/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 1360 by ph10, Tue Sep 3 10:25:39 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) 1997-2007 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69    /* Macro for setting individual bits in class bitmaps. */
70    
71    #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73    /* Maximum length value to check against when making sure that the integer that
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
95  *************************************************/  *************************************************/
# Line 72  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123  #define COMPILE_WORK_SIZE (4096)  #define NAMED_GROUP_LIST_SIZE  20
124    
125    /* The overrun tests check for a slightly smaller size so that they detect the
126    overrun before it actually does run off the end of the data block. */
127    
128    #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138    /* Repeated character flags. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
144  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
145  is invalid. */  is invalid. */
146    
147  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
148    
149    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150    in UTF-8 mode. */
151    
152  static const short int escapes[] = {  static const short int escapes[] = {
153       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
154       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
155     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
156       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
158  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
159     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
160       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
161  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
162       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
163         -ESC_D,                  -ESC_E,
164         0,                       -ESC_G,
165         -ESC_H,                  0,
166         0,                       -ESC_K,
167         0,                       0,
168         -ESC_N,                  0,
169         -ESC_P,                  -ESC_Q,
170         -ESC_R,                  -ESC_S,
171         0,                       0,
172         -ESC_V,                  -ESC_W,
173         -ESC_X,                  0,
174         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
175         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
176         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
177         CHAR_GRAVE_ACCENT,       7,
178         -ESC_b,                  0,
179         -ESC_d,                  ESC_e,
180         ESC_f,                   0,
181         -ESC_h,                  0,
182         0,                       -ESC_k,
183         0,                       0,
184         ESC_n,                   0,
185         -ESC_p,                  0,
186         ESC_r,                   -ESC_s,
187         ESC_tee,                 0,
188         -ESC_v,                  -ESC_w,
189         0,                       0,
190         -ESC_z
191  };  };
192    
193  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
194    
195    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196    
197  static const short int escapes[] = {  static const short int escapes[] = {
198  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
199  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 203  static const short int escapes[] = {
203  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
206  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
207  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
208  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
209  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
210  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
211  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
212  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
213  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
214  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
215  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
216  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
217  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
218  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
219  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
220  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 222  static const short int escapes[] = {
222  #endif  #endif
223    
224    
225  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
227  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
228    string is built from string macros so that it works in UTF-8 mode on EBCDIC
229  static const char *const posix_names[] = {  platforms. */
230    "alpha", "lower", "upper",  
231    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
232    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
233      int   op;                  /* Op when no arg, or -1 if arg mandatory */
234      int   op_arg;              /* Op when arg present, or -1 if not allowed */
235    } verbitem;
236    
237    static const char verbnames[] =
238      "\0"                       /* Empty name is a shorthand for MARK */
239      STRING_MARK0
240      STRING_ACCEPT0
241      STRING_COMMIT0
242      STRING_F0
243      STRING_FAIL0
244      STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248    static const verbitem verbs[] = {
249      { 0, -1,        OP_MARK },
250      { 4, -1,        OP_MARK },
251      { 6, OP_ACCEPT, -1 },
252      { 6, OP_COMMIT, -1 },
253      { 1, OP_FAIL,   -1 },
254      { 4, OP_FAIL,   -1 },
255      { 5, OP_PRUNE,  OP_PRUNE_ARG },
256      { 4, OP_SKIP,   OP_SKIP_ARG  },
257      { 4, OP_THEN,   OP_THEN_ARG  }
258    };
259    
260    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* 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. */
268    
269    static const char posix_names[] =
270      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
271      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
272      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
273      STRING_word0  STRING_xdigit;
274    
275  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
276    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 };
277    
278  /* 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
# Line 164  static const int posix_class_maps[] = { Line 302  static const int posix_class_maps[] = {
302    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
303  };  };
304    
305    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
306    substitutes must be in the order of the names, defined above, and there are
307    both positive and negative cases. NULL means no substitute. */
308    
309    #ifdef SUPPORT_UCP
310    static const pcre_uchar string_PNd[]  = {
311      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
312      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
313    static const pcre_uchar string_pNd[]  = {
314      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
315      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
316    static const pcre_uchar string_PXsp[] = {
317      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pXsp[] = {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_PXwd[] = {
323      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_pXwd[] = {
326      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    
329    static const pcre_uchar *substitutes[] = {
330      string_PNd,           /* \D */
331      string_pNd,           /* \d */
332      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
333      string_pXsp,          /* \s */
334      string_PXwd,          /* \W */
335      string_pXwd           /* \w */
336    };
337    
338    static const pcre_uchar string_pL[] =   {
339      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
340      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
341    static const pcre_uchar string_pLl[] =  {
342      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
343      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
344    static const pcre_uchar string_pLu[] =  {
345      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
346      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347    static const pcre_uchar string_pXan[] = {
348      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    static const pcre_uchar string_h[] =    {
351      CHAR_BACKSLASH, CHAR_h, '\0' };
352    static const pcre_uchar string_pXps[] = {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_PL[] =   {
356      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_PLl[] =  {
359      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_PLu[] =  {
362      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
363      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
364    static const pcre_uchar string_PXan[] = {
365      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
366      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
367    static const pcre_uchar string_H[] =    {
368      CHAR_BACKSLASH, CHAR_H, '\0' };
369    static const pcre_uchar string_PXps[] = {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    
373    static const pcre_uchar *posix_substitutes[] = {
374      string_pL,            /* alpha */
375      string_pLl,           /* lower */
376      string_pLu,           /* upper */
377      string_pXan,          /* alnum */
378      NULL,                 /* ascii */
379      string_h,             /* blank */
380      NULL,                 /* cntrl */
381      string_pNd,           /* digit */
382      NULL,                 /* graph */
383      NULL,                 /* print */
384      NULL,                 /* punct */
385      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
386      string_pXwd,          /* word */
387      NULL,                 /* xdigit */
388      /* Negated cases */
389      string_PL,            /* ^alpha */
390      string_PLl,           /* ^lower */
391      string_PLu,           /* ^upper */
392      string_PXan,          /* ^alnum */
393      NULL,                 /* ^ascii */
394      string_H,             /* ^blank */
395      NULL,                 /* ^cntrl */
396      string_PNd,           /* ^digit */
397      NULL,                 /* ^graph */
398      NULL,                 /* ^print */
399      NULL,                 /* ^punct */
400      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
401      string_PXwd,          /* ^word */
402      NULL                  /* ^xdigit */
403    };
404    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
405    #endif
406    
407  #define STRING(a)  # a  #define STRING(a)  # a
408  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 410  static const int posix_class_maps[] = {
410  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
411  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
412  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
413  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
414    the number of relocations needed when a shared library is loaded dynamically,
415  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
416    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
417    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
418    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
419    "unrecognized character follows \\",  
420    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
421    substring, so that the whole string ends with \0\0, which can be detected when
422    counting through. */
423    
424    static const char error_texts[] =
425      "no error\0"
426      "\\ at end of pattern\0"
427      "\\c at end of pattern\0"
428      "unrecognized character follows \\\0"
429      "numbers out of order in {} quantifier\0"
430    /* 5 */    /* 5 */
431    "number too big in {} quantifier",    "number too big in {} quantifier\0"
432    "missing terminating ] for character class",    "missing terminating ] for character class\0"
433    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
434    "range out of order in character class",    "range out of order in character class\0"
435    "nothing to repeat",    "nothing to repeat\0"
436    /* 10 */    /* 10 */
437    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
438    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
439    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
440    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
441    "missing )",    "missing )\0"
442    /* 15 */    /* 15 */
443    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
444    "erroffset passed as NULL",    "erroffset passed as NULL\0"
445    "unknown option bit(s) set",    "unknown option bit(s) set\0"
446    "missing ) after comment",    "missing ) after comment\0"
447    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
448    /* 20 */    /* 20 */
449    "regular expression too large",    "regular expression is too large\0"
450    "failed to get memory",    "failed to get memory\0"
451    "unmatched parentheses",    "unmatched parentheses\0"
452    "internal error: code overflow",    "internal error: code overflow\0"
453    "unrecognized character after (?<",    "unrecognized character after (?<\0"
454    /* 25 */    /* 25 */
455    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
456    "malformed number or name after (?(",    "malformed number or name after (?(\0"
457    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
458    "assertion expected after (?(",    "assertion expected after (?(\0"
459    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
460    /* 30 */    /* 30 */
461    "unknown POSIX class name",    "unknown POSIX class name\0"
462    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
463    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
464    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
465    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
466    /* 35 */    /* 35 */
467    "invalid condition (?(0)",    "invalid condition (?(0)\0"
468    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
469    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
470    "number after (?C is > 255",    "number after (?C is > 255\0"
471    "closing ) for (?C expected",    "closing ) for (?C expected\0"
472    /* 40 */    /* 40 */
473    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
474    "unrecognized character after (?P",    "unrecognized character after (?P\0"
475    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
476    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
477    "invalid UTF-8 string",    "invalid UTF-8 string\0"
478    /* 45 */    /* 45 */
479    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
480    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
481    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
482    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
483    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
484    /* 50 */    /* 50 */
485    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
486    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
487    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
488    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
489    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
490    /* 55 */    /* 55 */
491    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
492    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
493    "\\g is not followed by an (optionally braced) non-zero number"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
494  };    "a numbered reference must not be zero\0"
495      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496      /* 60 */
497      "(*VERB) not recognized or malformed\0"
498      "number is too big\0"
499      "subpattern name expected\0"
500      "digit expected after (?+\0"
501      "] is an invalid data character in JavaScript compatibility mode\0"
502      /* 65 */
503      "different names for subpatterns of the same number are not allowed\0"
504      "(*MARK) must have an argument\0"
505      "this version of PCRE is not compiled with Unicode property support\0"
506      "\\c must be followed by an ASCII character\0"
507      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
508      /* 70 */
509      "internal error: unknown opcode in find_fixedlength()\0"
510      "\\N is not supported in a class\0"
511      "too many forward references\0"
512      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
513      "invalid UTF-16 string\0"
514      /* 75 */
515      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516      "character value in \\u.... sequence is too large\0"
517      "invalid UTF-32 string\0"
518      "setting UTF is disabled by the application\0"
519      ;
520    
521  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
522  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 262  For convenience, we use the same bit def Line 534  For convenience, we use the same bit def
534    
535  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
536    
537  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
538  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
539    into a subtraction and unsigned comparison). */
540    
541    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
542    
543    #ifndef EBCDIC
544    
545    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
546    UTF-8 mode. */
547    
548    static const pcre_uint8 digitab[] =
549    {    {
550    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
551    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 298  static const unsigned char digitab[] = Line 580  static const unsigned char digitab[] =
580    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
581    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
582    
583  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
584  static const unsigned char digitab[] =  
585    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
586    
587    static const pcre_uint8 digitab[] =
588    {    {
589    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
590    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 334  static const unsigned char digitab[] = Line 619  static const unsigned char digitab[] =
619    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
620    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
621    
622  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
623    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
624    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
625    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 370  static const unsigned char ebcdic_charta Line 655  static const unsigned char ebcdic_charta
655  #endif  #endif
656    
657    
658  /* Definition to allow mutual recursion */  
659    /*************************************************
660    *            Find an error text                  *
661    *************************************************/
662    
663    /* The error texts are now all in one long string, to save on relocations. As
664    some of the text is of unknown length, we can't use a table of offsets.
665    Instead, just count through the strings. This is not a performance issue
666    because it happens only when there has been a compilation error.
667    
668    Argument:   the error number
669    Returns:    pointer to the error string
670    */
671    
672    static const char *
673    find_error_text(int n)
674    {
675    const char *s = error_texts;
676    for (; n > 0; n--)
677      {
678      while (*s++ != CHAR_NULL) {};
679      if (*s == CHAR_NULL) return "Error text not found (please report)";
680      }
681    return s;
682    }
683    
684    
685    /*************************************************
686    *           Expand the workspace                 *
687    *************************************************/
688    
689    /* This function is called during the second compiling phase, if the number of
690    forward references fills the existing workspace, which is originally a block on
691    the stack. A larger block is obtained from malloc() unless the ultimate limit
692    has been reached or the increase will be rather small.
693    
694    Argument: pointer to the compile data block
695    Returns:  0 if all went well, else an error number
696    */
697    
698    static int
699    expand_workspace(compile_data *cd)
700    {
701    pcre_uchar *newspace;
702    int newsize = cd->workspace_size * 2;
703    
704    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
705    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
706        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
707     return ERR72;
708    
709    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
710    if (newspace == NULL) return ERR21;
711    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
712    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
713    if (cd->workspace_size > COMPILE_WORK_SIZE)
714      (PUBL(free))((void *)cd->start_workspace);
715    cd->start_workspace = newspace;
716    cd->workspace_size = newsize;
717    return 0;
718    }
719    
720    
721    
722    /*************************************************
723    *            Check for counted repeat            *
724    *************************************************/
725    
726    /* This function is called when a '{' is encountered in a place where it might
727    start a quantifier. It looks ahead to see if it really is a quantifier or not.
728    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
729    where the ddds are digits.
730    
731    Arguments:
732      p         pointer to the first char after '{'
733    
734    Returns:    TRUE or FALSE
735    */
736    
737  static BOOL  static BOOL
738    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
739      int *, branch_chain *, compile_data *, int *);  {
740    if (!IS_DIGIT(*p)) return FALSE;
741    p++;
742    while (IS_DIGIT(*p)) p++;
743    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
744    
745    if (*p++ != CHAR_COMMA) return FALSE;
746    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
747    
748    if (!IS_DIGIT(*p)) return FALSE;
749    p++;
750    while (IS_DIGIT(*p)) p++;
751    
752    return (*p == CHAR_RIGHT_CURLY_BRACKET);
753    }
754    
755    
756    
# Line 383  static BOOL Line 759  static BOOL
759  *************************************************/  *************************************************/
760    
761  /* 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
762  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
763  encodes one of the more complicated things such as \d. A backreference to group  which will be placed in chptr. A backreference to group n is returned as
764  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  negative n. When UTF-8 is enabled, a positive value greater than 255 may
765  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  be returned in chptr.
766  ptr is pointing at the \. On exit, it is on the final character of the escape  On entry,ptr is pointing at the \. On exit, it is on the final character of the
767  sequence.  escape sequence.
768    
769  Arguments:  Arguments:
770    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
771      chptr          points to the data character
772    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
773    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
774    options        the options bits    options        the options bits
775    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
776    
777  Returns:         zero or positive => a data character  Returns:         zero => a data character
778                   negative => a special escape sequence                   positive => a special escape sequence
779                   on error, errorptr is set                   negative => a back reference
780                     on error, errorcodeptr is set
781  */  */
782    
783  static int  static int
784  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
785    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
786  {  {
787  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
788  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
789  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
790    pcre_uint32 c;
791    int escape = 0;
792    int i;
793    
794  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
795  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
796    
797  /* 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. */
798    
799  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
800    
801  /* 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
802  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.
803  Otherwise further processing may be required. */  Otherwise further processing may be required. */
804    
805  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
806  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
807  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
808    else if ((i = escapes[c - CHAR_0]) != 0)
809      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
810    
811  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
812  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
813  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
814    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
815  #endif  #endif
816    
817  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
818    
819  else  else
820    {    {
821    const uschar *oldptr;    const pcre_uchar *oldptr;
822    BOOL braced, negated;    BOOL braced, negated, overflow;
823      int s;
824    
825    switch (c)    switch (c)
826      {      {
827      /* 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
828      error. */      error. */
829    
830      case 'l':      case CHAR_l:
831      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
832      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
833      break;      break;
834    
835      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
836      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
837      This is a Perl 5.10 feature. */        {
838          /* In JavaScript, \u must be followed by four hexadecimal numbers.
839          Otherwise it is a lowercase u letter. */
840          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
841            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
842            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
843            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
844            {
845            c = 0;
846            for (i = 0; i < 4; ++i)
847              {
848              register pcre_uint32 cc = *(++ptr);
849    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
850              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
851              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
852    #else           /* EBCDIC coding */
853              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
854              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
855    #endif
856              }
857    
858    #if defined COMPILE_PCRE8
859            if (c > (utf ? 0x10ffffU : 0xffU))
860    #elif defined COMPILE_PCRE16
861            if (c > (utf ? 0x10ffffU : 0xffffU))
862    #elif defined COMPILE_PCRE32
863            if (utf && c > 0x10ffffU)
864    #endif
865              {
866              *errorcodeptr = ERR76;
867              }
868            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
869            }
870          }
871        else
872          *errorcodeptr = ERR37;
873        break;
874    
875        case CHAR_U:
876        /* In JavaScript, \U is an uppercase U letter. */
877        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
878        break;
879    
880        /* In a character class, \g is just a literal "g". Outside a character
881        class, \g must be followed by one of a number of specific things:
882    
883        (1) A number, either plain or braced. If positive, it is an absolute
884        backreference. If negative, it is a relative backreference. This is a Perl
885        5.10 feature.
886    
887        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
888        is part of Perl's movement towards a unified syntax for back references. As
889        this is synonymous with \k{name}, we fudge it up by pretending it really
890        was \k.
891    
892        (3) For Oniguruma compatibility we also support \g followed by a name or a
893        number either in angle brackets or in single quotes. However, these are
894        (possibly recursive) subroutine calls, _not_ backreferences. Just return
895        the ESC_g code (cf \k). */
896    
897        case CHAR_g:
898        if (isclass) break;
899        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
900          {
901          escape = ESC_g;
902          break;
903          }
904    
905        /* Handle the Perl-compatible cases */
906    
907      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
908        {        {
909          const pcre_uchar *p;
910          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
911            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
912          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
913            {
914            escape = ESC_k;
915            break;
916            }
917        braced = TRUE;        braced = TRUE;
918        ptr++;        ptr++;
919        }        }
920      else braced = FALSE;      else braced = FALSE;
921    
922      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
923        {        {
924        negated = TRUE;        negated = TRUE;
925        ptr++;        ptr++;
926        }        }
927      else negated = FALSE;      else negated = FALSE;
928    
929      c = 0;      /* The integer range is limited by the machine's int representation. */
930      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
931        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
932        while (IS_DIGIT(ptr[1]))
933          {
934          if (s > INT_MAX / 10 - 1) /* Integer overflow */
935            {
936            overflow = TRUE;
937            break;
938            }
939          s = s * 10 + (int)(*(++ptr) - CHAR_0);
940          }
941        if (overflow) /* Integer overflow */
942          {
943          while (IS_DIGIT(ptr[1]))
944            ptr++;
945          *errorcodeptr = ERR61;
946          break;
947          }
948    
949      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
950        {        {
951        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
952        return 0;        break;
953          }
954    
955        if (s == 0)
956          {
957          *errorcodeptr = ERR58;
958          break;
959        }        }
960    
961      if (negated)      if (negated)
962        {        {
963        if (c > bracount)        if (s > bracount)
964          {          {
965          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
966          return 0;          break;
967          }          }
968        c = bracount - (c - 1);        s = bracount - (s - 1);
969        }        }
970    
971      c = -(ESC_REF + c);      escape = -s;
972      break;      break;
973    
974      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
# Line 504  else Line 983  else
983      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
984      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
985    
986      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:
987      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
988    
989      if (!isclass)      if (!isclass)
990        {        {
991        oldptr = ptr;        oldptr = ptr;
992        c -= '0';        /* The integer range is limited by the machine's int representation. */
993        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
994          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
995        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
996            {
997            if (s > INT_MAX / 10 - 1) /* Integer overflow */
998              {
999              overflow = TRUE;
1000              break;
1001              }
1002            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1003            }
1004          if (overflow) /* Integer overflow */
1005            {
1006            while (IS_DIGIT(ptr[1]))
1007              ptr++;
1008            *errorcodeptr = ERR61;
1009            break;
1010            }
1011          if (s < 10 || s <= bracount)
1012          {          {
1013          c = -(ESC_REF + c);          escape = -s;
1014          break;          break;
1015          }          }
1016        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 525  else Line 1020  else
1020      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
1021      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1022    
1023      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1024        {        {
1025        ptr--;        ptr--;
1026        c = 0;        c = 0;
# Line 535  else Line 1030  else
1030      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1031      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1032      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1033      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1034      than 3 octal digits. */      but no more than 3 octal digits. */
1035    
1036      case '0':      case CHAR_0:
1037      c -= '0';      c -= CHAR_0;
1038      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1039          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1040      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1041        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1042    #endif
1043      break;      break;
1044    
1045      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
1046      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1047      treated as a data character. */      If not, { is treated as a data character. */
1048    
1049        case CHAR_x:
1050        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1051          {
1052          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1053          Otherwise it is a lowercase x letter. */
1054          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1055            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1056            {
1057            c = 0;
1058            for (i = 0; i < 2; ++i)
1059              {
1060              register pcre_uint32 cc = *(++ptr);
1061    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1062              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1063              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1064    #else           /* EBCDIC coding */
1065              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1066              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1067    #endif
1068              }
1069            }
1070          break;
1071          }
1072    
1073      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1074        {        {
1075        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1076    
1077        c = 0;        c = 0;
1078        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1079          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1080          {          {
1081          register int cc = *pt++;          register pcre_uint32 cc = *pt++;
1082          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1083          count++;  
1084    #ifdef COMPILE_PCRE32
1085  #ifndef EBCDIC  /* ASCII coding */          if (c >= 0x10000000l) { overflow = TRUE; break; }
1086          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #endif
1087          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  
1088    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1089            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1090            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1091  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1092          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1093          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1094    #endif
1095    
1096    #if defined COMPILE_PCRE8
1097            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1098    #elif defined COMPILE_PCRE16
1099            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1100    #elif defined COMPILE_PCRE32
1101            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1102  #endif  #endif
1103          }          }
1104    
1105        if (*pt == '}')        if (overflow)
1106          {          {
1107          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1108            *errorcodeptr = ERR34;
1109            }
1110    
1111          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1112            {
1113            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1114          ptr = pt;          ptr = pt;
1115          break;          break;
1116          }          }
# Line 585  else Line 1122  else
1122      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1123    
1124      c = 0;      c = 0;
1125      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1126        {        {
1127        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1128        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1129  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1130        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1131        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1132  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1133        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1134        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1135  #endif  #endif
1136        }        }
1137      break;      break;
1138    
1139      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1140      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1141        coding is ASCII-specific, but then the whole concept of \cx is
1142      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1143    
1144      case 'c':      case CHAR_c:
1145      c = *(++ptr);      c = *(++ptr);
1146      if (c == 0)      if (c == CHAR_NULL)
1147        {        {
1148        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1149        return 0;        break;
1150        }        }
1151    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1152  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1153      if (c >= 'a' && c <= 'z') c -= 32;        {
1154          *errorcodeptr = ERR68;
1155          break;
1156          }
1157        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1158      c ^= 0x40;      c ^= 0x40;
1159  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1160      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1161      c ^= 0xC0;      c ^= 0xC0;
1162  #endif  #endif
1163      break;      break;
1164    
1165      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1166      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1167      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
1168      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
1169      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1170    
1171      default:      default:
1172      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1179  else
1179      }      }
1180    }    }
1181    
1182  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1183  return c;  newline". PCRE does not support \N{name}. However, it does support
1184  }  quantification such as \N{2,3}. */
1185    
1186    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1187         !is_counted_repeat(ptr+2))
1188      *errorcodeptr = ERR37;
1189    
1190    /* If PCRE_UCP is set, we change the values for \d etc. */
1191    
1192    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1193      escape += (ESC_DU - ESC_D);
1194    
1195    /* Set the pointer to the final character before returning. */
1196    
1197    *ptrptr = ptr;
1198    *chptr = c;
1199    return escape;
1200    }
1201    
1202  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1203  /*************************************************  /*************************************************
# Line 656  escape sequence. Line 1212  escape sequence.
1212  Argument:  Argument:
1213    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1214    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
1215    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
1216      pdataptr       points to an unsigned int that is set to the detailed property value
1217    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1218    
1219  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
1220  */  */
1221    
1222  static int  static BOOL
1223  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1224      unsigned int *pdataptr, int *errorcodeptr)
1225  {  {
1226  int c, i, bot, top;  pcre_uchar c;
1227  const uschar *ptr = *ptrptr;  int i, bot, top;
1228  char name[32];  const pcre_uchar *ptr = *ptrptr;
1229    pcre_uchar name[32];
1230    
1231  c = *(++ptr);  c = *(++ptr);
1232  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1233    
1234  *negptr = FALSE;  *negptr = FALSE;
1235    
1236  /* \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
1237  negation. */  negation. */
1238    
1239  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1240    {    {
1241    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1242      {      {
1243      *negptr = TRUE;      *negptr = TRUE;
1244      ptr++;      ptr++;
1245      }      }
1246    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1247      {      {
1248      c = *(++ptr);      c = *(++ptr);
1249      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1250      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1251      name[i] = c;      name[i] = c;
1252      }      }
1253    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1254    name[i] = 0;    name[i] = 0;
1255    }    }
1256    
# Line 708  else Line 1267  else
1267  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1268    
1269  bot = 0;  bot = 0;
1270  top = _pcre_utt_size;  top = PRIV(utt_size);
1271    
1272  while (bot < top)  while (bot < top)
1273    {    {
1274      int r;
1275    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1276    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1277    if (c == 0)    if (r == 0)
1278      {      {
1279      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1280      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1281        return TRUE;
1282      }      }
1283    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1284    }    }
1285    
1286  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1287  *ptrptr = ptr;  *ptrptr = ptr;
1288  return -1;  return FALSE;
1289    
1290  ERROR_RETURN:  ERROR_RETURN:
1291  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1292  *ptrptr = ptr;  *ptrptr = ptr;
1293  return -1;  return FALSE;
1294  }  }
1295  #endif  #endif
1296    
# Line 737  return -1; Line 1298  return -1;
1298    
1299    
1300  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1301  *         Read repeat counts                     *  *         Read repeat counts                     *
1302  *************************************************/  *************************************************/
1303    
# Line 788  Returns:         pointer to '}' on succe Line 1316  Returns:         pointer to '}' on succe
1316                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1317  */  */
1318    
1319  static const uschar *  static const pcre_uchar *
1320  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)
1321  {  {
1322  int min = 0;  int min = 0;
1323  int max = -1;  int max = -1;
# Line 797  int max = -1; Line 1325  int max = -1;
1325  /* 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
1326  an integer overflow. */  an integer overflow. */
1327    
1328  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1329  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1330    {    {
1331    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1335  if (min < 0 || min > 65535)
1335  /* 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.
1336  Also, max must not be less than min. */  Also, max must not be less than min. */
1337    
1338  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1339    {    {
1340    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1341      {      {
1342      max = 0;      max = 0;
1343      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1344      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1345        {        {
1346        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1365  return p;
1365    
1366    
1367  /*************************************************  /*************************************************
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We know that if (?P< is encountered, the name will  
 be terminated by '>' because that is checked in the first pass.  
   
 Arguments:  
   ptr          current position in the pattern  
   count        current count of capturing parens so far encountered  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  
   BOOL xmode)  
 {  
 const uschar *thisname;  
   
 for (; *ptr != 0; ptr++)  
   {  
   int term;  
   
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == '\\')  
     {  
     if (*(++ptr) == 0) return -1;  
     if (*ptr == 'Q') for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != '\\');  
       if (*ptr == 0) return -1;  
       if (*(++ptr) == 'E') break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes */  
   
   if (*ptr == '[')  
     {  
     while (*(++ptr) != ']')  
       {  
       if (*ptr == '\\')  
         {  
         if (*(++ptr) == 0) return -1;  
         if (*ptr == 'Q') for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != '\\');  
           if (*ptr == 0) return -1;  
           if (*(++ptr) == 'E') break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == '#')  
     {  
     while (*(++ptr) != 0 && *ptr != '\n');  
     if (*ptr == 0) return -1;  
     continue;  
     }  
   
   /* An opening parens must now be a real metacharacter */  
   
   if (*ptr != '(') continue;  
   if (ptr[1] != '?')  
     {  
     count++;  
     if (name == NULL && count == lorn) return count;  
     continue;  
     }  
   
   ptr += 2;  
   if (*ptr == 'P') ptr++;                      /* Allow optional P */  
   
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
   
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
   
   if (name == NULL && count == lorn) return count;  
   term = *ptr++;  
   if (term == '<') term = '>';  
   thisname = ptr;  
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
   }  
   
 return -1;  
 }  
   
   
   
 /*************************************************  
1368  *      Find first significant op code            *  *      Find first significant op code            *
1369  *************************************************/  *************************************************/
1370    
1371  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1372  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
1373  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
1374  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
1375  assertions, and also the \b assertion; for others it does not.  does not.
1376    
1377  Arguments:  Arguments:
1378    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  
1379    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1380    
1381  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1382  */  */
1383    
1384  static const uschar*  static const pcre_uchar*
1385  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1386  {  {
1387  for (;;)  for (;;)
1388    {    {
1389    switch ((int)*code)    switch ((int)*code)
1390      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1391      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1392      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1393      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1394      if (!skipassert) return code;      if (!skipassert) return code;
1395      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1396      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1397      break;      break;
1398    
1399      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1403  for (;;)
1403    
1404      case OP_CALLOUT:      case OP_CALLOUT:
1405      case OP_CREF:      case OP_CREF:
1406        case OP_NCREF:
1407      case OP_RREF:      case OP_RREF:
1408        case OP_NRREF:
1409      case OP_DEF:      case OP_DEF:
1410      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1411      break;      break;
1412    
1413      default:      default:
# Line 1012  for (;;) Line 1421  for (;;)
1421    
1422    
1423  /*************************************************  /*************************************************
1424  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1425  *************************************************/  *************************************************/
1426    
1427  /* 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,
1428  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.
1429  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
1430    temporarily terminated with OP_END when this function is called.
1431    
1432    This function is called when a backward assertion is encountered, so that if it
1433    fails, the error message can point to the correct place in the pattern.
1434    However, we cannot do this when the assertion contains subroutine calls,
1435    because they can be forward references. We solve this by remembering this case
1436    and doing the check at the end; a flag specifies which mode we are running in.
1437    
1438  Arguments:  Arguments:
1439    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1440    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1441      atend    TRUE if called when the pattern is complete
1442  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1443               or -2 if \C was encountered  
1444    Returns:   the fixed length,
1445                 or -1 if there is no fixed length,
1446                 or -2 if \C was encountered (in UTF-8 mode only)
1447                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1448                 or -4 if an unknown opcode was encountered (internal error)
1449  */  */
1450    
1451  static int  static int
1452  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1453  {  {
1454  int length = -1;  int length = -1;
1455    
1456  register int branchlength = 0;  register int branchlength = 0;
1457  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1458    
1459  /* 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
1460  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1041  branch, check the length against that of Line 1462  branch, check the length against that of
1462  for (;;)  for (;;)
1463    {    {
1464    int d;    int d;
1465    register int op = *cc;    pcre_uchar *ce, *cs;
1466      register pcre_uchar op = *cc;
1467    
1468    switch (op)    switch (op)
1469      {      {
1470        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1471        OP_BRA (normal non-capturing bracket) because the other variants of these
1472        opcodes are all concerned with unlimited repeated groups, which of course
1473        are not of fixed length. */
1474    
1475      case OP_CBRA:      case OP_CBRA:
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478        case OP_ONCE_NC:
1479      case OP_COND:      case OP_COND:
1480      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1481      if (d < 0) return d;      if (d < 0) return d;
1482      branchlength += d;      branchlength += d;
1483      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1484      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1485      break;      break;
1486    
1487      /* 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.
1488      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
1489      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
1490        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1491        because they all imply an unlimited repeat. */
1492    
1493      case OP_ALT:      case OP_ALT:
1494      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1495      case OP_END:      case OP_END:
1496        case OP_ACCEPT:
1497        case OP_ASSERT_ACCEPT:
1498      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1499        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1500      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1502  for (;;)
1502      branchlength = 0;      branchlength = 0;
1503      break;      break;
1504    
1505        /* A true recursion implies not fixed length, but a subroutine call may
1506        be OK. If the subroutine is a forward reference, we can't deal with
1507        it until the end of the pattern, so return -3. */
1508    
1509        case OP_RECURSE:
1510        if (!atend) return -3;
1511        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1512        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1513        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1514        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1515        if (d < 0) return d;
1516        branchlength += d;
1517        cc += 1 + LINK_SIZE;
1518        break;
1519    
1520      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1521    
1522      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1524  for (;;)
1524      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1525      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1526      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1527      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1528        break;
1529    
1530      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1531    
1532      case OP_REVERSE:      case OP_MARK:
1533        case OP_PRUNE_ARG:
1534        case OP_SKIP_ARG:
1535        case OP_THEN_ARG:
1536        cc += cc[1] + PRIV(OP_lengths)[*cc];
1537        break;
1538    
1539        case OP_CALLOUT:
1540        case OP_CIRC:
1541        case OP_CIRCM:
1542        case OP_CLOSE:
1543        case OP_COMMIT:
1544      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1545      case OP_DEF:      case OP_DEF:
1546      case OP_OPT:      case OP_DOLL:
1547      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1548      case OP_EOD:      case OP_EOD:
1549      case OP_EODN:      case OP_EODN:
1550      case OP_CIRC:      case OP_FAIL:
1551      case OP_DOLL:      case OP_NCREF:
1552        case OP_NRREF:
1553      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1554        case OP_PRUNE:
1555        case OP_REVERSE:
1556        case OP_RREF:
1557        case OP_SET_SOM:
1558        case OP_SKIP:
1559        case OP_SOD:
1560        case OP_SOM:
1561        case OP_THEN:
1562      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1563      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1564      break;      break;
1565    
1566      /* Handle literal characters */      /* Handle literal characters */
1567    
1568      case OP_CHAR:      case OP_CHAR:
1569      case OP_CHARNC:      case OP_CHARI:
1570      case OP_NOT:      case OP_NOT:
1571        case OP_NOTI:
1572      branchlength++;      branchlength++;
1573      cc += 2;      cc += 2;
1574  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1575      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1576  #endif  #endif
1577      break;      break;
1578    
# Line 1119  for (;;) Line 1580  for (;;)
1580      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1581    
1582      case OP_EXACT:      case OP_EXACT:
1583      branchlength += GET2(cc,1);      case OP_EXACTI:
1584      cc += 4;      case OP_NOTEXACT:
1585  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1586      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1587        {      cc += 2 + IMM2_SIZE;
1588        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1589        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1590  #endif  #endif
1591      break;      break;
1592    
1593      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1594      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1595      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1596          cc += 2;
1597        cc += 1 + IMM2_SIZE + 1;
1598      break;      break;
1599    
1600      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1604  for (;;)
1604      cc += 2;      cc += 2;
1605      /* Fall through */      /* Fall through */
1606    
1607        case OP_HSPACE:
1608        case OP_VSPACE:
1609        case OP_NOT_HSPACE:
1610        case OP_NOT_VSPACE:
1611      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1612      case OP_DIGIT:      case OP_DIGIT:
1613      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1615  for (;;)
1615      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1616      case OP_WORDCHAR:      case OP_WORDCHAR:
1617      case OP_ANY:      case OP_ANY:
1618        case OP_ALLANY:
1619      branchlength++;      branchlength++;
1620      cc++;      cc++;
1621      break;      break;
1622    
1623      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1624        otherwise \C is coded as OP_ALLANY. */
1625    
1626      case OP_ANYBYTE:      case OP_ANYBYTE:
1627      return -2;      return -2;
1628    
1629      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1630    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1631      case OP_CLASS:      case OP_CLASS:
1632      case OP_NCLASS:      case OP_NCLASS:
1633      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1634        case OP_XCLASS:
1635        /* The original code caused an unsigned overflow in 64 bit systems,
1636        so now we use a conditional statement. */
1637        if (op == OP_XCLASS)
1638          cc += GET(cc, 1);
1639        else
1640          cc += PRIV(OP_lengths)[OP_CLASS];
1641    #else
1642        cc += PRIV(OP_lengths)[OP_CLASS];
1643    #endif
1644    
1645      switch (*cc)      switch (*cc)
1646        {        {
1647          case OP_CRPLUS:
1648          case OP_CRMINPLUS:
1649        case OP_CRSTAR:        case OP_CRSTAR:
1650        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1651        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1179  for (;;) Line 1654  for (;;)
1654    
1655        case OP_CRRANGE:        case OP_CRRANGE:
1656        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1657        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1658        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1659        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1660        break;        break;
1661    
1662        default:        default:
# Line 1191  for (;;) Line 1666  for (;;)
1666    
1667      /* Anything else is variable length */      /* Anything else is variable length */
1668    
1669      default:      case OP_ANYNL:
1670        case OP_BRAMINZERO:
1671        case OP_BRAPOS:
1672        case OP_BRAPOSZERO:
1673        case OP_BRAZERO:
1674        case OP_CBRAPOS:
1675        case OP_EXTUNI:
1676        case OP_KETRMAX:
1677        case OP_KETRMIN:
1678        case OP_KETRPOS:
1679        case OP_MINPLUS:
1680        case OP_MINPLUSI:
1681        case OP_MINQUERY:
1682        case OP_MINQUERYI:
1683        case OP_MINSTAR:
1684        case OP_MINSTARI:
1685        case OP_MINUPTO:
1686        case OP_MINUPTOI:
1687        case OP_NOTMINPLUS:
1688        case OP_NOTMINPLUSI:
1689        case OP_NOTMINQUERY:
1690        case OP_NOTMINQUERYI:
1691        case OP_NOTMINSTAR:
1692        case OP_NOTMINSTARI:
1693        case OP_NOTMINUPTO:
1694        case OP_NOTMINUPTOI:
1695        case OP_NOTPLUS:
1696        case OP_NOTPLUSI:
1697        case OP_NOTPOSPLUS:
1698        case OP_NOTPOSPLUSI:
1699        case OP_NOTPOSQUERY:
1700        case OP_NOTPOSQUERYI:
1701        case OP_NOTPOSSTAR:
1702        case OP_NOTPOSSTARI:
1703        case OP_NOTPOSUPTO:
1704        case OP_NOTPOSUPTOI:
1705        case OP_NOTQUERY:
1706        case OP_NOTQUERYI:
1707        case OP_NOTSTAR:
1708        case OP_NOTSTARI:
1709        case OP_NOTUPTO:
1710        case OP_NOTUPTOI:
1711        case OP_PLUS:
1712        case OP_PLUSI:
1713        case OP_POSPLUS:
1714        case OP_POSPLUSI:
1715        case OP_POSQUERY:
1716        case OP_POSQUERYI:
1717        case OP_POSSTAR:
1718        case OP_POSSTARI:
1719        case OP_POSUPTO:
1720        case OP_POSUPTOI:
1721        case OP_QUERY:
1722        case OP_QUERYI:
1723        case OP_REF:
1724        case OP_REFI:
1725        case OP_SBRA:
1726        case OP_SBRAPOS:
1727        case OP_SCBRA:
1728        case OP_SCBRAPOS:
1729        case OP_SCOND:
1730        case OP_SKIPZERO:
1731        case OP_STAR:
1732        case OP_STARI:
1733        case OP_TYPEMINPLUS:
1734        case OP_TYPEMINQUERY:
1735        case OP_TYPEMINSTAR:
1736        case OP_TYPEMINUPTO:
1737        case OP_TYPEPLUS:
1738        case OP_TYPEPOSPLUS:
1739        case OP_TYPEPOSQUERY:
1740        case OP_TYPEPOSSTAR:
1741        case OP_TYPEPOSUPTO:
1742        case OP_TYPEQUERY:
1743        case OP_TYPESTAR:
1744        case OP_TYPEUPTO:
1745        case OP_UPTO:
1746        case OP_UPTOI:
1747      return -1;      return -1;
1748    
1749        /* Catch unrecognized opcodes so that when new ones are added they
1750        are not forgotten, as has happened in the past. */
1751    
1752        default:
1753        return -4;
1754      }      }
1755    }    }
1756  /* Control never gets here */  /* Control never gets here */
# Line 1202  for (;;) Line 1760  for (;;)
1760    
1761    
1762  /*************************************************  /*************************************************
1763  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1764  *************************************************/  *************************************************/
1765    
1766  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1767  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1768    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1769    so that it can be called from pcre_study() when finding the minimum matching
1770    length.
1771    
1772  Arguments:  Arguments:
1773    code        points to start of expression    code        points to start of expression
1774    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
1775    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1776    
1777  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
1778  */  */
1779    
1780  static const uschar *  const pcre_uchar *
1781  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
1782  {  {
1783  for (;;)  for (;;)
1784    {    {
1785    register int c = *code;    register pcre_uchar c = *code;
1786    
1787    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1788    
1789    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1230  for (;;) Line 1792  for (;;)
1792    
1793    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1794    
1795      /* Handle recursion */
1796    
1797      else if (c == OP_REVERSE)
1798        {
1799        if (number < 0) return (pcre_uchar *)code;
1800        code += PRIV(OP_lengths)[c];
1801        }
1802    
1803    /* Handle capturing bracket */    /* Handle capturing bracket */
1804    
1805    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1806               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1807      {      {
1808      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
1809      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
1810      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1811      }      }
1812    
1813      /* Otherwise, we can get the item's length from the table, except that for
1814      repeated character types, we have to test for \p and \P, which have an extra
1815      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1816      must add in its length. */
1817    
1818      else
1819        {
1820        switch(c)
1821          {
1822          case OP_TYPESTAR:
1823          case OP_TYPEMINSTAR:
1824          case OP_TYPEPLUS:
1825          case OP_TYPEMINPLUS:
1826          case OP_TYPEQUERY:
1827          case OP_TYPEMINQUERY:
1828          case OP_TYPEPOSSTAR:
1829          case OP_TYPEPOSPLUS:
1830          case OP_TYPEPOSQUERY:
1831          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1832          break;
1833    
1834          case OP_TYPEUPTO:
1835          case OP_TYPEMINUPTO:
1836          case OP_TYPEEXACT:
1837          case OP_TYPEPOSUPTO:
1838          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
1839            code += 2;
1840          break;
1841    
1842          case OP_MARK:
1843          case OP_PRUNE_ARG:
1844          case OP_SKIP_ARG:
1845          case OP_THEN_ARG:
1846          code += code[1];
1847          break;
1848          }
1849    
1850        /* Add in the fixed length from the table */
1851    
1852        code += PRIV(OP_lengths)[c];
1853    
1854    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1855    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
1856    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
1857    
1858    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1859      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
1860        {        {
1861        case OP_CHAR:        case OP_CHAR:
1862        case OP_CHARNC:        case OP_CHARI:
1863        case OP_EXACT:        case OP_EXACT:
1864          case OP_EXACTI:
1865        case OP_UPTO:        case OP_UPTO:
1866          case OP_UPTOI:
1867        case OP_MINUPTO:        case OP_MINUPTO:
1868          case OP_MINUPTOI:
1869        case OP_POSUPTO:        case OP_POSUPTO:
1870          case OP_POSUPTOI:
1871        case OP_STAR:        case OP_STAR:
1872          case OP_STARI:
1873        case OP_MINSTAR:        case OP_MINSTAR:
1874          case OP_MINSTARI:
1875        case OP_POSSTAR:        case OP_POSSTAR:
1876          case OP_POSSTARI:
1877        case OP_PLUS:        case OP_PLUS:
1878          case OP_PLUSI:
1879        case OP_MINPLUS:        case OP_MINPLUS:
1880          case OP_MINPLUSI:
1881        case OP_POSPLUS:        case OP_POSPLUS:
1882          case OP_POSPLUSI:
1883        case OP_QUERY:        case OP_QUERY:
1884          case OP_QUERYI:
1885        case OP_MINQUERY:        case OP_MINQUERY:
1886          case OP_MINQUERYI:
1887        case OP_POSQUERY:        case OP_POSQUERY:
1888        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
1889          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
1890        break;        break;
1891        }        }
1892    #else
1893        (void)(utf);  /* Keep compiler happy by referencing function argument */
1894  #endif  #endif
1895      }      }
1896    }    }
# Line 1283  instance of OP_RECURSE. Line 1907  instance of OP_RECURSE.
1907    
1908  Arguments:  Arguments:
1909    code        points to start of expression    code        points to start of expression
1910    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
1911    
1912  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
1913  */  */
1914    
1915  static const uschar *  static const pcre_uchar *
1916  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
1917  {  {
1918  for (;;)  for (;;)
1919    {    {
1920    register int c = *code;    register pcre_uchar c = *code;
1921    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1922    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1923    
# Line 1303  for (;;) Line 1927  for (;;)
1927    
1928    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1929    
1930    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1931    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1932    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1933    bytes. */    must add in its length. */
1934    
1935    else    else
1936      {      {
1937      code += _pcre_OP_lengths[c];      switch(c)
1938  #ifdef SUPPORT_UTF8        {
1939      if (utf8) switch(c)        case OP_TYPESTAR:
1940          case OP_TYPEMINSTAR:
1941          case OP_TYPEPLUS:
1942          case OP_TYPEMINPLUS:
1943          case OP_TYPEQUERY:
1944          case OP_TYPEMINQUERY:
1945          case OP_TYPEPOSSTAR:
1946          case OP_TYPEPOSPLUS:
1947          case OP_TYPEPOSQUERY:
1948          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1949          break;
1950    
1951          case OP_TYPEPOSUPTO:
1952          case OP_TYPEUPTO:
1953          case OP_TYPEMINUPTO:
1954          case OP_TYPEEXACT:
1955          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
1956            code += 2;
1957          break;
1958    
1959          case OP_MARK:
1960          case OP_PRUNE_ARG:
1961          case OP_SKIP_ARG:
1962          case OP_THEN_ARG:
1963          code += code[1];
1964          break;
1965          }
1966    
1967        /* Add in the fixed length from the table */
1968    
1969        code += PRIV(OP_lengths)[c];
1970    
1971        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1972        by a multi-byte character. The length in the table is a minimum, so we have
1973        to arrange to skip the extra bytes. */
1974    
1975    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1976        if (utf) switch(c)
1977        {        {
1978        case OP_CHAR:        case OP_CHAR:
1979        case OP_CHARNC:        case OP_CHARI:
1980          case OP_NOT:
1981          case OP_NOTI:
1982        case OP_EXACT:        case OP_EXACT:
1983          case OP_EXACTI:
1984          case OP_NOTEXACT:
1985          case OP_NOTEXACTI:
1986        case OP_UPTO:        case OP_UPTO:
1987          case OP_UPTOI:
1988          case OP_NOTUPTO:
1989          case OP_NOTUPTOI:
1990        case OP_MINUPTO:        case OP_MINUPTO:
1991          case OP_MINUPTOI:
1992          case OP_NOTMINUPTO:
1993          case OP_NOTMINUPTOI:
1994        case OP_POSUPTO:        case OP_POSUPTO:
1995          case OP_POSUPTOI:
1996          case OP_NOTPOSUPTO:
1997          case OP_NOTPOSUPTOI:
1998        case OP_STAR:        case OP_STAR:
1999          case OP_STARI:
2000          case OP_NOTSTAR:
2001          case OP_NOTSTARI:
2002        case OP_MINSTAR:        case OP_MINSTAR:
2003          case OP_MINSTARI:
2004          case OP_NOTMINSTAR:
2005          case OP_NOTMINSTARI:
2006        case OP_POSSTAR:        case OP_POSSTAR:
2007          case OP_POSSTARI:
2008          case OP_NOTPOSSTAR:
2009          case OP_NOTPOSSTARI:
2010        case OP_PLUS:        case OP_PLUS:
2011          case OP_PLUSI:
2012          case OP_NOTPLUS:
2013          case OP_NOTPLUSI:
2014        case OP_MINPLUS:        case OP_MINPLUS:
2015          case OP_MINPLUSI:
2016          case OP_NOTMINPLUS:
2017          case OP_NOTMINPLUSI:
2018        case OP_POSPLUS:        case OP_POSPLUS:
2019          case OP_POSPLUSI:
2020          case OP_NOTPOSPLUS:
2021          case OP_NOTPOSPLUSI:
2022        case OP_QUERY:        case OP_QUERY:
2023          case OP_QUERYI:
2024          case OP_NOTQUERY:
2025          case OP_NOTQUERYI:
2026        case OP_MINQUERY:        case OP_MINQUERY:
2027          case OP_MINQUERYI:
2028          case OP_NOTMINQUERY:
2029          case OP_NOTMINQUERYI:
2030        case OP_POSQUERY:        case OP_POSQUERY:
2031        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2032          case OP_NOTPOSQUERY:
2033          case OP_NOTPOSQUERYI:
2034          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2035        break;        break;
2036        }        }
2037    #else
2038        (void)(utf);  /* Keep compiler happy by referencing function argument */
2039  #endif  #endif
2040      }      }
2041    }    }
# Line 1347  for (;;) Line 2051  for (;;)
2051  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2052  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
2053  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2054  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2055  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2056    bracket whose current branch will already have been scanned.
2057    
2058  Arguments:  Arguments:
2059    code        points to start of search    code        points to start of search
2060    endcode     points to where to stop    endcode     points to where to stop
2061    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2062      cd          contains pointers to tables etc.
2063      recurses    chain of recurse_check to catch mutual recursion
2064    
2065  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2066  */  */
2067    
2068    typedef struct recurse_check {
2069      struct recurse_check *prev;
2070      const pcre_uchar *group;
2071    } recurse_check;
2072    
2073  static BOOL  static BOOL
2074  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2075      BOOL utf, compile_data *cd, recurse_check *recurses)
2076  {  {
2077  register int c;  register pcre_uchar c;
2078  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2079    
2080    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2081       code < endcode;       code < endcode;
2082       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2083    {    {
2084    const uschar *ccode;    const pcre_uchar *ccode;
2085    
2086    c = *code;    c = *code;
2087    
2088    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2089      first_significant_code() with a TRUE final argument. */
2090    
2091      if (c == OP_ASSERT)
2092        {
2093        do code += GET(code, 1); while (*code == OP_ALT);
2094        c = *code;
2095        continue;
2096        }
2097    
2098      /* For a recursion/subroutine call, if its end has been reached, which
2099      implies a backward reference subroutine call, we can scan it. If it's a
2100      forward reference subroutine call, we can't. To detect forward reference
2101      we have to scan up the list that is kept in the workspace. This function is
2102      called only when doing the real compile, not during the pre-compile that
2103      measures the size of the compiled pattern. */
2104    
2105      if (c == OP_RECURSE)
2106      {      {
2107        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2108      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2109    
2110      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2111        when called to scan a completed pattern by setting cd->start_workspace to
2112        NULL. */
2113    
2114        if (cd->start_workspace != NULL)
2115          {
2116          const pcre_uchar *tcode;
2117          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2118            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2119          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2120          }
2121    
2122        /* If we are scanning a completed pattern, there are no forward references
2123        and all groups are complete. We need to detect whether this is a recursive
2124        call, as otherwise there will be an infinite loop. If it is a recursion,
2125        just skip over it. Simple recursions are easily detected. For mutual
2126        recursions we keep a chain on the stack. */
2127    
2128        else
2129          {
2130          recurse_check *r = recurses;
2131          const pcre_uchar *endgroup = scode;
2132    
2133          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2134          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2135    
2136          for (r = recurses; r != NULL; r = r->prev)
2137            if (r->group == scode) break;
2138          if (r != NULL) continue;   /* Mutual recursion */
2139          }
2140    
2141        /* Completed reference; scan the referenced group, remembering it on the
2142        stack chain to detect mutual recursions. */
2143    
2144      empty_branch = FALSE;      empty_branch = FALSE;
2145        this_recurse.prev = recurses;
2146        this_recurse.group = scode;
2147    
2148      do      do
2149        {        {
2150        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2151            {
2152          empty_branch = TRUE;          empty_branch = TRUE;
2153        code += GET(code, 1);          break;
2154            }
2155          scode += GET(scode, 1);
2156        }        }
2157      while (*code == OP_ALT);      while (*scode == OP_ALT);
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2158    
2159      /* Move past the KET and fudge things so that the increment in the "for"      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2160      above has no effect. */      continue;
2161        }
2162    
2163      c = OP_END;    /* Groups with zero repeats can of course be empty; skip them. */
2164      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2165      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2166          c == OP_BRAPOSZERO)
2167        {
2168        code += PRIV(OP_lengths)[c];
2169        do code += GET(code, 1); while (*code == OP_ALT);
2170        c = *code;
2171        continue;
2172        }
2173    
2174      /* A nested group that is already marked as "could be empty" can just be
2175      skipped. */
2176    
2177      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2178          c == OP_SCBRA || c == OP_SCBRAPOS)
2179        {
2180        do code += GET(code, 1); while (*code == OP_ALT);
2181        c = *code;
2182        continue;
2183        }
2184    
2185      /* For other groups, scan the branches. */
2186    
2187      if (c == OP_BRA  || c == OP_BRAPOS ||
2188          c == OP_CBRA || c == OP_CBRAPOS ||
2189          c == OP_ONCE || c == OP_ONCE_NC ||
2190          c == OP_COND)
2191        {
2192        BOOL empty_branch;
2193        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2194    
2195        /* If a conditional group has only one branch, there is a second, implied,
2196        empty branch, so just skip over the conditional, because it could be empty.
2197        Otherwise, scan the individual branches of the group. */
2198    
2199        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2200          code += GET(code, 1);
2201        else
2202          {
2203          empty_branch = FALSE;
2204          do
2205            {
2206            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2207              empty_branch = TRUE;
2208            code += GET(code, 1);
2209            }
2210          while (*code == OP_ALT);
2211          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2212          }
2213    
2214        c = *code;
2215      continue;      continue;
2216      }      }
2217    
# Line 1399  for (code = first_significant_code(code Line 2219  for (code = first_significant_code(code
2219    
2220    switch (c)    switch (c)
2221      {      {
2222      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2223        cannot be represented just by a bit map. This includes negated single
2224        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2225        actual length is stored in the compiled code, so we must update "code"
2226        here. */
2227    
2228  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2229      case OP_XCLASS:      case OP_XCLASS:
2230      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2231      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2232  #endif  #endif
2233    
2234      case OP_CLASS:      case OP_CLASS:
2235      case OP_NCLASS:      case OP_NCLASS:
2236      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2237    
2238  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2239      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2240  #endif  #endif
2241    
# Line 1437  for (code = first_significant_code(code Line 2261  for (code = first_significant_code(code
2261    
2262      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2263    
2264        case OP_ANY:
2265        case OP_ALLANY:
2266        case OP_ANYBYTE:
2267    
2268      case OP_PROP:      case OP_PROP:
2269      case OP_NOTPROP:      case OP_NOTPROP:
2270        case OP_ANYNL:
2271    
2272        case OP_NOT_HSPACE:
2273        case OP_HSPACE:
2274        case OP_NOT_VSPACE:
2275        case OP_VSPACE:
2276      case OP_EXTUNI:      case OP_EXTUNI:
2277    
2278      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2279      case OP_DIGIT:      case OP_DIGIT:
2280      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2281      case OP_WHITESPACE:      case OP_WHITESPACE:
2282      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2283      case OP_WORDCHAR:      case OP_WORDCHAR:
2284      case OP_ANY:  
     case OP_ANYBYTE:  
2285      case OP_CHAR:      case OP_CHAR:
2286      case OP_CHARNC:      case OP_CHARI:
2287      case OP_NOT:      case OP_NOT:
2288        case OP_NOTI:
2289    
2290      case OP_PLUS:      case OP_PLUS:
2291        case OP_PLUSI:
2292      case OP_MINPLUS:      case OP_MINPLUS:
2293      case OP_POSPLUS:      case OP_MINPLUSI:
2294      case OP_EXACT:  
2295      case OP_NOTPLUS:      case OP_NOTPLUS:
2296        case OP_NOTPLUSI:
2297      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2298        case OP_NOTMINPLUSI:
2299    
2300        case OP_POSPLUS:
2301        case OP_POSPLUSI:
2302      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2303        case OP_NOTPOSPLUSI:
2304    
2305        case OP_EXACT:
2306        case OP_EXACTI:
2307      case OP_NOTEXACT:      case OP_NOTEXACT:
2308        case OP_NOTEXACTI:
2309    
2310      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2311      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2312      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2313      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2314    
2315      return FALSE;      return FALSE;
2316    
2317        /* These are going to continue, as they may be empty, but we have to
2318        fudge the length for the \p and \P cases. */
2319    
2320        case OP_TYPESTAR:
2321        case OP_TYPEMINSTAR:
2322        case OP_TYPEPOSSTAR:
2323        case OP_TYPEQUERY:
2324        case OP_TYPEMINQUERY:
2325        case OP_TYPEPOSQUERY:
2326        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2327        break;
2328    
2329        /* Same for these */
2330    
2331        case OP_TYPEUPTO:
2332        case OP_TYPEMINUPTO:
2333        case OP_TYPEPOSUPTO:
2334        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2335          code += 2;
2336        break;
2337    
2338      /* End of branch */      /* End of branch */
2339    
2340      case OP_KET:      case OP_KET:
2341      case OP_KETRMAX:      case OP_KETRMAX:
2342      case OP_KETRMIN:      case OP_KETRMIN:
2343        case OP_KETRPOS:
2344      case OP_ALT:      case OP_ALT:
2345      return TRUE;      return TRUE;
2346    
2347      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2348      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2349        followed by a multibyte character. */
2350    
2351  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2352      case OP_STAR:      case OP_STAR:
2353        case OP_STARI:
2354        case OP_NOTSTAR:
2355        case OP_NOTSTARI:
2356    
2357      case OP_MINSTAR:      case OP_MINSTAR:
2358        case OP_MINSTARI:
2359        case OP_NOTMINSTAR:
2360        case OP_NOTMINSTARI:
2361    
2362      case OP_POSSTAR:      case OP_POSSTAR:
2363        case OP_POSSTARI:
2364        case OP_NOTPOSSTAR:
2365        case OP_NOTPOSSTARI:
2366    
2367      case OP_QUERY:      case OP_QUERY:
2368        case OP_QUERYI:
2369        case OP_NOTQUERY:
2370        case OP_NOTQUERYI:
2371    
2372      case OP_MINQUERY:      case OP_MINQUERY:
2373        case OP_MINQUERYI:
2374        case OP_NOTMINQUERY:
2375        case OP_NOTMINQUERYI:
2376    
2377      case OP_POSQUERY:      case OP_POSQUERY:
2378        case OP_POSQUERYI:
2379        case OP_NOTPOSQUERY:
2380        case OP_NOTPOSQUERYI:
2381    
2382        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2383        break;
2384    
2385      case OP_UPTO:      case OP_UPTO:
2386        case OP_UPTOI:
2387        case OP_NOTUPTO:
2388        case OP_NOTUPTOI:
2389    
2390      case OP_MINUPTO:      case OP_MINUPTO:
2391        case OP_MINUPTOI:
2392        case OP_NOTMINUPTO:
2393        case OP_NOTMINUPTOI:
2394    
2395      case OP_POSUPTO:      case OP_POSUPTO:
2396      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2397        case OP_NOTPOSUPTO:
2398        case OP_NOTPOSUPTOI:
2399    
2400        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2401      break;      break;
2402  #endif  #endif
2403    
2404        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2405        string. */
2406    
2407        case OP_MARK:
2408        case OP_PRUNE_ARG:
2409        case OP_SKIP_ARG:
2410        case OP_THEN_ARG:
2411        code += code[1];
2412        break;
2413    
2414        /* None of the remaining opcodes are required to match a character. */
2415    
2416        default:
2417        break;
2418      }      }
2419    }    }
2420    
# Line 1505  return TRUE; Line 2431  return TRUE;
2431  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
2432  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,
2433  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.
2434    This function is called only during the real compile, not during the
2435    pre-compile.
2436    
2437  Arguments:  Arguments:
2438    code        points to start of the recursion    code        points to start of the recursion
2439    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2440    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2441    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2442      cd          pointers to tables etc
2443    
2444  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2445  */  */
2446    
2447  static BOOL  static BOOL
2448  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2449    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2450  {  {
2451  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2452    {    {
2453    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2454        return FALSE;
2455    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2456    }    }
2457  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2464  return TRUE;
2464  *************************************************/  *************************************************/
2465    
2466  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2467  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2468  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2469  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2470    
2471    Originally, this function only recognized a sequence of letters between the
2472    terminators, but it seems that Perl recognizes any sequence of characters,
2473    though of course unknown POSIX names are subsequently rejected. Perl gives an
2474    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2475    didn't consider this to be a POSIX class. Likewise for [:1234:].
2476    
2477    The problem in trying to be exactly like Perl is in the handling of escapes. We
2478    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2479    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2480    below handles the special case of \], but does not try to do any other escape
2481    processing. This makes it different from Perl for cases such as [:l\ower:]
2482    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2483    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2484    I think.
2485    
2486    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2487    It seems that the appearance of a nested POSIX class supersedes an apparent
2488    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2489    a digit.
2490    
2491    In Perl, unescaped square brackets may also appear as part of class names. For
2492    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2493    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2494    seem right at all. PCRE does not allow closing square brackets in POSIX class
2495    names.
2496    
2497  Argument:  Arguments:
2498    ptr      pointer to the initial [    ptr      pointer to the initial [
2499    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2500    
2501  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2502  */  */
2503    
2504  static BOOL  static BOOL
2505  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2506  {  {
2507  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
2508  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2509  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != CHAR_NULL; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2510    {    {
2511    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2512    return TRUE;      ptr++;
2513      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2514      else
2515        {
2516        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2517          {
2518          *endptr = ptr;
2519          return TRUE;
2520          }
2521        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2522             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2523              ptr[1] == CHAR_EQUALS_SIGN) &&
2524            check_posix_syntax(ptr, endptr))
2525          return FALSE;
2526        }
2527    }    }
2528  return FALSE;  return FALSE;
2529  }  }
# Line 1579  Returns:     a value representing the na Line 2546  Returns:     a value representing the na
2546  */  */
2547    
2548  static int  static int
2549  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2550  {  {
2551    const char *pn = posix_names;
2552  register int yield = 0;  register int yield = 0;
2553  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2554    {    {
2555    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2556      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
2557      pn += posix_name_lengths[yield] + 1;
2558    yield++;    yield++;
2559    }    }
2560  return -1;  return -1;
# Line 1600  return -1; Line 2569  return -1;
2569  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2570  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2571  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2572  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2573  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2574  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2575  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2576  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2577    OP_END.
2578    
2579  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2580  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1615  value in the reference (which is a group Line 2585  value in the reference (which is a group
2585  Arguments:  Arguments:
2586    group      points to the start of the group    group      points to the start of the group
2587    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2588    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2589    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2590    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm   the hwm forward reference pointer at the start of the group
2591    
# Line 1623  Returns:     nothing Line 2593  Returns:     nothing
2593  */  */
2594    
2595  static void  static void
2596  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2597    uschar *save_hwm)    pcre_uchar *save_hwm)
2598  {  {
2599  uschar *ptr = group;  pcre_uchar *ptr = group;
2600  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2601    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2602    {    {
2603    int offset;    int offset;
2604    uschar *hc;    pcre_uchar *hc;
2605    
2606    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2607    reference. */    reference. */
2608    
2609    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2610      {      {
2611      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
2612      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
2613        {        {
2614        PUT(hc, 0, offset + adjust);        PUT(hc, 0, offset + adjust);
# Line 1650  while ((ptr = (uschar *)find_recurse(ptr Line 2621  while ((ptr = (uschar *)find_recurse(ptr
2621    
2622    if (hc >= cd->hwm)    if (hc >= cd->hwm)
2623      {      {
2624      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
2625      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2626      }      }
2627    
# Line 1675  Arguments: Line 2646  Arguments:
2646  Returns:         new code pointer  Returns:         new code pointer
2647  */  */
2648    
2649  static uschar *  static pcre_uchar *
2650  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2651  {  {
2652  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2653  *code++ = 255;  *code++ = 255;
2654  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2655  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2656  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2657  }  }
2658    
2659    
# Line 1704  Returns:             nothing Line 2675  Returns:             nothing
2675  */  */
2676    
2677  static void  static void
2678  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2679  {  {
2680  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2681  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2682  }  }
2683    
# Line 1718  PUT(previous_callout, 2 + LINK_SIZE, len Line 2689  PUT(previous_callout, 2 + LINK_SIZE, len
2689  *************************************************/  *************************************************/
2690    
2691  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
2692  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
2693  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
2694  start address.  start address. A character with multiple other cases is returned on its own
2695    with a special return value.
2696    
2697  Arguments:  Arguments:
2698    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1728  Arguments: Line 2700  Arguments:
2700    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2701    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2702    
2703  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2704                   0 when a range is returned
2705                  >0 the CASESET offset for char with multiple other cases
2706                    in this case, ocptr contains the original
2707  */  */
2708    
2709  static BOOL  static int
2710  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
2711    unsigned int *odptr)    pcre_uint32 *odptr)
2712  {  {
2713  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
2714    unsigned int co;
2715    
2716    /* Find the first character that has an other case. If it has multiple other
2717    cases, return its case offset value. */
2718    
2719  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2720    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    {
2721      if ((co = UCD_CASESET(c)) != 0)
2722        {
2723        *ocptr = c++;   /* Character that has the set */
2724        *cptr = c;      /* Rest of input range */
2725        return (int)co;
2726        }
2727      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2728      }
2729    
2730  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2731    
2732  *ocptr = othercase;  *ocptr = othercase;
2733  next = othercase + 1;  next = othercase + 1;
2734    
2735  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2736    {    {
2737    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2738    next++;    next++;
2739    }    }
2740    
2741  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2742  *cptr = c;  *cptr = c;             /* Rest of input range */
2743    return 0;
2744    }
2745    
2746  return TRUE;  
2747    
2748    /*************************************************
2749    *        Check a character and a property        *
2750    *************************************************/
2751    
2752    /* This function is called by check_auto_possessive() when a property item
2753    is adjacent to a fixed character.
2754    
2755    Arguments:
2756      c            the character
2757      ptype        the property type
2758      pdata        the data for the type
2759      negated      TRUE if it's a negated property (\P or \p{^)
2760    
2761    Returns:       TRUE if auto-possessifying is OK
2762    */
2763    
2764    static BOOL
2765    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)
2766    {
2767    #ifdef SUPPORT_UCP
2768    const pcre_uint32 *p;
2769    #endif
2770    
2771    const ucd_record *prop = GET_UCD(c);
2772    
2773    switch(ptype)
2774      {
2775      case PT_LAMP:
2776      return (prop->chartype == ucp_Lu ||
2777              prop->chartype == ucp_Ll ||
2778              prop->chartype == ucp_Lt) == negated;
2779    
2780      case PT_GC:
2781      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2782    
2783      case PT_PC:
2784      return (pdata == prop->chartype) == negated;
2785    
2786      case PT_SC:
2787      return (pdata == prop->script) == negated;
2788    
2789      /* These are specials */
2790    
2791      case PT_ALNUM:
2792      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2793              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2794    
2795      case PT_SPACE:    /* Perl space */
2796      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2797              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2798              == negated;
2799    
2800      case PT_PXSPACE:  /* POSIX space */
2801      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2802              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2803              c == CHAR_FF || c == CHAR_CR)
2804              == negated;
2805    
2806      case PT_WORD:
2807      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2808              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2809              c == CHAR_UNDERSCORE) == negated;
2810    
2811    #ifdef SUPPORT_UCP
2812      case PT_CLIST:
2813      p = PRIV(ucd_caseless_sets) + prop->caseset;
2814      for (;;)
2815        {
2816        if (c < *p) return !negated;
2817        if (c == *p++) return negated;
2818        }
2819      break;  /* Control never reaches here */
2820    #endif
2821      }
2822    
2823    return FALSE;
2824  }  }
2825  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2826    
# Line 1769  whether the next thing could possibly ma Line 2835  whether the next thing could possibly ma
2835  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2836    
2837  Arguments:  Arguments:
2838    op_code       the repeated op code    previous      pointer to the repeated opcode
2839    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2840    ptr           next character in pattern    ptr           next character in pattern
2841    options       options bits    options       options bits
2842    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns:        TRUE if possessifying is Line 2845  Returns:        TRUE if possessifying is
2845  */  */
2846    
2847  static BOOL  static BOOL
2848  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2849    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2850  {  {
2851  int next;  pcre_uint32 c = NOTACHAR;
2852    pcre_uint32 next;
2853    int escape;
2854    pcre_uchar op_code = *previous++;
2855    
2856  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2857    
# Line 1792  if ((options & PCRE_EXTENDED) != 0) Line 2859  if ((options & PCRE_EXTENDED) != 0)
2859    {    {
2860    for (;;)    for (;;)
2861      {      {
2862      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2863      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2864        {        {
2865        while (*(++ptr) != 0)        ptr++;
2866          while (*ptr != CHAR_NULL)
2867            {
2868          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2869            ptr++;
2870    #ifdef SUPPORT_UTF
2871            if (utf) FORWARDCHAR(ptr);
2872    #endif
2873            }
2874        }        }
2875      else break;      else break;
2876      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2879  if ((options & PCRE_EXTENDED) != 0)
2879  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2880  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2881    
2882  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2883    {    {
2884    int temperrorcode = 0;    int temperrorcode = 0;
2885    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,
2886        FALSE);
2887    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2888    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2889    }    }
2890    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2891    {    {
2892  #ifdef SUPPORT_UTF8    escape = 0;
2893    if (utf8) { GETCHARINC(next, ptr); } else  #ifdef SUPPORT_UTF
2894      if (utf) { GETCHARINC(next, ptr); } else
2895  #endif  #endif
2896    next = *ptr++;    next = *ptr++;
2897    }    }
   
2898  else return FALSE;  else return FALSE;
2899    
2900  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 2903  if ((options & PCRE_EXTENDED) != 0)
2903    {    {
2904    for (;;)    for (;;)
2905      {      {
2906      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2907      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2908        {        {
2909        while (*(++ptr) != 0)        ptr++;
2910          while (*ptr != CHAR_NULL)
2911            {
2912          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2913            ptr++;
2914    #ifdef SUPPORT_UTF
2915            if (utf) FORWARDCHAR(ptr);
2916    #endif
2917            }
2918        }        }
2919      else break;      else break;
2920      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2922  if ((options & PCRE_EXTENDED) != 0)
2922    
2923  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2924    
2925  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2926    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2927        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2928    
2929  /* Handle cases when the next item is a character. */  /* If the previous item is a character, get its value. */
2930    
2931  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
2932        op_code == OP_NOT || op_code == OP_NOTI)
2933    {    {
2934    case OP_CHAR:  #ifdef SUPPORT_UTF
2935  #ifdef SUPPORT_UTF8    GETCHARTEST(c, previous);
2936    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  #else
2937      c = *previous;
2938  #endif  #endif
2939    return item != next;    }
2940    
2941    /* For CHARNC (caseless character) we must check the other case. If we have  /* Now compare the next item with the previous opcode. First, handle cases when
2942    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
2943    
2944    case OP_CHARNC:  if (escape == 0)
2945  #ifdef SUPPORT_UTF8    {
2946    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    /* For a caseless UTF match, the next character may have more than one other
2947      case, which maps to the special PT_CLIST property. Check this first. */
2948    
2949    #ifdef SUPPORT_UCP
2950      if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
2951        {
2952        unsigned int ocs = UCD_CASESET(next);
2953        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
2954        }
2955  #endif  #endif
2956    if (item == next) return FALSE;  
2957  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
2958      {      {
2959      unsigned int othercase;      case OP_CHAR:
2960      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
2961    
2962        /* For CHARI (caseless character) we must check the other case. If we have
2963        Unicode property support, we can use it to test the other case of
2964        high-valued characters. We know that next can have only one other case,
2965        because multi-other-case characters are dealt with above. */
2966    
2967        case OP_CHARI:
2968        if (c == next) return FALSE;
2969    #ifdef SUPPORT_UTF
2970        if (utf)
2971          {
2972          pcre_uint32 othercase;
2973          if (next < 128) othercase = cd->fcc[next]; else
2974  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2975      othercase = _pcre_ucp_othercase((unsigned int)next);        othercase = UCD_OTHERCASE(next);
2976  #else  #else
2977      othercase = NOTACHAR;        othercase = NOTACHAR;
2978  #endif  #endif
2979      return (unsigned int)item != othercase;        return c != othercase;
2980      }        }
2981    else      else
2982  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
2983    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
2984    
2985    /* For OP_NOT, "item" must be a single-byte character. */      case OP_NOT:
2986        return c == next;
2987    
2988    case OP_NOT:      case OP_NOTI:
2989    if (next < 0) return FALSE;  /* Not a character */      if (c == next) return TRUE;
2990    if (item == next) return TRUE;  #ifdef SUPPORT_UTF
2991    if ((options & PCRE_CASELESS) == 0) return FALSE;      if (utf)
2992  #ifdef SUPPORT_UTF8        {
2993    if (utf8)        pcre_uint32 othercase;
2994      {        if (next < 128) othercase = cd->fcc[next]; else
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
2995  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2996      othercase = _pcre_ucp_othercase(next);        othercase = UCD_OTHERCASE(next);
2997  #else  #else
2998      othercase = NOTACHAR;        othercase = NOTACHAR;
2999  #endif  #endif
3000      return (unsigned int)item == othercase;        return c == othercase;
3001      }        }
3002    else      else
3003  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3004    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3005    
3006    case OP_DIGIT:      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3007    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3008    
3009    case OP_NOT_DIGIT:      case OP_DIGIT:
3010    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3011    
3012    case OP_WHITESPACE:      case OP_NOT_DIGIT:
3013    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3014    
3015    case OP_NOT_WHITESPACE:      case OP_WHITESPACE:
3016    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3017    
3018    case OP_WORDCHAR:      case OP_NOT_WHITESPACE:
3019    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3020    
3021    case OP_NOT_WORDCHAR:      case OP_WORDCHAR:
3022    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3023    
3024    default:      case OP_NOT_WORDCHAR:
3025    return FALSE;      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
   }  
3026    
3027        case OP_HSPACE:
3028        case OP_NOT_HSPACE:
3029        switch(next)
3030          {
3031          HSPACE_CASES:
3032          return op_code == OP_NOT_HSPACE;
3033    
3034  /* Handle the case when the next item is \d, \s, etc. */        default:
3035          return op_code != OP_NOT_HSPACE;
3036          }
3037    
3038        case OP_ANYNL:
3039        case OP_VSPACE:
3040        case OP_NOT_VSPACE:
3041        switch(next)
3042          {
3043          VSPACE_CASES:
3044          return op_code == OP_NOT_VSPACE;
3045    
3046          default:
3047          return op_code != OP_NOT_VSPACE;
3048          }
3049    
3050    #ifdef SUPPORT_UCP
3051        case OP_PROP:
3052        return check_char_prop(next, previous[0], previous[1], FALSE);
3053    
3054        case OP_NOTPROP:
3055        return check_char_prop(next, previous[0], previous[1], TRUE);
3056    #endif
3057    
3058        default:
3059        return FALSE;
3060        }
3061      }
3062    
3063    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3064    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3065    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3066    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3067    replaced by OP_PROP codes when PCRE_UCP is set. */
3068    
3069  switch(op_code)  switch(op_code)
3070    {    {
3071    case OP_CHAR:    case OP_CHAR:
3072    case OP_CHARNC:    case OP_CHARI:
3073  #ifdef SUPPORT_UTF8    switch(escape)
   if (utf8 && item > 127) { GETCHAR(item, utf8_char); }  
 #endif  
   switch(-next)  
3074      {      {
3075      case ESC_d:      case ESC_d:
3076      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3077    
3078      case ESC_D:      case ESC_D:
3079      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3080    
3081      case ESC_s:      case ESC_s:
3082      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3083    
3084      case ESC_S:      case ESC_S:
3085      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3086    
3087      case ESC_w:      case ESC_w:
3088      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3089    
3090      case ESC_W:      case ESC_W:
3091      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3092    
3093        case ESC_h:
3094        case ESC_H:
3095        switch(c)
3096          {
3097          HSPACE_CASES:
3098          return escape != ESC_h;
3099    
3100          default:
3101          return escape == ESC_h;
3102          }
3103    
3104        case ESC_v:
3105        case ESC_V:
3106        switch(c)
3107          {
3108          VSPACE_CASES:
3109          return escape != ESC_v;
3110    
3111          default:
3112          return escape == ESC_v;
3113          }
3114    
3115        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3116        their substitutions and process them. The result will always be either
3117        ESC_p or ESC_P. Then fall through to process those values. */
3118    
3119    #ifdef SUPPORT_UCP
3120        case ESC_du:
3121        case ESC_DU:
3122        case ESC_wu:
3123        case ESC_WU:
3124        case ESC_su:
3125        case ESC_SU:
3126          {
3127          int temperrorcode = 0;
3128          ptr = substitutes[escape - ESC_DU];
3129          escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3130          if (temperrorcode != 0) return FALSE;
3131          ptr++;    /* For compatibility */
3132          }
3133        /* Fall through */
3134    
3135        case ESC_p:
3136        case ESC_P:
3137          {
3138          unsigned int ptype = 0, pdata = 0;
3139          int errorcodeptr;
3140          BOOL negated;
3141    
3142          ptr--;      /* Make ptr point at the p or P */
3143          if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))
3144            return FALSE;
3145          ptr++;      /* Point past the final curly ket */
3146    
3147          /* If the property item is optional, we have to give up. (When generated
3148          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3149          to the original \d etc. At this point, ptr will point to a zero byte. */
3150    
3151          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3152            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3153              return FALSE;
3154    
3155          /* Do the property check. */
3156    
3157          return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3158          }
3159    #endif
3160    
3161      default:      default:
3162      return FALSE;      return FALSE;
3163      }      }
3164    
3165      /* In principle, support for Unicode properties should be integrated here as
3166      well. It means re-organizing the above code so as to get hold of the property
3167      values before switching on the op-code. However, I wonder how many patterns
3168      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3169      these op-codes are never generated.) */
3170    
3171    case OP_DIGIT:    case OP_DIGIT:
3172    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3173             escape == ESC_h || escape == ESC_v || escape == ESC_R;
3174    
3175    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3176    return next == -ESC_d;    return escape == ESC_d;
3177    
3178    case OP_WHITESPACE:    case OP_WHITESPACE:
3179    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
3180    
3181    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3182    return next == -ESC_s;    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
3183    
3184      case OP_HSPACE:
3185      return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3186             escape == ESC_w || escape == ESC_v || escape == ESC_R;
3187    
3188      case OP_NOT_HSPACE:
3189      return escape == ESC_h;
3190    
3191      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3192      case OP_ANYNL:
3193      case OP_VSPACE:
3194      return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3195    
3196      case OP_NOT_VSPACE:
3197      return escape == ESC_v || escape == ESC_R;
3198    
3199    case OP_WORDCHAR:    case OP_WORDCHAR:
3200    return next == -ESC_W || next == -ESC_s;    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3201             escape == ESC_v || escape == ESC_R;
3202    
3203      case OP_NOT_WORDCHAR:
3204      return escape == ESC_w || escape == ESC_d;
3205    
3206      default:
3207      return FALSE;
3208      }
3209    
3210    /* Control does not reach here */
3211    }
3212    
3213    
3214    
3215    /*************************************************
3216    *        Add a character or range to a class     *
3217    *************************************************/
3218    
3219    /* This function packages up the logic of adding a character or range of
3220    characters to a class. The character values in the arguments will be within the
3221    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3222    mutually recursive with the function immediately below.
3223    
3224    Arguments:
3225      classbits     the bit map for characters < 256
3226      uchardptr     points to the pointer for extra data
3227      options       the options word
3228      cd            contains pointers to tables etc.
3229      start         start of range character
3230      end           end of range character
3231    
3232    Returns:        the number of < 256 characters added
3233                    the pointer to extra data is updated
3234    */
3235    
3236    static int
3237    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3238      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3239    {
3240    pcre_uint32 c;
3241    int n8 = 0;
3242    
3243    /* If caseless matching is required, scan the range and process alternate
3244    cases. In Unicode, there are 8-bit characters that have alternate cases that
3245    are greater than 255 and vice-versa. Sometimes we can just extend the original
3246    range. */
3247    
3248    if ((options & PCRE_CASELESS) != 0)
3249      {
3250    #ifdef SUPPORT_UCP
3251      if ((options & PCRE_UTF8) != 0)
3252        {
3253        int rc;
3254        pcre_uint32 oc, od;
3255    
3256        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3257        c = start;
3258    
3259        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3260          {
3261          /* Handle a single character that has more than one other case. */
3262    
3263          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3264            PRIV(ucd_caseless_sets) + rc, oc);
3265    
3266          /* Do nothing if the other case range is within the original range. */
3267    
3268          else if (oc >= start && od <= end) continue;
3269    
3270          /* Extend the original range if there is overlap, noting that if oc < c, we
3271          can't have od > end because a subrange is always shorter than the basic
3272          range. Otherwise, use a recursive call to add the additional range. */
3273    
3274          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3275          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3276          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3277          }
3278        }
3279      else
3280    #endif  /* SUPPORT_UCP */
3281    
3282      /* Not UTF-mode, or no UCP */
3283    
3284      for (c = start; c <= end && c < 256; c++)
3285        {
3286        SETBIT(classbits, cd->fcc[c]);
3287        n8++;
3288        }
3289      }
3290    
3291    /* Now handle the original range. Adjust the final value according to the bit
3292    length - this means that the same lists of (e.g.) horizontal spaces can be used
3293    in all cases. */
3294    
3295    #if defined COMPILE_PCRE8
3296    #ifdef SUPPORT_UTF
3297      if ((options & PCRE_UTF8) == 0)
3298    #endif
3299      if (end > 0xff) end = 0xff;
3300    
3301    #elif defined COMPILE_PCRE16
3302    #ifdef SUPPORT_UTF
3303      if ((options & PCRE_UTF16) == 0)
3304    #endif
3305      if (end > 0xffff) end = 0xffff;
3306    
3307    #endif /* COMPILE_PCRE[8|16] */
3308    
3309    /* If all characters are less than 256, use the bit map. Otherwise use extra
3310    data. */
3311    
3312    if (end < 0x100)
3313      {
3314      for (c = start; c <= end; c++)
3315        {
3316        n8++;
3317        SETBIT(classbits, c);
3318        }
3319      }
3320    
3321    else
3322      {
3323      pcre_uchar *uchardata = *uchardptr;
3324    
3325    #ifdef SUPPORT_UTF
3326      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3327        {
3328        if (start < end)
3329          {
3330          *uchardata++ = XCL_RANGE;
3331          uchardata += PRIV(ord2utf)(start, uchardata);
3332          uchardata += PRIV(ord2utf)(end, uchardata);
3333          }
3334        else if (start == end)
3335          {
3336          *uchardata++ = XCL_SINGLE;
3337          uchardata += PRIV(ord2utf)(start, uchardata);
3338          }
3339        }
3340      else
3341    #endif  /* SUPPORT_UTF */
3342    
3343      /* Without UTF support, character values are constrained by the bit length,
3344      and can only be > 256 for 16-bit and 32-bit libraries. */
3345    
3346    #ifdef COMPILE_PCRE8
3347        {}
3348    #else
3349      if (start < end)
3350        {
3351        *uchardata++ = XCL_RANGE;
3352        *uchardata++ = start;
3353        *uchardata++ = end;
3354        }
3355      else if (start == end)
3356        {
3357        *uchardata++ = XCL_SINGLE;
3358        *uchardata++ = start;
3359        }
3360    #endif
3361    
3362      *uchardptr = uchardata;   /* Updata extra data pointer */
3363      }
3364    
3365    return n8;    /* Number of 8-bit characters */
3366    }
3367    
3368    
3369    
3370    
3371    /*************************************************
3372    *        Add a list of characters to a class     *
3373    *************************************************/
3374    
3375    /* This function is used for adding a list of case-equivalent characters to a
3376    class, and also for adding a list of horizontal or vertical whitespace. If the
3377    list is in order (which it should be), ranges of characters are detected and
3378    handled appropriately. This function is mutually recursive with the function
3379    above.
3380    
3381    Arguments:
3382      classbits     the bit map for characters < 256
3383      uchardptr     points to the pointer for extra data
3384      options       the options word
3385      cd            contains pointers to tables etc.
3386      p             points to row of 32-bit values, terminated by NOTACHAR
3387      except        character to omit; this is used when adding lists of
3388                      case-equivalent characters to avoid including the one we
3389                      already know about
3390    
3391    Returns:        the number of < 256 characters added
3392                    the pointer to extra data is updated
3393    */
3394    
3395    static int
3396    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3397      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3398    {
3399    int n8 = 0;
3400    while (p[0] < NOTACHAR)
3401      {
3402      int n = 0;
3403      if (p[0] != except)
3404        {
3405        while(p[n+1] == p[0] + n + 1) n++;
3406        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3407        }
3408      p += n + 1;
3409      }
3410    return n8;
3411    }
3412    
3413    
3414    
3415    /*************************************************
3416    *    Add characters not in a list to a class     *
3417    *************************************************/
3418    
3419    /* This function is used for adding the complement of a list of horizontal or
3420    vertical whitespace to a class. The list must be in order.
3421    
3422    Arguments:
3423      classbits     the bit map for characters < 256
3424      uchardptr     points to the pointer for extra data
3425      options       the options word
3426      cd            contains pointers to tables etc.
3427      p             points to row of 32-bit values, terminated by NOTACHAR
3428    
3429    case OP_NOT_WORDCHAR:  Returns:        the number of < 256 characters added
3430    return next == -ESC_w || next == -ESC_d;                  the pointer to extra data is updated
3431    */
3432    
3433    default:  static int
3434    return FALSE;  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3435      int options, compile_data *cd, const pcre_uint32 *p)
3436    {
3437    BOOL utf = (options & PCRE_UTF8) != 0;
3438    int n8 = 0;
3439    if (p[0] > 0)
3440      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3441    while (p[0] < NOTACHAR)
3442      {
3443      while (p[1] == p[0] + 1) p++;
3444      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3445        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3446      p++;
3447    }    }
3448    return n8;
 /* Control does not reach here */  
3449  }  }
3450    
3451    
# Line 2001  to find out the amount of memory needed, Line 3461  to find out the amount of memory needed,
3461  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
3462    
3463  Arguments:  Arguments:
3464    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
3465    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
3466    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
3467    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
3468    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
3469    reqbyteptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
3470    bcptr          points to current branch chain    reqcharptr        place to put the last required character
3471    cd             contains pointers to tables etc.    reqcharflagsptr   place to put the last required character flags, or a negative number
3472    lengthptr      NULL during the real compile phase    bcptr             points to current branch chain
3473                   points to length accumulator during pre-compile phase    cond_depth        conditional nesting depth
3474      cd                contains pointers to tables etc.
3475      lengthptr         NULL during the real compile phase
3476                        points to length accumulator during pre-compile phase
3477    
3478  Returns:         TRUE on success  Returns:            TRUE on success
3479                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
3480  */  */
3481    
3482  static BOOL  static BOOL
3483  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3484    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
3485      pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
3486      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
3487      branch_chain *bcptr, int cond_depth,
3488    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3489  {  {
3490  int repeat_type, op_type;  int repeat_type, op_type;
3491  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3492  int bravalue = 0;  int bravalue = 0;
3493  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3494  int firstbyte, reqbyte;  pcre_uint32 firstchar, reqchar;
3495  int zeroreqbyte, zerofirstbyte;  pcre_int32 firstcharflags, reqcharflags;
3496  int req_caseopt, reqvary, tempreqvary;  pcre_uint32 zeroreqchar, zerofirstchar;
3497  int options = *optionsptr;  pcre_int32 zeroreqcharflags, zerofirstcharflags;
3498    pcre_int32 req_caseopt, reqvary, tempreqvary;
3499    int options = *optionsptr;               /* May change dynamically */
3500  int after_manual_callout = 0;  int after_manual_callout = 0;
3501  int length_prevgroup = 0;  int length_prevgroup = 0;
3502  register int c;  register pcre_uint32 c;
3503  register uschar *code = *codeptr;  int escape;
3504  uschar *last_code = code;  register pcre_uchar *code = *codeptr;
3505  uschar *orig_code = code;  pcre_uchar *last_code = code;
3506  uschar *tempcode;  pcre_uchar *orig_code = code;
3507    pcre_uchar *tempcode;
3508  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3509  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3510  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3511  const uschar *tempptr;  const pcre_uchar *tempptr;
3512  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3513  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3514  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3515  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3516    pcre_uint8 classbits[32];
3517  #ifdef SUPPORT_UTF8  
3518  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3519  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3520  uschar *class_utf8data;  dynamically as we process the pattern. */
3521  uschar utf8_char[6];  
3522    #ifdef SUPPORT_UTF
3523    /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
3524    BOOL utf = (options & PCRE_UTF8) != 0;
3525    #ifndef COMPILE_PCRE32
3526    pcre_uchar utf_chars[6];
3527    #endif
3528  #else  #else
3529  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3530  uschar *utf8_char = NULL;  #endif
3531    
3532    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3533    class_uchardata always so that it can be passed to add_to_class() always,
3534    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3535    alternative calls for the different cases. */
3536    
3537    pcre_uchar *class_uchardata;
3538    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3539    BOOL xclass;
3540    pcre_uchar *class_uchardata_base;
3541  #endif  #endif
3542    
3543  #ifdef DEBUG  #ifdef PCRE_DEBUG
3544  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3545  #endif  #endif
3546    
# Line 2066  greedy_non_default = greedy_default ^ 1; Line 3551  greedy_non_default = greedy_default ^ 1;
3551    
3552  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3553  matching encountered yet". It gets changed to REQ_NONE if we hit something that  matching encountered yet". It gets changed to REQ_NONE if we hit something that
3554  matches a non-fixed char first char; reqbyte just remains unset if we never  matches a non-fixed char first char; reqchar just remains unset if we never
3555  find one.  find one.
3556    
3557  When we hit a repeat whose minimum is zero, we may have to adjust these values  When we hit a repeat whose minimum is zero, we may have to adjust these values
3558  to take the zero repeat into account. This is implemented by setting them to  to take the zero repeat into account. This is implemented by setting them to
3559  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3560  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3561    
3562  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
3563    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
3564    
3565  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3566  according to the current setting of the caseless flag. REQ_CASELESS is a bit  or zero, according to the current setting of the caseless flag. The
3567  value > 255. It is added into the firstbyte or reqbyte variables to record the  REQ_CASELESS leaves the lower 28 bit empty. It is added into the
3568  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3569    value. This is used only for ASCII characters. */
3570    
3571  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3572    
3573  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3574    
3575  for (;; ptr++)  for (;; ptr++)
3576    {    {
3577    BOOL negate_class;    BOOL negate_class;
3578      BOOL should_flip_negation;
3579    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3580    BOOL is_quantifier;    BOOL is_quantifier;
3581    BOOL is_recurse;    BOOL is_recurse;
3582    int class_charcount;    BOOL reset_bracount;
3583    int class_lastchar;    int class_has_8bitchar;
3584      int class_one_char;
3585    int newoptions;    int newoptions;
3586    int recno;    int recno;
3587      int refsign;
3588    int skipbytes;    int skipbytes;
3589    int subreqbyte;    pcre_uint32 subreqchar, subfirstchar;
3590    int subfirstbyte;    pcre_int32 subreqcharflags, subfirstcharflags;
3591    int terminator;    int terminator;
3592    int mclength;    unsigned int mclength;
3593    uschar mcbuffer[8];    unsigned int tempbracount;
3594      pcre_uint32 ec;
3595      pcre_uchar mcbuffer[8];
3596    
3597    /* Get next byte in the pattern */    /* Get next character in the pattern */
3598    
3599    c = *ptr;    c = *ptr;
3600    
3601      /* If we are at the end of a nested substitution, revert to the outer level
3602      string. Nesting only happens one level deep. */
3603