything else is variable length */      /* Anything else is variable length */
1915    
1916      default:      case OP_ANYNL:
1917        case OP_BRAMINZERO:
1918        case OP_BRAPOS:
1919        case OP_BRAPOSZERO:
1920        case OP_BRAZERO:
1921        case OP_CBRAPOS:
1922        case OP_EXTUNI:
1923        case OP_KETRMAX:
1924        case OP_KETRMIN:
1925        case OP_KETRPOS:
1926        case OP_MINPLUS:
1927        case OP_MINPLUSI:
1928        case OP_MINQUERY:
1929        case OP_MINQUERYI:
1930        case OP_MINSTAR:
1931        case OP_MINSTARI:
1932        case OP_MINUPTO:
1933        case OP_MINUPTOI:
1934        case OP_NOTMINPLUS:
1935        case OP_NOTMINPLUSI:
1936        case OP_NOTMINQUERY:
1937        case OP_NOTMINQUERYI:
1938        case OP_NOTMINSTAR:
1939        case OP_NOTMINSTARI:
1940        case OP_NOTMINUPTO:
1941        case OP_NOTMINUPTOI:
1942        case OP_NOTPLUS:
1943        case OP_NOTPLUSI:
1944        case OP_NOTPOSPLUS:
1945        case OP_NOTPOSPLUSI:
1946        case OP_NOTPOSQUERY:
1947        case OP_NOTPOSQUERYI:
1948        case OP_NOTPOSSTAR:
1949        case OP_NOTPOSSTARI:
1950        case OP_NOTPOSUPTO:
1951        case OP_NOTPOSUPTOI:
1952        case OP_NOTQUERY:
1953        case OP_NOTQUERYI:
1954        case OP_NOTSTAR:
1955        case OP_NOTSTARI:
1956        case OP_NOTUPTO:
1957        case OP_NOTUPTOI:
1958        case OP_PLUS:
1959        case OP_PLUSI:
1960        case OP_POSPLUS:
1961        case OP_POSPLUSI:
1962        case OP_POSQUERY:
1963        case OP_POSQUERYI:
1964        case OP_POSSTAR:
1965        case OP_POSSTARI:
1966        case OP_POSUPTO:
1967        case OP_POSUPTOI:
1968        case OP_QUERY:
1969        case OP_QUERYI:
1970        case OP_REF:
1971        case OP_REFI:
1972        case OP_DNREF:
1973        case OP_DNREFI:
1974        case OP_SBRA:
1975        case OP_SBRAPOS:
1976        case OP_SCBRA:
1977        case OP_SCBRAPOS:
1978        case OP_SCOND:
1979        case OP_SKIPZERO:
1980        case OP_STAR:
1981        case OP_STARI:
1982        case OP_TYPEMINPLUS:
1983        case OP_TYPEMINQUERY:
1984        case OP_TYPEMINSTAR:
1985        case OP_TYPEMINUPTO:
1986        case OP_TYPEPLUS:
1987        case OP_TYPEPOSPLUS:
1988        case OP_TYPEPOSQUERY:
1989        case OP_TYPEPOSSTAR:
1990        case OP_TYPEPOSUPTO:
1991        case OP_TYPEQUERY:
1992        case OP_TYPESTAR:
1993        case OP_TYPEUPTO:
1994        case OP_UPTO:
1995        case OP_UPTOI:
1996      return -1;      return -1;
1997    
1998        /* Catch unrecognized opcodes so that when new ones are added they
1999        are not forgotten, as has happened in the past. */
2000    
2001        default:
2002        return -4;
2003      }      }
2004    }    }
2005  /* Control never gets here */  /* Control never gets here */
# Line 1200  for (;;) Line 2007  for (;;)
2007    
2008    
2009    
   
2010  /*************************************************  /*************************************************
2011  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2012  *************************************************/  *************************************************/
2013    
2014  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2015  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2016    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2017    so that it can be called from pcre_study() when finding the minimum matching
2018    length.
2019    
2020  Arguments:  Arguments:
2021    code        points to start of expression    code        points to start of expression
2022    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2023    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2024    
2025  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
2026  */  */
2027    
2028  static const uschar *  const pcre_uchar *
2029  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2030  {  {
2031  for (;;)  for (;;)
2032    {    {
2033    register int c = *code;    register pcre_uchar c = *code;
2034    
2035    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2036    
2037    /* 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 1230  for (;;) Line 2040  for (;;)
2040    
2041    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2042    
2043      /* Handle recursion */
2044    
2045      else if (c == OP_REVERSE)
2046        {
2047        if (number < 0) return (pcre_uchar *)code;
2048        code += PRIV(OP_lengths)[c];
2049        }
2050    
2051    /* Handle capturing bracket */    /* Handle capturing bracket */
2052    
2053    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2054               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2055      {      {
2056      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2057      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2058      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2059      }      }
2060    
2061      /* Otherwise, we can get the item's length from the table, except that for
2062      repeated character types, we have to test for \p and \P, which have an extra
2063      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2064      must add in its length. */
2065    
2066      else
2067        {
2068        switch(c)
2069          {
2070          case OP_TYPESTAR:
2071          case OP_TYPEMINSTAR:
2072          case OP_TYPEPLUS:
2073          case OP_TYPEMINPLUS:
2074          case OP_TYPEQUERY:
2075          case OP_TYPEMINQUERY:
2076          case OP_TYPEPOSSTAR:
2077          case OP_TYPEPOSPLUS:
2078          case OP_TYPEPOSQUERY:
2079          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2080          break;
2081    
2082          case OP_TYPEUPTO:
2083          case OP_TYPEMINUPTO:
2084          case OP_TYPEEXACT:
2085          case OP_TYPEPOSUPTO:
2086          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2087            code += 2;
2088          break;
2089    
2090          case OP_MARK:
2091          case OP_PRUNE_ARG:
2092          case OP_SKIP_ARG:
2093          case OP_THEN_ARG:
2094          code += code[1];
2095          break;
2096          }
2097    
2098        /* Add in the fixed length from the table */
2099    
2100        code += PRIV(OP_lengths)[c];
2101    
2102    /* 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
2103    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
2104    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2105    
2106    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2107      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
     if (utf8) switch(c)  
2108        {        {
2109        case OP_CHAR:        case OP_CHAR:
2110        case OP_CHARNC:        case OP_CHARI:
2111        case OP_EXACT:        case OP_EXACT:
2112          case OP_EXACTI:
2113        case OP_UPTO:        case OP_UPTO:
2114          case OP_UPTOI:
2115        case OP_MINUPTO:        case OP_MINUPTO:
2116          case OP_MINUPTOI:
2117        case OP_POSUPTO:        case OP_POSUPTO:
2118          case OP_POSUPTOI:
2119        case OP_STAR:        case OP_STAR:
2120          case OP_STARI:
2121        case OP_MINSTAR:        case OP_MINSTAR:
2122          case OP_MINSTARI:
2123        case OP_POSSTAR:        case OP_POSSTAR:
2124          case OP_POSSTARI:
2125        case OP_PLUS:        case OP_PLUS:
2126          case OP_PLUSI:
2127        case OP_MINPLUS:        case OP_MINPLUS:
2128          case OP_MINPLUSI:
2129        case OP_POSPLUS:        case OP_POSPLUS:
2130          case OP_POSPLUSI:
2131        case OP_QUERY:        case OP_QUERY:
2132          case OP_QUERYI:
2133        case OP_MINQUERY:        case OP_MINQUERY:
2134          case OP_MINQUERYI:
2135        case OP_POSQUERY:        case OP_POSQUERY:
2136        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2137          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2138        break;        break;
2139        }        }
2140    #else
2141        (void)(utf);  /* Keep compiler happy by referencing function argument */
2142    #endif
2143      }      }
2144    }    }
2145  }  }
# Line 1281  instance of OP_RECURSE. Line 2155  instance of OP_RECURSE.
2155    
2156  Arguments:  Arguments:
2157    code        points to start of expression    code        points to start of expression
2158    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2159    
2160  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
2161  */  */
2162    
2163  static const uschar *  static const pcre_uchar *
2164  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2165  {  {
2166  for (;;)  for (;;)
2167    {    {
2168    register int c = *code;    register pcre_uchar c = *code;
2169    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2170    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2171    
# Line 1301  for (;;) Line 2175  for (;;)
2175    
2176    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2177    
2178    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
2179    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
2180    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2181    bytes. */    must add in its length. */
2182    
2183    else    else
2184      {      {
2185      code += _pcre_OP_lengths[c];      switch(c)
2186      if (utf8) switch(c)        {
2187          case OP_TYPESTAR:
2188          case OP_TYPEMINSTAR:
2189          case OP_TYPEPLUS:
2190          case OP_TYPEMINPLUS:
2191          case OP_TYPEQUERY:
2192          case OP_TYPEMINQUERY:
2193          case OP_TYPEPOSSTAR:
2194          case OP_TYPEPOSPLUS:
2195          case OP_TYPEPOSQUERY:
2196          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2197          break;
2198    
2199          case OP_TYPEPOSUPTO:
2200          case OP_TYPEUPTO:
2201          case OP_TYPEMINUPTO:
2202          case OP_TYPEEXACT:
2203          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2204            code += 2;
2205          break;
2206    
2207          case OP_MARK:
2208          case OP_PRUNE_ARG:
2209          case OP_SKIP_ARG:
2210          case OP_THEN_ARG:
2211          code += code[1];
2212          break;
2213          }
2214    
2215        /* Add in the fixed length from the table */
2216    
2217        code += PRIV(OP_lengths)[c];
2218    
2219        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2220        by a multi-byte character. The length in the table is a minimum, so we have
2221        to arrange to skip the extra bytes. */
2222    
2223    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2224        if (utf) switch(c)
2225        {        {
2226        case OP_CHAR:        case OP_CHAR:
2227        case OP_CHARNC:        case OP_CHARI:
2228          case OP_NOT:
2229          case OP_NOTI:
2230        case OP_EXACT:        case OP_EXACT:
2231          case OP_EXACTI:
2232          case OP_NOTEXACT:
2233          case OP_NOTEXACTI:
2234        case OP_UPTO:        case OP_UPTO:
2235          case OP_UPTOI:
2236          case OP_NOTUPTO:
2237          case OP_NOTUPTOI:
2238        case OP_MINUPTO:        case OP_MINUPTO:
2239          case OP_MINUPTOI:
2240          case OP_NOTMINUPTO:
2241          case OP_NOTMINUPTOI:
2242        case OP_POSUPTO:        case OP_POSUPTO:
2243          case OP_POSUPTOI:
2244          case OP_NOTPOSUPTO:
2245          case OP_NOTPOSUPTOI:
2246        case OP_STAR:        case OP_STAR:
2247          case OP_STARI:
2248          case OP_NOTSTAR:
2249          case OP_NOTSTARI:
2250        case OP_MINSTAR:        case OP_MINSTAR:
2251          case OP_MINSTARI:
2252          case OP_NOTMINSTAR:
2253          case OP_NOTMINSTARI:
2254        case OP_POSSTAR:        case OP_POSSTAR:
2255          case OP_POSSTARI:
2256          case OP_NOTPOSSTAR:
2257          case OP_NOTPOSSTARI:
2258        case OP_PLUS:        case OP_PLUS:
2259          case OP_PLUSI:
2260          case OP_NOTPLUS:
2261          case OP_NOTPLUSI:
2262        case OP_MINPLUS:        case OP_MINPLUS:
2263          case OP_MINPLUSI:
2264          case OP_NOTMINPLUS:
2265          case OP_NOTMINPLUSI:
2266        case OP_POSPLUS:        case OP_POSPLUS:
2267          case OP_POSPLUSI:
2268          case OP_NOTPOSPLUS:
2269          case OP_NOTPOSPLUSI:
2270        case OP_QUERY:        case OP_QUERY:
2271          case OP_QUERYI:
2272          case OP_NOTQUERY:
2273          case OP_NOTQUERYI:
2274        case OP_MINQUERY:        case OP_MINQUERY:
2275          case OP_MINQUERYI:
2276          case OP_NOTMINQUERY:
2277          case OP_NOTMINQUERYI:
2278        case OP_POSQUERY:        case OP_POSQUERY:
2279        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2280          case OP_NOTPOSQUERY:
2281          case OP_NOTPOSQUERYI:
2282          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2283        break;        break;
2284        }        }
2285    #else
2286        (void)(utf);  /* Keep compiler happy by referencing function argument */
2287    #endif
2288      }      }
2289    }    }
2290  }  }
# Line 1343  for (;;) Line 2299  for (;;)
2299  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2300  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
2301  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2302  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2303  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2304    bracket whose current branch will already have been scanned.
2305    
2306  Arguments:  Arguments:
2307    code        points to start of search    code        points to start of search
2308    endcode     points to where to stop    endcode     points to where to stop
2309    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2310      cd          contains pointers to tables etc.
2311      recurses    chain of recurse_check to catch mutual recursion
2312    
2313  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2314  */  */
2315    
2316    typedef struct recurse_check {
2317      struct recurse_check *prev;
2318      const pcre_uchar *group;
2319    } recurse_check;
2320    
2321  static BOOL  static BOOL
2322  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2323      BOOL utf, compile_data *cd, recurse_check *recurses)
2324  {  {
2325  register int c;  register pcre_uchar c;
2326  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2327    
2328    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2329       code < endcode;       code < endcode;
2330       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2331    {    {
2332    const uschar *ccode;    const pcre_uchar *ccode;
2333    
2334    c = *code;    c = *code;
2335    
2336    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2337      first_significant_code() with a TRUE final argument. */
2338    
2339      if (c == OP_ASSERT)
2340        {
2341        do code += GET(code, 1); while (*code == OP_ALT);
2342        c = *code;
2343        continue;
2344        }
2345    
2346      /* For a recursion/subroutine call, if its end has been reached, which
2347      implies a backward reference subroutine call, we can scan it. If it's a
2348      forward reference subroutine call, we can't. To detect forward reference
2349      we have to scan up the list that is kept in the workspace. This function is
2350      called only when doing the real compile, not during the pre-compile that
2351      measures the size of the compiled pattern. */
2352    
2353      if (c == OP_RECURSE)
2354      {      {
2355        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2356      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2357    
2358      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2359        when called to scan a completed pattern by setting cd->start_workspace to
2360        NULL. */
2361    
2362        if (cd->start_workspace != NULL)
2363          {
2364          const pcre_uchar *tcode;
2365          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2366            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2367          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2368          }
2369    
2370        /* If we are scanning a completed pattern, there are no forward references
2371        and all groups are complete. We need to detect whether this is a recursive
2372        call, as otherwise there will be an infinite loop. If it is a recursion,
2373        just skip over it. Simple recursions are easily detected. For mutual
2374        recursions we keep a chain on the stack. */
2375    
2376        else
2377          {
2378          recurse_check *r = recurses;
2379          const pcre_uchar *endgroup = scode;
2380    
2381          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2382          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2383    
2384          for (r = recurses; r != NULL; r = r->prev)
2385            if (r->group == scode) break;
2386          if (r != NULL) continue;   /* Mutual recursion */
2387          }
2388    
2389        /* Completed reference; scan the referenced group, remembering it on the
2390        stack chain to detect mutual recursions. */
2391    
2392      empty_branch = FALSE;      empty_branch = FALSE;
2393        this_recurse.prev = recurses;
2394        this_recurse.group = scode;
2395    
2396      do      do
2397        {        {
2398        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2399            {
2400          empty_branch = TRUE;          empty_branch = TRUE;
2401        code += GET(code, 1);          break;
2402            }
2403          scode += GET(scode, 1);
2404        }        }
2405      while (*code == OP_ALT);      while (*scode == OP_ALT);
2406      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2407        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2408        continue;
2409        }
2410    
2411      /* Groups with zero repeats can of course be empty; skip them. */
2412    
2413      /* Move past the KET and fudge things so that the increment in the "for"    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2414      above has no effect. */        c == OP_BRAPOSZERO)
2415        {
2416        code += PRIV(OP_lengths)[c];
2417        do code += GET(code, 1); while (*code == OP_ALT);
2418        c = *code;
2419        continue;
2420        }
2421    
2422      /* A nested group that is already marked as "could be empty" can just be
2423      skipped. */
2424    
2425      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2426          c == OP_SCBRA || c == OP_SCBRAPOS)
2427        {
2428        do code += GET(code, 1); while (*code == OP_ALT);
2429        c = *code;
2430        continue;
2431        }
2432    
2433      /* For other groups, scan the branches. */
2434    
2435      if (c == OP_BRA  || c == OP_BRAPOS ||
2436          c == OP_CBRA || c == OP_CBRAPOS ||
2437          c == OP_ONCE || c == OP_ONCE_NC ||
2438          c == OP_COND)
2439        {
2440        BOOL empty_branch;
2441        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2442    
2443        /* If a conditional group has only one branch, there is a second, implied,
2444        empty branch, so just skip over the conditional, because it could be empty.
2445        Otherwise, scan the individual branches of the group. */
2446    
2447      c = OP_END;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2448      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];        code += GET(code, 1);
2449        else
2450          {
2451          empty_branch = FALSE;
2452          do
2453            {
2454            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2455              empty_branch = TRUE;
2456            code += GET(code, 1);
2457            }
2458          while (*code == OP_ALT);
2459          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2460          }
2461    
2462        c = *code;
2463      continue;      continue;
2464      }      }
2465    
# Line 1395  for (code = first_significant_code(code Line 2467  for (code = first_significant_code(code
2467    
2468    switch (c)    switch (c)
2469      {      {
2470      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2471        cannot be represented just by a bit map. This includes negated single
2472        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2473        actual length is stored in the compiled code, so we must update "code"
2474        here. */
2475    
2476  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2477      case OP_XCLASS:      case OP_XCLASS:
2478      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2479      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2480  #endif  #endif
2481    
2482      case OP_CLASS:      case OP_CLASS:
2483      case OP_NCLASS:      case OP_NCLASS:
2484      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2485    
2486  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2487      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2488  #endif  #endif
2489    
# Line 1417  for (code = first_significant_code(code Line 2493  for (code = first_significant_code(code
2493        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2494        case OP_CRQUERY:        case OP_CRQUERY:
2495        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2496          case OP_CRPOSSTAR:
2497          case OP_CRPOSQUERY:
2498        break;        break;
2499    
2500        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2501        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2502        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2503          case OP_CRPOSPLUS:
2504        return FALSE;        return FALSE;
2505    
2506        case OP_CRRANGE:        case OP_CRRANGE:
2507        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2508          case OP_CRPOSRANGE:
2509        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2510        break;        break;
2511        }        }
# Line 1433  for (code = first_significant_code(code Line 2513  for (code = first_significant_code(code
2513    
2514      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2515    
2516        case OP_ANY:
2517        case OP_ALLANY:
2518        case OP_ANYBYTE:
2519    
2520      case OP_PROP:      case OP_PROP:
2521      case OP_NOTPROP:      case OP_NOTPROP:
2522        case OP_ANYNL:
2523    
2524        case OP_NOT_HSPACE:
2525        case OP_HSPACE:
2526        case OP_NOT_VSPACE:
2527        case OP_VSPACE:
2528      case OP_EXTUNI:      case OP_EXTUNI:
2529    
2530      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2531      case OP_DIGIT:      case OP_DIGIT:
2532      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2533      case OP_WHITESPACE:      case OP_WHITESPACE:
2534      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2535      case OP_WORDCHAR:      case OP_WORDCHAR:
2536      case OP_ANY:  
     case OP_ANYBYTE:  
2537      case OP_CHAR:      case OP_CHAR:
2538      case OP_CHARNC:      case OP_CHARI:
2539      case OP_NOT:      case OP_NOT:
2540        case OP_NOTI:
2541    
2542      case OP_PLUS:      case OP_PLUS:
2543        case OP_PLUSI:
2544      case OP_MINPLUS:      case OP_MINPLUS:
2545      case OP_POSPLUS:      case OP_MINPLUSI:
2546      case OP_EXACT:  
2547      case OP_NOTPLUS:      case OP_NOTPLUS:
2548        case OP_NOTPLUSI:
2549      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2550        case OP_NOTMINPLUSI:
2551    
2552        case OP_POSPLUS:
2553        case OP_POSPLUSI:
2554      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2555        case OP_NOTPOSPLUSI:
2556    
2557        case OP_EXACT:
2558        case OP_EXACTI:
2559      case OP_NOTEXACT:      case OP_NOTEXACT:
2560        case OP_NOTEXACTI:
2561    
2562      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2563      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2564      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2565      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2566    
2567      return FALSE;      return FALSE;
2568    
2569        /* These are going to continue, as they may be empty, but we have to
2570        fudge the length for the \p and \P cases. */
2571    
2572        case OP_TYPESTAR:
2573        case OP_TYPEMINSTAR:
2574        case OP_TYPEPOSSTAR:
2575        case OP_TYPEQUERY:
2576        case OP_TYPEMINQUERY:
2577        case OP_TYPEPOSQUERY:
2578        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2579        break;
2580    
2581        /* Same for these */
2582    
2583        case OP_TYPEUPTO:
2584        case OP_TYPEMINUPTO:
2585        case OP_TYPEPOSUPTO:
2586        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2587          code += 2;
2588        break;
2589    
2590      /* End of branch */      /* End of branch */
2591    
2592      case OP_KET:      case OP_KET:
2593      case OP_KETRMAX:      case OP_KETRMAX:
2594      case OP_KETRMIN:      case OP_KETRMIN:
2595        case OP_KETRPOS:
2596      case OP_ALT:      case OP_ALT:
2597      return TRUE;      return TRUE;
2598    
2599      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2600      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2601        followed by a multibyte character. */
2602    
2603  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2604      case OP_STAR:      case OP_STAR:
2605        case OP_STARI:
2606        case OP_NOTSTAR:
2607        case OP_NOTSTARI:
2608    
2609      case OP_MINSTAR:      case OP_MINSTAR:
2610        case OP_MINSTARI:
2611        case OP_NOTMINSTAR:
2612        case OP_NOTMINSTARI:
2613    
2614      case OP_POSSTAR:      case OP_POSSTAR:
2615        case OP_POSSTARI:
2616        case OP_NOTPOSSTAR:
2617        case OP_NOTPOSSTARI:
2618    
2619      case OP_QUERY:      case OP_QUERY:
2620        case OP_QUERYI:
2621        case OP_NOTQUERY:
2622        case OP_NOTQUERYI:
2623    
2624      case OP_MINQUERY:      case OP_MINQUERY:
2625        case OP_MINQUERYI:
2626        case OP_NOTMINQUERY:
2627        case OP_NOTMINQUERYI:
2628    
2629      case OP_POSQUERY:      case OP_POSQUERY:
2630        case OP_POSQUERYI:
2631        case OP_NOTPOSQUERY:
2632        case OP_NOTPOSQUERYI:
2633    
2634        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2635        break;
2636    
2637      case OP_UPTO:      case OP_UPTO:
2638        case OP_UPTOI:
2639        case OP_NOTUPTO:
2640        case OP_NOTUPTOI:
2641    
2642      case OP_MINUPTO:      case OP_MINUPTO:
2643        case OP_MINUPTOI:
2644        case OP_NOTMINUPTO:
2645        case OP_NOTMINUPTOI:
2646    
2647      case OP_POSUPTO:      case OP_POSUPTO:
2648      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2649        case OP_NOTPOSUPTO:
2650        case OP_NOTPOSUPTOI:
2651    
2652        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2653      break;      break;
2654  #endif  #endif
     }  
   }  
2655    
2656  return TRUE;      /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2657  }      string. */
2658    
2659        case OP_MARK:
2660        case OP_PRUNE_ARG:
2661        case OP_SKIP_ARG:
2662        case OP_THEN_ARG:
2663        code += code[1];
2664        break;
2665    
2666        /* None of the remaining opcodes are required to match a character. */
2667    
2668  /*************************************************      default:
2669        break;
2670        }
2671      }
2672    
2673    return TRUE;
2674    }
2675    
2676    
2677    
2678    /*************************************************
2679  *    Scan compiled regex for non-emptiness       *  *    Scan compiled regex for non-emptiness       *
2680  *************************************************/  *************************************************/
2681    
# Line 1501  return TRUE; Line 2683  return TRUE;
2683  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2684  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2685  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2686    This function is called only during the real compile, not during the
2687    pre-compile.
2688    
2689  Arguments:  Arguments:
2690    code        points to start of the recursion    code        points to start of the recursion
2691    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2692    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2693    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2694      cd          pointers to tables etc
2695    
2696  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2697  */  */
2698    
2699  static BOOL  static BOOL
2700  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2701    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2702  {  {
2703  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2704    {    {
2705    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2706        return FALSE;
2707    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2708    }    }
2709  return TRUE;  return TRUE;
2710    }
2711    
2712    
2713    
2714    /*************************************************
2715    *        Base opcode of repeated opcodes         *
2716    *************************************************/
2717    
2718    /* Returns the base opcode for repeated single character type opcodes. If the
2719    opcode is not a repeated character type, it returns with the original value.
2720    
2721    Arguments:  c opcode
2722    Returns:    base opcode for the type
2723    */
2724    
2725    static pcre_uchar
2726    get_repeat_base(pcre_uchar c)
2727    {
2728    return (c > OP_TYPEPOSUPTO)? c :
2729           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2730           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2731           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2732           (c >= OP_STARI)?      OP_STARI :
2733                                 OP_STAR;
2734    }
2735    
2736    
2737    
2738    #ifdef SUPPORT_UCP
2739    /*************************************************
2740    *        Check a character and a property        *
2741    *************************************************/
2742    
2743    /* This function is called by check_auto_possessive() when a property item
2744    is adjacent to a fixed character.
2745    
2746    Arguments:
2747      c            the character
2748      ptype        the property type
2749      pdata        the data for the type
2750      negated      TRUE if it's a negated property (\P or \p{^)
2751    
2752    Returns:       TRUE if auto-possessifying is OK
2753    */
2754    
2755    static BOOL
2756    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2757      BOOL negated)
2758    {
2759    const pcre_uint32 *p;
2760    const ucd_record *prop = GET_UCD(c);
2761    
2762    switch(ptype)
2763      {
2764      case PT_LAMP:
2765      return (prop->chartype == ucp_Lu ||
2766              prop->chartype == ucp_Ll ||
2767              prop->chartype == ucp_Lt) == negated;
2768    
2769      case PT_GC:
2770      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2771    
2772      case PT_PC:
2773      return (pdata == prop->chartype) == negated;
2774    
2775      case PT_SC:
2776      return (pdata == prop->script) == negated;
2777    
2778      /* These are specials */
2779    
2780      case PT_ALNUM:
2781      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2782              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2783    
2784      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2785      means that Perl space and POSIX space are now identical. PCRE was changed
2786      at release 8.34. */
2787    
2788      case PT_SPACE:    /* Perl space */
2789      case PT_PXSPACE:  /* POSIX space */
2790      switch(c)
2791        {
2792        HSPACE_CASES:
2793        VSPACE_CASES:
2794        return negated;
2795    
2796        default:
2797        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2798        }
2799      break;  /* Control never reaches here */
2800    
2801      case PT_WORD:
2802      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2803              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2804              c == CHAR_UNDERSCORE) == negated;
2805    
2806      case PT_CLIST:
2807      p = PRIV(ucd_caseless_sets) + prop->caseset;
2808      for (;;)
2809        {
2810        if (c < *p) return !negated;
2811        if (c == *p++) return negated;
2812        }
2813      break;  /* Control never reaches here */
2814      }
2815    
2816    return FALSE;
2817    }
2818    #endif  /* SUPPORT_UCP */
2819    
2820    
2821    
2822    /*************************************************
2823    *        Fill the character property list        *
2824    *************************************************/
2825    
2826    /* Checks whether the code points to an opcode that can take part in auto-
2827    possessification, and if so, fills a list with its properties.
2828    
2829    Arguments:
2830      code        points to start of expression
2831      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2832      fcc         points to case-flipping table
2833      list        points to output list
2834                  list[0] will be filled with the opcode
2835                  list[1] will be non-zero if this opcode
2836                    can match an empty character string
2837                  list[2..7] depends on the opcode
2838    
2839    Returns:      points to the start of the next opcode if *code is accepted
2840                  NULL if *code is not accepted
2841    */
2842    
2843    static const pcre_uchar *
2844    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2845      const pcre_uint8 *fcc, pcre_uint32 *list)
2846    {
2847    pcre_uchar c = *code;
2848    const pcre_uchar *end;
2849    const pcre_uint32 *clist_src;
2850    pcre_uint32 *clist_dest;
2851    pcre_uint32 chr;
2852    pcre_uchar base;
2853    
2854    list[0] = c;
2855    list[1] = FALSE;
2856    code++;
2857    
2858    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2859      {
2860      base = get_repeat_base(c);
2861      c -= (base - OP_STAR);
2862    
2863      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2864        code += IMM2_SIZE;
2865    
2866      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2867    
2868      switch(base)
2869        {
2870        case OP_STAR:
2871        list[0] = OP_CHAR;
2872        break;
2873    
2874        case OP_STARI:
2875        list[0] = OP_CHARI;
2876        break;
2877    
2878        case OP_NOTSTAR:
2879        list[0] = OP_NOT;
2880        break;
2881    
2882        case OP_NOTSTARI:
2883        list[0] = OP_NOTI;
2884        break;
2885    
2886        case OP_TYPESTAR:
2887        list[0] = *code;
2888        code++;
2889        break;
2890        }
2891      c = list[0];
2892      }
2893    
2894    switch(c)
2895      {
2896      case OP_NOT_DIGIT:
2897      case OP_DIGIT:
2898      case OP_NOT_WHITESPACE:
2899      case OP_WHITESPACE:
2900      case OP_NOT_WORDCHAR:
2901      case OP_WORDCHAR:
2902      case OP_ANY:
2903      case OP_ALLANY:
2904      case OP_ANYNL:
2905      case OP_NOT_HSPACE:
2906      case OP_HSPACE:
2907      case OP_NOT_VSPACE:
2908      case OP_VSPACE:
2909      case OP_EXTUNI:
2910      case OP_EODN:
2911      case OP_EOD:
2912      case OP_DOLL:
2913      case OP_DOLLM:
2914      return code;
2915    
2916      case OP_CHAR:
2917      case OP_NOT:
2918      GETCHARINCTEST(chr, code);
2919      list[2] = chr;
2920      list[3] = NOTACHAR;
2921      return code;
2922    
2923      case OP_CHARI:
2924      case OP_NOTI:
2925      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2926      GETCHARINCTEST(chr, code);
2927      list[2] = chr;
2928    
2929    #ifdef SUPPORT_UCP
2930      if (chr < 128 || (chr < 256 && !utf))
2931        list[3] = fcc[chr];
2932      else
2933        list[3] = UCD_OTHERCASE(chr);
2934    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2935      list[3] = (chr < 256) ? fcc[chr] : chr;
2936    #else
2937      list[3] = fcc[chr];
2938    #endif
2939    
2940      /* The othercase might be the same value. */
2941    
2942      if (chr == list[3])
2943        list[3] = NOTACHAR;
2944      else
2945        list[4] = NOTACHAR;
2946      return code;
2947    
2948    #ifdef SUPPORT_UCP
2949      case OP_PROP:
2950      case OP_NOTPROP:
2951      if (code[0] != PT_CLIST)
2952        {
2953        list[2] = code[0];
2954        list[3] = code[1];
2955        return code + 2;
2956        }
2957    
2958      /* Convert only if we have enough space. */
2959    
2960      clist_src = PRIV(ucd_caseless_sets) + code[1];
2961      clist_dest = list + 2;
2962      code += 2;
2963    
2964      do {
2965         if (clist_dest >= list + 8)
2966           {
2967           /* Early return if there is not enough space. This should never
2968           happen, since all clists are shorter than 5 character now. */
2969           list[2] = code[0];
2970           list[3] = code[1];
2971           return code;
2972           }
2973         *clist_dest++ = *clist_src;
2974         }
2975      while(*clist_src++ != NOTACHAR);
2976    
2977      /* All characters are stored. The terminating NOTACHAR
2978      is copied form the clist itself. */
2979    
2980      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2981      return code;
2982    #endif
2983    
2984      case OP_NCLASS:
2985      case OP_CLASS:
2986    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2987      case OP_XCLASS:
2988      if (c == OP_XCLASS)
2989        end = code + GET(code, 0) - 1;
2990      else
2991    #endif
2992        end = code + 32 / sizeof(pcre_uchar);
2993    
2994      switch(*end)
2995        {
2996        case OP_CRSTAR:
2997        case OP_CRMINSTAR:
2998        case OP_CRQUERY:
2999        case OP_CRMINQUERY:
3000        case OP_CRPOSSTAR:
3001        case OP_CRPOSQUERY:
3002        list[1] = TRUE;
3003        end++;
3004        break;
3005    
3006        case OP_CRPLUS:
3007        case OP_CRMINPLUS:
3008        case OP_CRPOSPLUS:
3009        end++;
3010        break;
3011    
3012        case OP_CRRANGE:
3013        case OP_CRMINRANGE:
3014        case OP_CRPOSRANGE:
3015        list[1] = (GET2(end, 1) == 0);
3016        end += 1 + 2 * IMM2_SIZE;
3017        break;
3018        }
3019      list[2] = end - code;
3020      return end;
3021      }
3022    return NULL;    /* Opcode not accepted */
3023    }
3024    
3025    
3026    
3027    /*************************************************
3028    *    Scan further character sets for match       *
3029    *************************************************/
3030    
3031    /* Checks whether the base and the current opcode have a common character, in
3032    which case the base cannot be possessified.
3033    
3034    Arguments:
3035      code        points to the byte code
3036      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3037      cd          static compile data
3038      base_list   the data list of the base opcode
3039    
3040    Returns:      TRUE if the auto-possessification is possible
3041    */
3042    
3043    static BOOL
3044    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3045      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3046    {
3047    pcre_uchar c;
3048    pcre_uint32 list[8];
3049    const pcre_uint32 *chr_ptr;
3050    const pcre_uint32 *ochr_ptr;
3051    const pcre_uint32 *list_ptr;
3052    const pcre_uchar *next_code;
3053    const pcre_uint8 *class_bitset;
3054    const pcre_uint32 *set1, *set2, *set_end;
3055    pcre_uint32 chr;
3056    BOOL accepted, invert_bits;
3057    
3058    /* Note: the base_list[1] contains whether the current opcode has greedy
3059    (represented by a non-zero value) quantifier. This is a different from
3060    other character type lists, which stores here that the character iterator
3061    matches to an empty string (also represented by a non-zero value). */
3062    
3063    for(;;)
3064      {
3065      /* All operations move the code pointer forward.
3066      Therefore infinite recursions are not possible. */
3067    
3068      c = *code;
3069    
3070      /* Skip over callouts */
3071    
3072      if (c == OP_CALLOUT)
3073        {
3074        code += PRIV(OP_lengths)[c];
3075        continue;
3076        }
3077    
3078      if (c == OP_ALT)
3079        {
3080        do code += GET(code, 1); while (*code == OP_ALT);
3081        c = *code;
3082        }
3083    
3084      switch(c)
3085        {
3086        case OP_END:
3087        case OP_KETRPOS:
3088        /* TRUE only in greedy case. The non-greedy case could be replaced by
3089        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3090        uses more memory, which we cannot get at this stage.) */
3091    
3092        return base_list[1] != 0;
3093    
3094        case OP_KET:
3095        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3096        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3097        cannot be converted to a possessive form. */
3098    
3099        if (base_list[1] == 0) return FALSE;
3100    
3101        switch(*(code - GET(code, 1)))
3102          {
3103          case OP_ASSERT:
3104          case OP_ASSERT_NOT:
3105          case OP_ASSERTBACK:
3106          case OP_ASSERTBACK_NOT:
3107          case OP_ONCE:
3108          case OP_ONCE_NC:
3109          /* Atomic sub-patterns and assertions can always auto-possessify their
3110          last iterator. */
3111          return TRUE;
3112          }
3113    
3114        code += PRIV(OP_lengths)[c];
3115        continue;
3116    
3117        case OP_ONCE:
3118        case OP_ONCE_NC:
3119        case OP_BRA:
3120        case OP_CBRA:
3121        next_code = code + GET(code, 1);
3122        code += PRIV(OP_lengths)[c];
3123    
3124        while (*next_code == OP_ALT)
3125          {
3126          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3127          code = next_code + 1 + LINK_SIZE;
3128          next_code += GET(next_code, 1);
3129          }
3130        continue;
3131    
3132        case OP_BRAZERO:
3133        case OP_BRAMINZERO:
3134    
3135        next_code = code + 1;
3136        if (*next_code != OP_BRA && *next_code != OP_CBRA
3137            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3138    
3139        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3140    
3141        /* The bracket content will be checked by the
3142        OP_BRA/OP_CBRA case above. */
3143        next_code += 1 + LINK_SIZE;
3144        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3145          return FALSE;
3146    
3147        code += PRIV(OP_lengths)[c];
3148        continue;
3149        }
3150    
3151      /* Check for a supported opcode, and load its properties. */
3152    
3153      code = get_chr_property_list(code, utf, cd->fcc, list);
3154      if (code == NULL) return FALSE;    /* Unsupported */
3155    
3156      /* If either opcode is a small character list, set pointers for comparing
3157      characters from that list with another list, or with a property. */
3158    
3159      if (base_list[0] == OP_CHAR)
3160        {
3161        chr_ptr = base_list + 2;
3162        list_ptr = list;
3163        }
3164      else if (list[0] == OP_CHAR)
3165        {
3166        chr_ptr = list + 2;
3167        list_ptr = base_list;
3168        }
3169    
3170      /* Character bitsets can also be compared to certain opcodes. */
3171    
3172      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3173    #ifdef COMPILE_PCRE8
3174          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3175          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3176    #endif
3177          )
3178        {
3179    #ifdef COMPILE_PCRE8
3180        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3181    #else
3182        if (base_list[0] == OP_CLASS)
3183    #endif
3184          {
3185          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3186          list_ptr = list;
3187          }
3188        else
3189          {
3190          set1 = (pcre_uint32 *)(code - list[2]);
3191          list_ptr = base_list;
3192          }
3193    
3194        invert_bits = FALSE;
3195        switch(list_ptr[0])
3196          {
3197          case OP_CLASS:
3198          case OP_NCLASS:
3199          set2 = (pcre_uint32 *)
3200            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3201          break;
3202    
3203          /* OP_XCLASS cannot be supported here, because its bitset
3204          is not necessarily complete. E.g: [a-\0x{200}] is stored
3205          as a character range, and the appropriate bits are not set. */
3206    
3207          case OP_NOT_DIGIT:
3208            invert_bits = TRUE;
3209            /* Fall through */
3210          case OP_DIGIT:
3211            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3212            break;
3213    
3214          case OP_NOT_WHITESPACE:
3215            invert_bits = TRUE;
3216            /* Fall through */
3217          case OP_WHITESPACE:
3218            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3219            break;
3220    
3221          case OP_NOT_WORDCHAR:
3222            invert_bits = TRUE;
3223            /* Fall through */
3224          case OP_WORDCHAR:
3225            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3226            break;
3227    
3228          default:
3229          return FALSE;
3230          }
3231    
3232        /* Compare 4 bytes to improve speed. */
3233        set_end = set1 + (32 / 4);
3234        if (invert_bits)
3235          {
3236          do
3237            {
3238            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3239            }
3240          while (set1 < set_end);
3241          }
3242        else
3243          {
3244          do
3245            {
3246            if ((*set1++ & *set2++) != 0) return FALSE;
3247            }
3248          while (set1 < set_end);
3249          }
3250    
3251        if (list[1] == 0) return TRUE;
3252        /* Might be an empty repeat. */
3253        continue;
3254        }
3255    
3256      /* Some property combinations also acceptable. Unicode property opcodes are
3257      processed specially; the rest can be handled with a lookup table. */
3258    
3259      else
3260        {
3261        pcre_uint32 leftop, rightop;
3262    
3263        leftop = base_list[0];
3264        rightop = list[0];
3265    
3266    #ifdef SUPPORT_UCP
3267        accepted = FALSE; /* Always set in non-unicode case. */
3268        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3269          {
3270          if (rightop == OP_EOD)
3271            accepted = TRUE;
3272          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3273            {
3274            int n;
3275            const pcre_uint8 *p;
3276            BOOL same = leftop == rightop;
3277            BOOL lisprop = leftop == OP_PROP;
3278            BOOL risprop = rightop == OP_PROP;
3279            BOOL bothprop = lisprop && risprop;
3280    
3281            /* There's a table that specifies how each combination is to be
3282            processed:
3283              0   Always return FALSE (never auto-possessify)
3284              1   Character groups are distinct (possessify if both are OP_PROP)
3285              2   Check character categories in the same group (general or particular)
3286              3   Return TRUE if the two opcodes are not the same
3287              ... see comments below
3288            */
3289    
3290            n = propposstab[base_list[2]][list[2]];
3291            switch(n)
3292              {
3293              case 0: break;
3294              case 1: accepted = bothprop; break;
3295              case 2: accepted = (base_list[3] == list[3]) != same; break;
3296              case 3: accepted = !same; break;