/[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 93 by nigel, Sat Feb 24 21:41:42 2007 UTC revision 1555 by ph10, Sun May 3 17:58:17 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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  #define NLBLOCK cd             /* Block containing newline information */  #ifdef HAVE_CONFIG_H
46  #define PSSTART start_pattern  /* Field containing processed string start */  #include "config.h"
47  #define PSEND   end_pattern    /* Field containing processed string end */  #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing pattern start */
51    #define PSEND   end_pattern    /* Field containing pattern 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    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275  static const uschar posix_name_lengths[] = {  
276    /* Tables of names of POSIX character classes and their lengths. The names are
277    now all in a single string, to reduce the number of relocations when a shared
278    library is dynamically loaded. The list of lengths is terminated by a zero
279    length entry. The first three must be alpha, lower, upper, as this is assumed
280    for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283    static const char posix_names[] =
284      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287      STRING_word0  STRING_xdigit;
288    
289    static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 164  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325    Unicode property escapes. */
326    
327    #ifdef SUPPORT_UCP
328    static const pcre_uchar string_PNd[]  = {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pNd[]  = {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_PXsp[] = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXsp[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXwd[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXwd[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *substitutes[] = {
348      string_PNd,           /* \D */
349      string_pNd,           /* \d */
350      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352      string_PXwd,          /* \W */
353      string_pXwd           /* \w */
354    };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362    static const pcre_uchar string_pL[] =   {
363      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    static const pcre_uchar string_pLl[] =  {
366      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368    static const pcre_uchar string_pLu[] =  {
369      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pXan[] = {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_h[] =    {
375      CHAR_BACKSLASH, CHAR_h, '\0' };
376    static const pcre_uchar string_pXps[] = {
377      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379    static const pcre_uchar string_PL[] =   {
380      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382    static const pcre_uchar string_PLl[] =  {
383      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PLu[] =  {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PXan[] = {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_H[] =    {
392      CHAR_BACKSLASH, CHAR_H, '\0' };
393    static const pcre_uchar string_PXps[] = {
394      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396    
397    static const pcre_uchar *posix_substitutes[] = {
398      string_pL,            /* alpha */
399      string_pLl,           /* lower */
400      string_pLu,           /* upper */
401      string_pXan,          /* alnum */
402      NULL,                 /* ascii */
403      string_h,             /* blank */
404      NULL,                 /* cntrl */
405      string_pNd,           /* digit */
406      NULL,                 /* graph */
407      NULL,                 /* print */
408      NULL,                 /* punct */
409      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411      NULL,                 /* xdigit */
412      /* Negated cases */
413      string_PL,            /* ^alpha */
414      string_PLl,           /* ^lower */
415      string_PLu,           /* ^upper */
416      string_PXan,          /* ^alnum */
417      NULL,                 /* ^ascii */
418      string_H,             /* ^blank */
419      NULL,                 /* ^cntrl */
420      string_PNd,           /* ^digit */
421      NULL,                 /* ^graph */
422      NULL,                 /* ^print */
423      NULL,                 /* ^punct */
424      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426      NULL                  /* ^xdigit */
427    };
428    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429    #endif
430    
431  #define STRING(a)  # a  #define STRING(a)  # a
432  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 434  static const int posix_class_maps[] = {
434  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
435  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
436  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
437  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
438    the number of relocations needed when a shared library is loaded dynamically,
439  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
440    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
442    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
443    "unrecognized character follows \\",  
444    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
445    substring, so that the whole string ends with \0\0, which can be detected when
446    counting through. */
447    
448    static const char error_texts[] =
449      "no error\0"
450      "\\ at end of pattern\0"
451      "\\c at end of pattern\0"
452      "unrecognized character follows \\\0"
453      "numbers out of order in {} quantifier\0"
454    /* 5 */    /* 5 */
455    "number too big in {} quantifier",    "number too big in {} quantifier\0"
456    "missing terminating ] for character class",    "missing terminating ] for character class\0"
457    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
458    "range out of order in character class",    "range out of order in character class\0"
459    "nothing to repeat",    "nothing to repeat\0"
460    /* 10 */    /* 10 */
461    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
462    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
463    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
464    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
465    "missing )",    "missing )\0"
466    /* 15 */    /* 15 */
467    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
468    "erroffset passed as NULL",    "erroffset passed as NULL\0"
469    "unknown option bit(s) set",    "unknown option bit(s) set\0"
470    "missing ) after comment",    "missing ) after comment\0"
471    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
472    /* 20 */    /* 20 */
473    "regular expression too large",    "regular expression is too large\0"
474    "failed to get memory",    "failed to get memory\0"
475    "unmatched parentheses",    "unmatched parentheses\0"
476    "internal error: code overflow",    "internal error: code overflow\0"
477    "unrecognized character after (?<",    "unrecognized character after (?<\0"
478    /* 25 */    /* 25 */
479    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
480    "malformed number or name after (?(",    "malformed number or name after (?(\0"
481    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
482    "assertion expected after (?(",    "assertion expected after (?(\0"
483    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
484    /* 30 */    /* 30 */
485    "unknown POSIX class name",    "unknown POSIX class name\0"
486    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
487    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
488    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large",    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)",    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
493    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494    "number after (?C is > 255",    "number after (?C is > 255\0"
495    "closing ) for (?C expected",    "closing ) for (?C expected\0"
496    /* 40 */    /* 40 */
497    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
498    "unrecognized character after (?P",    "unrecognized character after (?P\0"
499    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
500    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
501    "invalid UTF-8 string",    "invalid UTF-8 string\0"
502    /* 45 */    /* 45 */
503    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
504    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
505    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
506    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508    /* 50 */    /* 50 */
509    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
510    "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"
511    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
512    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
513    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
514    /* 55 */    /* 55 */
515    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
516    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
517    "\\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"
518  };    "a numbered reference must not be zero\0"
519      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520      /* 60 */
521      "(*VERB) not recognized or malformed\0"
522      "number is too big\0"
523      "subpattern name expected\0"
524      "digit expected after (?+\0"
525      "] is an invalid data character in JavaScript compatibility mode\0"
526      /* 65 */
527      "different names for subpatterns of the same number are not allowed\0"
528      "(*MARK) must have an argument\0"
529      "this version of PCRE is not compiled with Unicode property support\0"
530      "\\c must be followed by an ASCII character\0"
531      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532      /* 70 */
533      "internal error: unknown opcode in find_fixedlength()\0"
534      "\\N is not supported in a class\0"
535      "too many forward references\0"
536      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537      "invalid UTF-16 string\0"
538      /* 75 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553      ;
554    
555  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
556  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 568  For convenience, we use the same bit def
568    
569  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
570    
571  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
572  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
573    into a subtraction and unsigned comparison). */
574    
575    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
576    
577    #ifndef EBCDIC
578    
579    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
580    UTF-8 mode. */
581    
582    static const pcre_uint8 digitab[] =
583    {    {
584    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
585    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 614  static const unsigned char digitab[] =
614    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
615    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
616    
617  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
618  static const unsigned char digitab[] =  
619    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
620    
621    static const pcre_uint8 digitab[] =
622    {    {
623    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
624    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 631  static const unsigned char digitab[] =
631    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
632    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
633    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
634    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
635    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
636    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
637    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 653  static const unsigned char digitab[] =
653    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
654    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
655    
656  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
657    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
658    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
659    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 665  static const unsigned char ebcdic_charta
665    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
666    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
667    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
668    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
669    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
670    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
671    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 689  static const unsigned char ebcdic_charta
689  #endif  #endif
690    
691    
692  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730    /* This table is used to check whether auto-possessification is possible
731    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732    left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
785      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
790      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
870    /*************************************************
871    *            Find an error text                  *
872    *************************************************/
873    
874    /* The error texts are now all in one long string, to save on relocations. As
875    some of the text is of unknown length, we can't use a table of offsets.
876    Instead, just count through the strings. This is not a performance issue
877    because it happens only when there has been a compilation error.
878    
879    Argument:   the error number
880    Returns:    pointer to the error string
881    */
882    
883    static const char *
884    find_error_text(int n)
885    {
886    const char *s = error_texts;
887    for (; n > 0; n--)
888      {
889      while (*s++ != CHAR_NULL) {};
890      if (*s == CHAR_NULL) return "Error text not found (please report)";
891      }
892    return s;
893    }
894    
895    
896    
897    /*************************************************
898    *           Expand the workspace                 *
899    *************************************************/
900    
901    /* This function is called during the second compiling phase, if the number of
902    forward references fills the existing workspace, which is originally a block on
903    the stack. A larger block is obtained from malloc() unless the ultimate limit
904    has been reached or the increase will be rather small.
905    
906    Argument: pointer to the compile data block
907    Returns:  0 if all went well, else an error number
908    */
909    
910    static int
911    expand_workspace(compile_data *cd)
912    {
913    pcre_uchar *newspace;
914    int newsize = cd->workspace_size * 2;
915    
916    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
917    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
918        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
919     return ERR72;
920    
921    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
922    if (newspace == NULL) return ERR21;
923    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
924    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
925    if (cd->workspace_size > COMPILE_WORK_SIZE)
926      (PUBL(free))((void *)cd->start_workspace);
927    cd->start_workspace = newspace;
928    cd->workspace_size = newsize;
929    return 0;
930    }
931    
932    
933    
934    /*************************************************
935    *            Check for counted repeat            *
936    *************************************************/
937    
938    /* This function is called when a '{' is encountered in a place where it might
939    start a quantifier. It looks ahead to see if it really is a quantifier or not.
940    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
941    where the ddds are digits.
942    
943    Arguments:
944      p         pointer to the first char after '{'
945    
946    Returns:    TRUE or FALSE
947    */
948    
949  static BOOL  static BOOL
950    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
951      int *, branch_chain *, compile_data *, int *);  {
952    if (!IS_DIGIT(*p)) return FALSE;
953    p++;
954    while (IS_DIGIT(*p)) p++;
955    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956    
957    if (*p++ != CHAR_COMMA) return FALSE;
958    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
959    
960    if (!IS_DIGIT(*p)) return FALSE;
961    p++;
962    while (IS_DIGIT(*p)) p++;
963    
964    return (*p == CHAR_RIGHT_CURLY_BRACKET);
965    }
966    
967    
968    
# Line 383  static BOOL Line 971  static BOOL
971  *************************************************/  *************************************************/
972    
973  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
974  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
975  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.
976  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
977  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
978  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982      chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
986    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
987    
988  Returns:         zero or positive => a data character  Returns:         zero => a data character
989                   negative => a special escape sequence                   positive => a special escape sequence
990                   on error, errorptr is set                   negative => a back reference
991                     on error, errorcodeptr is set
992  */  */
993    
994  static int  static int
995  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
997  {  {
998  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
1000  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
1001    pcre_uint32 c;
1002    int escape = 0;
1003    int i;
1004    
1005  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1006  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1007    
1008  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1009    
1010  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011    
1012  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013  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.
1014  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1015    
1016  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1018  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1019    else if ((i = escapes[c - CHAR_0]) != 0)
1020  #else          /* EBCDIC coding */    { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  
1022  else if ((i = escapes[c - 0x48]) != 0)  c = i;  #else           /* EBCDIC coding */
1023    /* Not alphanumeric */
1024    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1025    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1026  #endif  #endif
1027    
1028  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1029    
1030  else  else
1031    {    {
1032    const uschar *oldptr;    const pcre_uchar *oldptr;
1033    BOOL braced, negated;    BOOL braced, negated, overflow;
1034      int s;
1035    
1036    switch (c)    switch (c)
1037      {      {
1038      /* 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
1039      error. */      error. */
1040    
1041      case 'l':      case CHAR_l:
1042      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1043      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1044      break;      break;
1045    
1046      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
1047      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1048      This is a Perl 5.10 feature. */        {
1049          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1050          Otherwise it is a lowercase u letter. */
1051          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1052            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1053            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1054            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1055            {
1056            c = 0;
1057            for (i = 0; i < 4; ++i)
1058              {
1059              register pcre_uint32 cc = *(++ptr);
1060    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1061              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1062              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1063    #else           /* EBCDIC coding */
1064              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1065              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066    #endif
1067              }
1068    
1069    #if defined COMPILE_PCRE8
1070            if (c > (utf ? 0x10ffffU : 0xffU))
1071    #elif defined COMPILE_PCRE16
1072            if (c > (utf ? 0x10ffffU : 0xffffU))
1073    #elif defined COMPILE_PCRE32
1074            if (utf && c > 0x10ffffU)
1075    #endif
1076              {
1077              *errorcodeptr = ERR76;
1078              }
1079            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080            }
1081          }
1082        else
1083          *errorcodeptr = ERR37;
1084        break;
1085    
1086        case CHAR_U:
1087        /* In JavaScript, \U is an uppercase U letter. */
1088        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1089        break;
1090    
1091        /* In a character class, \g is just a literal "g". Outside a character
1092        class, \g must be followed by one of a number of specific things:
1093    
1094        (1) A number, either plain or braced. If positive, it is an absolute
1095        backreference. If negative, it is a relative backreference. This is a Perl
1096        5.10 feature.
1097    
1098        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1099        is part of Perl's movement towards a unified syntax for back references. As
1100        this is synonymous with \k{name}, we fudge it up by pretending it really
1101        was \k.
1102    
1103        (3) For Oniguruma compatibility we also support \g followed by a name or a
1104        number either in angle brackets or in single quotes. However, these are
1105        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106        the ESC_g code (cf \k). */
1107    
1108        case CHAR_g:
1109        if (isclass) break;
1110        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111          {
1112          escape = ESC_g;
1113          break;
1114          }
1115    
1116        /* Handle the Perl-compatible cases */
1117    
1118      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1119        {        {
1120          const pcre_uchar *p;
1121          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1122            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124            {
1125            escape = ESC_k;
1126            break;
1127            }
1128        braced = TRUE;        braced = TRUE;
1129        ptr++;        ptr++;
1130        }        }
1131      else braced = FALSE;      else braced = FALSE;
1132    
1133      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1134        {        {
1135        negated = TRUE;        negated = TRUE;
1136        ptr++;        ptr++;
1137        }        }
1138      else negated = FALSE;      else negated = FALSE;
1139    
1140      c = 0;      /* The integer range is limited by the machine's int representation. */
1141      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1142        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1143        while (IS_DIGIT(ptr[1]))
1144          {
1145          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146            {
1147            overflow = TRUE;
1148            break;
1149            }
1150          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151          }
1152        if (overflow) /* Integer overflow */
1153          {
1154          while (IS_DIGIT(ptr[1]))
1155            ptr++;
1156          *errorcodeptr = ERR61;
1157          break;
1158          }
1159    
1160      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1161        {        {
1162        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1163        return 0;        break;
1164          }
1165    
1166        if (s == 0)
1167          {
1168          *errorcodeptr = ERR58;
1169          break;
1170        }        }
1171    
1172      if (negated)      if (negated)
1173        {        {
1174        if (c > bracount)        if (s > bracount)
1175          {          {
1176          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1177          return 0;          break;
1178          }          }
1179        c = bracount - (c - 1);        s = bracount - (s - 1);
1180        }        }
1181    
1182      c = -(ESC_REF + c);      escape = -s;
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1191      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1192      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1193      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1194      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1195      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198      case '1': case '2': case '3': case '4': case '5':      Inside a character class, \ followed by a digit is always either a literal
1199      case '6': case '7': case '8': case '9':      8 or 9 or an octal number. */
1200    
1201        case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1202        case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1203    
1204      if (!isclass)      if (!isclass)
1205        {        {
1206        oldptr = ptr;        oldptr = ptr;
1207        c -= '0';        /* The integer range is limited by the machine's int representation. */
1208        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1209          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1210        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1211            {
1212            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213              {
1214              overflow = TRUE;
1215              break;
1216              }
1217            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218            }
1219          if (overflow) /* Integer overflow */
1220            {
1221            while (IS_DIGIT(ptr[1]))
1222              ptr++;
1223            *errorcodeptr = ERR61;
1224            break;
1225            }
1226          if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          c = -(ESC_REF + c);          escape = -s;
1229          break;          break;
1230          }          }
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1235      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1236      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1244      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1245      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
1246      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,
1247      than 3 octal digits. */      but no more than 3 octal digits. */
1248    
1249      case '0':      case CHAR_0:
1250      c -= '0';      c -= CHAR_0;
1251      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1252          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1253      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1254        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1255    #endif
1256      break;      break;
1257    
1258      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1259      than 0xff in 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. */  
1260    
1261      case 'x':      case CHAR_o:
1262      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264        {        {
1265        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1266        c = 0;        c = 0;
1267        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269          {          {
1270          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1271          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272          count++;  #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274  #if !EBCDIC    /* ASCII coding */  #endif
1275          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1276          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1277  #else          /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1279          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282  #endif  #endif
1283          }          }
1284          if (overflow)
       if (*pt == '}')  
1285          {          {
1286          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1288          }          }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290        /* If the sequence of hex digits does not end with '}', then we don't          {
1291        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294        }        }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      /* Read just a single-byte hex-defined char */      case CHAR_x:
1301        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302          {
1303          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305            {
1306            c = 0;
1307            for (i = 0; i < 2; ++i)
1308              {
1309              register pcre_uint32 cc = *(++ptr);
1310    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1311              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1312              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1313    #else           /* EBCDIC coding */
1314              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1315              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1316    #endif
1317              }
1318            }
1319          }    /* End JavaScript handling */
1320    
1321      c = 0;      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        else
1329        {        {
1330        int cc;                               /* Some compilers don't like ++ */        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331        cc = *(++ptr);                        /* in initializers */          {
1332  #if !EBCDIC    /* ASCII coding */          ptr += 2;
1333        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));            {
1335  #else          /* EBCDIC coding */            *errorcodeptr = ERR86;
1336        if (cc <= 'z') cc += 64;              /* Convert to upper case */            break;
1337        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345    #ifdef COMPILE_PCRE32
1346              if (c >= 0x10000000l) { overflow = TRUE; break; }
1347  #endif  #endif
1348        }  
1349    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352    #else           /* EBCDIC coding */
1353              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1354              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355    #endif
1356    
1357    #if defined COMPILE_PCRE8
1358              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359    #elif defined COMPILE_PCRE16
1360              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361    #elif defined COMPILE_PCRE32
1362              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363    #endif
1364              }
1365    
1366            if (overflow)
1367              {
1368              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369              *errorcodeptr = ERR34;
1370              }
1371    
1372            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373              {
1374              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375              }
1376    
1377            /* If the sequence of hex digits does not end with '}', give an error.
1378            We used just to recognize this construct and fall through to the normal
1379            \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382            else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386    
1387          else
1388            {
1389            c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397    #else           /* EBCDIC coding */
1398              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400    #endif
1401              }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1407      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
1408        coding is ASCII-specific, but then the whole concept of \cx is
1409      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1410    
1411      case 'c':      case CHAR_c:
1412      c = *(++ptr);      c = *(++ptr);
1413      if (c == 0)      if (c == CHAR_NULL)
1414        {        {
1415        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1416        return 0;        break;
1417        }        }
1418    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1419  #if !EBCDIC    /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1420      if (c >= 'a' && c <= 'z') c -= 32;        {
1421          *errorcodeptr = ERR68;
1422          break;
1423          }
1424        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1425      c ^= 0x40;      c ^= 0x40;
1426  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1427      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1428      c ^= 0xC0;      c ^= 0xC0;
1429  #endif  #endif
1430      break;      break;
1431    
1432      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1433      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1434      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
1435      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
1436      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1437    
1438      default:      default:
1439      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1446  else
1446      }      }
1447    }    }
1448    
1449    /* Perl supports \N{name} for character names, as well as plain \N for "not
1450    newline". PCRE does not support \N{name}. However, it does support
1451    quantification such as \N{2,3}. */
1452    
1453    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454         !is_counted_repeat(ptr+2))
1455      *errorcodeptr = ERR37;
1456    
1457    /* If PCRE_UCP is set, we change the values for \d etc. */
1458    
1459    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460      escape += (ESC_DU - ESC_D);
1461    
1462    /* Set the pointer to the final character before returning. */
1463    
1464  *ptrptr = ptr;  *ptrptr = ptr;
1465  return c;  *chptr = c;
1466    return escape;
1467  }  }
1468    
1469    
# Line 656  escape sequence. Line 1481  escape sequence.
1481  Argument:  Argument:
1482    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1483    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1484    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1485      pdataptr       points to an unsigned int that is set to the detailed property value
1486    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1487    
1488  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1489  */  */
1490    
1491  static int  static BOOL
1492  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493      unsigned int *pdataptr, int *errorcodeptr)
1494  {  {
1495  int c, i, bot, top;  pcre_uchar c;
1496  const uschar *ptr = *ptrptr;  int i, bot, top;
1497  char name[32];  const pcre_uchar *ptr = *ptrptr;
1498    pcre_uchar name[32];
1499    
1500  c = *(++ptr);  c = *(++ptr);
1501  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1502    
1503  *negptr = FALSE;  *negptr = FALSE;
1504    
1505  /* \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
1506  negation. */  negation. */
1507    
1508  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1509    {    {
1510    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1511      {      {
1512      *negptr = TRUE;      *negptr = TRUE;
1513      ptr++;      ptr++;
1514      }      }
1515    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516      {      {
1517      c = *(++ptr);      c = *(++ptr);
1518      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1519      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520      name[i] = c;      name[i] = c;
1521      }      }
1522    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1523    name[i] = 0;    name[i] = 0;
1524    }    }
1525    
# Line 708  else Line 1536  else
1536  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1537    
1538  bot = 0;  bot = 0;
1539  top = _pcre_utt_size;  top = PRIV(utt_size);
1540    
1541  while (bot < top)  while (bot < top)
1542    {    {
1543      int r;
1544    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1545    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1546    if (c == 0)    if (r == 0)
1547      {      {
1548      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1549      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1550        return TRUE;
1551      }      }
1552    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1553    }    }
1554    
1555  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1556  *ptrptr = ptr;  *ptrptr = ptr;
1557  return -1;  return FALSE;
1558    
1559  ERROR_RETURN:  ERROR_RETURN:
1560  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1561  *ptrptr = ptr;  *ptrptr = ptr;
1562  return -1;  return FALSE;
1563  }  }
1564  #endif  #endif
1565    
1566    
1567    
   
 /*************************************************  
 *            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 == '}');  
 }  
   
   
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 788  Returns:         pointer to '}' on succe Line 1584  Returns:         pointer to '}' on succe
1584                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1585  */  */
1586    
1587  static const uschar *  static const pcre_uchar *
1588  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)
1589  {  {
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597        {
1598        *errorcodeptr = ERR5;
1599        return p;
1600        }
1601    }    }
1602    
1603  /* Read the maximum value if there is one, and again do a paranoid on its size.  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
 Also, max must not be less than min. */  
   
 if (*p == '}') max = min; else  
1604    {    {
1605    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612            {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616        }        }
1617      if (max < min)      if (max < min)
1618        {        {
# Line 826  if (*p == '}') max = min; else Line 1622  if (*p == '}') max = min; else
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 837  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We know that if (?P< is encountered, the name will  
 be terminated by '>' because that is checked in the first pass.  
   
 Arguments:  
   ptr          current position in the pattern  
   count        current count of capturing parens so far encountered  
   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  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  
   BOOL xmode)  
 {  
 const uschar *thisname;  
   
 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;  
 }  
   
   
   
 /*************************************************  
1633  *      Find first significant op code            *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
1636  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1637  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1638  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1639  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1640  assertions, and also the \b assertion; for others it does not.  does not.
1641    
1642  Arguments:  Arguments:
1643    code         pointer to the start of the group    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  
1644    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1645    
1646  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1647  */  */
1648    
1649  static const uschar*  static const pcre_uchar*
1650  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1651  {  {
1652  for (;;)  for (;;)
1653    {    {
1654    switch ((int)*code)    switch ((int)*code)
1655      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1656      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1657      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1658      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1659      if (!skipassert) return code;      if (!skipassert) return code;
1660      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1661      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1662      break;      break;
1663    
1664      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671        case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673        case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
1677    
1678      default:      default:
# Line 1010  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
1690    
1691  /* 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,
1692  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.
1693  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
1694    temporarily terminated with OP_END when this function is called.
1695    
1696    This function is called when a backward assertion is encountered, so that if it
1697    fails, the error message can point to the correct place in the pattern.
1698    However, we cannot do this when the assertion contains subroutine calls,
1699    because they can be forward references. We solve this by remembering this case
1700    and doing the check at the end; a flag specifies which mode we are running in.
1701    
1702  Arguments:  Arguments:
1703    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1704    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705      atend    TRUE if called when the pattern is complete
1706  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1707               or -2 if \C was encountered    recurses    chain of recurse_check to catch mutual recursion
1708    
1709    Returns:   the fixed length,
1710                 or -1 if there is no fixed length,
1711                 or -2 if \C was encountered (in UTF-8 mode only)
1712                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1713                 or -4 if an unknown opcode was encountered (internal error)
1714  */  */
1715    
1716  static int  static int
1717  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1718      recurse_check *recurses)
1719  {  {
1720  int length = -1;  int length = -1;
1721    recurse_check this_recurse;
1722  register int branchlength = 0;  register int branchlength = 0;
1723  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1724    
1725  /* 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
1726  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 1728  branch, check the length against that of
1728  for (;;)  for (;;)
1729    {    {
1730    int d;    int d;
1731    register int op = *cc;    pcre_uchar *ce, *cs;
1732      register pcre_uchar op = *cc;
1733    
1734    switch (op)    switch (op)
1735      {      {
1736        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1737        OP_BRA (normal non-capturing bracket) because the other variants of these
1738        opcodes are all concerned with unlimited repeated groups, which of course
1739        are not of fixed length. */
1740    
1741      case OP_CBRA:      case OP_CBRA:
1742      case OP_BRA:      case OP_BRA:
1743      case OP_ONCE:      case OP_ONCE:
1744        case OP_ONCE_NC:
1745      case OP_COND:      case OP_COND:
1746      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1747          recurses);
1748      if (d < 0) return d;      if (d < 0) return d;
1749      branchlength += d;      branchlength += d;
1750      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1751      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1752      break;      break;
1753    
1754      /* 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.
1755      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
1756      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
1757        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1758        because they all imply an unlimited repeat. */
1759    
1760      case OP_ALT:      case OP_ALT:
1761      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1762      case OP_END:      case OP_END:
1763        case OP_ACCEPT:
1764        case OP_ASSERT_ACCEPT:
1765      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1766        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1767      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1769  for (;;)
1769      branchlength = 0;      branchlength = 0;
1770      break;      break;
1771    
1772        /* A true recursion implies not fixed length, but a subroutine call may
1773        be OK. If the subroutine is a forward reference, we can't deal with
1774        it until the end of the pattern, so return -3. */
1775    
1776        case OP_RECURSE:
1777        if (!atend) return -3;
1778        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1779        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1780        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1781        else   /* Check for mutual recursion */
1782          {
1783          recurse_check *r = recurses;
1784          for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1785          if (r != NULL) return -1;   /* Mutual recursion */
1786          }
1787        this_recurse.prev = recurses;
1788        this_recurse.group = cs;
1789        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1790        if (d < 0) return d;
1791        branchlength += d;
1792        cc += 1 + LINK_SIZE;
1793        break;
1794    
1795      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1796    
1797      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1799  for (;;)
1799      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1800      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1801      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1802      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1803        break;
1804    
1805      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1806    
1807      case OP_REVERSE:      case OP_MARK:
1808        case OP_PRUNE_ARG:
1809        case OP_SKIP_ARG:
1810        case OP_THEN_ARG:
1811        cc += cc[1] + PRIV(OP_lengths)[*cc];
1812        break;
1813    
1814        case OP_CALLOUT:
1815        case OP_CIRC:
1816        case OP_CIRCM:
1817        case OP_CLOSE:
1818        case OP_COMMIT:
1819      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1820      case OP_DEF:      case OP_DEF:
1821      case OP_OPT:      case OP_DNCREF:
1822      case OP_CALLOUT:      case OP_DNRREF:
1823      case OP_SOD:      case OP_DOLL:
1824      case OP_SOM:      case OP_DOLLM:
1825      case OP_EOD:      case OP_EOD:
1826      case OP_EODN:      case OP_EODN:
1827      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1828      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1829        case OP_PRUNE:
1830        case OP_REVERSE:
1831        case OP_RREF:
1832        case OP_SET_SOM:
1833        case OP_SKIP:
1834        case OP_SOD:
1835        case OP_SOM:
1836        case OP_THEN:
1837      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1838      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1839      break;      break;
1840    
1841      /* Handle literal characters */      /* Handle literal characters */
1842    
1843      case OP_CHAR:      case OP_CHAR:
1844      case OP_CHARNC:      case OP_CHARI:
1845      case OP_NOT:      case OP_NOT:
1846        case OP_NOTI:
1847      branchlength++;      branchlength++;
1848      cc += 2;      cc += 2;
1849  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1850      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1851  #endif  #endif
1852      break;      break;
1853    
# Line 1119  for (;;) Line 1855  for (;;)
1855      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1856    
1857      case OP_EXACT:      case OP_EXACT:
1858      branchlength += GET2(cc,1);      case OP_EXACTI:
1859      cc += 4;      case OP_NOTEXACT:
1860  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1861      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1862        {      cc += 2 + IMM2_SIZE;
1863        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1864        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1865  #endif  #endif
1866      break;      break;
1867    
1868      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1869      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1870      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1871          cc += 2;
1872        cc += 1 + IMM2_SIZE + 1;
1873      break;      break;
1874    
1875      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1879  for (;;)
1879      cc += 2;      cc += 2;
1880      /* Fall through */      /* Fall through */
1881    
1882        case OP_HSPACE:
1883        case OP_VSPACE:
1884        case OP_NOT_HSPACE:
1885        case OP_NOT_VSPACE:
1886      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1887      case OP_DIGIT:      case OP_DIGIT:
1888      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1890  for (;;)
1890      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1891      case OP_WORDCHAR:      case OP_WORDCHAR:
1892      case OP_ANY:      case OP_ANY:
1893        case OP_ALLANY:
1894      branchlength++;      branchlength++;
1895      cc++;      cc++;
1896      break;      break;
1897    
1898      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1899        otherwise \C is coded as OP_ALLANY. */
1900    
1901      case OP_ANYBYTE:      case OP_ANYBYTE:
1902      return -2;      return -2;
1903    
1904      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1905    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1906      case OP_CLASS:      case OP_CLASS:
1907      case OP_NCLASS:      case OP_NCLASS:
1908      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1909        case OP_XCLASS:
1910        /* The original code caused an unsigned overflow in 64 bit systems,
1911        so now we use a conditional statement. */
1912        if (op == OP_XCLASS)
1913          cc += GET(cc, 1);
1914        else
1915          cc += PRIV(OP_lengths)[OP_CLASS];
1916    #else
1917        cc += PRIV(OP_lengths)[OP_CLASS];
1918    #endif
1919    
1920      switch (*cc)      switch (*cc)
1921        {        {
1922        case OP_CRSTAR:        case OP_CRSTAR:
1923        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1924          case OP_CRPLUS:
1925          case OP_CRMINPLUS:
1926        case OP_CRQUERY:        case OP_CRQUERY:
1927        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1928          case OP_CRPOSSTAR:
1929          case OP_CRPOSPLUS:
1930          case OP_CRPOSQUERY:
1931        return -1;        return -1;
1932    
1933        case OP_CRRANGE:        case OP_CRRANGE:
1934        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1935        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1936        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1937        cc += 5;        branchlength += (int)GET2(cc,1);
1938          cc += 1 + 2 * IMM2_SIZE;
1939        break;        break;
1940    
1941        default:        default:
# Line 1191  for (;;) Line 1945  for (;;)
1945    
1946      /* Anything else is variable length */      /* Anything else is variable length */
1947    
1948      default:      case OP_ANYNL:
1949        case OP_BRAMINZERO:
1950        case OP_BRAPOS:
1951        case OP_BRAPOSZERO:
1952        case OP_BRAZERO:
1953        case OP_CBRAPOS:
1954        case OP_EXTUNI:
1955        case OP_KETRMAX:
1956        case OP_KETRMIN:
1957        case OP_KETRPOS:
1958        case OP_MINPLUS:
1959        case OP_MINPLUSI:
1960        case OP_MINQUERY:
1961        case OP_MINQUERYI:
1962        case OP_MINSTAR:
1963        case OP_MINSTARI:
1964        case OP_MINUPTO:
1965        case OP_MINUPTOI:
1966        case OP_NOTMINPLUS:
1967        case OP_NOTMINPLUSI:
1968        case OP_NOTMINQUERY:
1969        case OP_NOTMINQUERYI:
1970        case OP_NOTMINSTAR:
1971        case OP_NOTMINSTARI:
1972        case OP_NOTMINUPTO:
1973        case OP_NOTMINUPTOI:
1974        case OP_NOTPLUS:
1975        case OP_NOTPLUSI:
1976        case OP_NOTPOSPLUS:
1977        case OP_NOTPOSPLUSI:
1978        case OP_NOTPOSQUERY:
1979        case OP_NOTPOSQUERYI:
1980        case OP_NOTPOSSTAR:
1981        case OP_NOTPOSSTARI:
1982        case OP_NOTPOSUPTO:
1983        case OP_NOTPOSUPTOI:
1984        case OP_NOTQUERY:
1985        case OP_NOTQUERYI:
1986        case OP_NOTSTAR:
1987        case OP_NOTSTARI:
1988        case OP_NOTUPTO:
1989        case OP_NOTUPTOI:
1990        case OP_PLUS:
1991        case OP_PLUSI:
1992        case OP_POSPLUS:
1993        case OP_POSPLUSI:
1994        case OP_POSQUERY:
1995        case OP_POSQUERYI:
1996        case OP_POSSTAR:
1997        case OP_POSSTARI:
1998        case OP_POSUPTO:
1999        case OP_POSUPTOI:
2000        case OP_QUERY:
2001        case OP_QUERYI:
2002        case OP_REF:
2003        case OP_REFI:
2004        case OP_DNREF:
2005        case OP_DNREFI:
2006        case OP_SBRA:
2007        case OP_SBRAPOS:
2008        case OP_SCBRA:
2009        case OP_SCBRAPOS:
2010        case OP_SCOND:
2011        case OP_SKIPZERO:
2012        case OP_STAR:
2013        case OP_STARI:
2014        case OP_TYPEMINPLUS:
2015        case OP_TYPEMINQUERY:
2016        case OP_TYPEMINSTAR:
2017        case OP_TYPEMINUPTO:
2018        case OP_TYPEPLUS:
2019        case OP_TYPEPOSPLUS:
2020        case OP_TYPEPOSQUERY:
2021        case OP_TYPEPOSSTAR:
2022        case OP_TYPEPOSUPTO:
2023        case OP_TYPEQUERY:
2024        case OP_TYPESTAR:
2025        case OP_TYPEUPTO:
2026        case OP_UPTO:
2027        case OP_UPTOI:
2028      return -1;      return -1;
2029    
2030        /* Catch unrecognized opcodes so that when new ones are added they
2031        are not forgotten, as has happened in the past. */
2032    
2033        default:
2034        return -4;
2035      }      }
2036    }    }
2037  /* Control never gets here */  /* Control never gets here */
# Line 1200  for (;;) Line 2039  for (;;)
2039    
2040    
2041    
   
2042  /*************************************************  /*************************************************
2043  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2044  *************************************************/  *************************************************/
2045    
2046  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2047  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2048    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2049    so that it can be called from pcre_study() when finding the minimum matching
2050    length.
2051    
2052  Arguments:  Arguments:
2053    code        points to start of expression    code        points to start of expression
2054    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2055    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2056    
2057  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
2058  */  */
2059    
2060  static const uschar *  const pcre_uchar *
2061  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2062  {  {
2063  for (;;)  for (;;)
2064    {    {
2065    register int c = *code;    register pcre_uchar c = *code;
2066    
2067    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2068    
2069    /* 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 2072  for (;;)
2072    
2073    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2074    
2075      /* Handle recursion */
2076    
2077      else if (c == OP_REVERSE)
2078        {
2079        if (number < 0) return (pcre_uchar *)code;
2080        code += PRIV(OP_lengths)[c];
2081        }
2082    
2083    /* Handle capturing bracket */    /* Handle capturing bracket */
2084    
2085    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2086               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2087      {      {
2088      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2089      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2090      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2091      }      }
2092    
2093      /* Otherwise, we can get the item's length from the table, except that for
2094      repeated character types, we have to test for \p and \P, which have an extra
2095      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2096      must add in its length. */
2097    
2098      else
2099        {
2100        switch(c)
2101          {
2102          case OP_TYPESTAR:
2103          case OP_TYPEMINSTAR:
2104          case OP_TYPEPLUS:
2105          case OP_TYPEMINPLUS:
2106          case OP_TYPEQUERY:
2107          case OP_TYPEMINQUERY:
2108          case OP_TYPEPOSSTAR:
2109          case OP_TYPEPOSPLUS:
2110          case OP_TYPEPOSQUERY:
2111          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2112          break;
2113    
2114          case OP_TYPEUPTO:
2115          case OP_TYPEMINUPTO:
2116          case OP_TYPEEXACT:
2117          case OP_TYPEPOSUPTO:
2118          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2119            code += 2;
2120          break;
2121    
2122          case OP_MARK:
2123          case OP_PRUNE_ARG:
2124          case OP_SKIP_ARG:
2125          case OP_THEN_ARG:
2126          code += code[1];
2127          break;
2128          }
2129    
2130        /* Add in the fixed length from the table */
2131    
2132        code += PRIV(OP_lengths)[c];
2133    
2134    /* 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
2135    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
2136    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2137    
2138    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2139      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
     if (utf8) switch(c)  
2140        {        {
2141        case OP_CHAR:        case OP_CHAR:
2142        case OP_CHARNC:        case OP_CHARI:
2143          case OP_NOT:
2144          case OP_NOTI:
2145        case OP_EXACT:        case OP_EXACT:
2146          case OP_EXACTI:
2147          case OP_NOTEXACT:
2148          case OP_NOTEXACTI:
2149        case OP_UPTO:        case OP_UPTO:
2150          case OP_UPTOI:
2151          case OP_NOTUPTO:
2152          case OP_NOTUPTOI:
2153        case OP_MINUPTO:        case OP_MINUPTO:
2154          case OP_MINUPTOI:
2155          case OP_NOTMINUPTO:
2156          case OP_NOTMINUPTOI:
2157        case OP_POSUPTO:        case OP_POSUPTO:
2158          case OP_POSUPTOI:
2159          case OP_NOTPOSUPTO:
2160          case OP_NOTPOSUPTOI:
2161        case OP_STAR:        case OP_STAR:
2162          case OP_STARI:
2163          case OP_NOTSTAR:
2164          case OP_NOTSTARI:
2165        case OP_MINSTAR:        case OP_MINSTAR:
2166          case OP_MINSTARI:
2167          case OP_NOTMINSTAR:
2168          case OP_NOTMINSTARI:
2169        case OP_POSSTAR:        case OP_POSSTAR:
2170          case OP_POSSTARI:
2171          case OP_NOTPOSSTAR:
2172          case OP_NOTPOSSTARI:
2173        case OP_PLUS:        case OP_PLUS:
2174          case OP_PLUSI:
2175          case OP_NOTPLUS:
2176          case OP_NOTPLUSI:
2177        case OP_MINPLUS:        case OP_MINPLUS:
2178          case OP_MINPLUSI:
2179          case OP_NOTMINPLUS:
2180          case OP_NOTMINPLUSI:
2181        case OP_POSPLUS:        case OP_POSPLUS:
2182          case OP_POSPLUSI:
2183          case OP_NOTPOSPLUS:
2184          case OP_NOTPOSPLUSI:
2185        case OP_QUERY:        case OP_QUERY:
2186          case OP_QUERYI:
2187          case OP_NOTQUERY:
2188          case OP_NOTQUERYI:
2189        case OP_MINQUERY:        case OP_MINQUERY:
2190          case OP_MINQUERYI:
2191          case OP_NOTMINQUERY:
2192          case OP_NOTMINQUERYI:
2193        case OP_POSQUERY:        case OP_POSQUERY:
2194        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2195          case OP_NOTPOSQUERY:
2196          case OP_NOTPOSQUERYI:
2197          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2198        break;        break;
2199        }        }
2200    #else
2201        (void)(utf);  /* Keep compiler happy by referencing function argument */
2202    #endif
2203      }      }
2204    }    }
2205  }  }
# Line 1281  instance of OP_RECURSE. Line 2215  instance of OP_RECURSE.
2215    
2216  Arguments:  Arguments:
2217    code        points to start of expression    code        points to start of expression
2218    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2219    
2220  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
2221  */  */
2222    
2223  static const uschar *  static const pcre_uchar *
2224  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2225  {  {
2226  for (;;)  for (;;)
2227    {    {
2228    register int c = *code;    register pcre_uchar c = *code;
2229    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2230    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2231    
# Line 1301  for (;;) Line 2235  for (;;)
2235    
2236    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2237    
2238    /* 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
2239    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
2240    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
2241    bytes. */    must add in its length. */
2242    
2243    else    else
2244      {      {
2245      code += _pcre_OP_lengths[c];      switch(c)
2246      if (utf8) switch(c)        {
2247          case OP_TYPESTAR:
2248          case OP_TYPEMINSTAR:
2249          case OP_TYPEPLUS:
2250          case OP_TYPEMINPLUS:
2251          case OP_TYPEQUERY:
2252          case OP_TYPEMINQUERY:
2253          case OP_TYPEPOSSTAR:
2254          case OP_TYPEPOSPLUS:
2255          case OP_TYPEPOSQUERY:
2256          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2257          break;
2258    
2259          case OP_TYPEPOSUPTO:
2260          case OP_TYPEUPTO:
2261          case OP_TYPEMINUPTO:
2262          case OP_TYPEEXACT:
2263          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2264            code += 2;
2265          break;
2266    
2267          case OP_MARK:
2268          case OP_PRUNE_ARG:
2269          case OP_SKIP_ARG:
2270          case OP_THEN_ARG:
2271          code += code[1];
2272          break;
2273          }
2274    
2275        /* Add in the fixed length from the table */
2276    
2277        code += PRIV(OP_lengths)[c];
2278    
2279        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2280        by a multi-byte character. The length in the table is a minimum, so we have
2281        to arrange to skip the extra bytes. */
2282    
2283    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2284        if (utf) switch(c)
2285        {        {
2286        case OP_CHAR:        case OP_CHAR:
2287        case OP_CHARNC:        case OP_CHARI:
2288          case OP_NOT:
2289          case OP_NOTI:
2290        case OP_EXACT:        case OP_EXACT:
2291          case OP_EXACTI:
2292          case OP_NOTEXACT:
2293          case OP_NOTEXACTI:
2294        case OP_UPTO:        case OP_UPTO:
2295          case OP_UPTOI:
2296          case OP_NOTUPTO:
2297          case OP_NOTUPTOI:
2298        case OP_MINUPTO:        case OP_MINUPTO:
2299          case OP_MINUPTOI:
2300          case OP_NOTMINUPTO:
2301          case OP_NOTMINUPTOI:
2302        case OP_POSUPTO:        case OP_POSUPTO:
2303          case OP_POSUPTOI:
2304          case OP_NOTPOSUPTO:
2305          case OP_NOTPOSUPTOI:
2306        case OP_STAR:        case OP_STAR:
2307          case OP_STARI:
2308          case OP_NOTSTAR:
2309          case OP_NOTSTARI:
2310        case OP_MINSTAR:        case OP_MINSTAR:
2311          case OP_MINSTARI:
2312          case OP_NOTMINSTAR:
2313          case OP_NOTMINSTARI:
2314        case OP_POSSTAR:        case OP_POSSTAR:
2315          case OP_POSSTARI:
2316          case OP_NOTPOSSTAR:
2317          case OP_NOTPOSSTARI:
2318        case OP_PLUS:        case OP_PLUS:
2319          case OP_PLUSI:
2320          case OP_NOTPLUS:
2321          case OP_NOTPLUSI:
2322        case OP_MINPLUS:        case OP_MINPLUS:
2323          case OP_MINPLUSI:
2324          case OP_NOTMINPLUS:
2325          case OP_NOTMINPLUSI:
2326        case OP_POSPLUS:        case OP_POSPLUS:
2327          case OP_POSPLUSI:
2328          case OP_NOTPOSPLUS:
2329          case OP_NOTPOSPLUSI:
2330        case OP_QUERY:        case OP_QUERY:
2331          case OP_QUERYI:
2332          case OP_NOTQUERY:
2333          case OP_NOTQUERYI:
2334        case OP_MINQUERY:        case OP_MINQUERY:
2335          case OP_MINQUERYI:
2336          case OP_NOTMINQUERY:
2337          case OP_NOTMINQUERYI:
2338        case OP_POSQUERY:        case OP_POSQUERY:
2339        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2340          case OP_NOTPOSQUERY:
2341          case OP_NOTPOSQUERYI:
2342          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2343        break;        break;
2344        }        }
2345    #else
2346        (void)(utf);  /* Keep compiler happy by referencing function argument */
2347    #endif
2348      }      }
2349    }    }
2350  }  }
# Line 1343  for (;;) Line 2359  for (;;)
2359  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()
2360  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
2361  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2362  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
2363  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
2364    bracket whose current branch will already have been scanned.
2365    
2366  Arguments:  Arguments:
2367    code        points to start of search    code        points to start of search
2368    endcode     points to where to stop    endcode     points to where to stop
2369    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2370      cd          contains pointers to tables etc.
2371      recurses    chain of recurse_check to catch mutual recursion
2372    
2373  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2374  */  */
2375    
2376  static BOOL  static BOOL
2377  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2378      BOOL utf, compile_data *cd, recurse_check *recurses)
2379  {  {
2380  register int c;  register pcre_uchar c;
2381  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2382    
2383    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2384       code < endcode;       code < endcode;
2385       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2386    {    {
2387    const uschar *ccode;    const pcre_uchar *ccode;
2388    
2389    c = *code;    c = *code;
2390    
2391    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2392      first_significant_code() with a TRUE final argument. */
2393    
2394      if (c == OP_ASSERT)
2395        {
2396        do code += GET(code, 1); while (*code == OP_ALT);
2397        c = *code;
2398        continue;
2399        }
2400    
2401      /* For a recursion/subroutine call, if its end has been reached, which
2402      implies a backward reference subroutine call, we can scan it. If it's a
2403      forward reference subroutine call, we can't. To detect forward reference
2404      we have to scan up the list that is kept in the workspace. This function is
2405      called only when doing the real compile, not during the pre-compile that
2406      measures the size of the compiled pattern. */
2407    
2408      if (c == OP_RECURSE)
2409      {      {
2410        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2411        const pcre_uchar *endgroup = scode;
2412      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2413    
2414      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2415        when called to scan a completed pattern by setting cd->start_workspace to
2416        NULL. */
2417    
2418        if (cd->start_workspace != NULL)
2419          {
2420          const pcre_uchar *tcode;
2421          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2422            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2423          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2424          }
2425    
2426        /* If the reference is to a completed group, we need to detect whether this
2427        is a recursive call, as otherwise there will be an infinite loop. If it is
2428        a recursion, just skip over it. Simple recursions are easily detected. For
2429        mutual recursions we keep a chain on the stack. */
2430    
2431        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2432        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2433        else
2434          {
2435          recurse_check *r = recurses;
2436          for (r = recurses; r != NULL; r = r->prev)
2437            if (r->group == scode) break;
2438          if (r != NULL) continue;   /* Mutual recursion */
2439          }
2440    
2441        /* Completed reference; scan the referenced group, remembering it on the
2442        stack chain to detect mutual recursions. */
2443    
2444      empty_branch = FALSE;      empty_branch = FALSE;
2445        this_recurse.prev = recurses;
2446        this_recurse.group = scode;
2447    
2448      do      do
2449        {        {
2450        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2451            {
2452          empty_branch = TRUE;          empty_branch = TRUE;
2453        code += GET(code, 1);          break;
2454            }
2455          scode += GET(scode, 1);
2456        }        }
2457      while (*code == OP_ALT);      while (*scode == OP_ALT);
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2458    
2459      /* Move past the KET and fudge things so that the increment in the "for"      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2460      above has no effect. */      continue;
2461        }
2462    
2463      /* Groups with zero repeats can of course be empty; skip them. */
2464    
2465      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2466          c == OP_BRAPOSZERO)
2467        {
2468        code += PRIV(OP_lengths)[c];
2469        do code += GET(code, 1); while (*code == OP_ALT);
2470        c = *code;
2471        continue;
2472        }
2473    
2474      /* A nested group that is already marked as "could be empty" can just be
2475      skipped. */
2476    
2477      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2478          c == OP_SCBRA || c == OP_SCBRAPOS)
2479        {
2480        do code += GET(code, 1); while (*code == OP_ALT);
2481        c = *code;
2482        continue;
2483        }
2484    
2485      /* For other groups, scan the branches. */
2486    
2487      if (c == OP_BRA  || c == OP_BRAPOS ||
2488          c == OP_CBRA || c == OP_CBRAPOS ||
2489          c == OP_ONCE || c == OP_ONCE_NC ||
2490          c == OP_COND)
2491        {
2492        BOOL empty_branch;
2493        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2494    
2495        /* If a conditional group has only one branch, there is a second, implied,
2496        empty branch, so just skip over the conditional, because it could be empty.
2497        Otherwise, scan the individual branches of the group. */
2498    
2499        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2500          code += GET(code, 1);
2501        else
2502          {
2503          empty_branch = FALSE;
2504          do
2505            {
2506            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2507              recurses)) empty_branch = TRUE;
2508            code += GET(code, 1);
2509            }
2510          while (*code == OP_ALT);
2511          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2512          }
2513    
2514      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2515      continue;      continue;
2516      }      }
2517    
# Line 1395  for (code = first_significant_code(code Line 2519  for (code = first_significant_code(code
2519    
2520    switch (c)    switch (c)
2521      {      {
2522      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2523        cannot be represented just by a bit map. This includes negated single
2524        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2525        actual length is stored in the compiled code, so we must update "code"
2526        here. */
2527    
2528  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2529      case OP_XCLASS:      case OP_XCLASS:
2530      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2531      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2532  #endif  #endif
2533    
2534      case OP_CLASS:      case OP_CLASS:
2535      case OP_NCLASS:      case OP_NCLASS:
2536      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2537    
2538  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2539      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2540  #endif  #endif
2541    
# Line 1417  for (code = first_significant_code(code Line 2545  for (code = first_significant_code(code
2545        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2546        case OP_CRQUERY:        case OP_CRQUERY:
2547        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2548          case OP_CRPOSSTAR:
2549          case OP_CRPOSQUERY:
2550        break;        break;
2551    
2552        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2553        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2554        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2555          case OP_CRPOSPLUS:
2556        return FALSE;        return FALSE;
2557    
2558        case OP_CRRANGE:        case OP_CRRANGE:
2559        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2560          case OP_CRPOSRANGE:
2561        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2562        break;        break;
2563        }        }
# Line 1433  for (code = first_significant_code(code Line 2565  for (code = first_significant_code(code
2565    
2566      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2567    
2568        case OP_ANY:
2569        case OP_ALLANY:
2570        case OP_ANYBYTE:
2571    
2572      case OP_PROP:      case OP_PROP:
2573      case OP_NOTPROP:      case OP_NOTPROP:
2574        case OP_ANYNL:
2575    
2576        case OP_NOT_HSPACE:
2577        case OP_HSPACE:
2578        case OP_NOT_VSPACE:
2579        case OP_VSPACE:
2580      case OP_EXTUNI:      case OP_EXTUNI:
2581    
2582      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2583      case OP_DIGIT:      case OP_DIGIT:
2584      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2585      case OP_WHITESPACE:      case OP_WHITESPACE:
2586      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2587      case OP_WORDCHAR:      case OP_WORDCHAR:
2588      case OP_ANY:  
     case OP_ANYBYTE:  
2589      case OP_CHAR:      case OP_CHAR:
2590      case OP_CHARNC:      case OP_CHARI:
2591      case OP_NOT:      case OP_NOT:
2592        case OP_NOTI:
2593    
2594      case OP_PLUS:      case OP_PLUS:
2595        case OP_PLUSI:
2596      case OP_MINPLUS:      case OP_MINPLUS:
2597      case OP_POSPLUS:      case OP_MINPLUSI:
2598      case OP_EXACT:  
2599      case OP_NOTPLUS:      case OP_NOTPLUS:
2600        case OP_NOTPLUSI:
2601      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2602        case OP_NOTMINPLUSI:
2603    
2604        case OP_POSPLUS:
2605        case OP_POSPLUSI:
2606      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2607        case OP_NOTPOSPLUSI:
2608    
2609        case OP_EXACT:
2610        case OP_EXACTI:
2611      case OP_NOTEXACT:      case OP_NOTEXACT:
2612        case OP_NOTEXACTI:
2613    
2614      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2615      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2616      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2617      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2618    
2619      return FALSE;      return FALSE;
2620    
2621        /* These are going to continue, as they may be empty, but we have to
2622        fudge the length for the \p and \P cases. */
2623    
2624        case OP_TYPESTAR:
2625        case OP_TYPEMINSTAR:
2626        case OP_TYPEPOSSTAR:
2627        case OP_TYPEQUERY:
2628        case OP_TYPEMINQUERY:
2629        case OP_TYPEPOSQUERY:
2630        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2631        break;
2632    
2633        /* Same for these */
2634    
2635        case OP_TYPEUPTO:
2636        case OP_TYPEMINUPTO:
2637        case OP_TYPEPOSUPTO:
2638        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2639          code += 2;
2640        break;
2641    
2642      /* End of branch */      /* End of branch */
2643    
2644      case OP_KET:      case OP_KET:
2645      case OP_KETRMAX:      case OP_KETRMAX:
2646      case OP_KETRMIN:      case OP_KETRMIN:
2647        case OP_KETRPOS:
2648      case OP_ALT:      case OP_ALT:
2649      return TRUE;      return TRUE;
2650    
2651      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2652      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2653        followed by a multibyte character. */
2654    
2655  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2656      case OP_STAR:      case OP_STAR:
2657        case OP_STARI:
2658        case OP_NOTSTAR:
2659        case OP_NOTSTARI:
2660    
2661      case OP_MINSTAR:      case OP_MINSTAR:
2662        case OP_MINSTARI:
2663        case OP_NOTMINSTAR:
2664        case OP_NOTMINSTARI:
2665    
2666      case OP_POSSTAR:      case OP_POSSTAR:
2667        case OP_POSSTARI:
2668        case OP_NOTPOSSTAR:
2669        case OP_NOTPOSSTARI:
2670    
2671      case OP_QUERY:      case OP_QUERY:
2672        case OP_QUERYI:
2673        case OP_NOTQUERY:
2674        case OP_NOTQUERYI:
2675    
2676      case OP_MINQUERY:      case OP_MINQUERY:
2677        case OP_MINQUERYI:
2678        case OP_NOTMINQUERY:
2679        case OP_NOTMINQUERYI:
2680    
2681      case OP_POSQUERY:      case OP_POSQUERY:
2682        case OP_POSQUERYI:
2683        case OP_NOTPOSQUERY:
2684        case OP_NOTPOSQUERYI:
2685    
2686        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2687        break;
2688    
2689      case OP_UPTO:      case OP_UPTO:
2690        case OP_UPTOI:
2691        case OP_NOTUPTO:
2692        case OP_NOTUPTOI:
2693    
2694      case OP_MINUPTO:      case OP_MINUPTO:
2695        case OP_MINUPTOI:
2696        case OP_NOTMINUPTO:
2697        case OP_NOTMINUPTOI:
2698    
2699      case OP_POSUPTO:      case OP_POSUPTO:
2700      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2701        case OP_NOTPOSUPTO:
2702        case OP_NOTPOSUPTOI:
2703    
2704        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2705      break;      break;
2706  #endif  #endif
2707      }  
2708    }      /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2709        string. */
2710    
2711        case OP_MARK:
2712        case OP_PRUNE_ARG:
2713        case OP_SKIP_ARG:
2714        case OP_THEN_ARG:
2715        code += code[1];
2716        break;
2717    
2718        /* None of the remaining opcodes are required to match a character. */
2719    
2720        default:
2721        break;
2722        }
2723      }
2724    
2725  return TRUE;  return TRUE;
2726  }  }
# Line 1501  return TRUE; Line 2735  return TRUE;
2735  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
2736  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,
2737  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.
2738    This function is called only during the real compile, not during the
2739    pre-compile.
2740    
2741  Arguments:  Arguments:
2742    code        points to start of the recursion    code        points to start of the recursion
2743    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2744    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2745    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2746      cd          pointers to tables etc
2747    
2748  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2749  */  */
2750    
2751  static BOOL  static BOOL
2752  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2753    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2754  {  {
2755  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2756    {    {
2757    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2758        return FALSE;
2759    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2760    }    }
2761  return TRUE;  return TRUE;
# Line 1526  return TRUE; Line 2764  return TRUE;
2764    
2765    
2766  /*************************************************  /*************************************************
2767    *        Base opcode of repeated opcodes         *
2768    *************************************************/
2769    
2770    /* Returns the base opcode for repeated single character type opcodes. If the
2771    opcode is not a repeated character type, it returns with the original value.
2772    
2773    Arguments:  c opcode
2774    Returns:    base opcode for the type
2775    */
2776    
2777    static pcre_uchar
2778    get_repeat_base(pcre_uchar c)
2779    {
2780    return (c > OP_TYPEPOSUPTO)? c :
2781           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2782           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2783           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2784           (c >= OP_STARI)?      OP_STARI :
2785                                 OP_STAR;
2786    }
2787    
2788    
2789    
2790    #ifdef SUPPORT_UCP
2791    /*************************************************
2792    *        Check a character and a property        *
2793    *************************************************/
2794    
2795    /* This function is called by check_auto_possessive() when a property item
2796    is adjacent to a fixed character.
2797    
2798    Arguments:
2799      c            the character
2800      ptype        the property type
2801      pdata        the data for the type
2802      negated      TRUE if it's a negated property (\P or \p{^)
2803    
2804    Returns:       TRUE if auto-possessifying is OK
2805    */
2806    
2807    static BOOL
2808    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2809      BOOL negated)
2810    {
2811    const pcre_uint32 *p;
2812    const ucd_record *prop = GET_UCD(c);
2813    
2814    switch(ptype)
2815      {
2816      case PT_LAMP:
2817      return (prop->chartype == ucp_Lu ||
2818              prop->chartype == ucp_Ll ||
2819              prop->chartype == ucp_Lt) == negated;
2820    
2821      case PT_GC:
2822      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2823    
2824      case PT_PC:
2825      return (pdata == prop->chartype) == negated;
2826    
2827      case PT_SC:
2828      return (pdata == prop->script) == negated;
2829    
2830      /* These are specials */
2831    
2832      case PT_ALNUM:
2833      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2834              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2835    
2836      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2837      means that Perl space and POSIX space are now identical. PCRE was changed
2838      at release 8.34. */
2839    
2840      case PT_SPACE:    /* Perl space */
2841      case PT_PXSPACE:  /* POSIX space */
2842      switch(c)
2843        {
2844        HSPACE_CASES:
2845        VSPACE_CASES:
2846        return negated;
2847    
2848        default:
2849        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2850        }
2851      break;  /* Control never reaches here */
2852    
2853      case PT_WORD:
2854      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2855              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2856              c == CHAR_UNDERSCORE) == negated;
2857    
2858      case PT_CLIST:
2859      p = PRIV(ucd_caseless_sets) + prop->caseset;
2860      for (;;)
2861        {
2862        if (c < *p) return !negated;
2863        if (c == *p++) return negated;
2864        }
2865      break;  /* Control never reaches here */
2866      }
2867    
2868    return FALSE;
2869    }
2870    #endif  /* SUPPORT_UCP */
2871    
2872    
2873    
2874    /*************************************************
2875    *        Fill the character property list        *
2876    *************************************************/
2877    
2878    /* Checks whether the code points to an opcode that can take part in auto-
2879    possessification, and if so, fills a list with its properties.
2880    
2881    Arguments:
2882      code        points to start of expression
2883      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2884      fcc         points to case-flipping table
2885      list        points to output list
2886                  list[0] will be filled with the opcode
2887                  list[1] will be non-zero if this opcode
2888                    can match an empty character string
2889                  list[2..7] depends on the opcode
2890    
2891    Returns:      points to the start of the next opcode if *code is accepted
2892                  NULL if *code is not accepted
2893    */
2894    
2895    static const pcre_uchar *
2896    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2897      const pcre_uint8 *fcc, pcre_uint32 *list)
2898    {
2899    pcre_uchar c = *code;
2900    pcre_uchar base;
2901    const pcre_uchar *end;
2902    pcre_uint32 chr;
2903    
2904    #ifdef SUPPORT_UCP
2905    pcre_uint32 *clist_dest;
2906    const pcre_uint32 *clist_src;
2907    #else
2908    utf = utf;  /* Suppress "unused parameter" compiler warning */
2909    #endif
2910    
2911    list[0] = c;
2912    list[1] = FALSE;
2913    code++;
2914    
2915    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2916      {
2917      base = get_repeat_base(c);
2918      c -= (base - OP_STAR);
2919    
2920      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2921        code += IMM2_SIZE;
2922    
2923      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2924    
2925      switch(base)
2926        {
2927        case OP_STAR:
2928        list[0] = OP_CHAR;
2929        break;
2930    
2931        case OP_STARI:
2932        list[0] = OP_CHARI;
2933        break;
2934    
2935        case OP_NOTSTAR:
2936        list[0] = OP_NOT;
2937        break;
2938    
2939        case OP_NOTSTARI:
2940        list[0] = OP_NOTI;
2941        break;
2942    
2943        case OP_TYPESTAR:
2944        list[0] = *code;
2945        code++;
2946        break;
2947        }
2948      c = list[0];
2949      }
2950    
2951    switch(c)
2952      {
2953      case OP_NOT_DIGIT:
2954      case OP_DIGIT:
2955      case OP_NOT_WHITESPACE:
2956      case OP_WHITESPACE:
2957      case OP_NOT_WORDCHAR:
2958      case OP_WORDCHAR:
2959      case OP_ANY:
2960      case OP_ALLANY:
2961      case OP_ANYNL:
2962      case OP_NOT_HSPACE:
2963      case OP_HSPACE:
2964      case OP_NOT_VSPACE:
2965      case OP_VSPACE:
2966      case OP_EXTUNI:
2967      case OP_EODN:
2968      case OP_EOD:
2969      case OP_DOLL:
2970      case OP_DOLLM:
2971      return code;
2972    
2973      case OP_CHAR:
2974      case OP_NOT:
2975      GETCHARINCTEST(chr, code);
2976      list[2] = chr;
2977      list[3] = NOTACHAR;
2978      return code;
2979    
2980      case OP_CHARI:
2981      case OP_NOTI:
2982      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2983      GETCHARINCTEST(chr, code);
2984      list[2] = chr;
2985    
2986    #ifdef SUPPORT_UCP
2987      if (chr < 128 || (chr < 256 && !utf))
2988        list[3] = fcc[chr];
2989      else
2990        list[3] = UCD_OTHERCASE(chr);
2991    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2992      list[3] = (chr < 256) ? fcc[chr] : chr;
2993    #else
2994      list[3] = fcc[chr];
2995    #endif
2996    
2997      /* The othercase might be the same value. */
2998    
2999      if (chr == list[3])
3000        list[3] = NOTACHAR;
3001      else
3002        list[4] = NOTACHAR;
3003      return code;
3004    
3005    #ifdef SUPPORT_UCP
3006      case OP_PROP:
3007      case OP_NOTPROP:
3008      if (code[0] != PT_CLIST)
3009        {
3010        list[2] = code[0];
3011        list[3] = code[1];
3012        return code + 2;
3013        }
3014    
3015      /* Convert only if we have enough space. */
3016    
3017      clist_src = PRIV(ucd_caseless_sets) + code[1];
3018      clist_dest = list + 2;
3019      code += 2;
3020    
3021      do {
3022         if (clist_dest >= list + 8)
3023           {
3024           /* Early return if there is not enough space. This should never
3025           happen, since all clists are shorter than 5 character now. */
3026           list[2] = code[0];
3027           list[3] = code[1];
3028           return code;
3029           }
3030         *clist_dest++ = *clist_src;
3031         }
3032      while(*clist_src++ != NOTACHAR);
3033    
3034      /* All characters are stored. The terminating NOTACHAR
3035      is copied form the clist itself. */
3036    
3037      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3038      return code;
3039    #endif
3040    
3041      case OP_NCLASS:
3042      case OP_CLASS:
3043    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3044      case OP_XCLASS:
3045      if (c == OP_XCLASS)
3046        end = code + GET(code, 0) - 1;
3047      else
3048    #endif
3049        end = code + 32 / sizeof(pcre_uchar);
3050    
3051      switch(*end)
3052        {
3053        case OP_CRSTAR:
3054        case OP_CRMINSTAR:
3055        case OP_CRQUERY:
3056        case OP_CRMINQUERY:
3057        case OP_CRPOSSTAR:
3058        case OP_CRPOSQUERY:
3059        list[1] = TRUE;
3060        end++;
3061        break;
3062    
3063        case OP_CRPLUS:
3064        case OP_CRMINPLUS:
3065        case OP_CRPOSPLUS:
3066        end++;
3067        break;
3068    
3069        case OP_CRRANGE:
3070        case OP_CRMINRANGE:
3071        case OP_CRPOSRANGE:
3072        list[1] = (GET2(end, 1) == 0);
3073        end += 1 + 2 * IMM2_SIZE;
3074        break;
3075        }
3076      list[2] = (pcre_uint32)(end - code);
3077      return end;
3078      }
3079    return NULL;    /* Opcode not accepted */
3080    }
3081    
3082    
3083    
3084    /*************************************************
3085    *    Scan further character sets for match       *
3086    *************************************************/
3087    
3088    /* Checks whether the base and the current opcode have a common character, in
3089    which case the base cannot be possessified.
3090    
3091    Arguments:
3092      code        points to the byte code
3093      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3094      cd          static compile data
3095      base_list   the data list of the base opcode
3096    
3097    Returns:      TRUE if the auto-possessification is possible
3098    */
3099    
3100    static BOOL
3101    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3102      const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3103    {
3104    pcre_uchar c;
3105    pcre_uint32 list[8];
3106    const pcre_uint32 *chr_ptr;
3107    const pcre_uint32 *ochr_ptr;
3108    const pcre_uint32 *list_ptr;
3109    const pcre_uchar *next_code;
3110    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3111    const pcre_uchar *xclass_flags;
3112    #endif
3113    const pcre_uint8 *class_bitset;
3114    const pcre_uint8 *set1, *set2, *set_end;
3115    pcre_uint32 chr;
3116    BOOL accepted, invert_bits;
3117    BOOL entered_a_group = FALSE;
3118    
3119    if (*rec_limit == 0) return FALSE;
3120    --(*rec_limit);
3121    
3122    /* Note: the base_list[1] contains whether the current opcode has greedy
3123    (represented by a non-zero value) quantifier. This is a different from
3124    other character type lists, which stores here that the character iterator
3125    matches to an empty string (also represented by a non-zero value). */
3126    
3127    for(;;)
3128      {
3129      /* All operations move the code pointer forward.
3130      Therefore infinite recursions are not possible. */
3131    
3132      c = *code;
3133    
3134      /* Skip over callouts */
3135    
3136      if (c == OP_CALLOUT)
3137        {
3138        code += PRIV(OP_lengths)[c];
3139        continue;
3140        }
3141    
3142      if (c == OP_ALT)
3143        {
3144        do code += GET(code, 1); while (*code == OP_ALT);
3145        c = *code;
3146        }
3147    
3148      switch(c)
3149        {
3150        case OP_END:
3151        case OP_KETRPOS:
3152        /* TRUE only in greedy case. The non-greedy case could be replaced by
3153        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3154        uses more memory, which we cannot get at this stage.) */
3155    
3156        return base_list[1] != 0;
3157    
3158        case OP_KET:
3159        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3160        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3161        cannot be converted to a possessive form. */
3162    
3163        if (base_list[1] == 0) return FALSE;
3164    
3165        switch(*(code - GET(code, 1)))
3166          {
3167          case OP_ASSERT:
3168          case OP_ASSERT_NOT:
3169          case OP_ASSERTBACK:
3170          case OP_ASSERTBACK_NOT:
3171          case OP_ONCE:
3172          case OP_ONCE_NC:
3173          /* Atomic sub-patterns and assertions can always auto-possessify their
3174          last iterator. However, if the group was entered as a result of checking
3175          a previous iterator, this is not possible. */
3176    
3177          return !entered_a_group;
3178          }
3179    
3180        code += PRIV(OP_lengths)[c];
3181        continue;
3182    
3183        case OP_ONCE:
3184        case OP_ONCE_NC:
3185        case OP_BRA:
3186        case OP_CBRA:
3187        next_code = code + GET(code, 1);
3188        code += PRIV(OP_lengths)[c];
3189    
3190        while (*next_code == OP_ALT)
3191          {
3192          if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3193            return FALSE;
3194          code = next_code + 1 + LINK_SIZE;
3195          next_code += GET(next_code, 1);
3196          }
3197    
3198        entered_a_group = TRUE;
3199        continue;
3200    
3201        case OP_BRAZERO:
3202        case OP_BRAMINZERO:
3203    
3204        next_code = code + 1;
3205        if (*next_code != OP_BRA && *next_code != OP_CBRA
3206            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3207    
3208        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3209    
3210        /* The bracket content will be checked by the
3211        OP_BRA/OP_CBRA case above. */
3212        next_code += 1 + LINK_SIZE;
3213        if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3214          return FALSE;
3215    
3216        code += PRIV(OP_lengths)[c];
3217        continue;
3218    
3219        default:
3220        break;
3221        }
3222    
3223      /* Check for a supported opcode, and load its properties. */
3224    
3225      code = get_chr_property_list(code, utf, cd->fcc, list);
3226      if (code == NULL) return FALSE;    /* Unsupported */
3227    
3228      /* If either opcode is a small character list, set pointers for comparing
3229      characters from that list with another list, or with a property. */
3230    
3231      if (base_list[0] == OP_CHAR)
3232        {
3233        chr_ptr = base_list + 2;
3234        list_ptr = list;
3235        }
3236      else if (list[0] == OP_CHAR)
3237        {
3238        chr_ptr = list + 2;
3239        list_ptr = base_list;
3240        }
3241    
3242      /* Character bitsets can also be compared to certain opcodes. */
3243    
3244      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3245    #ifdef COMPILE_PCRE8
3246          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3247          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3248    #endif
3249          )
3250        {
3251    #ifdef COMPILE_PCRE8
3252        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3253    #else
3254        if (base_list[0] == OP_CLASS)
3255    #endif
3256          {
3257          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3258          list_ptr = list;
3259          }
3260        else
3261          {
3262          set1 = (pcre_uint8 *)(code - list[2]);
3263          list_ptr = base_list;
3264          }
3265    
3266        invert_bits = FALSE;
3267        switch(list_ptr[0])
3268          {
3269          case OP_CLASS:
3270          case OP_NCLASS:
3271          set2 = (pcre_uint8 *)
3272            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3273          break;
3274    
3275    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3276          case OP_XCLASS:
3277          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3278          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3279          if ((*xclass_flags & XCL_MAP) == 0)
3280            {
3281            /* No bits are set for characters < 256. */
3282            if (list[1] == 0) return TRUE;
3283            /* Might be an empty repeat. */
3284            continue;
3285            }
3286          set2 = (pcre_uint8 *)(xclass_flags + 1);
3287          break;
3288    #endif
3289    
3290          case OP_NOT_DIGIT:
3291          invert_bits = TRUE;
3292          /* Fall through */
3293          case OP_DIGIT:
3294          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3295          break;
3296    
3297          case OP_NOT_WHITESPACE:
3298          invert_bits = TRUE;
3299          /* Fall through */
3300          case OP_WHITESPACE:
3301          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3302          break;
3303    
3304          case OP_NOT_WORDCHAR:
3305          invert_bits = TRUE;
3306          /* Fall through */
3307          case OP_WORDCHAR:
3308          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3309          break;
3310    
3311          default:
3312          return FALSE;
3313          }
3314    
3315        /* Because the sets are unaligned, we need
3316        to perform byte comparison here. */
3317        set_end = set1 + 32;
3318        if (invert_bits)
3319          {
3320          do
3321            {
3322            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3323            }
3324          while (set1 < set_end);
3325          }
3326        else
3327 &nbs