/[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 87 by nigel, Sat Feb 24 21:41:21 2007 UTC revision 1393 by ph10, Fri Nov 8 16:37:21 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #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  *************************************************/  *************************************************/
96    
97  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
98  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
99  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
100  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
101  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
102    so this number is very generous.
103    
104    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
106    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107    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 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  #define BRASTACK_SIZE 200  /* Repeated character flags. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
144  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
145  is invalid. */  is invalid. */
146    
147  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
148    
149    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150    in UTF-8 mode. */
151    
152  static const short int escapes[] = {  static const short int escapes[] = {
153       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
154       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
155     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
156       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q,      0, -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,      0,      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 96  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,      0,     '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,     0,      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 115  static const short int escapes[] = { Line 222  static const short int escapes[] = {
222  #endif  #endif
223    
224    
225  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
227  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
228    string is built from string macros so that it works in UTF-8 mode on EBCDIC
229  static const char *const posix_names[] = {  platforms. */
230    "alpha", "lower", "upper",  
231    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
232    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
233      int   op;                  /* Op when no arg, or -1 if arg mandatory */
234      int   op_arg;              /* Op when arg present, or -1 if not allowed */
235    } verbitem;
236    
237    static const char verbnames[] =
238      "\0"                       /* Empty name is a shorthand for MARK */
239      STRING_MARK0
240      STRING_ACCEPT0
241      STRING_COMMIT0
242      STRING_F0
243      STRING_FAIL0
244      STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248    static const verbitem verbs[] = {
249      { 0, -1,        OP_MARK },
250      { 4, -1,        OP_MARK },
251      { 6, OP_ACCEPT, -1 },
252      { 6, OP_COMMIT, -1 },
253      { 1, OP_FAIL,   -1 },
254      { 4, OP_FAIL,   -1 },
255      { 5, OP_PRUNE,  OP_PRUNE_ARG },
256      { 4, OP_SKIP,   OP_SKIP_ARG  },
257      { 4, OP_THEN,   OP_THEN_ARG  }
258    };
259    
260    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Tables of names of POSIX character classes and their lengths. The names are
264    now all in a single string, to reduce the number of relocations when a shared
265    library is dynamically loaded. The list of lengths is terminated by a zero
266    length entry. The first three must be alpha, lower, upper, as this is assumed
267    for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270    static const char posix_names[] =
271      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
272      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
273      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
274      STRING_word0  STRING_xdigit;
275    
276  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
285  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
286  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 154  static const int posix_class_maps[] = { Line 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312    Unicode property escapes. */
313    
314  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
315  are passed to the outside world. */  static const pcre_uchar string_PNd[]  = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pNd[]  = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    static const pcre_uchar string_PXsp[] = {
322      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
323      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324    static const pcre_uchar string_pXsp[] = {
325      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
326      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
327    static const pcre_uchar string_PXwd[] = {
328      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_pXwd[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    
334    static const pcre_uchar *substitutes[] = {
335      string_PNd,           /* \D */
336      string_pNd,           /* \d */
337      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339      string_PXwd,          /* \W */
340      string_pXwd           /* \w */
341    };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349    static const pcre_uchar string_pL[] =   {
350      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352    static const pcre_uchar string_pLl[] =  {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_pLu[] =  {
356      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_pXan[] = {
359      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_h[] =    {
362      CHAR_BACKSLASH, CHAR_h, '\0' };
363    static const pcre_uchar string_pXps[] = {
364      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
365      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
366    static const pcre_uchar string_PL[] =   {
367      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
368      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
369    static const pcre_uchar string_PLl[] =  {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    static const pcre_uchar string_PLu[] =  {
373      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
374      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
375    static const pcre_uchar string_PXan[] = {
376      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
377      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
378    static const pcre_uchar string_H[] =    {
379      CHAR_BACKSLASH, CHAR_H, '\0' };
380    static const pcre_uchar string_PXps[] = {
381      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
382      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
383    
384    static const pcre_uchar *posix_substitutes[] = {
385      string_pL,            /* alpha */
386      string_pLl,           /* lower */
387      string_pLu,           /* upper */
388      string_pXan,          /* alnum */
389      NULL,                 /* ascii */
390      string_h,             /* blank */
391      NULL,                 /* cntrl */
392      string_pNd,           /* digit */
393      NULL,                 /* graph */
394      NULL,                 /* print */
395      NULL,                 /* punct */
396      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398      NULL,                 /* xdigit */
399      /* Negated cases */
400      string_PL,            /* ^alpha */
401      string_PLl,           /* ^lower */
402      string_PLu,           /* ^upper */
403      string_PXan,          /* ^alnum */
404      NULL,                 /* ^ascii */
405      string_H,             /* ^blank */
406      NULL,                 /* ^cntrl */
407      string_PNd,           /* ^digit */
408      NULL,                 /* ^graph */
409      NULL,                 /* ^print */
410      NULL,                 /* ^punct */
411      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413      NULL                  /* ^xdigit */
414    };
415    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
416    #endif
417    
418  static const char *error_texts[] = {  #define STRING(a)  # a
419    "no error",  #define XSTRING(s) STRING(s)
420    "\\ at end of pattern",  
421    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
422    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
423    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
424    are no longer used. This used to be a table of strings, but in order to reduce
425    the number of relocations needed when a shared library is loaded dynamically,
426    it is now one long string. We cannot use a table of offsets, because the
427    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
428    simply count through to the one we want - this isn't a performance issue
429    because these strings are used only when there is a compilation error.
430    
431    Each substring ends with \0 to insert a null character. This includes the final
432    substring, so that the whole string ends with \0\0, which can be detected when
433    counting through. */
434    
435    static const char error_texts[] =
436      "no error\0"
437      "\\ at end of pattern\0"
438      "\\c at end of pattern\0"
439      "unrecognized character follows \\\0"
440      "numbers out of order in {} quantifier\0"
441    /* 5 */    /* 5 */
442    "number too big in {} quantifier",    "number too big in {} quantifier\0"
443    "missing terminating ] for character class",    "missing terminating ] for character class\0"
444    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
445    "range out of order in character class",    "range out of order in character class\0"
446    "nothing to repeat",    "nothing to repeat\0"
447    /* 10 */    /* 10 */
448    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
449    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
450    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
451    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
452    "missing )",    "missing )\0"
453    /* 15 */    /* 15 */
454    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
455    "erroffset passed as NULL",    "erroffset passed as NULL\0"
456    "unknown option bit(s) set",    "unknown option bit(s) set\0"
457    "missing ) after comment",    "missing ) after comment\0"
458    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
459    /* 20 */    /* 20 */
460    "regular expression too large",    "regular expression is too large\0"
461    "failed to get memory",    "failed to get memory\0"
462    "unmatched parentheses",    "unmatched parentheses\0"
463    "internal error: code overflow",    "internal error: code overflow\0"
464    "unrecognized character after (?<",    "unrecognized character after (?<\0"
465    /* 25 */    /* 25 */
466    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
467    "malformed number after (?(",    "malformed number or name after (?(\0"
468    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
469    "assertion expected after (?(",    "assertion expected after (?(\0"
470    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
471    /* 30 */    /* 30 */
472    "unknown POSIX class name",    "unknown POSIX class name\0"
473    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
474    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
475    "spare error",    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large",    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)",    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
480    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
481    "number after (?C is > 255",    "number after (?C is > 255\0"
482    "closing ) for (?C expected",    "closing ) for (?C expected\0"
483    /* 40 */    /* 40 */
484    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
485    "unrecognized character after (?P",    "unrecognized character after (?P\0"
486    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
487    "two named groups have the same name",    "two named subpatterns have the same name\0"
488    "invalid UTF-8 string",    "invalid UTF-8 string\0"
489    /* 45 */    /* 45 */
490    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
491    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
492    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
493  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
494      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
495      /* 50 */
496      "repeated subpattern is too long\0"    /** DEAD **/
497      "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
498      "internal error: overran compiling workspace\0"
499      "internal error: previously-checked referenced subpattern not found\0"
500      "DEFINE group contains more than one branch\0"
501      /* 55 */
502      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
503      "inconsistent NEWLINE options\0"
504      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
505      "a numbered reference must not be zero\0"
506      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507      /* 60 */
508      "(*VERB) not recognized or malformed\0"
509      "number is too big\0"
510      "subpattern name expected\0"
511      "digit expected after (?+\0"
512      "] is an invalid data character in JavaScript compatibility mode\0"
513      /* 65 */
514      "different names for subpatterns of the same number are not allowed\0"
515      "(*MARK) must have an argument\0"
516      "this version of PCRE is not compiled with Unicode property support\0"
517      "\\c must be followed by an ASCII character\0"
518      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
519      /* 70 */
520      "internal error: unknown opcode in find_fixedlength()\0"
521      "\\N is not supported in a class\0"
522      "too many forward references\0"
523      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524      "invalid UTF-16 string\0"
525      /* 75 */
526      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527      "character value in \\u.... sequence is too large\0"
528      "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      ;
537    
538  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
539  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 235  For convenience, we use the same bit def Line 551  For convenience, we use the same bit def
551    
552  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
553    
554  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
555  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
556    into a subtraction and unsigned comparison). */
557    
558    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
559    
560    #ifndef EBCDIC
561    
562    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
563    UTF-8 mode. */
564    
565    static const pcre_uint8 digitab[] =
566    {    {
567    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
568    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 271  static const unsigned char digitab[] = Line 597  static const unsigned char digitab[] =
597    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
598    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
599    
600  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
601  static const unsigned char digitab[] =  
602    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
603    
604    static const pcre_uint8 digitab[] =
605    {    {
606    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
607    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 285  static const unsigned char digitab[] = Line 614  static const unsigned char digitab[] =
614    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
615    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
616    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
617    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
618    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
619    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
620    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 307  static const unsigned char digitab[] = Line 636  static const unsigned char digitab[] =
636    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
637    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
638    
639  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
640    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
641    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
642    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 319  static const unsigned char ebcdic_charta Line 648  static const unsigned char ebcdic_charta
648    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
649    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
650    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
651    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
652    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
653    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
654    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 343  static const unsigned char ebcdic_charta Line 672  static const unsigned char ebcdic_charta
672  #endif  #endif
673    
674    
675  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
676    between adjacent character-type opcodes. The left-hand (repeated) opcode is
677    used to select the row, and the right-hand opcode is use to select the column.
678    A value of 1 means that auto-possessification is OK. For example, the second
679    value in the first row means that \D+\d can be turned into \D++\d.
680    
681    The Unicode property types (\P and \p) have to be present to fill out the table
682    because of what their opcode values are, but the table values should always be
683    zero because property types are handled separately in the code. The last four
684    columns apply to items that cannot be repeated, so there is no need to have
685    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
686    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
687    
688    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
689    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
690    
691    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
692    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
693      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
694      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
695      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
696      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
697      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
698      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
699      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
704      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
705      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
706      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
707      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
708      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
709      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
710    };
711    
712    
713    /* This table is used to check whether auto-possessification is possible
714    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
715    left-hand (repeated) opcode is used to select the row, and the right-hand
716    opcode is used to select the column. The values are as follows:
717    
718      0   Always return FALSE (never auto-possessify)
719      1   Character groups are distinct (possessify if both are OP_PROP)
720      2   Check character categories in the same group (general or particular)
721      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
722    
723      4   Check left general category vs right particular category
724      5   Check right general category vs left particular category
725    
726      6   Left alphanum vs right general category
727      7   Left space vs right general category
728      8   Left word vs right general category
729    
730      9   Right alphanum vs left general category
731     10   Right space vs left general category
732     11   Right word vs left general category
733    
734     12   Left alphanum vs right particular category
735     13   Left space vs right particular category
736     14   Left word vs right particular category
737    
738     15   Right alphanum vs left particular category
739     16   Right space vs left particular category
740     17   Right word vs left particular category
741    */
742    
743    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
744    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
745      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
746      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
747      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
748      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
749      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
750      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
751      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
753      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
754      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
756    };
757    
758    /* This table is used to check whether auto-possessification is possible
759    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
760    specifies a general category and the other specifies a particular category. The
761    row is selected by the general category and the column by the particular
762    category. The value is 1 if the particular category is not part of the general
763    category. */
764    
765    static const pcre_uint8 catposstab[7][30] = {
766    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
767      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
768      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
769      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
770      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
771      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
772      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
773      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
774    };
775    
776    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
777    a general or particular category. The properties in each row are those
778    that apply to the character set in question. Duplication means that a little
779    unnecessary work is done when checking, but this keeps things much simpler
780    because they can all use the same code. For more details see the comment where
781    this table is used.
782    
783    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
784    "space", but from Perl 5.18 it's included, so both categories are treated the
785    same here. */
786    
787    static const pcre_uint8 posspropstab[3][4] = {
788      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
789      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
790      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
791    };
792    
793    /* This table is used when converting repeating opcodes into possessified
794    versions as a result of an explicit possessive quantifier such as ++. A zero
795    value means there is no possessified version - in those cases the item in
796    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
797    because all relevant opcodes are less than that. */
798    
799    static const pcre_uint8 opcode_possessify[] = {
800      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
802    
803      0,                       /* NOTI */
804      OP_POSSTAR, 0,           /* STAR, MINSTAR */
805      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
806      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
807      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
808      0,                       /* EXACT */
809      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
810    
811      OP_POSSTARI, 0,          /* STARI, MINSTARI */
812      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
813      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
814      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
815      0,                       /* EXACTI */
816      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
817    
818      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
819      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
820      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
821      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
822      0,                       /* NOTEXACT */
823      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
826      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
827      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
828      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
829      0,                       /* NOTEXACTI */
830      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
833      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
834      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
835      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
836      0,                       /* TYPEEXACT */
837      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
840      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
841      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
842      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
843      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
844    
845      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
846      0, 0,                    /* REF, REFI */
847      0, 0,                    /* DNREF, DNREFI */
848      0, 0                     /* RECURSE, CALLOUT */
849    };
850    
851    
852    
853    /*************************************************
854    *            Find an error text                  *
855    *************************************************/
856    
857    /* The error texts are now all in one long string, to save on relocations. As
858    some of the text is of unknown length, we can't use a table of offsets.
859    Instead, just count through the strings. This is not a performance issue
860    because it happens only when there has been a compilation error.
861    
862    Argument:   the error number
863    Returns:    pointer to the error string
864    */
865    
866    static const char *
867    find_error_text(int n)
868    {
869    const char *s = error_texts;
870    for (; n > 0; n--)
871      {
872      while (*s++ != CHAR_NULL) {};
873      if (*s == CHAR_NULL) return "Error text not found (please report)";
874      }
875    return s;
876    }
877    
878    
879    
880    /*************************************************
881    *           Expand the workspace                 *
882    *************************************************/
883    
884    /* This function is called during the second compiling phase, if the number of
885    forward references fills the existing workspace, which is originally a block on
886    the stack. A larger block is obtained from malloc() unless the ultimate limit
887    has been reached or the increase will be rather small.
888    
889    Argument: pointer to the compile data block
890    Returns:  0 if all went well, else an error number
891    */
892    
893    static int
894    expand_workspace(compile_data *cd)
895    {
896    pcre_uchar *newspace;
897    int newsize = cd->workspace_size * 2;
898    
899    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
900    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
901        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
902     return ERR72;
903    
904    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
905    if (newspace == NULL) return ERR21;
906    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
907    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
908    if (cd->workspace_size > COMPILE_WORK_SIZE)
909      (PUBL(free))((void *)cd->start_workspace);
910    cd->start_workspace = newspace;
911    cd->workspace_size = newsize;
912    return 0;
913    }
914    
915    
916    
917    /*************************************************
918    *            Check for counted repeat            *
919    *************************************************/
920    
921    /* This function is called when a '{' is encountered in a place where it might
922    start a quantifier. It looks ahead to see if it really is a quantifier or not.
923    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
924    where the ddds are digits.
925    
926    Arguments:
927      p         pointer to the first char after '{'
928    
929    Returns:    TRUE or FALSE
930    */
931    
932  static BOOL  static BOOL
933    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,  is_counted_repeat(const pcre_uchar *p)
934      int *, int *, branch_chain *, compile_data *);  {
935    if (!IS_DIGIT(*p)) return FALSE;
936    p++;
937    while (IS_DIGIT(*p)) p++;
938    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
939    
940    if (*p++ != CHAR_COMMA) return FALSE;
941    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943    if (!IS_DIGIT(*p)) return FALSE;
944    p++;
945    while (IS_DIGIT(*p)) p++;
946    
947    return (*p == CHAR_RIGHT_CURLY_BRACKET);
948    }
949    
950    
951    
# Line 356  static BOOL Line 954  static BOOL
954  *************************************************/  *************************************************/
955    
956  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
957  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
958  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  will be placed in chptr. A backreference to group n is returned as negative n.
959  a positive value greater than 255 may be returned. On entry, ptr is pointing at  When UTF-8 is enabled, a positive value greater than 255 may be returned in
960  the \. On exit, it is on the final character of the escape sequence.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
961    character of the escape sequence.
962    
963  Arguments:  Arguments:
964    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
965      chptr          points to a returned data character
966    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
967    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
968    options        the options bits    options        the options bits
969    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
970    
971  Returns:         zero or positive => a data character  Returns:         zero => a data character
972                   negative => a special escape sequence                   positive => a special escape sequence
973                   on error, errorptr is set                   negative => a back reference
974                     on error, errorcodeptr is set
975  */  */
976    
977  static int  static int
978  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
979    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
980  {  {
981  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
982  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
983  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
984    pcre_uint32 c;
985    int escape = 0;
986    int i;
987    
988  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
989  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
990    
991  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
992    
993  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
994    
995  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
996  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
997  Otherwise further processing may be required. */  Otherwise further processing may be required. */
998    
999  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1000  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1001  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1002    else if ((i = escapes[c - CHAR_0]) != 0)
1003  #else          /* EBCDIC coding */    { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1004  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  
1005  else if ((i = escapes[c - 0x48]) != 0)  c = i;  #else           /* EBCDIC coding */
1006    /* Not alphanumeric */
1007    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1008    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1009  #endif  #endif
1010    
1011  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1012    
1013  else  else
1014    {    {
1015    const uschar *oldptr;    const pcre_uchar *oldptr;
1016      BOOL braced, negated, overflow;
1017      int s;
1018    
1019    switch (c)    switch (c)
1020      {      {
1021      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
1022      error. */      error. */
1023    
1024      case 'l':      case CHAR_l:
1025      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1026      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1027      break;      break;
1028    
1029        case CHAR_u:
1030        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1031          {
1032          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1033          Otherwise it is a lowercase u letter. */
1034          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1035            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1037            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1038            {
1039            c = 0;
1040            for (i = 0; i < 4; ++i)
1041              {
1042              register pcre_uint32 cc = *(++ptr);
1043    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1044              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1045              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1046    #else           /* EBCDIC coding */
1047              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1048              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1049    #endif
1050              }
1051    
1052    #if defined COMPILE_PCRE8
1053            if (c > (utf ? 0x10ffffU : 0xffU))
1054    #elif defined COMPILE_PCRE16
1055            if (c > (utf ? 0x10ffffU : 0xffffU))
1056    #elif defined COMPILE_PCRE32
1057            if (utf && c > 0x10ffffU)
1058    #endif
1059              {
1060              *errorcodeptr = ERR76;
1061              }
1062            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1063            }
1064          }
1065        else
1066          *errorcodeptr = ERR37;
1067        break;
1068    
1069        case CHAR_U:
1070        /* In JavaScript, \U is an uppercase U letter. */
1071        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1072        break;
1073    
1074        /* In a character class, \g is just a literal "g". Outside a character
1075        class, \g must be followed by one of a number of specific things:
1076    
1077        (1) A number, either plain or braced. If positive, it is an absolute
1078        backreference. If negative, it is a relative backreference. This is a Perl
1079        5.10 feature.
1080    
1081        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1082        is part of Perl's movement towards a unified syntax for back references. As
1083        this is synonymous with \k{name}, we fudge it up by pretending it really
1084        was \k.
1085    
1086        (3) For Oniguruma compatibility we also support \g followed by a name or a
1087        number either in angle brackets or in single quotes. However, these are
1088        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1089        the ESC_g code (cf \k). */
1090    
1091        case CHAR_g:
1092        if (isclass) break;
1093        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1094          {
1095          escape = ESC_g;
1096          break;
1097          }
1098    
1099        /* Handle the Perl-compatible cases */
1100    
1101        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1102          {
1103          const pcre_uchar *p;
1104          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1105            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1106          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1107            {
1108            escape = ESC_k;
1109            break;
1110            }
1111          braced = TRUE;
1112          ptr++;
1113          }
1114        else braced = FALSE;
1115    
1116        if (ptr[1] == CHAR_MINUS)
1117          {
1118          negated = TRUE;
1119          ptr++;
1120          }
1121        else negated = FALSE;
1122    
1123        /* The integer range is limited by the machine's int representation. */
1124        s = 0;
1125        overflow = FALSE;
1126        while (IS_DIGIT(ptr[1]))
1127          {
1128          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1129            {
1130            overflow = TRUE;
1131            break;
1132            }
1133          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1134          }
1135        if (overflow) /* Integer overflow */
1136          {
1137          while (IS_DIGIT(ptr[1]))
1138            ptr++;
1139          *errorcodeptr = ERR61;
1140          break;
1141          }
1142    
1143        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1144          {
1145          *errorcodeptr = ERR57;
1146          break;
1147          }
1148    
1149        if (s == 0)
1150          {
1151          *errorcodeptr = ERR58;
1152          break;
1153          }
1154    
1155        if (negated)
1156          {
1157          if (s > bracount)
1158            {
1159            *errorcodeptr = ERR15;
1160            break;
1161            }
1162          s = bracount - (s - 1);
1163          }
1164    
1165        escape = -s;
1166        break;
1167    
1168      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1169      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1170      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1171        recommended to avoid the ambiguities in the old syntax.
1172    
1173      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1174      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1175      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1176      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1177      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1178      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1179      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1180    
1181        Inside a character class, \ followed by a digit is always either a literal
1182        8 or 9 or an octal number. */
1183    
1184      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:
1185      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1186    
1187      if (!isclass)      if (!isclass)
1188        {        {
1189        oldptr = ptr;        oldptr = ptr;
1190        c -= '0';        /* The integer range is limited by the machine's int representation. */
1191        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1192          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1193        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1194          {          {
1195          c = -(ESC_REF + c);          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1196              {
1197              overflow = TRUE;
1198              break;
1199              }
1200            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1201            }
1202          if (overflow) /* Integer overflow */
1203            {
1204            while (IS_DIGIT(ptr[1]))
1205              ptr++;
1206            *errorcodeptr = ERR61;
1207            break;
1208            }
1209          if (s < 8 || s <= bracount)  /* Check for back reference */
1210            {
1211            escape = -s;
1212          break;          break;
1213          }          }
1214        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1215        }        }
1216    
1217      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1218      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1219      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1220        changed so as not to insert the binary zero. */
1221    
1222      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
       {  
       ptr--;  
       c = 0;  
       break;  
       }  
1223    
1224      /* \0 always starts an octal number, but we may drop through to here with a      /* Fall through with a digit less than 8 */
     larger first octal digit. */  
1225    
1226      case '0':      /* \0 always starts an octal number, but we may drop through to here with a
1227      c -= '0';      larger first octal digit. The original code used just to take the least
1228      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      significant 8 bits of octal numbers (I think this is what early Perls used
1229          c = c * 8 + *(++ptr) - '0';      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1230      c &= 255;     /* Take least significant 8 bits */      but no more than 3 octal digits. */
1231    
1232        case CHAR_0:
1233        c -= CHAR_0;
1234        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1235            c = c * 8 + *(++ptr) - CHAR_0;
1236    #ifdef COMPILE_PCRE8
1237        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1238    #endif
1239      break;      break;
1240    
1241      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1242      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1243    
1244      case 'x':      case CHAR_o:
1245      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1246        {        {
1247        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1248        c = 0;        c = 0;
1249        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1250          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1251          {          {
1252          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1253          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1254          count++;  #ifdef COMPILE_PCRE32
1255            if (c >= 0x20000000l) { overflow = TRUE; break; }
1256  #if !EBCDIC    /* ASCII coding */  #endif
1257          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1258          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1259  #else          /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1260          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1261          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1262    #elif defined COMPILE_PCRE32
1263            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1264  #endif  #endif
1265          }          }
1266          if (overflow)
       if (*pt == '}')  
1267          {          {
1268          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1269          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1270          }          }
1271          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1272        /* If the sequence of hex digits does not end with '}', then we don't          {
1273        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1274            }
1275          else *errorcodeptr = ERR80;
1276        }        }
1277        break;
1278    
1279      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1280        numbers. Otherwise it is a lowercase x letter. */
1281    
1282      c = 0;      case CHAR_x:
1283      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1284        {        {
1285        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1286        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1287  #if !EBCDIC    /* ASCII coding */          {
1288        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1289        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1290  #else          /* EBCDIC coding */            {
1291        if (cc <= 'z') cc += 64;              /* Convert to upper case */            register pcre_uint32 cc = *(++ptr);
1292        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1293              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1294              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1295    #else           /* EBCDIC coding */
1296              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1297              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1298  #endif  #endif
1299        }            }
1300            }
1301          }    /* End JavaScript handling */
1302    
1303        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1304        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1305        digits. If not, { used to be treated as a data character. However, Perl
1306        seems to read hex digits up to the first non-such, and ignore the rest, so
1307        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1308        now gives an error. */
1309    
1310        else
1311          {
1312          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1313            {
1314            ptr += 2;
1315            c = 0;
1316            overflow = FALSE;
1317            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1318              {
1319              register pcre_uint32 cc = *ptr++;
1320              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1321    
1322    #ifdef COMPILE_PCRE32
1323              if (c >= 0x10000000l) { overflow = TRUE; break; }
1324    #endif
1325    
1326    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1327              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1328              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1329    #else           /* EBCDIC coding */
1330              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1331              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1332    #endif
1333    
1334    #if defined COMPILE_PCRE8
1335              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1336    #elif defined COMPILE_PCRE16
1337              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1338    #elif defined COMPILE_PCRE32
1339              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1340    #endif
1341              }
1342    
1343            if (overflow)
1344              {
1345              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1346              *errorcodeptr = ERR34;
1347              }
1348    
1349            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1350              {
1351              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1352              }
1353    
1354            /* If the sequence of hex digits does not end with '}', give an error.
1355            We used just to recognize this construct and fall through to the normal
1356            \x handling, but nowadays Perl gives an error, which seems much more
1357            sensible, so we do too. */
1358    
1359            else *errorcodeptr = ERR79;
1360            }   /* End of \x{} processing */
1361    
1362          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1363    
1364          else
1365            {
1366            c = 0;
1367            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1368              {
1369              pcre_uint32 cc;                          /* Some compilers don't like */
1370              cc = *(++ptr);                           /* ++ in initializers */
1371    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1372              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1373              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1374    #else           /* EBCDIC coding */
1375              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1376              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1377    #endif
1378              }
1379            }     /* End of \xdd handling */
1380          }       /* End of Perl-style \x handling */
1381      break;      break;
1382    
1383      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1384        An error is given if the byte following \c is not an ASCII character. This
1385        coding is ASCII-specific, but then the whole concept of \cx is
1386        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1387    
1388      case 'c':      case CHAR_c:
1389      c = *(++ptr);      c = *(++ptr);
1390      if (c == 0)      if (c == CHAR_NULL)
1391        {        {
1392        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1393        return 0;        break;
1394        }        }
1395    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1396      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
1397      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
1398      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
1399          break;
1400  #if !EBCDIC    /* ASCII coding */        }
1401      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1402      c ^= 0x40;      c ^= 0x40;
1403  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1404      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1405      c ^= 0xC0;      c ^= 0xC0;
1406  #endif  #endif
1407      break;      break;
1408    
1409      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1410      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1411      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
1412      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
1413      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1414    
1415      default:      default:
1416      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 563  else Line 1423  else
1423      }      }
1424    }    }
1425    
1426    /* Perl supports \N{name} for character names, as well as plain \N for "not
1427    newline". PCRE does not support \N{name}. However, it does support
1428    quantification such as \N{2,3}. */
1429    
1430    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1431         !is_counted_repeat(ptr+2))
1432      *errorcodeptr = ERR37;
1433    
1434    /* If PCRE_UCP is set, we change the values for \d etc. */
1435    
1436    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1437      escape += (ESC_DU - ESC_D);
1438    
1439    /* Set the pointer to the final character before returning. */
1440    
1441  *ptrptr = ptr;  *ptrptr = ptr;
1442  return c;  *chptr = c;
1443    return escape;
1444  }  }
1445    
1446    
# Line 582  escape sequence. Line 1458  escape sequence.
1458  Argument:  Argument:
1459    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1460    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1461    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1462      pdataptr       points to an unsigned int that is set to the detailed property value
1463    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1464    
1465  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1466  */  */
1467    
1468  static int  static BOOL
1469  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1470      unsigned int *pdataptr, int *errorcodeptr)
1471  {  {
1472  int c, i, bot, top;  pcre_uchar c;
1473  const uschar *ptr = *ptrptr;  int i, bot, top;
1474  char name[32];  const pcre_uchar *ptr = *ptrptr;
1475    pcre_uchar name[32];
1476    
1477  c = *(++ptr);  c = *(++ptr);
1478  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1479    
1480  *negptr = FALSE;  *negptr = FALSE;
1481    
1482  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1483  negation. */  negation. */
1484    
1485  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1486    {    {
1487    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1488      {      {
1489      *negptr = TRUE;      *negptr = TRUE;
1490      ptr++;      ptr++;
1491      }      }
1492    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1493      {      {
1494      c = *(++ptr);      c = *(++ptr);
1495      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1496      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1497      name[i] = c;      name[i] = c;
1498      }      }
1499    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1500    name[i] = 0;    name[i] = 0;
1501    }    }
1502    
# Line 634  else Line 1513  else
1513  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1514    
1515  bot = 0;  bot = 0;
1516  top = _pcre_utt_size;  top = PRIV(utt_size);
1517    
1518  while (bot < top)  while (bot < top)
1519    {    {
1520      int r;
1521    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1522    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1523    if (c == 0)    if (r == 0)
1524      {      {
1525      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1526      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1527        return TRUE;
1528      }      }
1529    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1530    }    }
1531    
1532  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1533  *ptrptr = ptr;  *ptrptr = ptr;
1534  return -1;  return FALSE;
1535    
1536  ERROR_RETURN:  ERROR_RETURN:
1537  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1538  *ptrptr = ptr;  *ptrptr = ptr;
1539  return -1;  return FALSE;
1540  }  }
1541  #endif  #endif
1542    
1543    
1544    
   
1545  /*************************************************  /*************************************************
1546  *            Check for counted repeat            *  *         Read repeat counts                     *
1547  *************************************************/  *************************************************/
1548    
1549  /* This function is called when a '{' is encountered in a place where it might  /* Read an item of the form {n,m} and return the values. This is called only
1550  start a quantifier. It looks ahead to see if it really is a quantifier or not.  after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1551  It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  so the syntax is guaranteed to be correct, but we need to check the values.
 where the ddds are digits.  
1552    
1553  Arguments:  Arguments:
1554    p         pointer to the first char after '{'    p              pointer to first char after '{'
1555      minp           pointer to int for min
1556  Returns:    TRUE or FALSE    maxp           pointer to int for max
1557  */                   returned as -1 if no max
1558      errorcodeptr   points to error code variable
 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 == '}');  
 }  
   
   
   
 /*************************************************  
 *         Read repeat counts                     *  
 *************************************************/  
   
 /* Read an item of the form {n,m} and return the values. This is called only  
 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,  
 so the syntax is guaranteed to be correct, but we need to check the values.  
   
 Arguments:  
   p              pointer to first char after '{'  
   minp           pointer to int for min  
   maxp           pointer to int for max  
                  returned as -1 if no max  
   errorcodeptr   points to error code variable  
1559    
1560  Returns:         pointer to '}' on success;  Returns:         pointer to '}' on success;
1561                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1562  */  */
1563    
1564  static const uschar *  static const pcre_uchar *
1565  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1566  {  {
1567  int min = 0;  int min = 0;
1568  int max = -1;  int max = -1;
# Line 723  int max = -1; Line 1570  int max = -1;
1570  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1571  an integer overflow. */  an integer overflow. */
1572    
1573  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1574  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1575    {    {
1576    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 733  if (min < 0 || min > 65535) Line 1580  if (min < 0 || min > 65535)
1580  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1581  Also, max must not be less than min. */  Also, max must not be less than min. */
1582    
1583  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1584    {    {
1585    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1586      {      {
1587      max = 0;      max = 0;
1588      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1589      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1590        {        {
1591        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 768  return p; Line 1615  return p;
1615    
1616  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1617  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1618  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1619  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1620  assertions, and also the \b assertion; for others it does not.  does not.
1621    
1622  Arguments:  Arguments:
1623    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  
1624    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1625    
1626  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1627  */  */
1628    
1629  static const uschar*  static const pcre_uchar*
1630  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1631  {  {
1632  for (;;)  for (;;)
1633    {    {
1634    switch ((int)*code)    switch ((int)*code)
1635      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1636      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1637      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1638      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1639      if (!skipassert) return code;      if (!skipassert) return code;
1640      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1641      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1642      break;      break;
1643    
1644      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 811  for (;;) Line 1648  for (;;)
1648    
1649      case OP_CALLOUT:      case OP_CALLOUT:
1650      case OP_CREF:      case OP_CREF:
1651      case OP_BRANUMBER:      case OP_DNCREF:
1652      code += _pcre_OP_lengths[*code];      case OP_RREF:
1653        case OP_DNRREF:
1654        case OP_DEF:
1655        code += PRIV(OP_lengths)[*code];
1656      break;      break;
1657    
1658      default:      default:
# Line 824  for (;;) Line 1664  for (;;)
1664    
1665    
1666    
   
1667  /*************************************************  /*************************************************
1668  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1669  *************************************************/  *************************************************/
1670    
1671  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1672  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1673  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1674    temporarily terminated with OP_END when this function is called.
1675    
1676    This function is called when a backward assertion is encountered, so that if it
1677    fails, the error message can point to the correct place in the pattern.
1678    However, we cannot do this when the assertion contains subroutine calls,
1679    because they can be forward references. We solve this by remembering this case
1680    and doing the check at the end; a flag specifies which mode we are running in.
1681    
1682  Arguments:  Arguments:
1683    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1684    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1685      atend    TRUE if called when the pattern is complete
1686  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1687               or -2 if \C was encountered  
1688    Returns:   the fixed length,
1689                 or -1 if there is no fixed length,
1690                 or -2 if \C was encountered (in UTF-8 mode only)
1691                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1692                 or -4 if an unknown opcode was encountered (internal error)
1693  */  */
1694    
1695  static int  static int
1696  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1697  {  {
1698  int length = -1;  int length = -1;
1699    
1700  register int branchlength = 0;  register int branchlength = 0;
1701  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1702    
1703  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1704  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 855  branch, check the length against that of Line 1706  branch, check the length against that of
1706  for (;;)  for (;;)
1707    {    {
1708    int d;    int d;
1709    register int op = *cc;    pcre_uchar *ce, *cs;
1710    if (op >= OP_BRA) op = OP_BRA;    register pcre_uchar op = *cc;
1711    
1712    switch (op)    switch (op)
1713      {      {
1714        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1715        OP_BRA (normal non-capturing bracket) because the other variants of these
1716        opcodes are all concerned with unlimited repeated groups, which of course
1717        are not of fixed length. */
1718    
1719        case OP_CBRA:
1720      case OP_BRA:      case OP_BRA:
1721      case OP_ONCE:      case OP_ONCE:
1722        case OP_ONCE_NC:
1723      case OP_COND:      case OP_COND:
1724      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1725      if (d < 0) return d;      if (d < 0) return d;
1726      branchlength += d;      branchlength += d;
1727      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1728      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1729      break;      break;
1730    
1731      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1732      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1733      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1734        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1735        because they all imply an unlimited repeat. */
1736    
1737      case OP_ALT:      case OP_ALT:
1738      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1739      case OP_END:      case OP_END:
1740        case OP_ACCEPT:
1741        case OP_ASSERT_ACCEPT:
1742      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1743        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1744      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 886  for (;;) Line 1746  for (;;)
1746      branchlength = 0;      branchlength = 0;
1747      break;      break;
1748    
1749        /* A true recursion implies not fixed length, but a subroutine call may
1750        be OK. If the subroutine is a forward reference, we can't deal with
1751        it until the end of the pattern, so return -3. */
1752    
1753        case OP_RECURSE:
1754        if (!atend) return -3;
1755        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1756        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1757        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1758        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1759        if (d < 0) return d;
1760        branchlength += d;
1761        cc += 1 + LINK_SIZE;
1762        break;
1763    
1764      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1765    
1766      case OP_ASSERT:      case OP_ASSERT:
# Line 893  for (;;) Line 1768  for (;;)
1768      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1769      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1770      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1771      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1772        break;
1773    
1774      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1775    
1776      case OP_REVERSE:      case OP_MARK:
1777      case OP_BRANUMBER:      case OP_PRUNE_ARG:
1778      case OP_CREF:      case OP_SKIP_ARG:
1779      case OP_OPT:      case OP_THEN_ARG:
1780        cc += cc[1] + PRIV(OP_lengths)[*cc];
1781        break;
1782    
1783      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1784      case OP_CIRC:      case OP_CIRC:
1785        case OP_CIRCM:
1786        case OP_CLOSE:
1787        case OP_COMMIT:
1788        case OP_CREF:
1789        case OP_DEF:
1790        case OP_DNCREF:
1791        case OP_DNRREF:
1792      case OP_DOLL:      case OP_DOLL:
1793        case OP_DOLLM:
1794        case OP_EOD:
1795        case OP_EODN:
1796        case OP_FAIL:
1797      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1798        case OP_PRUNE:
1799        case OP_REVERSE:
1800        case OP_RREF:
1801        case OP_SET_SOM:
1802        case OP_SKIP:
1803        case OP_SOD:
1804        case OP_SOM:
1805        case OP_THEN:
1806      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1807      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1808      break;      break;
1809    
1810      /* Handle literal characters */      /* Handle literal characters */
1811    
1812      case OP_CHAR:      case OP_CHAR:
1813      case OP_CHARNC:      case OP_CHARI:
1814        case OP_NOT:
1815        case OP_NOTI:
1816      branchlength++;      branchlength++;
1817      cc += 2;      cc += 2;
1818  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1819      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1820  #endif  #endif
1821      break;      break;
1822    
# Line 931  for (;;) Line 1824  for (;;)
1824      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1825    
1826      case OP_EXACT:      case OP_EXACT:
1827      branchlength += GET2(cc,1);      case OP_EXACTI:
1828      cc += 4;      case OP_NOTEXACT:
1829  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1830      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1831        {      cc += 2 + IMM2_SIZE;
1832        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1833        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834  #endif  #endif
1835      break;      break;
1836    
1837      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1838      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1839      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1840          cc += 2;
1841        cc += 1 + IMM2_SIZE + 1;
1842      break;      break;
1843    
1844      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 953  for (;;) Line 1848  for (;;)
1848      cc += 2;      cc += 2;
1849      /* Fall through */      /* Fall through */
1850    
1851        case OP_HSPACE:
1852        case OP_VSPACE:
1853        case OP_NOT_HSPACE:
1854        case OP_NOT_VSPACE:
1855      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1856      case OP_DIGIT:      case OP_DIGIT:
1857      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 960  for (;;) Line 1859  for (;;)
1859      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1860      case OP_WORDCHAR:      case OP_WORDCHAR:
1861      case OP_ANY:      case OP_ANY:
1862        case OP_ALLANY:
1863      branchlength++;      branchlength++;
1864      cc++;      cc++;
1865      break;      break;
1866    
1867      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1868        otherwise \C is coded as OP_ALLANY. */
1869    
1870      case OP_ANYBYTE:      case OP_ANYBYTE:
1871      return -2;      return -2;
1872    
1873      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1874    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1875      case OP_CLASS:      case OP_CLASS:
1876      case OP_NCLASS:      case OP_NCLASS:
1877      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1878        case OP_XCLASS:
1879        /* The original code caused an unsigned overflow in 64 bit systems,
1880        so now we use a conditional statement. */
1881        if (op == OP_XCLASS)
1882          cc += GET(cc, 1);
1883        else
1884          cc += PRIV(OP_lengths)[OP_CLASS];
1885    #else
1886        cc += PRIV(OP_lengths)[OP_CLASS];
1887    #endif
1888    
1889      switch (*cc)      switch (*cc)
1890        {        {
1891        case OP_CRSTAR:        case OP_CRSTAR:
1892        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1893          case OP_CRPLUS:
1894          case OP_CRMINPLUS:
1895        case OP_CRQUERY:        case OP_CRQUERY:
1896        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1897          case OP_CRPOSSTAR:
1898          case OP_CRPOSPLUS:
1899          case OP_CRPOSQUERY:
1900        return -1;        return -1;
1901    
1902        case OP_CRRANGE:        case OP_CRRANGE:
1903        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1904        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1905        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1906        cc += 5;        branchlength += (int)GET2(cc,1);
1907          cc += 1 + 2 * IMM2_SIZE;
1908        break;        break;
1909    
1910        default:        default:
# Line 1003  for (;;) Line 1914  for (;;)
1914    
1915      /* Anything else is variable length */      /* Anything else is variable length */
1916    
1917      default:      case OP_ANYNL:
1918        case OP_BRAMINZERO:
1919        case OP_BRAPOS:
1920        case OP_BRAPOSZERO:
1921        case OP_BRAZERO:
1922        case OP_CBRAPOS:
1923        case OP_EXTUNI:
1924        case OP_KETRMAX:
1925        case OP_KETRMIN:
1926        case OP_KETRPOS:
1927        case OP_MINPLUS:
1928        case OP_MINPLUSI:
1929        case OP_MINQUERY:
1930        case OP_MINQUERYI:
1931        case OP_MINSTAR:
1932        case OP_MINSTARI:
1933        case OP_MINUPTO:
1934        case OP_MINUPTOI:
1935        case OP_NOTMINPLUS:
1936        case OP_NOTMINPLUSI:
1937        case OP_NOTMINQUERY:
1938        case OP_NOTMINQUERYI:
1939        case OP_NOTMINSTAR:
1940        case OP_NOTMINSTARI:
1941        case OP_NOTMINUPTO:
1942        case OP_NOTMINUPTOI:
1943        case OP_NOTPLUS:
1944        case OP_NOTPLUSI:
1945        case OP_NOTPOSPLUS:
1946        case OP_NOTPOSPLUSI:
1947        case OP_NOTPOSQUERY:
1948        case OP_NOTPOSQUERYI:
1949        case OP_NOTPOSSTAR:
1950        case OP_NOTPOSSTARI:
1951        case OP_NOTPOSUPTO:
1952        case OP_NOTPOSUPTOI:
1953        case OP_NOTQUERY:
1954        case OP_NOTQUERYI:
1955        case OP_NOTSTAR:
1956        case OP_NOTSTARI:
1957        case OP_NOTUPTO:
1958        case OP_NOTUPTOI:
1959        case OP_PLUS:
1960        case OP_PLUSI:
1961        case OP_POSPLUS:
1962        case OP_POSPLUSI:
1963        case OP_POSQUERY:
1964        case OP_POSQUERYI:
1965        case OP_POSSTAR:
1966        case OP_POSSTARI:
1967        case OP_POSUPTO:
1968        case OP_POSUPTOI:
1969        case OP_QUERY:
1970        case OP_QUERYI:
1971        case OP_REF:
1972        case OP_REFI:
1973        case OP_DNREF:
1974        case OP_DNREFI:
1975        case OP_SBRA:
1976        case OP_SBRAPOS:
1977        case OP_SCBRA:
1978        case OP_SCBRAPOS:
1979        case OP_SCOND:
1980        case OP_SKIPZERO:
1981        case OP_STAR:
1982        case OP_STARI:
1983        case OP_TYPEMINPLUS:
1984        case OP_TYPEMINQUERY:
1985        case OP_TYPEMINSTAR:
1986        case OP_TYPEMINUPTO:
1987        case OP_TYPEPLUS:
1988        case OP_TYPEPOSPLUS:
1989        case OP_TYPEPOSQUERY:
1990        case OP_TYPEPOSSTAR:
1991        case OP_TYPEPOSUPTO:
1992        case OP_TYPEQUERY:
1993        case OP_TYPESTAR:
1994        case OP_TYPEUPTO:
1995        case OP_UPTO:
1996        case OP_UPTOI:
1997      return -1;      return -1;
1998    
1999        /* Catch unrecognized opcodes so that when new ones are added they
2000        are not forgotten, as has happened in the past. */
2001    
2002        default:
2003        return -4;
2004      }      }
2005    }    }
2006  /* Control never gets here */  /* Control never gets here */
# Line 1012  for (;;) Line 2008  for (;;)
2008    
2009    
2010    
   
2011  /*************************************************  /*************************************************
2012  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2013  *************************************************/  *************************************************/
2014    
2015  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2016  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2017    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2018    so that it can be called from pcre_study() when finding the minimum matching
2019    length.
2020    
2021  Arguments:  Arguments:
2022    code        points to start of expression    code        points to start of expression
2023    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2024    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2025    
2026  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
2027  */  */
2028    
2029  static const uschar *  const pcre_uchar *
2030  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2031  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2032  for (;;)  for (;;)
2033    {    {
2034    register int c = *code;    register pcre_uchar c = *code;
2035    
2036    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2037    else if (c > OP_BRA)  
2038      /* XCLASS is used for classes that cannot be represented just by a bit
2039      map. This includes negated single high-valued characters. The length in
2040      the table is zero; the actual length is stored in the compiled code. */
2041    
2042      if (c == OP_XCLASS) code += GET(code, 1);
2043    
2044      /* Handle recursion */
2045    
2046      else if (c == OP_REVERSE)
2047        {
2048        if (number < 0) return (pcre_uchar *)code;
2049        code += PRIV(OP_lengths)[c];
2050        }
2051    
2052      /* Handle capturing bracket */
2053    
2054      else if (c == OP_CBRA || c == OP_SCBRA ||
2055               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2056      {      {
2057      int n = c - OP_BRA;      int n = (int)GET2(code, 1+LINK_SIZE);
2058      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      if (n == number) return (pcre_uchar *)code;
2059      if (n == number) return (uschar *)code;      code += PRIV(OP_lengths)[c];
     code += _pcre_OP_lengths[OP_BRA];  
2060      }      }
2061    
2062      /* Otherwise, we can get the item's length from the table, except that for
2063      repeated character types, we have to test for \p and \P, which have an extra
2064      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2065      must add in its length. */
2066    
2067    else    else
2068      {      {
2069      code += _pcre_OP_lengths[c];      switch(c)
2070          {
2071          case OP_TYPESTAR:
2072          case OP_TYPEMINSTAR:
2073          case OP_TYPEPLUS:
2074          case OP_TYPEMINPLUS:
2075          case OP_TYPEQUERY:
2076          case OP_TYPEMINQUERY:
2077          case OP_TYPEPOSSTAR:
2078          case OP_TYPEPOSPLUS:
2079          case OP_TYPEPOSQUERY:
2080          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2081          break;
2082    
2083  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
2084          case OP_TYPEMINUPTO:
2085          case OP_TYPEEXACT:
2086          case OP_TYPEPOSUPTO:
2087          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2088            code += 2;
2089          break;
2090    
2091      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
2092      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
2093      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
2094      can use relatively efficient code. */        case OP_THEN_ARG:
2095          code += code[1];
2096          break;
2097          }
2098    
2099        /* Add in the fixed length from the table */
2100    
2101        code += PRIV(OP_lengths)[c];
2102    
2103      if (utf8) switch(c)    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2104      a multi-byte character. The length in the table is a minimum, so we have to
2105      arrange to skip the extra bytes. */
2106    
2107    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2108        if (utf) switch(c)
2109        {        {
2110        case OP_CHAR:        case OP_CHAR:
2111        case OP_CHARNC:        case OP_CHARI:
2112        case OP_EXACT:        case OP_EXACT:
2113          case OP_EXACTI:
2114        case OP_UPTO:        case OP_UPTO:
2115          case OP_UPTOI:
2116        case OP_MINUPTO:        case OP_MINUPTO:
2117          case OP_MINUPTOI:
2118          case OP_POSUPTO:
2119          case OP_POSUPTOI:
2120        case OP_STAR:        case OP_STAR:
2121          case OP_STARI:
2122        case OP_MINSTAR:        case OP_MINSTAR:
2123          case OP_MINSTARI:
2124          case OP_POSSTAR:
2125          case OP_POSSTARI:
2126        case OP_PLUS:        case OP_PLUS:
2127          case OP_PLUSI:
2128        case OP_MINPLUS:        case OP_MINPLUS:
2129          case OP_MINPLUSI:
2130          case OP_POSPLUS:
2131          case OP_POSPLUSI:
2132        case OP_QUERY:        case OP_QUERY:
2133          case OP_QUERYI:
2134        case OP_MINQUERY:        case OP_MINQUERY:
2135        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2136        break;        case OP_POSQUERY:
2137          case OP_POSQUERYI:
2138        /* XCLASS is used for classes that cannot be represented just by a bit        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
2139        break;        break;
2140        }        }
2141    #else
2142        (void)(utf);  /* Keep compiler happy by referencing function argument */
2143  #endif  #endif
2144      }      }
2145    }    }
# Line 1097  instance of OP_RECURSE. Line 2156  instance of OP_RECURSE.
2156    
2157  Arguments:  Arguments:
2158    code        points to start of expression    code        points to start of expression
2159    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2160    
2161  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2162  */  */
2163    
2164  static const uschar *  static const pcre_uchar *
2165  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2166  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2167  for (;;)  for (;;)
2168    {    {
2169    register int c = *code;    register pcre_uchar c = *code;
2170    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2171    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2172    else if (c > OP_BRA)  
2173      {    /* XCLASS is used for classes that cannot be represented just by a bit
2174      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
2175      }    the table is zero; the actual length is stored in the compiled code. */
2176    
2177      if (c == OP_XCLASS) code += GET(code, 1);
2178    
2179      /* Otherwise, we can get the item's length from the table, except that for
2180      repeated character types, we have to test for \p and \P, which have an extra
2181      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2182      must add in its length. */
2183    
2184    else    else
2185      {      {
2186      code += _pcre_OP_lengths[c];      switch(c)
2187          {
2188          case OP_TYPESTAR:
2189          case OP_TYPEMINSTAR:
2190          case OP_TYPEPLUS:
2191          case OP_TYPEMINPLUS:
2192          case OP_TYPEQUERY:
2193          case OP_TYPEMINQUERY:
2194          case OP_TYPEPOSSTAR:
2195          case OP_TYPEPOSPLUS:
2196          case OP_TYPEPOSQUERY:
2197          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2198          break;
2199    
2200          case OP_TYPEPOSUPTO:
2201          case OP_TYPEUPTO:
2202          case OP_TYPEMINUPTO:
2203          case OP_TYPEEXACT:
2204          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2205            code += 2;
2206          break;
2207    
2208  #ifdef SUPPORT_UTF8        case OP_MARK:
2209          case OP_PRUNE_ARG:
2210          case OP_SKIP_ARG:
2211          case OP_THEN_ARG:
2212          code += code[1];
2213          break;
2214          }
2215    
2216        /* Add in the fixed length from the table */
2217    
2218        code += PRIV(OP_lengths)[c];
2219    
2220      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2221      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2222      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
2223    
2224      if (utf8) switch(c)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2225        if (utf) switch(c)
2226        {        {
2227        case OP_CHAR:        case OP_CHAR:
2228        case OP_CHARNC:        case OP_CHARI:
2229          case OP_NOT:
2230          case OP_NOTI:
2231        case OP_EXACT:        case OP_EXACT:
2232          case OP_EXACTI:
2233          case OP_NOTEXACT:
2234          case OP_NOTEXACTI:
2235        case OP_UPTO:        case OP_UPTO:
2236          case OP_UPTOI:
2237          case OP_NOTUPTO:
2238          case OP_NOTUPTOI:
2239        case OP_MINUPTO:        case OP_MINUPTO:
2240          case OP_MINUPTOI:
2241          case OP_NOTMINUPTO:
2242          case OP_NOTMINUPTOI:
2243          case OP_POSUPTO:
2244          case OP_POSUPTOI:
2245          case OP_NOTPOSUPTO:
2246          case OP_NOTPOSUPTOI:
2247        case OP_STAR:        case OP_STAR:
2248          case OP_STARI:
2249          case OP_NOTSTAR:
2250          case OP_NOTSTARI:
2251        case OP_MINSTAR:        case OP_MINSTAR:
2252          case OP_MINSTARI:
2253          case OP_NOTMINSTAR:
2254          case OP_NOTMINSTARI:
2255          case OP_POSSTAR:
2256          case OP_POSSTARI:
2257          case OP_NOTPOSSTAR:
2258          case OP_NOTPOSSTARI:
2259        case OP_PLUS:        case OP_PLUS:
2260          case OP_PLUSI:
2261          case OP_NOTPLUS:
2262          case OP_NOTPLUSI:
2263        case OP_MINPLUS:        case OP_MINPLUS:
2264          case OP_MINPLUSI:
2265          case OP_NOTMINPLUS:
2266          case OP_NOTMINPLUSI:
2267          case OP_POSPLUS:
2268          case OP_POSPLUSI:
2269          case OP_NOTPOSPLUS:
2270          case OP_NOTPOSPLUSI:
2271        case OP_QUERY:        case OP_QUERY:
2272          case OP_QUERYI:
2273          case OP_NOTQUERY:
2274          case OP_NOTQUERYI:
2275        case OP_MINQUERY:        case OP_MINQUERY:
2276        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2277        break;        case OP_NOTMINQUERY:
2278          case OP_NOTMINQUERYI:
2279        /* XCLASS is used for classes that cannot be represented just by a bit        case OP_POSQUERY:
2280        map. This includes negated single high-valued characters. The length in        case OP_POSQUERYI:
2281        the table is zero; the actual length is stored in the compiled code. */        case OP_NOTPOSQUERY:
2282          case OP_NOTPOSQUERYI:
2283        case OP_XCLASS:        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       code += GET(code, 1) + 1;  
2284        break;        break;
2285        }        }
2286    #else
2287        (void)(utf);  /* Keep compiler happy by referencing function argument */
2288  #endif  #endif
2289      }      }
2290    }    }
# Line 1165  for (;;) Line 2297  for (;;)
2297  *************************************************/  *************************************************/
2298    
2299  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
2300  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2301  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
2302  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
2303  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2304    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2305    bracket whose current branch will already have been scanned.
2306    
2307  Arguments:  Arguments:
2308    code        points to start of search    code        points to start of search
2309    endcode     points to where to stop    endcode     points to where to stop
2310    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2311      cd          contains pointers to tables etc.
2312      recurses    chain of recurse_check to catch mutual recursion
2313    
2314  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2315  */  */
2316    
2317    typedef struct recurse_check {
2318      struct recurse_check *prev;
2319      const pcre_uchar *group;
2320    } recurse_check;
2321    
2322  static BOOL  static BOOL
2323  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2324      BOOL utf, compile_data *cd, recurse_check *recurses)
2325  {  {
2326  register int c;  register pcre_uchar c;
2327  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  recurse_check this_recurse;
2328    
2329    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2330       code < endcode;       code < endcode;
2331       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2332    {    {
2333    const uschar *ccode;    const pcre_uchar *ccode;
2334    
2335    c = *code;    c = *code;
2336    
2337    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2338      first_significant_code() with a TRUE final argument. */
2339    
2340      if (c == OP_ASSERT)
2341        {
2342        do code += GET(code, 1); while (*code == OP_ALT);
2343        c = *code;
2344        continue;
2345        }
2346    
2347      /* For a recursion/subroutine call, if its end has been reached, which
2348      implies a backward reference subroutine call, we can scan it. If it's a
2349      forward reference subroutine call, we can't. To detect forward reference
2350      we have to scan up the list that is kept in the workspace. This function is
2351      called only when doing the real compile, not during the pre-compile that
2352      measures the size of the compiled pattern. */
2353    
2354      if (c == OP_RECURSE)
2355      {      {
2356        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2357      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2358    
2359      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2360        when called to scan a completed pattern by setting cd->start_workspace to
2361        NULL. */
2362    
2363        if (cd->start_workspace != NULL)
2364          {
2365          const pcre_uchar *tcode;
2366          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2367            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2368          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2369          }
2370    
2371        /* If we are scanning a completed pattern, there are no forward references
2372        and all groups are complete. We need to detect whether this is a recursive
2373        call, as otherwise there will be an infinite loop. If it is a recursion,
2374        just skip over it. Simple recursions are easily detected. For mutual
2375        recursions we keep a chain on the stack. */
2376    
2377        else
2378          {
2379          recurse_check *r = recurses;
2380          const pcre_uchar *endgroup = scode;
2381    
2382          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2383          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2384    
2385          for (r = recurses; r != NULL; r = r->prev)
2386            if (r->group == scode) break;
2387          if (r != NULL) continue;   /* Mutual recursion */
2388          }
2389    
2390        /* Completed reference; scan the referenced group, remembering it on the
2391        stack chain to detect mutual recursions. */
2392    
2393      empty_branch = FALSE;      empty_branch = FALSE;
2394        this_recurse.prev = recurses;
2395        this_recurse.group = scode;
2396    
2397      do      do
2398        {        {
2399        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2400            {
2401          empty_branch = TRUE;          empty_branch = TRUE;
2402            break;
2403            }
2404          scode += GET(scode, 1);
2405          }
2406        while (*scode == OP_ALT);
2407    
2408        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2409        continue;
2410        }
2411    
2412      /* Groups with zero repeats can of course be empty; skip them. */
2413    
2414      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2415          c == OP_BRAPOSZERO)
2416        {
2417        code += PRIV(OP_lengths)[c];
2418        do code += GET(code, 1); while (*code == OP_ALT);
2419        c = *code;
2420        continue;
2421        }
2422    
2423      /* A nested group that is already marked as "could be empty" can just be
2424      skipped. */
2425    
2426      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2427          c == OP_SCBRA || c == OP_SCBRAPOS)
2428        {
2429        do code += GET(code, 1); while (*code == OP_ALT);
2430        c = *code;
2431        continue;
2432        }
2433    
2434      /* For other groups, scan the branches. */
2435    
2436      if (c == OP_BRA  || c == OP_BRAPOS ||
2437          c == OP_CBRA || c == OP_CBRAPOS ||
2438          c == OP_ONCE || c == OP_ONCE_NC ||
2439          c == OP_COND)
2440        {
2441        BOOL empty_branch;
2442        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2443    
2444        /* If a conditional group has only one branch, there is a second, implied,
2445        empty branch, so just skip over the conditional, because it could be empty.
2446        Otherwise, scan the individual branches of the group. */
2447    
2448        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2449        code += GET(code, 1);        code += GET(code, 1);
2450        else
2451          {
2452          empty_branch = FALSE;
2453          do
2454            {
2455            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2456              empty_branch = TRUE;
2457            code += GET(code, 1);
2458            }
2459          while (*code == OP_ALT);
2460          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2461        }        }
2462      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2463      c = *code;      c = *code;
2464        continue;
2465      }      }
2466    
2467    else switch (c)    /* Handle the other opcodes */
2468    
2469      switch (c)
2470      {      {
2471      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2472        cannot be represented just by a bit map. This includes negated single
2473        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2474        actual length is stored in the compiled code, so we must update "code"
2475        here. */
2476    
2477  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2478      case OP_XCLASS:      case OP_XCLASS:
2479      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2480      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2481  #endif  #endif
2482    
2483      case OP_CLASS:      case OP_CLASS:
2484      case OP_NCLASS:      case OP_NCLASS:
2485      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2486    
2487  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2488      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2489  #endif  #endif
2490    
# Line 1234  for (code = first_significant_code(code Line 2494  for (code = first_significant_code(code
2494        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2495        case OP_CRQUERY:        case OP_CRQUERY:
2496        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2497          case OP_CRPOSSTAR:
2498          case OP_CRPOSQUERY:
2499        break;        break;
2500    
2501        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2502        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2503        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2504          case OP_CRPOSPLUS:
2505        return FALSE;        return FALSE;
2506    
2507        case OP_CRRANGE:        case OP_CRRANGE:
2508        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2509          case OP_CRPOSRANGE:
2510        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2511        break;        break;
2512        }        }
# Line 1250  for (code = first_significant_code(code Line 2514  for (code = first_significant_code(code
2514    
2515      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2516    
2517        case OP_ANY:
2518        case OP_ALLANY:
2519        case OP_ANYBYTE:
2520    
2521      case OP_PROP:      case OP_PROP:
2522      case OP_NOTPROP:      case OP_NOTPROP:
2523        case OP_ANYNL:
2524    
2525        case OP_NOT_HSPACE:
2526        case OP_HSPACE:
2527        case OP_NOT_VSPACE:
2528        case OP_VSPACE:
2529      case OP_EXTUNI:      case OP_EXTUNI:
2530    
2531      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2532      case OP_DIGIT:      case OP_DIGIT:
2533      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2534      case OP_WHITESPACE:      case OP_WHITESPACE:
2535      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2536      case OP_WORDCHAR:      case OP_WORDCHAR:
2537      case OP_ANY:  
     case OP_ANYBYTE:  
2538      case OP_CHAR:      case OP_CHAR:
2539      case OP_CHARNC:      case OP_CHARI:
2540      case OP_NOT:      case OP_NOT:
2541        case OP_NOTI:
2542    
2543      case OP_PLUS:      case OP_PLUS:
2544        case OP_PLUSI:
2545      case OP_MINPLUS:      case OP_MINPLUS:
2546      case OP_EXACT:      case OP_MINPLUSI:
2547    
2548      case OP_NOTPLUS:      case OP_NOTPLUS:
2549        case OP_NOTPLUSI:
2550      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2551        case OP_NOTMINPLUSI:
2552    
2553        case OP_POSPLUS:
2554        case OP_POSPLUSI:
2555        case OP_NOTPOSPLUS:
2556        case OP_NOTPOSPLUSI:
2557    
2558        case OP_EXACT:
2559        case OP_EXACTI:
2560      case OP_NOTEXACT:      case OP_NOTEXACT:
2561        case OP_NOTEXACTI:
2562    
2563      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2564      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2565        case OP_TYPEPOSPLUS:
2566      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2567    
2568      return FALSE;      return FALSE;
2569    
2570        /* These are going to continue, as they may be empty, but we have to
2571        fudge the length for the \p and \P cases. */
2572    
2573        case OP_TYPESTAR:
2574        case OP_TYPEMINSTAR:
2575        case OP_TYPEPOSSTAR:
2576        case OP_TYPEQUERY:
2577        case OP_TYPEMINQUERY:
2578        case OP_TYPEPOSQUERY:
2579        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2580        break;
2581    
2582        /* Same for these */
2583    
2584        case OP_TYPEUPTO:
2585        case OP_TYPEMINUPTO:
2586        case OP_TYPEPOSUPTO:
2587        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2588          code += 2;
2589        break;
2590    
2591      /* End of branch */      /* End of branch */
2592    
2593      case OP_KET:      case OP_KET:
2594      case OP_KETRMAX:      case OP_KETRMAX:
2595      case OP_KETRMIN:      case OP_KETRMIN:
2596        case OP_KETRPOS:
2597      case OP_ALT:      case OP_ALT:
2598      return TRUE;      return TRUE;
2599    
2600      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2601      followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2602        followed by a multibyte character. */
2603    
2604  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2605      case OP_STAR:      case OP_STAR:
2606        case OP_STARI:
2607        case OP_NOTSTAR:
2608        case OP_NOTSTARI:
2609    
2610      case OP_MINSTAR:      case OP_MINSTAR:
2611        case OP_MINSTARI:
2612        case OP_NOTMINSTAR:
2613        case OP_NOTMINSTARI:
2614    
2615        case OP_POSSTAR:
2616        case OP_POSSTARI:
2617        case OP_NOTPOSSTAR:
2618        case OP_NOTPOSSTARI:
2619    
2620      case OP_QUERY:      case OP_QUERY:
2621        case OP_QUERYI:
2622        case OP_NOTQUERY:
2623        case OP_NOTQUERYI:
2624    
2625      case OP_MINQUERY:      case OP_MINQUERY:
2626        case OP_MINQUERYI:
2627        case OP_NOTMINQUERY:
2628        case OP_NOTMINQUERYI:
2629    
2630        case OP_POSQUERY:
2631        case OP_POSQUERYI:
2632        case OP_NOTPOSQUERY:
2633        case OP_NOTPOSQUERYI:
2634    
2635        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2636        break;
2637    
2638      case OP_UPTO:      case OP_UPTO:
2639        case OP_UPTOI:
2640        case OP_NOTUPTO:
2641        case OP_NOTUPTOI:
2642    
2643      case OP_MINUPTO:      case OP_MINUPTO:
2644      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2645        case OP_NOTMINUPTO:
2646        case OP_NOTMINUPTOI:
2647    
2648        case OP_POSUPTO:
2649        case OP_POSUPTOI:
2650        case OP_NOTPOSUPTO:
2651        case OP_NOTPOSUPTOI:
2652    
2653        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2654      break;      break;
2655  #endif  #endif
2656    
2657        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2658        string. */
2659    
2660        case OP_MARK:
2661        case OP_PRUNE_ARG:
2662        case OP_SKIP_ARG:
2663        case OP_THEN_ARG:
2664        code += code[1];
2665        break;
2666    
2667        /* None of the remaining opcodes are required to match a character. */
2668    
2669        default:
2670        break;
2671      }      }
2672    }    }
2673    
# Line 1312  return TRUE; Line 2684  return TRUE;
2684  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
2685  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,
2686  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.
2687    This function is called only during the real compile, not during the
2688    pre-compile.
2689    
2690  Arguments:  Arguments:
2691    code        points to start of the recursion    code        points to start of the recursion
2692    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2693    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2694    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695      cd          pointers to tables etc
2696    
2697  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2698  */  */
2699    
2700  static BOOL  static BOOL
2701  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2702    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2703  {  {
2704  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2705    {    {
2706    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2707        return FALSE;
2708    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2709    }    }
2710  return TRUE;  return TRUE;
# Line 1337  return TRUE; Line 2713  return TRUE;
2713    
2714    
2715  /*************************************************  /*************************************************
2716  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2717  *************************************************/  *************************************************/
2718    
2719  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2720  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 "=]".  
   
 Argument:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
   cd       pointer to compile data  
2721    
2722  Returns:   TRUE or FALSE  Arguments:  c opcode
2723    Returns:    base opcode for the type
2724  */  */
2725    
2726  static BOOL  static pcre_uchar
2727  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  get_repeat_base(pcre_uchar c)
2728  {  {
2729  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2730  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2731  if (*(++ptr) == '^') ptr++;         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2732  while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2733  if (*ptr == terminator && ptr[1] == ']')         (c >= OP_STARI)?      OP_STARI :
2734    {                               OP_STAR;
   *endptr = ptr;  
   return TRUE;  
   }  
 return FALSE;  
2735  }  }
2736    
2737    
2738    
2739    #ifdef SUPPORT_UCP
2740  /*************************************************  /*************************************************
2741  *          Check POSIX class name                *  *        Check a character and a property        *
2742  *************************************************/  *************************************************/
2743    
2744  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2745  such as [:alnum:].  is adjacent to a fixed character.
2746    
2747  Arguments:  Arguments:
2748    ptr        points to the first letter    c            the character
2749    len        the length of the name    ptype        the property type
2750      pdata        the data for the type
2751      negated      TRUE if it's a negated property (\P or \p{^)
2752    
2753  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2754  */  */
2755    
2756  static int  static BOOL
2757  check_posix_name(const uschar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2758      BOOL negated)
2759  {  {
2760  register int yield = 0;  const pcre_uint32 *p;
2761  while (posix_name_lengths[yield] != 0)  const ucd_record *prop = GET_UCD(c);
2762    
2763    switch(ptype)
2764    {    {
2765    if (len == posix_name_lengths[yield] &&    case PT_LAMP:
2766      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;    return (prop->chartype == ucp_Lu ||
2767    yield++;            prop->chartype == ucp_Ll ||
2768    }            prop->chartype == ucp_Lt) == negated;
 return -1;  
 }  
2769    
2770      case PT_GC:
2771      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2772    
2773  /*************************************************    case PT_PC:
2774  *    Adjust OP_RECURSE items in repeated group   *    return (pdata == prop->chartype) == negated;
 *************************************************/  
2775    
2776  /* OP_RECURSE items contain an offset from the start of the regex to the group    case PT_SC:
2777  that is referenced. This means that groups can be replicated for fixed    return (pdata == prop->script) == negated;
 repetition simply by copying (because the recursion is allowed to refer to  
 earlier groups that are outside the current group). However, when a group is  
 optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  
 it, after it has been compiled. This means that any OP_RECURSE items within it  
 that refer to the group itself or any contained groups have to have their  
 offsets adjusted. That is the job of this function. Before it is called, the  
 partially compiled regex must be temporarily terminated with OP_END.  
2778    
2779  Arguments:    /* These are specials */
   group      points to the start of the group  
   adjust     the amount by which the group is to be moved  
   utf8       TRUE in UTF-8 mode  
   cd         contains pointers to tables etc.  
2780    
2781  Returns:     nothing    case PT_ALNUM:
2782  */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2783              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2784    
2785  static void    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2786  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)    means that Perl space and POSIX space are now identical. PCRE was changed
2787  {    at release 8.34. */
2788  uschar *ptr = group;  
2789  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)    case PT_SPACE:    /* Perl space */
2790    {    case PT_PXSPACE:  /* POSIX space */
2791    int offset = GET(ptr, 1);    switch(c)
2792    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      {
2793    ptr += 1 + LINK_SIZE;      HSPACE_CASES:
2794        VSPACE_CASES:
2795        return negated;
2796    
2797        default:
2798        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2799        }
2800      break;  /* Control never reaches here */
2801    
2802      case PT_WORD:
2803      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2804              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2805              c == CHAR_UNDERSCORE) == negated;
2806    
2807      case PT_CLIST:
2808      p = PRIV(ucd_caseless_sets) + prop->caseset;
2809      for (;;)
2810        {
2811        if (c < *p) return !negated;
2812        if (c == *p++) return negated;
2813        }
2814      break;  /* Control never reaches here */
2815    }    }
2816    
2817    return FALSE;
2818  }  }
2819    #endif  /* SUPPORT_UCP */
2820    
2821    
2822    
2823  /*************************************************  /*************************************************
2824  *        Insert an automatic callout point       *  *        Fill the character property list        *
2825  *************************************************/  *************************************************/
2826    
2827  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2828  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2829    
2830  Arguments:  Arguments:
2831    code           current code pointer    code        points to start of expression
2832    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2833    cd             pointers to tables etc    fcc         points to case-flipping table
2834      list        points to output list
2835                  list[0] will be filled with the opcode
2836                  list[1] will be non-zero if this opcode
2837                    can match an empty character string
2838                  list[2..7] depends on the opcode
2839    
2840  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2841                  NULL if *code is not accepted
2842  */  */
2843    
2844  static uschar *  static const pcre_uchar *
2845  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2846      const pcre_uint8 *fcc, pcre_uint32 *list)
2847  {  {
2848  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2849  *code++ = 255;  const pcre_uchar *end;
2850  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  const pcre_uint32 *clist_src;
2851  PUT(code, LINK_SIZE, 0);                /* Default length */  pcre_uint32 *clist_dest;
2852  return code + 2*LINK_SIZE;  pcre_uint32 chr;
2853  }  pcre_uchar base;
2854    
2855    list[0] = c;
2856    list[1] = FALSE;
2857    code++;
2858    
2859    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2860      {
2861      base = get_repeat_base(c);
2862      c -= (base - OP_STAR);
2863    
2864      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2865        code += IMM2_SIZE;
2866    
2867  /*************************************************    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 *         Complete a callout item                *  
 *************************************************/  
2868    
2869  /* A callout item contains the length of the next item in the pattern, which    switch(base)
2870  we can't fill in till after we have reached the relevant point. This is used      {
2871  for both automatic and manual callouts.      case OP_STAR:
2872        list[0] = OP_CHAR;
2873        break;
2874    
2875  Arguments:      case OP_STARI:
2876    previous_callout   points to previous callout item      list[0] = OP_CHARI;
2877    ptr                current pattern pointer      break;
   cd                 pointers to tables etc  
2878    
2879  Returns:             nothing      case OP_NOTSTAR:
2880  */      list[0] = OP_NOT;
2881        break;
2882    
2883  static void      case OP_NOTSTARI:
2884  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)      list[0] = OP_NOTI;
2885  {      break;
 int length = ptr - cd->start_pattern - GET(previous_callout, 2);  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2886    
2887        case OP_TYPESTAR:
2888        list[0] = *code;
2889        code++;
2890        break;
2891        }
2892      c = list[0];
2893      }
2894    
2895    switch(c)
2896      {
2897      case OP_NOT_DIGIT:
2898      case OP_DIGIT:
2899      case OP_NOT_WHITESPACE:
2900      case OP_WHITESPACE:
2901      case OP_NOT_WORDCHAR:
2902      case OP_WORDCHAR:
2903      case OP_ANY:
2904      case OP_ALLANY:
2905      case OP_ANYNL:
2906      case OP_NOT_HSPACE:
2907      case OP_HSPACE:
2908      case OP_NOT_VSPACE:
2909      case OP_VSPACE:
2910      case OP_EXTUNI:
2911      case OP_EODN:
2912      case OP_EOD:
2913      case OP_DOLL:
2914      case OP_DOLLM:
2915      return code;
2916    
2917      case OP_CHAR:
2918      case OP_NOT:
2919      GETCHARINCTEST(chr, code);
2920      list[2] = chr;
2921      list[3] = NOTACHAR;
2922      return code;
2923    
2924      case OP_CHARI:
2925      case OP_NOTI:
2926      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2927      GETCHARINCTEST(chr, code);
2928      list[2] = chr;
2929    
2930  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2931  /*************************************************    if (chr < 128 || (chr < 256 && !utf))
2932  *           Get othercase range                  *      list[3] = fcc[chr];
2933  *************************************************/    else
2934        list[3] = UCD_OTHERCASE(chr);
2935    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2936      list[3] = (chr < 256) ? fcc[chr] : chr;
2937    #else
2938      list[3] = fcc[chr];
2939    #endif
2940    
2941  /* This function is passed the start and end of a class range, in UTF-8 mode    /* The othercase might be the same value. */
 with UCP support. It searches up the characters, looking for internal ranges of  
 characters in the "other" case. Each call returns the next one, updating the  
 start address.  
2942    
2943  Arguments:    if (chr == list[3])
2944    cptr        points to starting character value; updated      list[3] = NOTACHAR;
2945    d           end value    else
2946    ocptr       where to put start of othercase range      list[4] = NOTACHAR;
2947    odptr       where to put end of othercase range    return code;
2948    
2949  Yield:        TRUE when range returned; FALSE when no more  #ifdef SUPPORT_UCP
2950  */    case OP_PROP:
2951      case OP_NOTPROP:
2952      if (code[0] != PT_CLIST)
2953        {
2954        list[2] = code[0];
2955        list[3] = code[1];
2956        return code + 2;
2957        }
2958    
2959  static BOOL    /* Convert only if we have enough space. */
 get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  
 {  
 int c, othercase, next;  
2960    
2961  for (c = *cptr; c <= d; c++)    clist_src = PRIV(ucd_caseless_sets) + code[1];
2962    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    clist_dest = list + 2;
2963      code += 2;
2964    
2965      do {
2966         if (clist_dest >= list + 8)
2967           {
2968           /* Early return if there is not enough space. This should never
2969           happen, since all clists are shorter than 5 character now. */
2970           list[2] = code[0];
2971           list[3] = code[1];
2972           return code;
2973           }
2974         *clist_dest++ = *clist_src;
2975         }
2976      while(*clist_src++ != NOTACHAR);
2977    
2978  if (c > d) return FALSE;    /* All characters are stored. The terminating NOTACHAR
2979      is copied form the clist itself. */
2980    
2981  *ocptr = othercase;    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2982  next = othercase + 1;    return code;
2983    #endif
2984    
2985  for (++c; c <= d; c++)    case OP_NCLASS:
2986    {    case OP_CLASS:
2987    if (_pcre_ucp_othercase(c) != next) break;  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2988    next++;    case OP_XCLASS:
2989    }    if (c == OP_XCLASS)
2990        end = code + GET(code, 0) - 1;
2991      else
2992    #endif
2993        end = code + 32 / sizeof(pcre_uchar);
2994    
2995      switch(*end)
2996        {
2997        case OP_CRSTAR:
2998        case OP_CRMINSTAR:
2999        case OP_CRQUERY:
3000        case OP_CRMINQUERY:
3001        case OP_CRPOSSTAR:
3002        case OP_CRPOSQUERY:
3003        list[1] = TRUE;
3004        end++;
3005        break;
3006    
3007  *odptr = next - 1;      case OP_CRPLUS:
3008  *cptr = c;      case OP_CRMINPLUS:
3009        case OP_CRPOSPLUS:
3010        end++;
3011        break;
3012    
3013  return TRUE;      case OP_CRRANGE:
3014        case OP_CRMINRANGE:
3015        case OP_CRPOSRANGE:
3016        list[1] = (GET2(end, 1) == 0);
3017        end += 1 + 2 * IMM2_SIZE;
3018        break;
3019        }
3020      list[2] = end - code;
3021      return end;
3022      }
3023    return NULL;    /* Opcode not accepted */
3024  }  }
3025  #endif  /* SUPPORT_UCP */  
3026    
3027    
3028  /*************************************************  /*************************************************
3029  *           Compile one branch                   *  *    Scan further character sets for match       *
3030  *************************************************/  *************************************************/
3031    
3032  /* Scan the pattern, compiling it into the code vector. If the options are  /* Checks whether the base and the current opcode have a common character, in
3033  changed during the branch, the pointer is used to change the external options  which case the base cannot be possessified.
 bits.  
3034    
3035  Arguments:  Arguments:
3036    optionsptr     pointer to the option bits    code        points to the byte code
3037    brackets       points to number of extracting brackets used    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3038    codeptr        points to the pointer to the current code point    cd          static compile data
3039    ptrptr         points to the current pattern pointer    base_list   the data list of the base opcode
   errorcodeptr   points to error code variable  
   firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)  
   reqbyteptr     set to the last literal character required, else < 0  
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
3040    
3041  Returns:         TRUE on success  Returns:      TRUE if the auto-possessification is possible
                  FALSE, with *errorcodeptr set non-zero on error  
3042  */  */
3043    
3044  static BOOL  static BOOL
3045  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3046    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const pcre_uint32 *base_list, const pcre_uchar *base_end)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
3047  {  {
3048  int repeat_type, op_type;  pcre_uchar c;
3049  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  pcre_uint32 list[8];
3050  int bravalue = 0;  const pcre_uint32 *chr_ptr;
3051  int greedy_default, greedy_non_default;  const pcre_uint32 *ochr_ptr;
3052  int firstbyte, reqbyte;  const pcre_uint32 *list_ptr;
3053  int zeroreqbyte, zerofirstbyte;  const pcre_uchar *next_code;
3054  int req_caseopt, reqvary, tempreqvary;  const pcre_uint8 *class_bitset;
3055  int condcount = 0;  const pcre_uint32 *set1, *set2, *set_end;
3056  int options = *optionsptr;  pcre_uint32 chr;
3057  int after_manual_callout = 0;  BOOL accepted, invert_bits;
3058  register int c;  
3059  register uschar *code = *codeptr;  /* Note: the base_list[1] contains whether the current opcode has greedy
3060  uschar *tempcode;  (represented by a non-zero value) quantifier. This is a different from
3061  BOOL inescq = FALSE;  other character type lists, which stores here that the character iterator
3062  BOOL groupsetfirstbyte = FALSE;  matches to an empty string (also represented by a non-zero value). */
 const uschar *ptr = *ptrptr;  
 const uschar *tempptr;  
 uschar *previous = NULL;  
 uschar *previous_callout = NULL;  
 uschar classbits[32];  
   
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
 #endif  
   
 /* Set up the default and non-default settings for greediness */  
   
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
   
 /* Initialize no first byte, no required byte. REQ_UNSET means "no char  
 matching encountered yet". It gets changed to REQ_NONE if we hit something that  
 matches a non-fixed char first char; reqbyte just remains unset if we never  
 find one.  
3063    
3064  When we hit a repeat whose minimum is zero, we may have to adjust these values  for(;;)
3065  to take the zero repeat into account. This is implemented by setting them to    {
3066  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual    /* All operations move the code pointer forward.
3067  item types that can be repeated set these backoff variables appropriately. */    Therefore infinite recursions are not possible. */
3068    
3069  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;    c = *code;
3070    
3071  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,    /* Skip over callouts */
 according to the current setting of the caseless flag. REQ_CASELESS is a bit  
 value > 255. It is added into the firstbyte or reqbyte variables to record the  
 case status of the value. This is used only for ASCII characters. */  
3072    
3073  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;    if (c == OP_CALLOUT)
3074        {
3075        code += PRIV(OP_lengths)[c];
3076        continue;
3077        }
3078    
3079  /* Switch on next character until the end of the branch */    if (c == OP_ALT)
3080        {
3081        do code += GET(code, 1); while (*code == OP_ALT);
3082        c = *code;
3083        }
3084    
3085  for (;; ptr++)    switch(c)
3086    {      {
3087    BOOL negate_class;      case OP_END:
3088    BOOL possessive_quantifier;      case OP_KETRPOS:
3089    BOOL is_quantifier;      /* TRUE only in greedy case. The non-greedy case could be replaced by
3090    int class_charcount;      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3091    int class_lastchar;      uses more memory, which we cannot get at this stage.) */
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
3092    
3093    /* Next byte in the pattern */      return base_list[1] != 0;
3094    
3095    c = *ptr;      case OP_KET:
3096        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3097        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3098        cannot be converted to a possessive form. */
3099    
3100        if (base_list[1] == 0) return FALSE;
3101    
3102        switch(*(code - GET(code, 1)))
3103          {
3104          case OP_ASSERT:
3105          case OP_ASSERT_NOT:
3106          case OP_ASSERTBACK:
3107          case OP_ASSERTBACK_NOT:
3108          case OP_ONCE:
3109          case OP_ONCE_NC:
3110          /* Atomic sub-patterns and assertions can always auto-possessify their
3111          last iterator. */
3112          return TRUE;
3113          }
3114    
3115    /* If in \Q...\E, check for the end; if not, we have a literal */      code += PRIV(OP_lengths)[c];
3116        continue;
3117    
3118    if (inescq && c != 0)      case OP_ONCE:
3119      {      case OP_ONCE_NC:
3120      if (c == '\\' && ptr[1] == 'E')      case OP_BRA:
3121        {      case OP_CBRA:
3122        inescq = FALSE;      next_code = code + GET(code, 1);
3123        ptr++;      code += PRIV(OP_lengths)[c];
3124        continue;  
3125        }      while (*next_code == OP_ALT)
3126      else        {
3127        {        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3128        if (previous_callout != NULL)        code = next_code + 1 + LINK_SIZE;
3129          {        next_code += GET(next_code, 1);
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
         {  
         previous_callout = code;  
         code = auto_callout(code, ptr, cd);  
         }  
       goto NORMAL_CHAR;  
3130        }        }
3131      }      continue;
3132    
3133    /* Fill in length of a previous callout, except when the next thing is      case OP_BRAZERO:
3134    a quantifier. */      case OP_BRAMINZERO:
3135    
3136    is_quantifier = c == '*' || c == '+' || c == '?' ||      next_code = code + 1;
3137      (c == '{' && is_counted_repeat(ptr+1));      if (*next_code != OP_BRA && *next_code != OP_CBRA
3138            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3139    
3140        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3141    
3142        /* The bracket content will be checked by the
3143        OP_BRA/OP_CBRA case above. */
3144        next_code += 1 + LINK_SIZE;
3145        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3146          return FALSE;
3147    
3148    if (!is_quantifier && previous_callout != NULL &&      code += PRIV(OP_lengths)[c];
3149         after_manual_callout-- <= 0)      continue;
     {  
     complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
3150      }      }
3151    
3152    /* In extended mode, skip white space and comments */    /* Check for a supported opcode, and load its properties. */
3153    
3154    if ((options & PCRE_EXTENDED) != 0)    code = get_chr_property_list(code, utf, cd->fcc, list);
3155      {    if (code == NULL) return FALSE;    /* Unsupported */
     if ((cd->ctypes[c] & ctype_space) != 0) continue;  
     if (c == '#')  
       {  
       /* The space before the ; is to avoid a warning on a silly compiler  
       on the Macintosh. */  
       while ((c = *(++ptr)) != 0 && c != NEWLINE) ;  
       if (c != 0) continue;   /* Else fall through to handle end of string */  
       }  
     }  
3156    
3157    /* No auto callout for quantifiers. */    /* If either opcode is a small character list, set pointers for comparing
3158      characters from that list with another list, or with a property. */
3159    
3160    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if (base_list[0] == OP_CHAR)
3161      {      {
3162      previous_callout = code;      chr_ptr = base_list + 2;
3163      code = auto_callout(code, ptr, cd);      list_ptr = list;
3164      }      }
3165      else if (list[0] == OP_CHAR)
   switch(c)  
3166      {      {
3167      /* The branch terminates at end of string, |, or ). */      chr_ptr = list + 2;
3168        list_ptr = base_list;
3169      case 0:      }
     case '|':  
     case ')':  
     *firstbyteptr = firstbyte;  
     *reqbyteptr = reqbyte;  
     *codeptr = code;  
     *ptrptr = ptr;  
     return TRUE;  
3170    
3171      /* Handle single-character metacharacters. In multiline mode, ^ disables    /* Character bitsets can also be compared to certain opcodes. */
     the setting of any following char as a first character. */  
3172    
3173      case '^':    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3174      if ((options & PCRE_MULTILINE) != 0)  #ifdef COMPILE_PCRE8
3175          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3176          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3177    #endif
3178          )
3179        {
3180    #ifdef COMPILE_PCRE8
3181        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3182    #else
3183        if (base_list[0] == OP_CLASS)
3184    #endif
3185        {        {
3186        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        set1 = (pcre_uint32 *)(base_end - base_list[2]);
3187          list_ptr = list;
3188          }
3189        else
3190          {
3191          set1 = (pcre_uint32 *)(code - list[2]);
3192          list_ptr = base_list;
3193        }        }
     previous = NULL;  
     *code++ = OP_CIRC;  
     break;  
   
     case '$':  
     previous = NULL;  
     *code++ = OP_DOLL;  
     break;  
   
     /* There can never be a first char if '.' is first, whatever happens about  
     repeats. The value of reqbyte doesn't change either. */  
3194    
3195      case '.':      invert_bits = FALSE;
3196      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      switch(list_ptr[0])
3197      zerofirstbyte = firstbyte;        {
3198      zeroreqbyte = reqbyte;        case OP_CLASS:
3199      previous = code;        case OP_NCLASS:
3200      *code++ = OP_ANY;        set2 = (pcre_uint32 *)
3201      break;          ((list_ptr == list ? code : base_end) - list_ptr[2]);
3202          break;
3203    
3204      /* Character classes. If the included characters are all < 256, we build a        /* OP_XCLASS cannot be supported here, because its bitset
3205      32-byte bitmap of the permitted characters, except in the special case        is not necessarily complete. E.g: [a-\0x{200}] is stored
3206      where there is only one such character. For negated classes, we build the        as a character range, and the appropriate bits are not set. */
     map as usual, then invert it at the end. However, we use a different opcode  
     so that data characters > 255 can be handled correctly.  
3207    
3208      If the class contains characters outside the 0-255 range, a different        case OP_NOT_DIGIT:
3209      opcode is compiled. It may optionally have a bit map for characters < 256,          invert_bits = TRUE;
3210      but those above are are explicitly listed afterwards. A flag byte tells          /* Fall through */
3211      whether the bitmap is present, and whether this is a negated class or not.        case OP_DIGIT:
3212      */          set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3213            break;
3214    
3215      case '[':        case OP_NOT_WHITESPACE:
3216      previous = code;          invert_bits = TRUE;
3217            /* Fall through */
3218          case OP_WHITESPACE:
3219            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3220            break;
3221    
3222      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if        case OP_NOT_WORDCHAR:
3223      they are encountered at the top level, so we'll do that too. */          invert_bits = TRUE;
3224            /* Fall through */
3225          case OP_WORDCHAR:
3226            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3227            break;
3228    
3229      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&        default:
3230          check_posix_syntax(ptr, &tempptr, cd))        return FALSE;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
3231        }        }
3232    
3233      /* If the first character is '^', set the negation flag and skip it. */      /* Compare 4 bytes to improve speed. */
3234        set_end = set1 + (32 / 4);
3235      if ((c = *(++ptr)) == '^')      if (invert_bits)
3236        {        {
3237        negate_class = TRUE;        do
3238        c = *(++ptr);          {
3239            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3240            }
3241          while (set1 < set_end);
3242        }        }
3243      else      else
3244        {        {
3245        negate_class = FALSE;        do
3246            {
3247            if ((*set1++ & *set2++) != 0) return FALSE;
3248            }
3249          while (set1 < set_end);
3250        }        }
3251    
3252      /* Keep a count of chars with values < 256 so that we can optimize the case      if (list[1] == 0) return TRUE;
3253      of just a single character (as long as it's < 256). For higher valued UTF-8      /* Might be an empty repeat. */
3254      characters, we don't yet do any optimization. */      continue;
3255        }
     class_charcount = 0;  
     class_lastchar = -1;  
   
 #ifdef SUPPORT_UTF8  
     class_utf8 = FALSE;                       /* No chars >= 256 */  
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
 #endif  
3256    
3257      /* Initialize the 32-char bit map to all zeros. We have to build the    /* Some property combinations also acceptable. Unicode property opcodes are
3258      map in a temporary bit of store, in case the class contains only 1    processed specially; the rest can be handled with a lookup table. */
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
3259    
3260      memset(classbits, 0, 32 * sizeof(uschar));    else
3261        {
3262        pcre_uint32 leftop, rightop;
3263    
3264      /* Process characters until ] is reached. By writing this as a "do" it      leftop = base_list[0];
3265      means that an initial ] is taken as a data character. The first pass      rightop = list[0];
     through the regex checked the overall syntax, so we don't need to be very  
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3266    
3267      do  #ifdef SUPPORT_UCP
3268        accepted = FALSE; /* Always set in non-unicode case. */
3269        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3270        {        {
3271  #ifdef SUPPORT_UTF8        if (rightop == OP_EOD)
3272        if (utf8 && c > 127)          accepted = TRUE;
3273          {                           /* Braces are required because the */        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3274          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          {
3275          }          int n;
3276  #endif          const pcre_uint8 *p;
3277            BOOL same = leftop == rightop;
3278        /* Inside \Q...\E everything is literal except \E */          BOOL lisprop = leftop == OP_PROP;
3279            BOOL risprop = rightop == OP_PROP;
3280            BOOL bothprop = lisprop && risprop;
3281    
3282            /* There's a table that specifies how each combination is to be
3283            processed:
3284              0   Always return FALSE (never auto-possessify)
3285              1   Character groups are distinct (possessify if both are OP_PROP)
3286              2   Check character categories in the same group (general or particular)
3287              3   Return TRUE if the two opcodes are not the same
3288              ... see comments below
3289            */
3290    
3291            n = propposstab[base_list[2]][list[2]];
3292            switch(n)
3293              {
3294              case 0: break;
3295              case 1: accepted = bothprop; break;
3296              case 2: accepted = (base_list[3] == list[3]) != same; break;
3297              case 3: accepted = !same; break;
3298    
3299        if (inescq)            case 4:  /* Left general category, right particular category */
3300          {            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3301          if (c == '\\' && ptr[1] == 'E')            break;
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
3302    
3303        /* Handle POSIX class names. Perl allows a negation extension of the            case 5:  /* Right general category, left particular category */
3304        form [:^name:]. A square bracket that doesn't match the syntax is            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3305        treated as a literal. We also recognize the POSIX constructions            break;
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
3306    
3307        if (c == '[' &&            /* This code is logically tricky. Think hard before fiddling with it.
3308            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            The posspropstab table has four entries per row. Each row relates to
3309            check_posix_syntax(ptr, &tempptr, cd))            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3310          {            Only WORD actually needs all four entries, but using repeats for the
3311          BOOL local_negate = FALSE;            others means they can all use the same code below.
3312          int posix_class, taboffset, tabopt;  
3313          register const uschar *cbits = cd->cbits;            The first two entries in each row are Unicode general categories, and
3314          uschar pbits[32];            apply always, because all the characters they include are part of the
3315              PCRE character set. The third and fourth entries are a general and a
3316              particular category, respectively, that include one or more relevant
3317              characters. One or the other is used, depending on whether the check
3318              is for a general or a particular category. However, in both cases the
3319              category contains more characters than the specials that are defined
3320              for the property being tested against. Therefore, it cannot be used
3321              in a NOTPROP case.
3322    
3323              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3324              Underscore is covered by ucp_P or ucp_Po. */
3325    
3326              case 6:  /* Left alphanum vs right general category */
3327              case 7:  /* Left space vs right general category */
3328              case 8:  /* Left word vs right general category */
3329              p = posspropstab[n-6];
3330              accepted = risprop && lisprop ==
3331                (list[3] != p[0] &&
3332                 list[3] != p[1] &&
3333                (list[3] != p[2] || !lisprop));
3334              break;
3335    
3336          if (ptr[1] != ':')            case 9:   /* Right alphanum vs left general category */
3337            {            case 10:  /* Right space vs left general category */
3338            *errorcodeptr = ERR31;            case 11:  /* Right word vs left general category */
3339            goto FAILED;            p = posspropstab[n-9];
3340            }            accepted = lisprop && risprop ==
3341                (base_list[3] != p[0] &&
3342                 base_list[3] != p[1] &&
3343                (base_list[3] != p[2] || !risprop));
3344              break;
3345    
3346          ptr += 2;            case 12:  /* Left alphanum vs right particular category */
3347          if (*ptr == '^')            case 13:  /* Left space vs right particular category */
3348            {            case 14:  /* Left word vs right particular category */
3349            local_negate = TRUE;            p = posspropstab[n-12];
3350            ptr++;            accepted = risprop && lisprop ==
3351            }              (catposstab[p[0]][list[3]] &&
3352                 catposstab[p[1]][list[3]] &&
3353                (list[3] != p[3] || !lisprop));
3354              break;
3355    
3356          posix_class = check_posix_name(ptr, tempptr - ptr);            case 15:  /* Right alphanum vs left particular category */
3357          if (posix_class < 0)            case 16:  /* Right space vs left particular category */
3358            {            case 17:  /* Right word vs left particular category */
3359            *errorcodeptr = ERR30;            p = posspropstab[n-15];
3360            goto FAILED;            accepted = lisprop && risprop ==
3361                (catposstab[p[0]][base_list[3]] &&
3362                 catposstab[p[1]][base_list[3]] &&
3363                (base_list[3] != p[3] || !risprop));
3364              break;
3365            }            }
3366            }
3367          }
3368    
3369          /* If matching is caseless, upper and lower are converted to      else
3370          alpha. This relies on the fact that the class table starts with  #endif  /* SUPPORT_UCP */
         alpha, lower, upper as the first 3 entries. */  
3371    
3372          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3373            posix_class = 0;             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3374               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3375    
3376          /* We build the bit map for the POSIX class in a chunk of local store      if (!accepted)
3377          because we may be adding and subtracting from it, and we don't want to        return FALSE;
         subtract bits that may be in the main map already. At the end we or the  
         result into the bit map that is being built. */  
3378    
3379          posix_class *= 3;      if (list[1] == 0) return TRUE;
3380        /* Might be an empty repeat. */
3381        continue;
3382        }
3383    
3384          /* Copy in the first table (always present) */    /* Control reaches here only if one of the items is a small character list.
3385      All characters are checked against the other side. */
3386    
3387          memcpy(pbits, cbits + posix_class_maps[posix_class],    do
3388            32 * sizeof(uschar));      {
3389        chr = *chr_ptr;
3390    
3391          /* If there is a second table, add or remove it as required. */      switch(list_ptr[0])
3392          {
3393          case OP_CHAR:
3394          ochr_ptr = list_ptr + 2;
3395          do
3396            {
3397            if (chr == *ochr_ptr) return FALSE;
3398            ochr_ptr++;
3399            }
3400          while(*ochr_ptr != NOTACHAR);
3401          break;
3402    
3403          taboffset = posix_class_maps[posix_class + 1];        case OP_NOT:
3404          tabopt = posix_class_maps[posix_class + 2];        ochr_ptr = list_ptr + 2;
3405          do
3406            {
3407            if (chr == *ochr_ptr)
3408              break;
3409            ochr_ptr++;
3410            }
3411          while(*ochr_ptr != NOTACHAR);
3412          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3413          break;
3414