/[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 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 1454 by ph10, Sun Feb 9 18:55:03 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 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    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69    /* Macro for setting individual bits in class bitmaps. */
70    
71    #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73    /* Maximum length value to check against when making sure that the integer that
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
95  *************************************************/  *************************************************/
# Line 72  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123  #define COMPILE_WORK_SIZE (4096)  #define NAMED_GROUP_LIST_SIZE  20
124    
125    /* The overrun tests check for a slightly smaller size so that they detect the
126    overrun before it actually does run off the end of the data block. */
127    
128    #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138    /* Repeated character flags. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
144  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
145  is invalid. */  is invalid. */
146    
147  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
148    
149    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150    in UTF-8 mode. */
151    
152  static const short int escapes[] = {  static const short int escapes[] = {
153       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
154       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
155     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
156       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      ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
555  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 567  For convenience, we use the same bit def
567    
568  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
569    
570  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
571  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
572    into a subtraction and unsigned comparison). */
573    
574    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
575    
576    #ifndef EBCDIC
577    
578    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
579    UTF-8 mode. */
580    
581    static const pcre_uint8 digitab[] =
582    {    {
583    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
584    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 613  static const unsigned char digitab[] =
613    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
614    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
615    
616  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
617  static const unsigned char digitab[] =  
618    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
619    
620    static const pcre_uint8 digitab[] =
621    {    {
622    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
623    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 334  static const unsigned char digitab[] = Line 652  static const unsigned char digitab[] =
652    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
653    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
654    
655  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
656    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
657    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
658    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 370  static const unsigned char ebcdic_charta Line 688  static const unsigned char ebcdic_charta
688  #endif  #endif
689    
690    
691  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
692    between adjacent character-type opcodes. The left-hand (repeated) opcode is
693    used to select the row, and the right-hand opcode is use to select the column.
694    A value of 1 means that auto-possessification is OK. For example, the second
695    value in the first row means that \D+\d can be turned into \D++\d.
696    
697    The Unicode property types (\P and \p) have to be present to fill out the table
698    because of what their opcode values are, but the table values should always be
699    zero because property types are handled separately in the code. The last four
700    columns apply to items that cannot be repeated, so there is no need to have
701    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
702    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
703    
704    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
705    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
706    
707    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
708    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
709      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
710      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
711      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
712      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
713      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
714      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
715      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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 },  /* \C */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
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, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
721      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
722      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
723      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
724      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
725      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
726    };
727    
728    
729    /* This table is used to check whether auto-possessification is possible
730    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
731    left-hand (repeated) opcode is used to select the row, and the right-hand
732    opcode is used to select the column. The values are as follows:
733    
734      0   Always return FALSE (never auto-possessify)
735      1   Character groups are distinct (possessify if both are OP_PROP)
736      2   Check character categories in the same group (general or particular)
737      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
738    
739      4   Check left general category vs right particular category
740      5   Check right general category vs left particular category
741    
742      6   Left alphanum vs right general category
743      7   Left space vs right general category
744      8   Left word vs right general category
745    
746      9   Right alphanum vs left general category
747     10   Right space vs left general category
748     11   Right word vs left general category
749    
750     12   Left alphanum vs right particular category
751     13   Left space vs right particular category
752     14   Left word vs right particular category
753    
754     15   Right alphanum vs left particular category
755     16   Right space vs left particular category
756     17   Right word vs left particular category
757    */
758    
759    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
760    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
761      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
762      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
763      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
764      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
765      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
766      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
767      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
769      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
770      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
772    };
773    
774    /* This table is used to check whether auto-possessification is possible
775    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
776    specifies a general category and the other specifies a particular category. The
777    row is selected by the general category and the column by the particular
778    category. The value is 1 if the particular category is not part of the general
779    category. */
780    
781    static const pcre_uint8 catposstab[7][30] = {
782    /* 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 */
783      { 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 */
784      { 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 */
785      { 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 */
786      { 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 */
787      { 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 */
788      { 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 */
789      { 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 */
790    };
791    
792    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
793    a general or particular category. The properties in each row are those
794    that apply to the character set in question. Duplication means that a little
795    unnecessary work is done when checking, but this keeps things much simpler
796    because they can all use the same code. For more details see the comment where
797    this table is used.
798    
799    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
800    "space", but from Perl 5.18 it's included, so both categories are treated the
801    same here. */
802    
803    static const pcre_uint8 posspropstab[3][4] = {
804      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
805      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
806      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807    };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
869    /*************************************************
870    *            Find an error text                  *
871    *************************************************/
872    
873    /* The error texts are now all in one long string, to save on relocations. As
874    some of the text is of unknown length, we can't use a table of offsets.
875    Instead, just count through the strings. This is not a performance issue
876    because it happens only when there has been a compilation error.
877    
878    Argument:   the error number
879    Returns:    pointer to the error string
880    */
881    
882    static const char *
883    find_error_text(int n)
884    {
885    const char *s = error_texts;
886    for (; n > 0; n--)
887      {
888      while (*s++ != CHAR_NULL) {};
889      if (*s == CHAR_NULL) return "Error text not found (please report)";
890      }
891    return s;
892    }
893    
894    
895    
896    /*************************************************
897    *           Expand the workspace                 *
898    *************************************************/
899    
900    /* This function is called during the second compiling phase, if the number of
901    forward references fills the existing workspace, which is originally a block on
902    the stack. A larger block is obtained from malloc() unless the ultimate limit
903    has been reached or the increase will be rather small.
904    
905    Argument: pointer to the compile data block
906    Returns:  0 if all went well, else an error number
907    */
908    
909    static int
910    expand_workspace(compile_data *cd)
911    {
912    pcre_uchar *newspace;
913    int newsize = cd->workspace_size * 2;
914    
915    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
916    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
917        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
918     return ERR72;
919    
920    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
921    if (newspace == NULL) return ERR21;
922    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
923    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
924    if (cd->workspace_size > COMPILE_WORK_SIZE)
925      (PUBL(free))((void *)cd->start_workspace);
926    cd->start_workspace = newspace;
927    cd->workspace_size = newsize;
928    return 0;
929    }
930    
931    
932    
933    /*************************************************
934    *            Check for counted repeat            *
935    *************************************************/
936    
937    /* This function is called when a '{' is encountered in a place where it might
938    start a quantifier. It looks ahead to see if it really is a quantifier or not.
939    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
940    where the ddds are digits.
941    
942    Arguments:
943      p         pointer to the first char after '{'
944    
945    Returns:    TRUE or FALSE
946    */
947    
948  static BOOL  static BOOL
949    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
950      int *, branch_chain *, compile_data *, int *);  {
951    if (!IS_DIGIT(*p)) return FALSE;
952    p++;
953    while (IS_DIGIT(*p)) p++;
954    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
955    
956    if (*p++ != CHAR_COMMA) return FALSE;
957    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
958    
959    if (!IS_DIGIT(*p)) return FALSE;
960    p++;
961    while (IS_DIGIT(*p)) p++;
962    
963    return (*p == CHAR_RIGHT_CURLY_BRACKET);
964    }
965    
966    
967    
# Line 383  static BOOL Line 970  static BOOL
970  *************************************************/  *************************************************/
971    
972  /* 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
973  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
974  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.
975  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
976  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
977  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981      chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
985    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
986    
987  Returns:         zero or positive => a data character  Returns:         zero => a data character
988                   negative => a special escape sequence                   positive => a special escape sequence
989                   on error, errorptr is set                   negative => a back reference
990                     on error, errorcodeptr is set
991  */  */
992    
993  static int  static int
994  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
995    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
996  {  {
997  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
998  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
999  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
1000    pcre_uint32 c;
1001    int escape = 0;
1002    int i;
1003    
1004  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1005  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1006    
1007  /* 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. */
1008    
1009  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1010    
1011  /* 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
1012  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.
1013  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1014    
1015  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1016  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1017  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1018    else if ((i = escapes[c - CHAR_0]) != 0)
1019      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1020    
1021  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1022  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
1023  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1024    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1025  #endif  #endif
1026    
1027  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1028    
1029  else  else
1030    {    {
1031    const uschar *oldptr;    const pcre_uchar *oldptr;
1032    BOOL braced, negated;    BOOL braced, negated, overflow;
1033      int s;
1034    
1035    switch (c)    switch (c)
1036      {      {
1037      /* 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
1038      error. */      error. */
1039    
1040      case 'l':      case CHAR_l:
1041      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1042      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1043      break;      break;
1044    
1045      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
1046      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1047      This is a Perl 5.10 feature. */        {
1048          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1049          Otherwise it is a lowercase u letter. */
1050          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1051            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1052            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1053            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1054            {
1055            c = 0;
1056            for (i = 0; i < 4; ++i)
1057              {
1058              register pcre_uint32 cc = *(++ptr);
1059    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1060              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1061              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1062    #else           /* EBCDIC coding */
1063              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1064              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1065    #endif
1066              }
1067    
1068    #if defined COMPILE_PCRE8
1069            if (c > (utf ? 0x10ffffU : 0xffU))
1070    #elif defined COMPILE_PCRE16
1071            if (c > (utf ? 0x10ffffU : 0xffffU))
1072    #elif defined COMPILE_PCRE32
1073            if (utf && c > 0x10ffffU)
1074    #endif
1075              {
1076              *errorcodeptr = ERR76;
1077              }
1078            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1079            }
1080          }
1081        else
1082          *errorcodeptr = ERR37;
1083        break;
1084    
1085        case CHAR_U:
1086        /* In JavaScript, \U is an uppercase U letter. */
1087        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1088        break;
1089    
1090        /* In a character class, \g is just a literal "g". Outside a character
1091        class, \g must be followed by one of a number of specific things:
1092    
1093        (1) A number, either plain or braced. If positive, it is an absolute
1094        backreference. If negative, it is a relative backreference. This is a Perl
1095        5.10 feature.
1096    
1097        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1098        is part of Perl's movement towards a unified syntax for back references. As
1099        this is synonymous with \k{name}, we fudge it up by pretending it really
1100        was \k.
1101    
1102        (3) For Oniguruma compatibility we also support \g followed by a name or a
1103        number either in angle brackets or in single quotes. However, these are
1104        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1105        the ESC_g code (cf \k). */
1106    
1107        case CHAR_g:
1108        if (isclass) break;
1109        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1110          {
1111          escape = ESC_g;
1112          break;
1113          }
1114    
1115        /* Handle the Perl-compatible cases */
1116    
1117      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1118        {        {
1119          const pcre_uchar *p;
1120          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1121            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1122          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1123            {
1124            escape = ESC_k;
1125            break;
1126            }
1127        braced = TRUE;        braced = TRUE;
1128        ptr++;        ptr++;
1129        }        }
1130      else braced = FALSE;      else braced = FALSE;
1131    
1132      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1133        {        {
1134        negated = TRUE;        negated = TRUE;
1135        ptr++;        ptr++;
1136        }        }
1137      else negated = FALSE;      else negated = FALSE;
1138    
1139      c = 0;      /* The integer range is limited by the machine's int representation. */
1140      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1141        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1142        while (IS_DIGIT(ptr[1]))
1143          {
1144          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1145            {
1146            overflow = TRUE;
1147            break;
1148            }
1149          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1150          }
1151        if (overflow) /* Integer overflow */
1152          {
1153          while (IS_DIGIT(ptr[1]))
1154            ptr++;
1155          *errorcodeptr = ERR61;
1156          break;
1157          }
1158    
1159      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1160        {        {
1161        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1162        return 0;        break;
1163          }
1164    
1165        if (s == 0)
1166          {
1167          *errorcodeptr = ERR58;
1168          break;
1169        }        }
1170    
1171      if (negated)      if (negated)
1172        {        {
1173        if (c > bracount)        if (s > bracount)
1174          {          {
1175          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1176          return 0;          break;
1177          }          }
1178        c = bracount - (c - 1);        s = bracount - (s - 1);
1179        }        }
1180    
1181      c = -(ESC_REF + c);      escape = -s;
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      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
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      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
1190      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
1191      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
1192      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
1193      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
1194      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
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1202    
1203      if (!isclass)      if (!isclass)
1204        {        {
1205        oldptr = ptr;        oldptr = ptr;
1206        c -= '0';        /* The integer range is limited by the machine's int representation. */
1207        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1208          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1209        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1210            {
1211            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1212              {
1213              overflow = TRUE;
1214              break;
1215              }
1216            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1217            }
1218          if (overflow) /* Integer overflow */
1219            {
1220            while (IS_DIGIT(ptr[1]))
1221              ptr++;
1222            *errorcodeptr = ERR61;
1223            break;
1224            }
1225          if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          c = -(ESC_REF + c);          escape = -s;
1228          break;          break;
1229          }          }
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* 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
1234      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
1235      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
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \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
1243      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
1244      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
1245      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,
1246      than 3 octal digits. */      but no more than 3 octal digits. */
1247    
1248      case '0':      case CHAR_0:
1249      c -= '0';      c -= CHAR_0;
1250      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1251          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1252      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1253        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1254    #endif
1255      break;      break;
1256    
1257      /* \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
1258      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. */  
1259    
1260      case 'x':      case CHAR_o:
1261      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262        {        {
1263        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1264        c = 0;        c = 0;
1265        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267          {          {
1268          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1269          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270          count++;  #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272  #ifndef EBCDIC  /* ASCII coding */  #endif
1273          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1274          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1275  #else           /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1277          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280  #endif  #endif
1281          }          }
1282          if (overflow)
       if (*pt == '}')  
1283          {          {
1284          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1286          }          }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288        /* If the sequence of hex digits does not end with '}', then we don't          {
1289        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292        }        }
1293        break;
1294    
1295      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298      c = 0;      case CHAR_x:
1299      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300        {        {
1301        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303  #ifndef EBCDIC  /* ASCII coding */          {
1304        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1305        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1306              {
1307              register pcre_uint32 cc = *(++ptr);
1308    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1309              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1310              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1311  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1312        if (cc <= 'z') cc += 64;              /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1313        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1314  #endif  #endif
1315        }            }
1316            }
1317          }    /* End JavaScript handling */
1318    
1319        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        else
1327          {
1328          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329            {
1330            ptr += 2;
1331            c = 0;
1332            overflow = FALSE;
1333            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338    #ifdef COMPILE_PCRE32
1339              if (c >= 0x10000000l) { overflow = TRUE; break; }
1340    #endif
1341    
1342    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345    #else           /* EBCDIC coding */
1346              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1347              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348    #endif
1349    
1350    #if defined COMPILE_PCRE8
1351              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352    #elif defined COMPILE_PCRE16
1353              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354    #elif defined COMPILE_PCRE32
1355              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356    #endif
1357              }
1358    
1359            if (overflow)
1360              {
1361              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362              *errorcodeptr = ERR34;
1363              }
1364    
1365            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366              {
1367              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368              }
1369    
1370            /* If the sequence of hex digits does not end with '}', give an error.
1371            We used just to recognize this construct and fall through to the normal
1372            \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375            else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379    
1380          else
1381            {
1382            c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390    #else           /* EBCDIC coding */
1391              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393    #endif
1394              }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* 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.
1400      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
1401        coding is ASCII-specific, but then the whole concept of \cx is
1402      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1403    
1404      case 'c':      case CHAR_c:
1405      c = *(++ptr);      c = *(++ptr);
1406      if (c == 0)      if (c == CHAR_NULL)
1407        {        {
1408        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1409        return 0;        break;
1410        }        }
1411    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1412  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1413      if (c >= 'a' && c <= 'z') c -= 32;        {
1414          *errorcodeptr = ERR68;
1415          break;
1416          }
1417        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1418      c ^= 0x40;      c ^= 0x40;
1419  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1420      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1421      c ^= 0xC0;      c ^= 0xC0;
1422  #endif  #endif
1423      break;      break;
1424    
1425      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1426      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1427      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
1428      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
1429      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1430    
1431      default:      default:
1432      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1439  else
1439      }      }
1440    }    }
1441    
1442    /* Perl supports \N{name} for character names, as well as plain \N for "not
1443    newline". PCRE does not support \N{name}. However, it does support
1444    quantification such as \N{2,3}. */
1445    
1446    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1447         !is_counted_repeat(ptr+2))
1448      *errorcodeptr = ERR37;
1449    
1450    /* If PCRE_UCP is set, we change the values for \d etc. */
1451    
1452    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1453      escape += (ESC_DU - ESC_D);
1454    
1455    /* Set the pointer to the final character before returning. */
1456    
1457  *ptrptr = ptr;  *ptrptr = ptr;
1458  return c;  *chptr = c;
1459    return escape;
1460  }  }
1461    
1462    
# Line 656  escape sequence. Line 1474  escape sequence.
1474  Argument:  Argument:
1475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1476    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
1477    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
1478      pdataptr       points to an unsigned int that is set to the detailed property value
1479    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1480    
1481  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
1482  */  */
1483    
1484  static int  static BOOL
1485  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1486      unsigned int *pdataptr, int *errorcodeptr)
1487  {  {
1488  int c, i, bot, top;  pcre_uchar c;
1489  const uschar *ptr = *ptrptr;  int i, bot, top;
1490  char name[32];  const pcre_uchar *ptr = *ptrptr;
1491    pcre_uchar name[32];
1492    
1493  c = *(++ptr);  c = *(++ptr);
1494  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1495    
1496  *negptr = FALSE;  *negptr = FALSE;
1497    
1498  /* \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
1499  negation. */  negation. */
1500    
1501  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1502    {    {
1503    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1504      {      {
1505      *negptr = TRUE;      *negptr = TRUE;
1506      ptr++;      ptr++;
1507      }      }
1508    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1509      {      {
1510      c = *(++ptr);      c = *(++ptr);
1511      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1512      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1513      name[i] = c;      name[i] = c;
1514      }      }
1515    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1516    name[i] = 0;    name[i] = 0;
1517    }    }
1518    
# Line 708  else Line 1529  else
1529  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1530    
1531  bot = 0;  bot = 0;
1532  top = _pcre_utt_size;  top = PRIV(utt_size);
1533    
1534  while (bot < top)  while (bot < top)
1535    {    {
1536      int r;
1537    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1538    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1539    if (c == 0)    if (r == 0)
1540      {      {
1541      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1542      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1543        return TRUE;
1544      }      }
1545    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1546    }    }
1547    
1548  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1549  *ptrptr = ptr;  *ptrptr = ptr;
1550  return -1;  return FALSE;
1551    
1552  ERROR_RETURN:  ERROR_RETURN:
1553  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1554  *ptrptr = ptr;  *ptrptr = ptr;
1555  return -1;  return FALSE;
1556  }  }
1557  #endif  #endif
1558    
1559    
1560    
   
 /*************************************************  
 *            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 == '}');  
 }  
   
   
   
1561  /*************************************************  /*************************************************
1562  *         Read repeat counts                     *  *         Read repeat counts                     *
1563  *************************************************/  *************************************************/
# Line 788  Returns:         pointer to '}' on succe Line 1577  Returns:         pointer to '}' on succe
1577                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1578  */  */
1579    
1580  static const uschar *  static const pcre_uchar *
1581  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)
1582  {  {
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
# Line 797  int max = -1; Line 1586  int max = -1;
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1587  an integer overflow. */  an integer overflow. */
1588    
1589  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1590  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1591    {    {
1592    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1596  if (min < 0 || min > 65535)
1596  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1597  Also, max must not be less than min. */  Also, max must not be less than min. */
1598    
1599  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1600    {    {
1601    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1602      {      {
1603      max = 0;      max = 0;
1604      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1605      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1606        {        {
1607        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1626  return p;
1626    
1627    
1628  /*************************************************  /*************************************************
1629  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1630  *************************************************/  *************************************************/
1631    
1632  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1633  subpatterns, and counting them. If it finds a named pattern that matches the  for a fixed first character, or an anchoring op code etc. It skips over things
1634  name it is given, it returns its number. Alternatively, if the name is NULL, it  that do not influence this. For some calls, it makes sense to skip negative
1635  returns when it reaches a given numbered subpattern. This is used for forward  forward and all backward assertions, and also the \b assertion; for others it
1636  references to subpatterns. We know that if (?P< is encountered, the name will  does not.
 be terminated by '>' because that is checked in the first pass.  
   
 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;  
 }  
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, a change of option is important.  
 For some calls, it makes sense to skip negative forward and all backward  
 assertions, and also the \b assertion; for others it does not.  
1637    
1638  Arguments:  Arguments:
1639    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  
1640    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1641    
1642  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1643  */  */
1644    
1645  static const uschar*  static const pcre_uchar*
1646  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1647  {  {
1648  for (;;)  for (;;)
1649    {    {
1650    switch ((int)*code)    switch ((int)*code)
1651      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1652      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1653      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1654      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1655      if (!skipassert) return code;      if (!skipassert) return code;
1656      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1657      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1658      break;      break;
1659    
1660      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1664  for (;;)
1664    
1665      case OP_CALLOUT:      case OP_CALLOUT:
1666      case OP_CREF:      case OP_CREF:
1667        case OP_DNCREF:
1668      case OP_RREF:      case OP_RREF:
1669        case OP_DNRREF:
1670      case OP_DEF:      case OP_DEF:
1671      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1672      break;      break;
1673    
1674      default:      default:
# Line 1010  for (;;) Line 1680  for (;;)
1680    
1681    
1682    
   
1683  /*************************************************  /*************************************************
1684  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1685  *************************************************/  *************************************************/
1686    
1687  /* 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,
1688  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.
1689  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
1690    temporarily terminated with OP_END when this function is called.
1691    
1692    This function is called when a backward assertion is encountered, so that if it
1693    fails, the error message can point to the correct place in the pattern.
1694    However, we cannot do this when the assertion contains subroutine calls,
1695    because they can be forward references. We solve this by remembering this case
1696    and doing the check at the end; a flag specifies which mode we are running in.
1697    
1698  Arguments:  Arguments:
1699    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1700    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1701      atend    TRUE if called when the pattern is complete
1702  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1703               or -2 if \C was encountered  
1704    Returns:   the fixed length,
1705                 or -1 if there is no fixed length,
1706                 or -2 if \C was encountered (in UTF-8 mode only)
1707                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1708                 or -4 if an unknown opcode was encountered (internal error)
1709  */  */
1710    
1711  static int  static int
1712  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1713  {  {
1714  int length = -1;  int length = -1;
1715    
1716  register int branchlength = 0;  register int branchlength = 0;
1717  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1718    
1719  /* 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
1720  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 1722  branch, check the length against that of
1722  for (;;)  for (;;)
1723    {    {
1724    int d;    int d;
1725    register int op = *cc;    pcre_uchar *ce, *cs;
1726      register pcre_uchar op = *cc;
1727    
1728    switch (op)    switch (op)
1729      {      {
1730        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1731        OP_BRA (normal non-capturing bracket) because the other variants of these
1732        opcodes are all concerned with unlimited repeated groups, which of course
1733        are not of fixed length. */
1734    
1735      case OP_CBRA:      case OP_CBRA:
1736      case OP_BRA:      case OP_BRA:
1737      case OP_ONCE:      case OP_ONCE:
1738        case OP_ONCE_NC:
1739      case OP_COND:      case OP_COND:
1740      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1741      if (d < 0) return d;      if (d < 0) return d;
1742      branchlength += d;      branchlength += d;
1743      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1744      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1745      break;      break;
1746    
1747      /* 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.
1748      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
1749      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
1750        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1751        because they all imply an unlimited repeat. */
1752    
1753      case OP_ALT:      case OP_ALT:
1754      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1755      case OP_END:      case OP_END:
1756        case OP_ACCEPT:
1757        case OP_ASSERT_ACCEPT:
1758      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1759        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1760      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1762  for (;;)
1762      branchlength = 0;      branchlength = 0;
1763      break;      break;
1764    
1765        /* A true recursion implies not fixed length, but a subroutine call may
1766        be OK. If the subroutine is a forward reference, we can't deal with
1767        it until the end of the pattern, so return -3. */
1768    
1769        case OP_RECURSE:
1770        if (!atend) return -3;
1771        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1772        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1773        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1774        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1775        if (d < 0) return d;
1776        branchlength += d;
1777        cc += 1 + LINK_SIZE;
1778        break;
1779    
1780      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1781    
1782      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1784  for (;;)
1784      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1785      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1786      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1787      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1788        break;
1789    
1790      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1791    
1792      case OP_REVERSE:      case OP_MARK:
1793        case OP_PRUNE_ARG:
1794        case OP_SKIP_ARG:
1795        case OP_THEN_ARG:
1796        cc += cc[1] + PRIV(OP_lengths)[*cc];
1797        break;
1798    
1799        case OP_CALLOUT:
1800        case OP_CIRC:
1801        case OP_CIRCM:
1802        case OP_CLOSE:
1803        case OP_COMMIT:
1804      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1805      case OP_DEF:      case OP_DEF:
1806      case OP_OPT:      case OP_DNCREF:
1807      case OP_CALLOUT:      case OP_DNRREF:
1808      case OP_SOD:      case OP_DOLL:
1809      case OP_SOM:      case OP_DOLLM:
1810      case OP_EOD:      case OP_EOD:
1811      case OP_EODN:      case OP_EODN:
1812      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1813      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1814        case OP_PRUNE:
1815        case OP_REVERSE:
1816        case OP_RREF:
1817        case OP_SET_SOM:
1818        case OP_SKIP:
1819        case OP_SOD:
1820        case OP_SOM:
1821        case OP_THEN:
1822      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1823      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1824      break;      break;
1825    
1826      /* Handle literal characters */      /* Handle literal characters */
1827    
1828      case OP_CHAR:      case OP_CHAR:
1829      case OP_CHARNC:      case OP_CHARI:
1830      case OP_NOT:      case OP_NOT:
1831        case OP_NOTI:
1832      branchlength++;      branchlength++;
1833      cc += 2;      cc += 2;
1834  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1835      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1836  #endif  #endif
1837      break;      break;
1838    
# Line 1119  for (;;) Line 1840  for (;;)
1840      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1841    
1842      case OP_EXACT:      case OP_EXACT:
1843      branchlength += GET2(cc,1);      case OP_EXACTI:
1844      cc += 4;      case OP_NOTEXACT:
1845  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1846      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1847        {      cc += 2 + IMM2_SIZE;
1848        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1849        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1850  #endif  #endif
1851      break;      break;
1852    
1853      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1854      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1855      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1856          cc += 2;
1857        cc += 1 + IMM2_SIZE + 1;
1858      break;      break;
1859    
1860      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1864  for (;;)
1864      cc += 2;      cc += 2;
1865      /* Fall through */      /* Fall through */
1866    
1867        case OP_HSPACE:
1868        case OP_VSPACE:
1869        case OP_NOT_HSPACE:
1870        case OP_NOT_VSPACE:
1871      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1872      case OP_DIGIT:      case OP_DIGIT:
1873      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1875  for (;;)
1875      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1876      case OP_WORDCHAR:      case OP_WORDCHAR:
1877      case OP_ANY:      case OP_ANY:
1878        case OP_ALLANY:
1879      branchlength++;      branchlength++;
1880      cc++;      cc++;
1881      break;      break;
1882    
1883      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1884        otherwise \C is coded as OP_ALLANY. */
1885    
1886      case OP_ANYBYTE:      case OP_ANYBYTE:
1887      return -2;      return -2;
1888    
1889      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1890    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1891      case OP_CLASS:      case OP_CLASS:
1892      case OP_NCLASS:      case OP_NCLASS:
1893      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1894        case OP_XCLASS:
1895        /* The original code caused an unsigned overflow in 64 bit systems,
1896        so now we use a conditional statement. */
1897        if (op == OP_XCLASS)
1898          cc += GET(cc, 1);
1899        else
1900          cc += PRIV(OP_lengths)[OP_CLASS];
1901    #else
1902        cc += PRIV(OP_lengths)[OP_CLASS];
1903    #endif
1904    
1905      switch (*cc)      switch (*cc)
1906        {        {
1907        case OP_CRSTAR:        case OP_CRSTAR:
1908        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1909          case OP_CRPLUS:
1910          case OP_CRMINPLUS:
1911        case OP_CRQUERY:        case OP_CRQUERY:
1912        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1913          case OP_CRPOSSTAR:
1914          case OP_CRPOSPLUS:
1915          case OP_CRPOSQUERY:
1916        return -1;        return -1;
1917    
1918        case OP_CRRANGE:        case OP_CRRANGE:
1919        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1920        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1921        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1922        cc += 5;        branchlength += (int)GET2(cc,1);
1923          cc += 1 + 2 * IMM2_SIZE;
1924        break;        break;
1925    
1926        default:        default:
# Line 1191  for (;;) Line 1930  for (;;)
1930    
1931      /* Anything else is variable length */      /* Anything else is variable length */
1932    
1933      default:      case OP_ANYNL:
1934        case OP_BRAMINZERO:
1935        case OP_BRAPOS:
1936        case OP_BRAPOSZERO:
1937        case OP_BRAZERO:
1938        case OP_CBRAPOS:
1939        case OP_EXTUNI:
1940        case OP_KETRMAX:
1941        case OP_KETRMIN:
1942        case OP_KETRPOS:
1943        case OP_MINPLUS:
1944        case OP_MINPLUSI:
1945        case OP_MINQUERY:
1946        case OP_MINQUERYI:
1947        case OP_MINSTAR:
1948        case OP_MINSTARI:
1949        case OP_MINUPTO:
1950        case OP_MINUPTOI:
1951        case OP_NOTMINPLUS:
1952        case OP_NOTMINPLUSI:
1953        case OP_NOTMINQUERY:
1954        case OP_NOTMINQUERYI:
1955        case OP_NOTMINSTAR:
1956        case OP_NOTMINSTARI:
1957        case OP_NOTMINUPTO:
1958        case OP_NOTMINUPTOI:
1959        case OP_NOTPLUS:
1960        case OP_NOTPLUSI:
1961        case OP_NOTPOSPLUS:
1962        case OP_NOTPOSPLUSI:
1963        case OP_NOTPOSQUERY:
1964        case OP_NOTPOSQUERYI:
1965        case OP_NOTPOSSTAR:
1966        case OP_NOTPOSSTARI:
1967        case OP_NOTPOSUPTO:
1968        case OP_NOTPOSUPTOI:
1969        case OP_NOTQUERY:
1970        case OP_NOTQUERYI:
1971        case OP_NOTSTAR:
1972        case OP_NOTSTARI:
1973        case OP_NOTUPTO:
1974        case OP_NOTUPTOI:
1975        case OP_PLUS:
1976        case OP_PLUSI:
1977        case OP_POSPLUS:
1978        case OP_POSPLUSI:
1979        case OP_POSQUERY:
1980        case OP_POSQUERYI:
1981        case OP_POSSTAR:
1982        case OP_POSSTARI:
1983        case OP_POSUPTO:
1984        case OP_POSUPTOI:
1985        case OP_QUERY:
1986        case OP_QUERYI:
1987        case OP_REF:
1988        case OP_REFI:
1989        case OP_DNREF:
1990        case OP_DNREFI:
1991        case OP_SBRA:
1992        case OP_SBRAPOS:
1993        case OP_SCBRA:
1994        case OP_SCBRAPOS:
1995        case OP_SCOND:
1996        case OP_SKIPZERO:
1997        case OP_STAR:
1998        case OP_STARI:
1999        case OP_TYPEMINPLUS:
2000        case OP_TYPEMINQUERY:
2001        case OP_TYPEMINSTAR:
2002        case OP_TYPEMINUPTO:
2003        case OP_TYPEPLUS:
2004        case OP_TYPEPOSPLUS:
2005        case OP_TYPEPOSQUERY:
2006        case OP_TYPEPOSSTAR:
2007        case OP_TYPEPOSUPTO:
2008        case OP_TYPEQUERY:
2009        case OP_TYPESTAR:
2010        case OP_TYPEUPTO:
2011        case OP_UPTO:
2012        case OP_UPTOI:
2013      return -1;      return -1;
2014    
2015        /* Catch unrecognized opcodes so that when new ones are added they
2016        are not forgotten, as has happened in the past. */
2017    
2018        default:
2019        return -4;
2020      }      }
2021    }    }
2022  /* Control never gets here */  /* Control never gets here */
# Line 1200  for (;;) Line 2024  for (;;)
2024    
2025    
2026    
   
2027  /*************************************************  /*************************************************
2028  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2029  *************************************************/  *************************************************/
2030    
2031  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2032  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2033    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2034    so that it can be called from pcre_study() when finding the minimum matching
2035    length.
2036    
2037  Arguments:  Arguments:
2038    code        points to start of expression    code        points to start of expression
2039    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2040    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2041    
2042  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
2043  */  */
2044    
2045  static const uschar *  const pcre_uchar *
2046  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2047  {  {
2048  for (;;)  for (;;)
2049    {    {
2050    register int c = *code;    register pcre_uchar c = *code;
2051    
2052    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2053    
2054    /* 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 2057  for (;;)
2057    
2058    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2059    
2060      /* Handle recursion */
2061    
2062      else if (c == OP_REVERSE)
2063        {
2064        if (number < 0) return (pcre_uchar *)code;
2065        code += PRIV(OP_lengths)[c];
2066        }
2067    
2068    /* Handle capturing bracket */    /* Handle capturing bracket */
2069    
2070    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2071               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2072      {      {
2073      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2074      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2075      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2076      }      }
2077    
2078      /* Otherwise, we can get the item's length from the table, except that for
2079      repeated character types, we have to test for \p and \P, which have an extra
2080      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2081      must add in its length. */
2082    
2083      else
2084        {
2085        switch(c)
2086          {
2087          case OP_TYPESTAR:
2088          case OP_TYPEMINSTAR:
2089          case OP_TYPEPLUS:
2090          case OP_TYPEMINPLUS:
2091          case OP_TYPEQUERY:
2092          case OP_TYPEMINQUERY:
2093          case OP_TYPEPOSSTAR:
2094          case OP_TYPEPOSPLUS:
2095          case OP_TYPEPOSQUERY:
2096          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2097          break;
2098    
2099          case OP_TYPEUPTO:
2100          case OP_TYPEMINUPTO:
2101          case OP_TYPEEXACT:
2102          case OP_TYPEPOSUPTO:
2103          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2104            code += 2;
2105          break;
2106    
2107          case OP_MARK:
2108          case OP_PRUNE_ARG:
2109          case OP_SKIP_ARG:
2110          case OP_THEN_ARG:
2111          code += code[1];
2112          break;
2113          }
2114    
2115        /* Add in the fixed length from the table */
2116    
2117        code += PRIV(OP_lengths)[c];
2118    
2119    /* 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
2120    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
2121    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2122    
2123    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2124      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2125        {        {
2126        case OP_CHAR:        case OP_CHAR:
2127        case OP_CHARNC:        case OP_CHARI:
2128        case OP_EXACT:        case OP_EXACT:
2129          case OP_EXACTI:
2130        case OP_UPTO:        case OP_UPTO:
2131          case OP_UPTOI:
2132        case OP_MINUPTO:        case OP_MINUPTO:
2133          case OP_MINUPTOI:
2134        case OP_POSUPTO:        case OP_POSUPTO:
2135          case OP_POSUPTOI:
2136        case OP_STAR:        case OP_STAR:
2137          case OP_STARI:
2138        case OP_MINSTAR:        case OP_MINSTAR:
2139          case OP_MINSTARI:
2140        case OP_POSSTAR:        case OP_POSSTAR:
2141          case OP_POSSTARI:
2142        case OP_PLUS:        case OP_PLUS:
2143          case OP_PLUSI:
2144        case OP_MINPLUS:        case OP_MINPLUS:
2145          case OP_MINPLUSI:
2146        case OP_POSPLUS:        case OP_POSPLUS:
2147          case OP_POSPLUSI:
2148        case OP_QUERY:        case OP_QUERY:
2149          case OP_QUERYI:
2150        case OP_MINQUERY:        case OP_MINQUERY:
2151          case OP_MINQUERYI:
2152        case OP_POSQUERY:        case OP_POSQUERY:
2153        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2154          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2155        break;        break;
2156        }        }
2157    #else
2158        (void)(utf);  /* Keep compiler happy by referencing function argument */
2159  #endif  #endif
2160      }      }
2161    }    }
# Line 1283  instance of OP_RECURSE. Line 2172  instance of OP_RECURSE.
2172    
2173  Arguments:  Arguments:
2174    code        points to start of expression    code        points to start of expression
2175    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2176    
2177  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
2178  */  */
2179    
2180  static const uschar *  static const pcre_uchar *
2181  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2182  {  {
2183  for (;;)  for (;;)
2184    {    {
2185    register int c = *code;    register pcre_uchar c = *code;
2186    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2187    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2188    
# Line 1303  for (;;) Line 2192  for (;;)
2192    
2193    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2194    
2195    /* 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
2196    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
2197    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
2198    bytes. */    must add in its length. */
2199    
2200    else    else
2201      {      {
2202      code += _pcre_OP_lengths[c];      switch(c)
2203  #ifdef SUPPORT_UTF8        {
2204      if (utf8) switch(c)        case OP_TYPESTAR:
2205          case OP_TYPEMINSTAR:
2206          case OP_TYPEPLUS:
2207          case OP_TYPEMINPLUS:
2208          case OP_TYPEQUERY:
2209          case OP_TYPEMINQUERY:
2210          case OP_TYPEPOSSTAR:
2211          case OP_TYPEPOSPLUS:
2212          case OP_TYPEPOSQUERY:
2213          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2214          break;
2215    
2216          case OP_TYPEPOSUPTO:
2217          case OP_TYPEUPTO:
2218          case OP_TYPEMINUPTO:
2219          case OP_TYPEEXACT:
2220          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2221            code += 2;
2222          break;
2223    
2224          case OP_MARK:
2225          case OP_PRUNE_ARG:
2226          case OP_SKIP_ARG:
2227          case OP_THEN_ARG:
2228          code += code[1];
2229          break;
2230          }
2231    
2232        /* Add in the fixed length from the table */
2233    
2234        code += PRIV(OP_lengths)[c];
2235    
2236        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2237        by a multi-byte character. The length in the table is a minimum, so we have
2238        to arrange to skip the extra bytes. */
2239    
2240    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2241        if (utf) switch(c)
2242        {        {
2243        case OP_CHAR:        case OP_CHAR:
2244        case OP_CHARNC:        case OP_CHARI:
2245          case OP_NOT:
2246          case OP_NOTI:
2247        case OP_EXACT:        case OP_EXACT:
2248          case OP_EXACTI:
2249          case OP_NOTEXACT:
2250          case OP_NOTEXACTI:
2251        case OP_UPTO:        case OP_UPTO:
2252          case OP_UPTOI:
2253          case OP_NOTUPTO:
2254          case OP_NOTUPTOI:
2255        case OP_MINUPTO:        case OP_MINUPTO:
2256          case OP_MINUPTOI:
2257          case OP_NOTMINUPTO:
2258          case OP_NOTMINUPTOI:
2259        case OP_POSUPTO:        case OP_POSUPTO:
2260          case OP_POSUPTOI:
2261          case OP_NOTPOSUPTO:
2262          case OP_NOTPOSUPTOI:
2263        case OP_STAR:        case OP_STAR:
2264          case OP_STARI:
2265          case OP_NOTSTAR:
2266          case OP_NOTSTARI:
2267        case OP_MINSTAR:        case OP_MINSTAR:
2268          case OP_MINSTARI:
2269          case OP_NOTMINSTAR:
2270          case OP_NOTMINSTARI:
2271        case OP_POSSTAR:        case OP_POSSTAR:
2272          case OP_POSSTARI:
2273          case OP_NOTPOSSTAR:
2274          case OP_NOTPOSSTARI:
2275        case OP_PLUS:        case OP_PLUS:
2276          case OP_PLUSI:
2277          case OP_NOTPLUS:
2278          case OP_NOTPLUSI:
2279        case OP_MINPLUS:        case OP_MINPLUS:
2280          case OP_MINPLUSI:
2281          case OP_NOTMINPLUS:
2282          case OP_NOTMINPLUSI:
2283        case OP_POSPLUS:        case OP_POSPLUS:
2284          case OP_POSPLUSI:
2285          case OP_NOTPOSPLUS:
2286          case OP_NOTPOSPLUSI:
2287        case OP_QUERY:        case OP_QUERY:
2288          case OP_QUERYI:
2289          case OP_NOTQUERY:
2290          case OP_NOTQUERYI:
2291        case OP_MINQUERY:        case OP_MINQUERY:
2292          case OP_MINQUERYI:
2293          case OP_NOTMINQUERY:
2294          case OP_NOTMINQUERYI:
2295        case OP_POSQUERY:        case OP_POSQUERY:
2296        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2297          case OP_NOTPOSQUERY:
2298          case OP_NOTPOSQUERYI:
2299          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2300        break;        break;
2301        }        }
2302    #else
2303        (void)(utf);  /* Keep compiler happy by referencing function argument */
2304  #endif  #endif
2305      }      }
2306    }    }
# Line 1347  for (;;) Line 2316  for (;;)
2316  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()
2317  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
2318  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2319  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
2320  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
2321    bracket whose current branch will already have been scanned.
2322    
2323  Arguments:  Arguments:
2324    code        points to start of search    code        points to start of search
2325    endcode     points to where to stop    endcode     points to where to stop
2326    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2327      cd          contains pointers to tables etc.
2328      recurses    chain of recurse_check to catch mutual recursion
2329    
2330  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2331  */  */
2332    
2333    typedef struct recurse_check {
2334      struct recurse_check *prev;
2335      const pcre_uchar *group;
2336    } recurse_check;
2337    
2338  static BOOL  static BOOL
2339  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2340      BOOL utf, compile_data *cd, recurse_check *recurses)
2341  {  {
2342  register int c;  register pcre_uchar c;
2343  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2344    
2345    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2346       code < endcode;       code < endcode;
2347       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2348    {    {
2349    const uschar *ccode;    const pcre_uchar *ccode;
2350    
2351    c = *code;    c = *code;
2352    
2353    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2354      first_significant_code() with a TRUE final argument. */
2355    
2356      if (c == OP_ASSERT)
2357        {
2358        do code += GET(code, 1); while (*code == OP_ALT);
2359        c = *code;
2360        continue;
2361        }
2362    
2363      /* For a recursion/subroutine call, if its end has been reached, which
2364      implies a backward reference subroutine call, we can scan it. If it's a
2365      forward reference subroutine call, we can't. To detect forward reference
2366      we have to scan up the list that is kept in the workspace. This function is
2367      called only when doing the real compile, not during the pre-compile that
2368      measures the size of the compiled pattern. */
2369    
2370      if (c == OP_RECURSE)
2371      {      {
2372        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2373      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2374    
2375      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2376        when called to scan a completed pattern by setting cd->start_workspace to
2377        NULL. */
2378    
2379        if (cd->start_workspace != NULL)
2380          {
2381          const pcre_uchar *tcode;
2382          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2383            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2384          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2385          }
2386    
2387        /* If we are scanning a completed pattern, there are no forward references
2388        and all groups are complete. We need to detect whether this is a recursive
2389        call, as otherwise there will be an infinite loop. If it is a recursion,
2390        just skip over it. Simple recursions are easily detected. For mutual
2391        recursions we keep a chain on the stack. */
2392    
2393        else
2394          {
2395          recurse_check *r = recurses;
2396          const pcre_uchar *endgroup = scode;
2397    
2398          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2399          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2400    
2401          for (r = recurses; r != NULL; r = r->prev)
2402            if (r->group == scode) break;
2403          if (r != NULL) continue;   /* Mutual recursion */
2404          }
2405    
2406        /* Completed reference; scan the referenced group, remembering it on the
2407        stack chain to detect mutual recursions. */
2408    
2409      empty_branch = FALSE;      empty_branch = FALSE;
2410        this_recurse.prev = recurses;
2411        this_recurse.group = scode;
2412    
2413      do      do
2414        {        {
2415        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2416            {
2417          empty_branch = TRUE;          empty_branch = TRUE;
2418        code += GET(code, 1);          break;
2419            }
2420          scode += GET(scode, 1);
2421        }        }
2422      while (*code == OP_ALT);      while (*scode == OP_ALT);
2423      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2424        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2425        continue;
2426        }
2427    
2428      /* Groups with zero repeats can of course be empty; skip them. */
2429    
2430      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2431          c == OP_BRAPOSZERO)
2432        {
2433        code += PRIV(OP_lengths)[c];
2434        do code += GET(code, 1); while (*code == OP_ALT);
2435        c = *code;
2436        continue;
2437        }
2438    
2439      /* A nested group that is already marked as "could be empty" can just be
2440      skipped. */
2441    
2442      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2443          c == OP_SCBRA || c == OP_SCBRAPOS)
2444        {
2445        do code += GET(code, 1); while (*code == OP_ALT);
2446        c = *code;
2447        continue;
2448        }
2449    
2450      /* For other groups, scan the branches. */
2451    
2452      /* Move past the KET and fudge things so that the increment in the "for"    if (c == OP_BRA  || c == OP_BRAPOS ||
2453      above has no effect. */        c == OP_CBRA || c == OP_CBRAPOS ||
2454          c == OP_ONCE || c == OP_ONCE_NC ||
2455          c == OP_COND)
2456        {
2457        BOOL empty_branch;
2458        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2459    
2460        /* If a conditional group has only one branch, there is a second, implied,
2461        empty branch, so just skip over the conditional, because it could be empty.
2462        Otherwise, scan the individual branches of the group. */
2463    
2464        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2465          code += GET(code, 1);
2466        else
2467          {
2468          empty_branch = FALSE;
2469          do
2470            {
2471            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2472              empty_branch = TRUE;
2473            code += GET(code, 1);
2474            }
2475          while (*code == OP_ALT);
2476          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2477          }
2478    
2479      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2480      continue;      continue;
2481      }      }
2482    
# Line 1399  for (code = first_significant_code(code Line 2484  for (code = first_significant_code(code
2484    
2485    switch (c)    switch (c)
2486      {      {
2487      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2488        cannot be represented just by a bit map. This includes negated single
2489        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2490        actual length is stored in the compiled code, so we must update "code"
2491        here. */
2492    
2493  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2494      case OP_XCLASS:      case OP_XCLASS:
2495      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2496      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2497  #endif  #endif
2498    
2499      case OP_CLASS:      case OP_CLASS:
2500      case OP_NCLASS:      case OP_NCLASS:
2501      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2502    
2503  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2504      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2505  #endif  #endif
2506    
# Line 1421  for (code = first_significant_code(code Line 2510  for (code = first_significant_code(code
2510        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2511        case OP_CRQUERY:        case OP_CRQUERY:
2512        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2513          case OP_CRPOSSTAR:
2514          case OP_CRPOSQUERY:
2515        break;        break;
2516    
2517        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2518        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2519        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2520          case OP_CRPOSPLUS:
2521        return FALSE;        return FALSE;
2522    
2523        case OP_CRRANGE:        case OP_CRRANGE:
2524        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2525          case OP_CRPOSRANGE:
2526        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2527        break;        break;
2528        }        }
# Line 1437  for (code = first_significant_code(code Line 2530  for (code = first_significant_code(code
2530    
2531      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2532    
2533        case OP_ANY:
2534        case OP_ALLANY:
2535        case OP_ANYBYTE:
2536    
2537      case OP_PROP:      case OP_PROP:
2538      case OP_NOTPROP:      case OP_NOTPROP:
2539        case OP_ANYNL:
2540    
2541        case OP_NOT_HSPACE:
2542        case OP_HSPACE:
2543        case OP_NOT_VSPACE:
2544        case OP_VSPACE:
2545      case OP_EXTUNI:      case OP_EXTUNI:
2546    
2547      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2548      case OP_DIGIT:      case OP_DIGIT:
2549      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2550      case OP_WHITESPACE:      case OP_WHITESPACE:
2551      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2552      case OP_WORDCHAR:      case OP_WORDCHAR:
2553      case OP_ANY:  
     case OP_ANYBYTE:  
2554      case OP_CHAR:      case OP_CHAR:
2555      case OP_CHARNC:      case OP_CHARI:
2556      case OP_NOT:      case OP_NOT:
2557        case OP_NOTI:
2558    
2559      case OP_PLUS:      case OP_PLUS:
2560        case OP_PLUSI:
2561      case OP_MINPLUS:      case OP_MINPLUS:
2562      case OP_POSPLUS:      case OP_MINPLUSI:
2563      case OP_EXACT:  
2564      case OP_NOTPLUS:      case OP_NOTPLUS:
2565        case OP_NOTPLUSI:
2566      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2567        case OP_NOTMINPLUSI:
2568    
2569        case OP_POSPLUS:
2570        case OP_POSPLUSI:
2571      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2572        case OP_NOTPOSPLUSI:
2573    
2574        case OP_EXACT:
2575        case OP_EXACTI:
2576      case OP_NOTEXACT:      case OP_NOTEXACT:
2577        case OP_NOTEXACTI:
2578    
2579      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2580      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2581      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2582      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2583    
2584      return FALSE;      return FALSE;
2585    
2586        /* These are going to continue, as they may be empty, but we have to
2587        fudge the length for the \p and \P cases. */
2588    
2589        case OP_TYPESTAR:
2590        case OP_TYPEMINSTAR:
2591        case OP_TYPEPOSSTAR:
2592        case OP_TYPEQUERY:
2593        case OP_TYPEMINQUERY:
2594        case OP_TYPEPOSQUERY:
2595        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2596        break;
2597    
2598        /* Same for these */
2599    
2600        case OP_TYPEUPTO:
2601        case OP_TYPEMINUPTO:
2602        case OP_TYPEPOSUPTO:
2603        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2604          code += 2;
2605        break;
2606    
2607      /* End of branch */      /* End of branch */
2608    
2609      case OP_KET:      case OP_KET:
2610      case OP_KETRMAX:      case OP_KETRMAX:
2611      case OP_KETRMIN:      case OP_KETRMIN:
2612        case OP_KETRPOS:
2613      case OP_ALT:      case OP_ALT:
2614      return TRUE;      return TRUE;
2615    
2616      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2617      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2618        followed by a multibyte character. */
2619    
2620  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2621      case OP_STAR:      case OP_STAR:
2622        case OP_STARI:
2623        case OP_NOTSTAR:
2624        case OP_NOTSTARI:
2625    
2626      case OP_MINSTAR:      case OP_MINSTAR:
2627        case OP_MINSTARI:
2628        case OP_NOTMINSTAR:
2629        case OP_NOTMINSTARI:
2630    
2631      case OP_POSSTAR:      case OP_POSSTAR:
2632        case OP_POSSTARI:
2633        case OP_NOTPOSSTAR:
2634        case OP_NOTPOSSTARI:
2635    
2636      case OP_QUERY:      case OP_QUERY:
2637        case OP_QUERYI:
2638        case OP_NOTQUERY:
2639        case OP_NOTQUERYI:
2640    
2641      case OP_MINQUERY:      case OP_MINQUERY:
2642        case OP_MINQUERYI:
2643        case OP_NOTMINQUERY:
2644        case OP_NOTMINQUERYI:
2645    
2646      case OP_POSQUERY:      case OP_POSQUERY:
2647        case OP_POSQUERYI:
2648        case OP_NOTPOSQUERY:
2649        case OP_NOTPOSQUERYI:
2650    
2651        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2652        break;
2653    
2654      case OP_UPTO:      case OP_UPTO:
2655        case OP_UPTOI:
2656        case OP_NOTUPTO:
2657        case OP_NOTUPTOI:
2658    
2659      case OP_MINUPTO:      case OP_MINUPTO:
2660        case OP_MINUPTOI:
2661        case OP_NOTMINUPTO:
2662        case OP_NOTMINUPTOI:
2663    
2664      case OP_POSUPTO:      case OP_POSUPTO:
2665      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2666        case OP_NOTPOSUPTO:
2667        case OP_NOTPOSUPTOI:
2668    
2669        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2670      break;      break;
2671  #endif  #endif
2672    
2673        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2674        string. */
2675    
2676        case OP_MARK:
2677        case OP_PRUNE_ARG:
2678        case OP_SKIP_ARG:
2679        case OP_THEN_ARG:
2680        code += code[1];
2681        break;
2682    
2683        /* None of the remaining opcodes are required to match a character. */
2684    
2685        default:
2686        break;
2687      }      }
2688    }    }
2689    
# Line 1505  return TRUE; Line 2700  return TRUE;
2700  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
2701  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,
2702  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.
2703    This function is called only during the real compile, not during the
2704    pre-compile.
2705    
2706  Arguments:  Arguments:
2707    code        points to start of the recursion    code        points to start of the recursion
2708    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2709    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2710    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2711      cd          pointers to tables etc
2712    
2713  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2714  */  */
2715    
2716  static BOOL  static BOOL
2717  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2718    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2719  {  {
2720  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2721    {    {
2722    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2723        return FALSE;
2724    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2725    }    }
2726  return TRUE;  return TRUE;
# Line 1530  return TRUE; Line 2729  return TRUE;
2729    
2730    
2731  /*************************************************  /*************************************************
2732  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2733  *************************************************/  *************************************************/
2734    
2735  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2736  encountered in a character class. It checks whether this is followed by an  opcode is not a repeated character type, it returns with the original value.
 optional ^ and then a sequence of letters, terminated by a matching ":]" or  
 ".]" or "=]".  
2737    
2738  Argument:  Arguments:  c opcode
2739    Returns:    base opcode for the type
2740    */
2741    
2742    static pcre_uchar
2743    get_repeat_base(pcre_uchar c)
2744    {
2745    return (c > OP_TYPEPOSUPTO)? c :
2746           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2747           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2748           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2749           (c >= OP_STARI)?      OP_STARI :
2750                                 OP_STAR;
2751    }
2752    
2753    
2754    
2755    #ifdef SUPPORT_UCP
2756    /*************************************************
2757    *        Check a character and a property        *
2758    *************************************************/
2759    
2760    /* This function is called by check_auto_possessive() when a property item
2761    is adjacent to a fixed character.
2762    
2763    Arguments:
2764      c            the character
2765      ptype        the property type
2766      pdata        the data for the type
2767      negated      TRUE if it's a negated property (\P or \p{^)
2768    
2769    Returns:       TRUE if auto-possessifying is OK
2770    */
2771    
2772    static BOOL
2773    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2774      BOOL negated)
2775    {
2776    const pcre_uint32 *p;
2777    const ucd_record *prop = GET_UCD(c);
2778    
2779    switch(ptype)
2780      {
2781      case PT_LAMP:
2782      return (prop->chartype == ucp_Lu ||
2783              prop->chartype == ucp_Ll ||
2784              prop->chartype == ucp_Lt) == negated;
2785    
2786      case PT_GC:
2787      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2788    
2789      case PT_PC:
2790      return (pdata == prop->chartype) == negated;
2791    
2792      case PT_SC:
2793      return (pdata == prop->script) == negated;
2794    
2795      /* These are specials */
2796    
2797      case PT_ALNUM:
2798      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2799              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2800    
2801      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2802      means that Perl space and POSIX space are now identical. PCRE was changed
2803      at release 8.34. */
2804    
2805      case PT_SPACE:    /* Perl space */
2806      case PT_PXSPACE:  /* POSIX space */
2807      switch(c)
2808        {
2809        HSPACE_CASES:
2810        VSPACE_CASES:
2811        return negated;
2812    
2813        default:
2814        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2815        }
2816      break;  /* Control never reaches here */
2817    
2818      case PT_WORD:
2819      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2820              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2821              c == CHAR_UNDERSCORE) == negated;
2822    
2823      case PT_CLIST:
2824      p = PRIV(ucd_caseless_sets) + prop->caseset;
2825      for (;;)
2826        {
2827        if (c < *p) return !negated;
2828        if (c == *p++) return negated;
2829        }
2830      break;  /* Control never reaches here */
2831      }
2832    
2833    return FALSE;
2834    }
2835    #endif  /* SUPPORT_UCP */
2836    
2837    
2838    
2839    /*************************************************
2840    *        Fill the character property list        *
2841    *************************************************/
2842    
2843    /* Checks whether the code points to an opcode that can take part in auto-
2844    possessification, and if so, fills a list with its properties.
2845    
2846    Arguments:
2847      code        points to start of expression
2848      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2849      fcc         points to case-flipping table
2850      list        points to output list
2851                  list[0] will be filled with the opcode
2852                  list[1] will be non-zero if this opcode
2853                    can match an empty character string
2854                  list[2..7] depends on the opcode
2855    
2856    Returns:      points to the start of the next opcode if *code is accepted
2857                  NULL if *code is not accepted
2858    */
2859    
2860    static const pcre_uchar *
2861    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2862      const pcre_uint8 *fcc, pcre_uint32 *list)
2863    {
2864    pcre_uchar c = *code;
2865    pcre_uchar base;
2866    const pcre_uchar *end;
2867    pcre_uint32 chr;
2868    
2869    #ifdef SUPPORT_UCP
2870    pcre_uint32 *clist_dest;
2871    const pcre_uint32 *clist_src;
2872    #else
2873    utf = utf;  /* Suppress "unused parameter" compiler warning */
2874    #endif
2875    
2876    list[0] = c;
2877    list[1] = FALSE;
2878    code++;
2879    
2880    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2881      {
2882      base = get_repeat_base(c);
2883      c -= (base - OP_STAR);
2884    
2885      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2886        code += IMM2_SIZE;
2887    
2888      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2889    
2890      switch(base)
2891        {
2892        case OP_STAR:
2893        list[0] = OP_CHAR;
2894        break;
2895    
2896        case OP_STARI:
2897        list[0] = OP_CHARI;
2898        break;
2899    
2900        case OP_NOTSTAR:
2901        list[0] = OP_NOT;
2902        break;
2903    
2904        case OP_NOTSTARI:
2905        list[0] = OP_NOTI;
2906        break;
2907    
2908        case OP_TYPESTAR:
2909        list[0] = *code;
2910        code++;
2911        break;
2912        }
2913      c = list[0];
2914      }
2915    
2916    switch(c)
2917      {
2918      case OP_NOT_DIGIT:
2919      case OP_DIGIT:
2920      case OP_NOT_WHITESPACE:
2921      case OP_WHITESPACE:
2922      case OP_NOT_WORDCHAR:
2923      case OP_WORDCHAR:
2924      case OP_ANY:
2925      case OP_ALLANY:
2926      case OP_ANYNL:
2927      case OP_NOT_HSPACE:
2928      case OP_HSPACE:
2929      case OP_NOT_VSPACE:
2930      case OP_VSPACE:
2931      case OP_EXTUNI:
2932      case OP_EODN:
2933      case OP_EOD:
2934      case OP_DOLL:
2935      case OP_DOLLM:
2936      return code;
2937    
2938      case OP_CHAR:
2939      case OP_NOT:
2940      GETCHARINCTEST(chr, code);
2941      list[2] = chr;
2942      list[3] = NOTACHAR;
2943      return code;
2944    
2945      case OP_CHARI:
2946      case OP_NOTI:
2947      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2948      GETCHARINCTEST(chr, code);
2949      list[2] = chr;
2950    
2951    #ifdef SUPPORT_UCP
2952      if (chr < 128 || (chr < 256 && !utf))
2953        list[3] = fcc[chr];
2954      else
2955        list[3] = UCD_OTHERCASE(chr);
2956    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2957      list[3] = (chr < 256) ? fcc[chr] : chr;
2958    #else
2959      list[3] = fcc[chr];
2960    #endif
2961    
2962      /* The othercase might be the same value. */
2963    
2964      if (chr == list[3])
2965        list[3] = NOTACHAR;
2966      else
2967        list[4] = NOTACHAR;
2968      return code;
2969    
2970    #ifdef SUPPORT_UCP
2971      case OP_PROP:
2972      case OP_NOTPROP:
2973      if (code[0] != PT_CLIST)
2974        {
2975        list[2] = code[0];
2976        list[3] = code[1];
2977        return code + 2;
2978        }
2979    
2980      /* Convert only if we have enough space. */
2981    
2982      clist_src = PRIV(ucd_caseless_sets) + code[1];
2983      clist_dest = list + 2;
2984      code += 2;
2985    
2986      do {
2987         if (clist_dest >= list + 8)
2988           {
2989           /* Early return if there is not enough space. This should never
2990           happen, since all clists are shorter than 5 character now. */
2991           list[2] = code[0];
2992           list[3] = code[1];
2993           return code;
2994           }
2995         *clist_dest++ = *clist_src;
2996         }
2997      while(*clist_src++ != NOTACHAR);
2998    
2999      /* All characters are stored. The terminating NOTACHAR
3000      is copied form the clist itself. */
3001    
3002      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3003      return code;
3004    #endif
3005    
3006      case OP_NCLASS:
3007      case OP_CLASS:
3008    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3009      case OP_XCLASS:
3010      if (c == OP_XCLASS)
3011        end = code + GET(code, 0) - 1;
3012      else
3013    #endif
3014        end = code + 32 / sizeof(pcre_uchar);
3015    
3016      switch(*end)
3017        {
3018        case OP_CRSTAR:
3019        case OP_CRMINSTAR:
3020        case OP_CRQUERY:
3021        case OP_CRMINQUERY:
3022        case OP_CRPOSSTAR:
3023        case OP_CRPOSQUERY:
3024        list[1] = TRUE;
3025        end++;
3026        break;
3027    
3028        case OP_CRPLUS:
3029        case OP_CRMINPLUS:
3030        case OP_CRPOSPLUS:
3031        end++;
3032        break;
3033    
3034        case OP_CRRANGE:
3035        case OP_CRMINRANGE:
3036        case OP_CRPOSRANGE:
3037        list[1] = (GET2(end, 1) == 0);
3038        end += 1 + 2 * IMM2_SIZE;
3039        break;
3040        }
3041      list[2] = end - code;
3042      return end;
3043      }
3044    return NULL;    /* Opcode not accepted */
3045    }
3046    
3047    
3048    
3049    /*************************************************
3050    *    Scan further character sets for match       *
3051    *************************************************/
3052    
3053    /* Checks whether the base and the current opcode have a common character, in
3054    which case the base cannot be possessified.
3055    
3056    Arguments:
3057      code        points to the byte code
3058      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3059      cd          static compile data
3060      base_list   the data list of the base opcode
3061    
3062    Returns:      TRUE if the auto-possessification is possible
3063    */
3064    
3065    static BOOL
3066    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3067      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3068    {
3069    pcre_uchar c;
3070    pcre_uint32 list[8];
3071    const pcre_uint32 *chr_ptr;
3072    const pcre_uint32 *ochr_ptr;
3073    const pcre_uint32 *list_ptr;
3074    const pcre_uchar *next_code;
3075    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3076    const pcre_uchar *xclass_flags;
3077    #endif
3078    const pcre_uint8 *class_bitset;
3079    const pcre_uint8 *set1, *set2, *set_end;
3080    pcre_uint32 chr;
3081    BOOL accepted, invert_bits;
3082    
3083    /* Note: the base_list[1] contains whether the current opcode has greedy
3084    (represented by a non-zero value) quantifier. This is a different from
3085    other character type lists, which stores here that the character iterator
3086    matches to an empty string (also represented by a non-zero value). */
3087    
3088    for(;;)
3089      {
3090      /* All operations move the code pointer forward.
3091      Therefore infinite recursions are not possible. */
3092    
3093      c = *code;
3094    
3095      /* Skip over callouts */
3096    
3097      if (c == OP_CALLOUT)
3098        {
3099        code += PRIV(OP_lengths)[c];
3100        continue;
3101        }
3102    
3103      if (c == OP_ALT)
3104        {
3105        do code += GET(code, 1); while (*code == OP_ALT);
3106        c = *code;
3107        }
3108    
3109      switch(c)
3110        {
3111        case OP_END:
3112        case OP_KETRPOS:
3113        /* TRUE only in greedy case. The non-greedy case could be replaced by
3114        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3115        uses more memory, which we cannot get at this stage.) */
3116    
3117        return base_list[1] != 0;
3118    
3119        case OP_KET:
3120        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3121        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3122        cannot be converted to a possessive form. */
3123    
3124        if (base_list[1] == 0) return FALSE;
3125    
3126        switch(*(code - GET(code, 1)))
3127          {
3128          case OP_ASSERT:
3129          case OP_ASSERT_NOT:
3130          case OP_ASSERTBACK:
3131          case OP_ASSERTBACK_NOT:
3132          case OP_ONCE:
3133          case OP_ONCE_NC:
3134          /* Atomic sub-patterns and assertions can always auto-possessify their
3135          last iterator. */
3136          return TRUE;
3137          }
3138    
3139        code += PRIV(OP_lengths)[c];
3140        continue;
3141    
3142        case OP_ONCE:
3143        case OP_ONCE_NC:
3144        case OP_BRA:
3145        case OP_CBRA:
3146        next_code = code + GET(code, 1);
3147        code += PRIV(OP_lengths)[c];
3148    
3149        while (*next_code == OP_ALT)
3150          {
3151          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3152          code = next_code + 1 + LINK_SIZE;
3153          next_code += GET(next_code, 1);
3154          }
3155        continue;
3156    
3157        case OP_BRAZERO:
3158        case OP_BRAMINZERO:
3159    
3160        next_code = code + 1;
3161        if (*next_code != OP_BRA && *next_code != OP_CBRA
3162            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3163    
3164        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3165    
3166        /* The bracket content will be checked by the
3167        OP_BRA/OP_CBRA case above. */
3168        next_code += 1 + LINK_SIZE;
3169        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3170          return FALSE;
3171    
3172        code += PRIV(OP_lengths)[c];
3173        continue;
3174        }
3175    
3176      /* Check for a supported opcode, and load its properties. */
3177    
3178      code = get_chr_property_list(code, utf, cd->fcc, list);
3179      if (code == NULL) return FALSE;    /* Unsupported */
3180    
3181      /* If either opcode is a small character list, set pointers for comparing
3182      characters from that list with another list, or with a property. */
3183    
3184      if (base_list[0] == OP_CHAR)
3185        {
3186        chr_ptr = base_list + 2;
3187        list_ptr = list;
3188        }
3189      else if (list[0] == OP_CHAR)
3190        {
3191        chr_ptr = list + 2;
3192        list_ptr = base_list;
3193        }
3194    
3195      /* Character bitsets can also be compared to certain opcodes. */
3196    
3197      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3198    #ifdef COMPILE_PCRE8
3199          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3200          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3201    #endif
3202          )
3203        {
3204    #ifdef COMPILE_PCRE8
3205        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3206    #else
3207        if (base_list[0] == OP_CLASS)
3208    #endif
3209          {
3210          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3211          list_ptr = list;
3212          }
3213        else
3214          {
3215          set1 = (pcre_uint8 *)(code - list[2]);
3216          list_ptr = base_list;
3217          }
3218    
3219        invert_bits = FALSE;
3220        switch(list_ptr[0])
3221          {
3222          case OP_CLASS:
3223          case OP_NCLASS:
3224          set2 = (pcre_uint8 *)
3225            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3226          break;
3227    
3228    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3229          case OP_XCLASS:
3230          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3231          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3232          if ((*xclass_flags & XCL_MAP) == 0)
3233            {
3234            /* No bits are set for characters < 256. */
3235            if (list[1] == 0) return TRUE;
3236            /* Might be an empty repeat. */
3237            continue;
3238            }
3239          set2 = (pcre_uint8 *)(xclass_flags + 1);
3240          break;
3241    #endif
3242    
3243          case OP_NOT_DIGIT:
3244          invert_bits = TRUE;
3245          /* Fall through */
3246          case OP_DIGIT:
3247          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3248          break;
3249    
3250          case OP_NOT_WHITESPACE:
3251          invert_bits = TRUE;
3252          /* Fall through */
3253          case OP_WHITESPACE:
3254          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3255          break;
3256    
3257          case OP_NOT_WORDCHAR:
3258          invert_bits = TRUE;
3259          /* Fall through */
3260          case OP_WORDCHAR:
3261          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3262          break;
3263    
3264          default:
3265          return FALSE;
3266          }
3267    
3268        /* Because the sets are unaligned, we need
3269        to perform byte comparison here. */
3270        set_end = set1 + 32;
3271        if (invert_bits)
3272          {
3273          do
3274            {
3275            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3276            }
3277          while (set1 < set_end);
3278          }
3279        else
3280          {
3281          do
3282            {
3283            if ((*set1++ & *set2++) != 0) return FALSE;
3284            }
3285          while (set1 < set_end);
3286          }
3287    
3288        if (list[1] == 0) return TRUE;
3289        /* Might be an empty repeat. */
3290        continue;
3291        }
3292    
3293      /* Some property combinations also acceptable. Unicode property opcodes are
3294      processed specially; the rest can be handled with a lookup table. */
3295    
3296      else
3297        {
3298        pcre_uint32 leftop, rightop;
3299    
3300        leftop = base_list[0];
3301        rightop = list[0];
3302    
3303    #ifdef SUPPORT_UCP
3304        accepted = FALSE; /* Always set in non-unicode case. */
3305        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3306          {
3307          if (rightop == OP_EOD)
3308            accepted = TRUE;
3309          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3310            {
3311            int n;
3312            const pcre_uint8 *p;
3313            BOOL same = leftop == rightop;
3314            BOOL lisprop = leftop == OP_PROP;
3315            BOOL risprop = rightop == OP_PROP;
3316            BOOL bothprop = lisprop && risprop;
3317    
3318            /* There's a table that specifies how each combination is to be
3319            processed:
3320              0   Always return FALSE (never auto-possessify)
3321              1   Character groups are distinct (possessify if both are OP_PROP)
3322              2   Check character categories in the same group (general or particular)
3323              3   Return TRUE if the two opcodes are not the same
3324              ... see comments below
3325            */
3326    
3327            n = propposstab[base_list[2]][list[2]];
3328            switch(n)
3329              {
3330              case 0: break;
3331              case 1: accepted = bothprop; break;
3332              case 2: accepted = (base_list[3] == list[3]) != same; break;
3333              case 3: accepted = !same; break;
3334    
3335              case 4:  /* Left general category, right particular category */
3336              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3337              break;
3338    
3339              case 5:  /* Right general category, left particular category */
3340              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3341              break;
3342    
3343              /* This code is logically tricky. Think hard before fiddling with it.
3344              The posspropstab table has four entries per row. Each row relates to
3345              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3346              Only WORD actually needs all four entries, but using repeats for the
3347              others means they can all use the same code below.
3348    
3349              The first two entries in each row are Unicode general categories, and
3350              apply always, because all the characters they include are part of the
3351              PCRE character set. The third and fourth entries are a general and a
3352              particular category, respectively, that include one or more relevant
3353              characters. One or the other is used, depending on whether the check
3354              is for a general or a particular category. However, in both cases the
3355              category contains more characters than the specials that are defined
3356              for the property being tested against. Therefore, it cannot be used
3357              in a NOTPROP case.
3358    
3359              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3360              Underscore is covered by ucp_P or ucp_Po. */
3361    
3362              case 6:  /* Left alphanum vs right general category */
3363              case 7:  /* Left space vs right general category */
3364              case 8:  /* Left word vs right general category */
3365              p = posspropstab[n-6];
3366              accepted = risprop && lisprop ==
3367                (list[3] != p[0] &&
3368                 list[3] != p[1] &&
3369                (list[3] != p[2] || !lisprop));
3370              break;
3371    
3372              case 9:   /* Right alphanum vs left general category */
3373              case 10:  /* Right space vs left general category */
3374              case 11:  /* Right word vs left general category */
3375              p = posspropstab[n-9];
3376              accepted = lisprop && risprop ==
3377                (base_list[3] != p[0] &&
3378                 base_list[3] != p[1] &&
3379                (base_list[3] != p[2] || !risprop));
3380              break;
3381    
3382              case 12:  /* Left alphanum vs right particular category */
3383              case 13:  /* Left space vs right particular category */
3384              case 14:  /* Left word vs right particular category */
3385              p = posspropstab[n-12];
3386              accepted = risprop && lisprop ==
3387                (catposstab[p[0]][list[3]] &&
3388                 catposstab[p[1]][list[3]] &&
3389                (list[3] != p[3] || !lisprop));
3390              break;
3391    
3392              case 15:  /* Right alphanum vs left particular category */
3393              case 16:  /* Right space vs left particular category */
3394              case 17:  /* Right word vs left particular category */
3395              p = posspropstab[n-15];
3396              accepted = lisprop && risprop ==
3397                (catposstab[p[0]][base_list[3]] &&
3398                 catposstab[p[1]][base_list[3]] &&
3399                (base_list[3] != p[3] || !risprop));
3400              break;
3401              }
3402            }
3403          }
3404    
3405        else
3406    #endif  /* SUPPORT_UCP */
3407    
3408        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3409               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3410               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3411    
3412        if (!accepted)
3413          return FALSE;
3414    
3415        if (list[1] == 0) return TRUE;
3416        /* Might be an empty repeat. */
3417        continue;
3418        }
3419    
3420      /* Control reaches here only if one of the items is a small character list.
3421      All characters are checked against the other side. */
3422    
3423      do
3424        {
3425        chr = *chr_ptr;
3426    
3427        switch(list_ptr[0])
3428          {
3429          case OP_CHAR:
3430          ochr_ptr = list_ptr + 2;
3431          do
3432            {
3433            if (chr == *ochr_ptr) return FALSE;
3434            ochr_ptr++;
3435            }
3436          while(*ochr_ptr != NOTACHAR);
3437          break;
3438    
3439          case OP_NOT:
3440          ochr_ptr = list_ptr + 2;
3441          do
3442            {
3443            if (chr == *ochr_ptr)
3444              break;
3445            ochr_ptr++;
3446            }
3447          while(*ochr_ptr != NOTACHAR);
3448          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3449          break;
3450    
3451          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3452          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3453    
3454          case OP_DIGIT:
3455          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3456          break;
3457    
3458          case OP_NOT_DIGIT:
3459          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3460          break;
3461    
3462          case OP_WHITESPACE:
3463          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3464          break;
3465    
3466          case OP_NOT_WHITESPACE:
3467          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3468          break;
3469    
3470          case OP_WORDCHAR:
3471          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3472          break;
3473    
3474          case OP_NOT_WORDCHAR:
3475          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3476          break;
3477    
3478          case OP_HSPACE:
3479          switch(chr)
3480            {
3481            HSPACE_CASES: return FALSE;
3482            default: break;
3483            }
3484          break;
3485    
3486          case OP_NOT_HSPACE:
3487          switch(chr)
3488            {
3489            HSPACE_CASES: break;
3490            default: return FALSE;
3491