/[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 334 by ph10, Fri Apr 11 15:48:14 2008 UTC revision 1498 by ph10, Fri Aug 8 15:22:51 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 47  supporting internal functions that are n Line 47  supporting internal functions that are n
47  #endif  #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 pattern start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing pattern end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59  #ifdef DEBUG  COMPILE_PCREx macro will already be appropriately set. */
60  #include "pcre_printint.src"  
61    #ifdef PCRE_DEBUG
62    /* 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. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #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  /* 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  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 72  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #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         *
# Line 87  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  #define COMPILE_WORK_SIZE (4096)  /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125    /* The overrun tests check for a slightly smaller size so that they detect the
126    overrun before it actually does run off the end of the data block. */
127    
128    #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138    /* 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  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -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  -ESC_h,      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, -ESC_v, -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 130  static const short int escapes[] = { Line 212  static const short int escapes[] = {
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 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
215  /*  D0 */   '}',     0, -ESC_K,       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,-ESC_V, -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,
# Line 142  static const short int escapes[] = { Line 224  static const short int escapes[] = {
224    
225  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
227  the number of relocations when a shared library is dynamically linked. */  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    platforms. */
230    
231  typedef struct verbitem {  typedef struct verbitem {
232    int   len;    int   len;                 /* Length of verb name */
233    int   op;    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;  } verbitem;
236    
237  static const char verbnames[] =  static const char verbnames[] =
238    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
239    "COMMIT\0"    STRING_MARK0
240    "F\0"    STRING_ACCEPT0
241    "FAIL\0"    STRING_COMMIT0
242    "PRUNE\0"    STRING_F0
243    "SKIP\0"    STRING_FAIL0
244    "THEN";    STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248  static const verbitem verbs[] = {  static const verbitem verbs[] = {
249    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
250    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
251    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
252    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
253    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
254    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
255    { 4, OP_THEN  }    { 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);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287      STRING_word0  STRING_xdigit;
288    
289  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 212  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325    Unicode property escapes. */
326    
327    #ifdef SUPPORT_UCP
328    static const pcre_uchar string_PNd[]  = {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pNd[]  = {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_PXsp[] = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXsp[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXwd[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXwd[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    
347    static const pcre_uchar *substitutes[] = {
348      string_PNd,           /* \D */
349      string_pNd,           /* \d */
350      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352      string_PXwd,          /* \W */
353      string_pXwd           /* \w */
354    };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362    static const pcre_uchar string_pL[] =   {
363      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    static const pcre_uchar string_pLl[] =  {
366      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368    static const pcre_uchar string_pLu[] =  {
369      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pXan[] = {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_h[] =    {
375      CHAR_BACKSLASH, CHAR_h, '\0' };
376    static const pcre_uchar string_pXps[] = {
377      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379    static const pcre_uchar string_PL[] =   {
380      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382    static const pcre_uchar string_PLl[] =  {
383      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PLu[] =  {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PXan[] = {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_H[] =    {
392      CHAR_BACKSLASH, CHAR_H, '\0' };
393    static const pcre_uchar string_PXps[] = {
394      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396    
397    static const pcre_uchar *posix_substitutes[] = {
398      string_pL,            /* alpha */
399      string_pLl,           /* lower */
400      string_pLu,           /* upper */
401      string_pXan,          /* alnum */
402      NULL,                 /* ascii */
403      string_h,             /* blank */
404      NULL,                 /* cntrl */
405      string_pNd,           /* digit */
406      NULL,                 /* graph */
407      NULL,                 /* print */
408      NULL,                 /* punct */
409      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411      NULL,                 /* xdigit */
412      /* Negated cases */
413      string_PL,            /* ^alpha */
414      string_PLl,           /* ^lower */
415      string_PLu,           /* ^upper */
416      string_PXan,          /* ^alnum */
417      NULL,                 /* ^ascii */
418      string_H,             /* ^blank */
419      NULL,                 /* ^cntrl */
420      string_PNd,           /* ^digit */
421      NULL,                 /* ^graph */
422      NULL,                 /* ^print */
423      NULL,                 /* ^punct */
424      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426      NULL                  /* ^xdigit */
427    };
428    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429    #endif
430    
431  #define STRING(a)  # a  #define STRING(a)  # a
432  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 439  the number of relocations needed when a
439  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
440  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
442  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
443    
444    Each substring ends with \0 to insert a null character. This includes the final
445    substring, so that the whole string ends with \0\0, which can be detected when
446    counting through. */
447    
448  static const char error_texts[] =  static const char error_texts[] =
449    "no error\0"    "no error\0"
# Line 265  static const char error_texts[] = Line 484  static const char error_texts[] =
484    /* 30 */    /* 30 */
485    "unknown POSIX class name\0"    "unknown POSIX class name\0"
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
493    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494    "number after (?C is > 255\0"    "number after (?C is > 255\0"
495    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
496    /* 40 */    /* 40 */
# Line 288  static const char error_texts[] = Line 507  static const char error_texts[] =
507    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508    /* 50 */    /* 50 */
509    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
510    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
512    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
513    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
514    /* 55 */    /* 55 */
515    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
516    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
517    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+";    "digit expected after (?+\0"
525      "] is an invalid data character in JavaScript compatibility mode\0"
526      /* 65 */
527      "different names for subpatterns of the same number are not allowed\0"
528      "(*MARK) must have an argument\0"
529      "this version of PCRE is not compiled with Unicode property support\0"
530      "\\c must be followed by an ASCII character\0"
531      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532      /* 70 */
533      "internal error: unknown opcode in find_fixedlength()\0"
534      "\\N is not supported in a class\0"
535      "too many forward references\0"
536      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537      "invalid UTF-16 string\0"
538      /* 75 */
539      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553      ;
554    
555  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
556  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 568  For convenience, we use the same bit def
568    
569  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
570    
571  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
572  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
573    into a subtraction and unsigned comparison). */
574    
575    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
576    
577    #ifndef EBCDIC
578    
579    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
580    UTF-8 mode. */
581    
582    static const pcre_uint8 digitab[] =
583    {    {
584    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
585    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 357  static const unsigned char digitab[] = Line 614  static const unsigned char digitab[] =
614    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
615    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
616    
617  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
618  static const unsigned char digitab[] =  
619    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
620    
621    static const pcre_uint8 digitab[] =
622    {    {
623    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
624    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 393  static const unsigned char digitab[] = Line 653  static const unsigned char digitab[] =
653    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
654    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
655    
656  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
657    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
658    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
659    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 429  static const unsigned char ebcdic_charta Line 689  static const unsigned char ebcdic_charta
689  #endif  #endif
690    
691    
692  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730  static BOOL  /* This table is used to check whether auto-possessification is possible
731    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732      int *, int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
785      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
786      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
787      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
788      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
789      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
790      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
# Line 454  static const char * Line 884  static const char *
884  find_error_text(int n)  find_error_text(int n)
885  {  {
886  const char *s = error_texts;  const char *s = error_texts;
887  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
888      {
889      while (*s++ != CHAR_NULL) {};
890      if (*s == CHAR_NULL) return "Error text not found (please report)";
891      }
892  return s;  return s;
893  }  }
894    
895    
896    
897    /*************************************************
898    *           Expand the workspace                 *
899    *************************************************/
900    
901    /* This function is called during the second compiling phase, if the number of
902    forward references fills the existing workspace, which is originally a block on
903    the stack. A larger block is obtained from malloc() unless the ultimate limit
904    has been reached or the increase will be rather small.
905    
906    Argument: pointer to the compile data block
907    Returns:  0 if all went well, else an error number
908    */
909    
910    static int
911    expand_workspace(compile_data *cd)
912    {
913    pcre_uchar *newspace;
914    int newsize = cd->workspace_size * 2;
915    
916    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
917    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
918        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
919     return ERR72;
920    
921    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
922    if (newspace == NULL) return ERR21;
923    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
924    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
925    if (cd->workspace_size > COMPILE_WORK_SIZE)
926      (PUBL(free))((void *)cd->start_workspace);
927    cd->start_workspace = newspace;
928    cd->workspace_size = newsize;
929    return 0;
930    }
931    
932    
933    
934    /*************************************************
935    *            Check for counted repeat            *
936    *************************************************/
937    
938    /* This function is called when a '{' is encountered in a place where it might
939    start a quantifier. It looks ahead to see if it really is a quantifier or not.
940    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
941    where the ddds are digits.
942    
943    Arguments:
944      p         pointer to the first char after '{'
945    
946    Returns:    TRUE or FALSE
947    */
948    
949    static BOOL
950    is_counted_repeat(const pcre_uchar *p)
951    {
952    if (!IS_DIGIT(*p)) return FALSE;
953    p++;
954    while (IS_DIGIT(*p)) p++;
955    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956    
957    if (*p++ != CHAR_COMMA) return FALSE;
958    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
959    
960    if (!IS_DIGIT(*p)) return FALSE;
961    p++;
962    while (IS_DIGIT(*p)) p++;
963    
964    return (*p == CHAR_RIGHT_CURLY_BRACKET);
965    }
966    
967    
968    
969  /*************************************************  /*************************************************
970  *            Handle escapes                      *  *            Handle escapes                      *
971  *************************************************/  *************************************************/
972    
973  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
974  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
975  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
976  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
977  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982      chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
986    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
987    
988  Returns:         zero or positive => a data character  Returns:         zero => a data character
989                   negative => a special escape sequence                   positive => a special escape sequence
990                     negative => a back reference
991                   on error, errorcodeptr is set                   on error, errorcodeptr is set
992  */  */
993    
994  static int  static int
995  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
997  {  {
998  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
1000  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
1001    pcre_uint32 c;
1002    int escape = 0;
1003    int i;
1004    
1005  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1006  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1007    
1008  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1009    
1010  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011    
1012  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
1014  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1015    
1016  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  /* Not alphanumeric */
1018  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1019    else if ((i = escapes[c - CHAR_0]) != 0)
1020      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021    
1022  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1023  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
1024  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1025    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1026  #endif  #endif
1027    
1028  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1029    
1030  else  else
1031    {    {
1032    const uschar *oldptr;    const pcre_uchar *oldptr;
1033    BOOL braced, negated;    BOOL braced, negated, overflow;
1034      int s;
1035    
1036    switch (c)    switch (c)
1037      {      {
1038      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
1039      error. */      error. */
1040    
1041      case 'l':      case CHAR_l:
1042      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1043      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1044      break;      break;
1045    
1046      /* \g must be followed by one of a number of specific things:      case CHAR_u:
1047        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1048          {
1049          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1050          Otherwise it is a lowercase u letter. */
1051          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1052            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1053            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1054            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1055            {
1056            c = 0;
1057            for (i = 0; i < 4; ++i)
1058              {
1059              register pcre_uint32 cc = *(++ptr);
1060    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1061              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1062              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1063    #else           /* EBCDIC coding */
1064              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1065              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066    #endif
1067              }
1068    
1069    #if defined COMPILE_PCRE8
1070            if (c > (utf ? 0x10ffffU : 0xffU))
1071    #elif defined COMPILE_PCRE16
1072            if (c > (utf ? 0x10ffffU : 0xffffU))
1073    #elif defined COMPILE_PCRE32
1074            if (utf && c > 0x10ffffU)
1075    #endif
1076              {
1077              *errorcodeptr = ERR76;
1078              }
1079            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080            }
1081          }
1082        else
1083          *errorcodeptr = ERR37;
1084        break;
1085    
1086        case CHAR_U:
1087        /* In JavaScript, \U is an uppercase U letter. */
1088        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1089        break;
1090    
1091        /* In a character class, \g is just a literal "g". Outside a character
1092        class, \g must be followed by one of a number of specific things:
1093    
1094      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
1095      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
1096      5.10 feature.      5.10 feature.
1097    
1098      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1099      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
1100      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
1101      was \k.      was \k.
1102    
1103      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1104      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1105      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1107    
1108      case 'g':      case CHAR_g:
1109      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
1110        {      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111        c = -ESC_g;        {
1112        break;        escape = ESC_g;
1113        }        break;
1114          }
1115    
1116      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
1117    
1118      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119        {        {
1120        const uschar *p;        const pcre_uchar *p;
1121        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1122          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123        if (*p != 0 && *p != '}')        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124          {          {
1125          c = -ESC_k;          escape = ESC_k;
1126          break;          break;
1127          }          }
1128        braced = TRUE;        braced = TRUE;
# Line 571  else Line 1130  else
1130        }        }
1131      else braced = FALSE;      else braced = FALSE;
1132    
1133      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1134        {        {
1135        negated = TRUE;        negated = TRUE;
1136        ptr++;        ptr++;
1137        }        }
1138      else negated = FALSE;      else negated = FALSE;
1139    
1140      c = 0;      /* The integer range is limited by the machine's int representation. */
1141      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1142        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1143        while (IS_DIGIT(ptr[1]))
     if (c < 0)   /* Integer overflow */  
1144        {        {
1145          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146            {
1147            overflow = TRUE;
1148            break;
1149            }
1150          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151          }
1152        if (overflow) /* Integer overflow */
1153          {
1154          while (IS_DIGIT(ptr[1]))
1155            ptr++;
1156        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
1157        break;        break;
1158        }        }
1159    
1160      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1161        {        {
1162        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1163        break;        break;
1164        }        }
1165    
1166      if (c == 0)      if (s == 0)
1167        {        {
1168        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1169        break;        break;
1170        }        }
1171    
1172      if (negated)      if (negated)
1173        {        {
1174        if (c > bracount)        if (s > bracount)
1175          {          {
1176          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1177          break;          break;
1178          }          }
1179        c = bracount - (c - 1);        s = bracount - (s - 1);
1180        }        }
1181    
1182      c = -(ESC_REF + c);      escape = -s;
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1191      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1192      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1193      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1194      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1195      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      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:
1202      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1203    
1204      if (!isclass)      if (!isclass)
1205        {        {
1206        oldptr = ptr;        oldptr = ptr;
1207        c -= '0';        /* The integer range is limited by the machine's int representation. */
1208        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1209          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1210        if (c < 0)    /* Integer overflow */        while (IS_DIGIT(ptr[1]))
1211            {
1212            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213              {
1214              overflow = TRUE;
1215              break;
1216              }
1217            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218            }
1219          if (overflow) /* Integer overflow */
1220          {          {
1221            while (IS_DIGIT(ptr[1]))
1222              ptr++;
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          c = -(ESC_REF + c);          escape = -s;
1229          break;          break;
1230          }          }
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1235      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1236      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1244      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1245      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1246      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1247      than 3 octal digits. */      but no more than 3 octal digits. */
1248    
1249      case '0':      case CHAR_0:
1250      c -= '0';      c -= CHAR_0;
1251      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1252          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1253      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1254        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1255    #endif
1256      break;      break;
1257    
1258      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1259      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1260    
1261      case 'x':      case CHAR_o:
1262      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264        {        {
1265        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1266        c = 0;        c = 0;
1267        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269          {          {
1270          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1271          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272          count++;  #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274  #ifndef EBCDIC  /* ASCII coding */  #endif
1275          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1276          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1277  #else           /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1279          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282  #endif  #endif
1283          }          }
1284          if (overflow)
       if (*pt == '}')  
1285          {          {
1286          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1288          }          }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290        /* If the sequence of hex digits does not end with '}', then we don't          {
1291        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294        }        }
1295        break;
1296    
1297      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      c = 0;      case CHAR_x:
1301      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
1303        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305  #ifndef EBCDIC  /* ASCII coding */          {
1306        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1307        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1308              {
1309              register pcre_uint32 cc = *(++ptr);
1310    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1311              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1312              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1313  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1314        if (cc <= 'z') cc += 64;              /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1315        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1316  #endif  #endif
1317        }            }
1318            }
1319          }    /* End JavaScript handling */
1320    
1321        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        else
1329          {
1330          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331            {
1332            ptr += 2;
1333            if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345    #ifdef COMPILE_PCRE32
1346              if (c >= 0x10000000l) { overflow = TRUE; break; }
1347    #endif
1348    
1349    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352    #else           /* EBCDIC coding */
1353              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1354              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355    #endif
1356    
1357    #if defined COMPILE_PCRE8
1358              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359    #elif defined COMPILE_PCRE16
1360              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361    #elif defined COMPILE_PCRE32
1362              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363    #endif
1364              }
1365    
1366            if (overflow)
1367              {
1368              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369              *errorcodeptr = ERR34;
1370              }
1371    
1372            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373              {
1374              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375              }
1376    
1377            /* If the sequence of hex digits does not end with '}', give an error.
1378            We used just to recognize this construct and fall through to the normal
1379            \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382            else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386    
1387          else
1388            {
1389            c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397    #else           /* EBCDIC coding */
1398              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400    #endif
1401              }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1407      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1408        coding is ASCII-specific, but then the whole concept of \cx is
1409      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1410    
1411      case 'c':      case CHAR_c:
1412      c = *(++ptr);      c = *(++ptr);
1413      if (c == 0)      if (c == CHAR_NULL)
1414        {        {
1415        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1416        break;        break;
1417        }        }
1418    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1419  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1420      if (c >= 'a' && c <= 'z') c -= 32;        {
1421          *errorcodeptr = ERR68;
1422          break;
1423          }
1424        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1425      c ^= 0x40;      c ^= 0x40;
1426  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1427      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1428      c ^= 0xC0;      c ^= 0xC0;
1429  #endif  #endif
1430      break;      break;
# Line 763  else Line 1446  else
1446      }      }
1447    }    }
1448    
1449    /* Perl supports \N{name} for character names, as well as plain \N for "not
1450    newline". PCRE does not support \N{name}. However, it does support
1451    quantification such as \N{2,3}. */
1452    
1453    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454         !is_counted_repeat(ptr+2))
1455      *errorcodeptr = ERR37;
1456    
1457    /* If PCRE_UCP is set, we change the values for \d etc. */
1458    
1459    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460      escape += (ESC_DU - ESC_D);
1461    
1462    /* Set the pointer to the final character before returning. */
1463    
1464  *ptrptr = ptr;  *ptrptr = ptr;
1465  return c;  *chptr = c;
1466    return escape;
1467  }  }
1468    
1469    
# Line 782  escape sequence. Line 1481  escape sequence.
1481  Argument:  Argument:
1482    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1483    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1484    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1485      pdataptr       points to an unsigned int that is set to the detailed property value
1486    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1487    
1488  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1489  */  */
1490    
1491  static int  static BOOL
1492  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493      unsigned int *pdataptr, int *errorcodeptr)
1494  {  {
1495  int c, i, bot, top;  pcre_uchar c;
1496  const uschar *ptr = *ptrptr;  int i, bot, top;
1497  char name[32];  const pcre_uchar *ptr = *ptrptr;
1498    pcre_uchar name[32];
1499    
1500  c = *(++ptr);  c = *(++ptr);
1501  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1502    
1503  *negptr = FALSE;  *negptr = FALSE;
1504    
1505  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1506  negation. */  negation. */
1507    
1508  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1509    {    {
1510    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1511      {      {
1512      *negptr = TRUE;      *negptr = TRUE;
1513      ptr++;      ptr++;
1514      }      }
1515    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516      {      {
1517      c = *(++ptr);      c = *(++ptr);
1518      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1519      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520      name[i] = c;      name[i] = c;
1521      }      }
1522    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1523    name[i] = 0;    name[i] = 0;
1524    }    }
1525    
# Line 834  else Line 1536  else
1536  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1537    
1538  bot = 0;  bot = 0;
1539  top = _pcre_utt_size;  top = PRIV(utt_size);
1540    
1541  while (bot < top)  while (bot < top)
1542    {    {
1543      int r;
1544    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1545    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1546    if (c == 0)    if (r == 0)
1547      {      {
1548      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1549      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1550        return TRUE;
1551      }      }
1552    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1553    }    }
1554    
1555  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1556  *ptrptr = ptr;  *ptrptr = ptr;
1557  return -1;  return FALSE;
1558    
1559  ERROR_RETURN:  ERROR_RETURN:
1560  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1561  *ptrptr = ptr;  *ptrptr = ptr;
1562  return -1;  return FALSE;
1563  }  }
1564  #endif  #endif
1565    
1566    
1567    
   
 /*************************************************  
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 914  Returns:         pointer to '}' on succe Line 1584  Returns:         pointer to '}' on succe
1584                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1585  */  */
1586    
1587  static const uschar *  static const pcre_uchar *
1588  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1589  {  {
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597        {
1598        *errorcodeptr = ERR5;
1599        return p;
1600        }
1601    }    }
1602    
1603  /* Read the maximum value if there is one, and again do a paranoid on its size.  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
 Also, max must not be less than min. */  
   
 if (*p == '}') max = min; else  
1604    {    {
1605    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612            {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616        }        }
1617      if (max < min)      if (max < min)
1618        {        {
# Line 952  if (*p == '}') max = min; else Line 1622  if (*p == '}') max = min; else
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 963  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
1633  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
1636  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1637  subpatterns, and counting them. If it finds a named pattern that matches the  for a fixed first character, or an anchoring op code etc. It skips over things
1638  name it is given, it returns its number. Alternatively, if the name is NULL, it  that do not influence this. For some calls, it makes sense to skip negative
1639  returns when it reaches a given numbered subpattern. This is used for forward  forward and all backward assertions, and also the \b assertion; for others it
1640  references to subpatterns. We know that if (?P< is encountered, the name will  does not.
 be terminated by '>' because that is checked in the first pass.  
1641    
1642  Arguments:  Arguments:
1643    ptr          current position in the pattern    code         pointer to the start of the group
1644    count        current count of capturing parens so far encountered    skipassert   TRUE if certain assertions are to be skipped
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
1645    
1646  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1647  */  */
1648    
1649  static int  static const pcre_uchar*
1650  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode)  
1651  {  {
1652  const uschar *thisname;  for (;;)
   
 for (; *ptr != 0; ptr++)  
1653    {    {
1654    int term;    switch ((int)*code)
   
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == '\\')  
1655      {      {
1656      if (*(++ptr) == 0) return -1;      case OP_ASSERT_NOT:
1657      if (*ptr == 'Q') for (;;)      case OP_ASSERTBACK:
1658        {      case OP_ASSERTBACK_NOT:
1659        while (*(++ptr) != 0 && *ptr != '\\');      if (!skipassert) return code;
1660        if (*ptr == 0) return -1;      do code += GET(code, 1); while (*code == OP_ALT);
1661        if (*(++ptr) == 'E') break;      code += PRIV(OP_lengths)[*code];
1662        }      break;
     continue;  
     }  
   
   /* Skip over character classes */  
   
   if (*ptr == '[')  
     {  
     while (*(++ptr) != ']')  
       {  
       if (*ptr == 0) return -1;  
       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] != '?' && ptr[1] != '*')  
     {  
     count++;  
     if (name == NULL && count == lorn) return count;  
     continue;  
     }  
   
   ptr += 2;  
   if (*ptr == 'P') ptr++;                      /* Allow optional P */  
   
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
   
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
   
   if (name == NULL && count == lorn) return count;  
   term = *ptr++;  
   if (term == '<') term = '>';  
   thisname = ptr;  
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
   }  
   
 return -1;  
 }  
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, a change of option is important.  
 For some calls, it makes sense to skip negative forward and all backward  
 assertions, and also the \b assertion; for others it does not.  
   
 Arguments:  
   code         pointer to the start of the group  
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const uschar*  
 first_significant_code(const uschar *code, int *options, int optbit,  
   BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
     {  
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1663    
1664      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1665      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
# Line 1123  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671        case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673        case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
1677    
1678      default:      default:
# Line 1137  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
1690    
1691  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1692  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1693  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1694    temporarily terminated with OP_END when this function is called.
1695    
1696    This function is called when a backward assertion is encountered, so that if it
1697    fails, the error message can point to the correct place in the pattern.
1698    However, we cannot do this when the assertion contains subroutine calls,
1699    because they can be forward references. We solve this by remembering this case
1700    and doing the check at the end; a flag specifies which mode we are running in.
1701    
1702  Arguments:  Arguments:
1703    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1704    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705      atend    TRUE if called when the pattern is complete
1706  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1707               or -2 if \C was encountered  
1708    Returns:   the fixed length,
1709                 or -1 if there is no fixed length,
1710                 or -2 if \C was encountered (in UTF-8 mode only)
1711                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1712                 or -4 if an unknown opcode was encountered (internal error)
1713  */  */
1714    
1715  static int  static int
1716  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1717  {  {
1718  int length = -1;  int length = -1;
1719    
1720  register int branchlength = 0;  register int branchlength = 0;
1721  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1722    
1723  /* 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
1724  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1168  branch, check the length against that of Line 1726  branch, check the length against that of
1726  for (;;)  for (;;)
1727    {    {
1728    int d;    int d;
1729    register int op = *cc;    pcre_uchar *ce, *cs;
1730      register pcre_uchar op = *cc;
1731    
1732    switch (op)    switch (op)
1733      {      {
1734        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1735        OP_BRA (normal non-capturing bracket) because the other variants of these
1736        opcodes are all concerned with unlimited repeated groups, which of course
1737        are not of fixed length. */
1738    
1739      case OP_CBRA:      case OP_CBRA:
1740      case OP_BRA:      case OP_BRA:
1741      case OP_ONCE:      case OP_ONCE:
1742        case OP_ONCE_NC:
1743      case OP_COND:      case OP_COND:
1744      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1745      if (d < 0) return d;      if (d < 0) return d;
1746      branchlength += d;      branchlength += d;
1747      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1748      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1749      break;      break;
1750    
1751      /* 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.
1752      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
1753      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
1754        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1755        because they all imply an unlimited repeat. */
1756    
1757      case OP_ALT:      case OP_ALT:
1758      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1759      case OP_END:      case OP_END:
1760        case OP_ACCEPT:
1761        case OP_ASSERT_ACCEPT:
1762      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1763        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1764      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1766  for (;;)
1766      branchlength = 0;      branchlength = 0;
1767      break;      break;
1768    
1769        /* A true recursion implies not fixed length, but a subroutine call may
1770        be OK. If the subroutine is a forward reference, we can't deal with
1771        it until the end of the pattern, so return -3. */
1772    
1773        case OP_RECURSE:
1774        if (!atend) return -3;
1775        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1776        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1777        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1778        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1779        if (d < 0) return d;
1780        branchlength += d;
1781        cc += 1 + LINK_SIZE;
1782        break;
1783    
1784      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1785    
1786      case OP_ASSERT:      case OP_ASSERT:
# Line 1205  for (;;) Line 1788  for (;;)
1788      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1789      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1790      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1791      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1792        break;
1793    
1794      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1795    
1796      case OP_REVERSE:      case OP_MARK:
1797        case OP_PRUNE_ARG:
1798        case OP_SKIP_ARG:
1799        case OP_THEN_ARG:
1800        cc += cc[1] + PRIV(OP_lengths)[*cc];
1801        break;
1802    
1803        case OP_CALLOUT:
1804        case OP_CIRC:
1805        case OP_CIRCM:
1806        case OP_CLOSE:
1807        case OP_COMMIT:
1808      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1809      case OP_DEF:      case OP_DEF:
1810      case OP_OPT:      case OP_DNCREF:
1811      case OP_CALLOUT:      case OP_DNRREF:
1812      case OP_SOD:      case OP_DOLL:
1813      case OP_SOM:      case OP_DOLLM:
1814      case OP_EOD:      case OP_EOD:
1815      case OP_EODN:      case OP_EODN:
1816      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1817      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1818        case OP_PRUNE:
1819        case OP_REVERSE:
1820        case OP_RREF:
1821        case OP_SET_SOM:
1822        case OP_SKIP:
1823        case OP_SOD:
1824        case OP_SOM:
1825        case OP_THEN:
1826      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1827      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1828      break;      break;
1829    
1830      /* Handle literal characters */      /* Handle literal characters */
1831    
1832      case OP_CHAR:      case OP_CHAR:
1833      case OP_CHARNC:      case OP_CHARI:
1834      case OP_NOT:      case OP_NOT:
1835        case OP_NOTI:
1836      branchlength++;      branchlength++;
1837      cc += 2;      cc += 2;
1838  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1839      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1840  #endif  #endif
1841      break;      break;
1842    
# Line 1245  for (;;) Line 1844  for (;;)
1844      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1845    
1846      case OP_EXACT:      case OP_EXACT:
1847      branchlength += GET2(cc,1);      case OP_EXACTI:
1848      cc += 4;      case OP_NOTEXACT:
1849  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1850      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1851        {      cc += 2 + IMM2_SIZE;
1852        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1853        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1854  #endif  #endif
1855      break;      break;
1856    
1857      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1858      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1859      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1860      cc += 4;        cc += 2;
1861        cc += 1 + IMM2_SIZE + 1;
1862      break;      break;
1863    
1864      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1268  for (;;) Line 1868  for (;;)
1868      cc += 2;      cc += 2;
1869      /* Fall through */      /* Fall through */
1870    
1871        case OP_HSPACE:
1872        case OP_VSPACE:
1873        case OP_NOT_HSPACE:
1874        case OP_NOT_VSPACE:
1875      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1876      case OP_DIGIT:      case OP_DIGIT:
1877      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1879  for (;;)
1879      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1880      case OP_WORDCHAR:      case OP_WORDCHAR:
1881      case OP_ANY:      case OP_ANY:
1882        case OP_ALLANY:
1883      branchlength++;      branchlength++;
1884      cc++;      cc++;
1885      break;      break;
1886    
1887      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1888        otherwise \C is coded as OP_ALLANY. */
1889    
1890      case OP_ANYBYTE:      case OP_ANYBYTE:
1891      return -2;      return -2;
1892    
1893      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1894    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1895      case OP_CLASS:      case OP_CLASS:
1896      case OP_NCLASS:      case OP_NCLASS:
1897      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1898        case OP_XCLASS:
1899        /* The original code caused an unsigned overflow in 64 bit systems,
1900        so now we use a conditional statement. */
1901        if (op == OP_XCLASS)
1902          cc += GET(cc, 1);
1903        else
1904          cc += PRIV(OP_lengths)[OP_CLASS];
1905    #else
1906        cc += PRIV(OP_lengths)[OP_CLASS];
1907    #endif
1908    
1909      switch (*cc)      switch (*cc)
1910        {        {
1911        case OP_CRSTAR:        case OP_CRSTAR:
1912        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRQUERY:        case OP_CRQUERY:
1916        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1917          case OP_CRPOSSTAR:
1918          case OP_CRPOSPLUS:
1919          case OP_CRPOSQUERY:
1920        return -1;        return -1;
1921    
1922        case OP_CRRANGE:        case OP_CRRANGE:
1923        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1924        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1925        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926        cc += 5;        branchlength += (int)GET2(cc,1);
1927          cc += 1 + 2 * IMM2_SIZE;
1928        break;        break;
1929    
1930        default:        default:
# Line 1318  for (;;) Line 1934  for (;;)
1934    
1935      /* Anything else is variable length */      /* Anything else is variable length */
1936    
1937      default:      case OP_ANYNL:
1938        case OP_BRAMINZERO:
1939        case OP_BRAPOS:
1940        case OP_BRAPOSZERO:
1941        case OP_BRAZERO:
1942        case OP_CBRAPOS:
1943        case OP_EXTUNI:
1944        case OP_KETRMAX:
1945        case OP_KETRMIN:
1946        case OP_KETRPOS:
1947        case OP_MINPLUS:
1948        case OP_MINPLUSI:
1949        case OP_MINQUERY:
1950        case OP_MINQUERYI:
1951        case OP_MINSTAR:
1952        case OP_MINSTARI:
1953        case OP_MINUPTO:
1954        case OP_MINUPTOI:
1955        case OP_NOTMINPLUS:
1956        case OP_NOTMINPLUSI:
1957        case OP_NOTMINQUERY:
1958        case OP_NOTMINQUERYI:
1959        case OP_NOTMINSTAR:
1960        case OP_NOTMINSTARI:
1961        case OP_NOTMINUPTO:
1962        case OP_NOTMINUPTOI:
1963        case OP_NOTPLUS:
1964        case OP_NOTPLUSI:
1965        case OP_NOTPOSPLUS:
1966        case OP_NOTPOSPLUSI:
1967        case OP_NOTPOSQUERY:
1968        case OP_NOTPOSQUERYI:
1969        case OP_NOTPOSSTAR:
1970        case OP_NOTPOSSTARI:
1971        case OP_NOTPOSUPTO:
1972        case OP_NOTPOSUPTOI:
1973        case OP_NOTQUERY:
1974        case OP_NOTQUERYI:
1975        case OP_NOTSTAR:
1976        case OP_NOTSTARI:
1977        case OP_NOTUPTO:
1978        case OP_NOTUPTOI:
1979        case OP_PLUS:
1980        case OP_PLUSI:
1981        case OP_POSPLUS:
1982        case OP_POSPLUSI:
1983        case OP_POSQUERY:
1984        case OP_POSQUERYI:
1985        case OP_POSSTAR:
1986        case OP_POSSTARI:
1987        case OP_POSUPTO:
1988        case OP_POSUPTOI:
1989        case OP_QUERY:
1990        case OP_QUERYI:
1991        case OP_REF:
1992        case OP_REFI:
1993        case OP_DNREF:
1994        case OP_DNREFI:
1995        case OP_SBRA:
1996        case OP_SBRAPOS:
1997        case OP_SCBRA:
1998        case OP_SCBRAPOS:
1999        case OP_SCOND:
2000        case OP_SKIPZERO:
2001        case OP_STAR:
2002        case OP_STARI:
2003        case OP_TYPEMINPLUS:
2004        case OP_TYPEMINQUERY:
2005        case OP_TYPEMINSTAR:
2006        case OP_TYPEMINUPTO:
2007        case OP_TYPEPLUS:
2008        case OP_TYPEPOSPLUS:
2009        case OP_TYPEPOSQUERY:
2010        case OP_TYPEPOSSTAR:
2011        case OP_TYPEPOSUPTO:
2012        case OP_TYPEQUERY:
2013        case OP_TYPESTAR:
2014        case OP_TYPEUPTO:
2015        case OP_UPTO:
2016        case OP_UPTOI:
2017      return -1;      return -1;
2018    
2019        /* Catch unrecognized opcodes so that when new ones are added they
2020        are not forgotten, as has happened in the past. */
2021    
2022        default:
2023        return -4;
2024      }      }
2025    }    }
2026  /* Control never gets here */  /* Control never gets here */
# Line 1327  for (;;) Line 2028  for (;;)
2028    
2029    
2030    
   
2031  /*************************************************  /*************************************************
2032  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2033  *************************************************/  *************************************************/
2034    
2035  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2036  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2037    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2038    so that it can be called from pcre_study() when finding the minimum matching
2039    length.
2040    
2041  Arguments:  Arguments:
2042    code        points to start of expression    code        points to start of expression
2043    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2044    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2045    
2046  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
2047  */  */
2048    
2049  static const uschar *  const pcre_uchar *
2050  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2051  {  {
2052  for (;;)  for (;;)
2053    {    {
2054    register int c = *code;    register pcre_uchar c = *code;
2055    
2056    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2057    
2058    /* 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 1357  for (;;) Line 2061  for (;;)
2061    
2062    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2063    
2064      /* Handle recursion */
2065    
2066      else if (c == OP_REVERSE)
2067        {
2068        if (number < 0) return (pcre_uchar *)code;
2069        code += PRIV(OP_lengths)[c];
2070        }
2071    
2072    /* Handle capturing bracket */    /* Handle capturing bracket */
2073    
2074    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2075               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2076      {      {
2077      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2078      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2079      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2080      }      }
2081    
2082    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2083    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2084    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2085      must add in its length. */
2086    
2087    else    else
2088      {      {
# Line 1390  for (;;) Line 2104  for (;;)
2104        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2105        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2106        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2107        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2108            code += 2;
2109          break;
2110    
2111          case OP_MARK:
2112          case OP_PRUNE_ARG:
2113          case OP_SKIP_ARG:
2114          case OP_THEN_ARG:
2115          code += code[1];
2116        break;        break;
2117        }        }
2118    
2119      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2120    
2121      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2122    
2123    /* 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
2124    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
2125    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2126    
2127  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2128      if (utf8) switch(c)      if (utf) switch(c)
2129        {        {
2130        case OP_CHAR:        case OP_CHAR:
2131        case OP_CHARNC:        case OP_CHARI:
2132        case OP_EXACT:        case OP_EXACT:
2133          case OP_EXACTI:
2134        case OP_UPTO:        case OP_UPTO:
2135          case OP_UPTOI:
2136        case OP_MINUPTO:        case OP_MINUPTO:
2137          case OP_MINUPTOI:
2138        case OP_POSUPTO:        case OP_POSUPTO:
2139          case OP_POSUPTOI:
2140        case OP_STAR:        case OP_STAR:
2141          case OP_STARI:
2142        case OP_MINSTAR:        case OP_MINSTAR:
2143          case OP_MINSTARI:
2144        case OP_POSSTAR:        case OP_POSSTAR:
2145          case OP_POSSTARI:
2146        case OP_PLUS:        case OP_PLUS:
2147          case OP_PLUSI:
2148        case OP_MINPLUS:        case OP_MINPLUS:
2149          case OP_MINPLUSI:
2150        case OP_POSPLUS:        case OP_POSPLUS:
2151          case OP_POSPLUSI:
2152        case OP_QUERY:        case OP_QUERY:
2153          case OP_QUERYI:
2154        case OP_MINQUERY:        case OP_MINQUERY:
2155          case OP_MINQUERYI:
2156        case OP_POSQUERY:        case OP_POSQUERY:
2157        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2158          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2159        break;        break;
2160        }        }
2161    #else
2162        (void)(utf);  /* Keep compiler happy by referencing function argument */
2163  #endif  #endif
2164      }      }
2165    }    }
# Line 1439  instance of OP_RECURSE. Line 2176  instance of OP_RECURSE.
2176    
2177  Arguments:  Arguments:
2178    code        points to start of expression    code        points to start of expression
2179    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2180    
2181  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
2182  */  */
2183    
2184  static const uschar *  static const pcre_uchar *
2185  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2186  {  {
2187  for (;;)  for (;;)
2188    {    {
2189    register int c = *code;    register pcre_uchar c = *code;
2190    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2191    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2192    
# Line 1461  for (;;) Line 2198  for (;;)
2198    
2199    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2200    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2201    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2202      must add in its length. */
2203    
2204    else    else
2205      {      {
# Line 1483  for (;;) Line 2221  for (;;)
2221        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2222        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2223        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2224        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2225            code += 2;
2226          break;
2227    
2228          case OP_MARK:
2229          case OP_PRUNE_ARG:
2230          case OP_SKIP_ARG:
2231          case OP_THEN_ARG:
2232          code += code[1];
2233        break;        break;
2234        }        }
2235    
2236      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2237    
2238      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2239    
2240      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2241      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2242      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2243    
2244  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2245      if (utf8) switch(c)      if (utf) switch(c)
2246        {        {
2247        case OP_CHAR:        case OP_CHAR:
2248        case OP_CHARNC:        case OP_CHARI:
2249          case OP_NOT:
2250          case OP_NOTI:
2251        case OP_EXACT:        case OP_EXACT:
2252          case OP_EXACTI:
2253          case OP_NOTEXACT:
2254          case OP_NOTEXACTI:
2255        case OP_UPTO:        case OP_UPTO:
2256          case OP_UPTOI:
2257          case OP_NOTUPTO:
2258          case OP_NOTUPTOI:
2259        case OP_MINUPTO:        case OP_MINUPTO:
2260          case OP_MINUPTOI:
2261          case OP_NOTMINUPTO:
2262          case OP_NOTMINUPTOI:
2263        case OP_POSUPTO:        case OP_POSUPTO:
2264          case OP_POSUPTOI:
2265          case OP_NOTPOSUPTO:
2266          case OP_NOTPOSUPTOI:
2267        case OP_STAR:        case OP_STAR:
2268          case OP_STARI:
2269          case OP_NOTSTAR:
2270          case OP_NOTSTARI:
2271        case OP_MINSTAR:        case OP_MINSTAR:
2272          case OP_MINSTARI:
2273          case OP_NOTMINSTAR:
2274          case OP_NOTMINSTARI:
2275        case OP_POSSTAR:        case OP_POSSTAR:
2276          case OP_POSSTARI:
2277          case OP_NOTPOSSTAR:
2278          case OP_NOTPOSSTARI:
2279        case OP_PLUS:        case OP_PLUS:
2280          case OP_PLUSI:
2281          case OP_NOTPLUS:
2282          case OP_NOTPLUSI:
2283        case OP_MINPLUS:        case OP_MINPLUS:
2284          case OP_MINPLUSI:
2285          case OP_NOTMINPLUS:
2286          case OP_NOTMINPLUSI:
2287        case OP_POSPLUS:        case OP_POSPLUS:
2288          case OP_POSPLUSI:
2289          case OP_NOTPOSPLUS:
2290          case OP_NOTPOSPLUSI:
2291        case OP_QUERY:        case OP_QUERY:
2292          case OP_QUERYI:
2293          case OP_NOTQUERY:
2294          case OP_NOTQUERYI:
2295        case OP_MINQUERY:        case OP_MINQUERY:
2296          case OP_MINQUERYI:
2297          case OP_NOTMINQUERY:
2298          case OP_NOTMINQUERYI:
2299        case OP_POSQUERY:        case OP_POSQUERY:
2300        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2301          case OP_NOTPOSQUERY:
2302          case OP_NOTPOSQUERYI:
2303          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2304        break;        break;
2305        }        }
2306    #else
2307        (void)(utf);  /* Keep compiler happy by referencing function argument */
2308  #endif  #endif
2309      }      }
2310    }    }
# Line 1538  bracket whose current branch will alread Line 2327  bracket whose current branch will alread
2327  Arguments:  Arguments:
2328    code        points to start of search    code        points to start of search
2329    endcode     points to where to stop    endcode     points to where to stop
2330    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2331      cd          contains pointers to tables etc.
2332      recurses    chain of recurse_check to catch mutual recursion
2333    
2334  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2335  */  */
2336    
2337    typedef struct recurse_check {
2338      struct recurse_check *prev;
2339      const pcre_uchar *group;
2340    } recurse_check;
2341    
2342  static BOOL  static BOOL
2343  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2344      BOOL utf, compile_data *cd, recurse_check *recurses)
2345  {  {
2346  register int c;  register pcre_uchar c;
2347  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2348    
2349    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2350       code < endcode;       code < endcode;
2351       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2352    {    {
2353    const uschar *ccode;    const pcre_uchar *ccode;
2354    
2355    c = *code;    c = *code;
2356    
# Line 1565  for (code = first_significant_code(code Line 2364  for (code = first_significant_code(code
2364      continue;      continue;
2365      }      }
2366    
2367      /* For a recursion/subroutine call, if its end has been reached, which
2368      implies a backward reference subroutine call, we can scan it. If it's a
2369      forward reference subroutine call, we can't. To detect forward reference
2370      we have to scan up the list that is kept in the workspace. This function is
2371      called only when doing the real compile, not during the pre-compile that
2372      measures the size of the compiled pattern. */
2373    
2374      if (c == OP_RECURSE)
2375        {
2376        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2377        const pcre_uchar *endgroup = scode;
2378        BOOL empty_branch;
2379    
2380        /* Test for forward reference or uncompleted reference. This is disabled
2381        when called to scan a completed pattern by setting cd->start_workspace to
2382        NULL. */
2383    
2384        if (cd->start_workspace != NULL)
2385          {
2386          const pcre_uchar *tcode;
2387          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2388            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2389          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2390          }
2391    
2392        /* If the reference is to a completed group, we need to detect whether this
2393        is a recursive call, as otherwise there will be an infinite loop. If it is
2394        a recursion, just skip over it. Simple recursions are easily detected. For
2395        mutual recursions we keep a chain on the stack. */
2396    
2397        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2398        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2399        else
2400          {
2401          recurse_check *r = recurses;
2402          for (r = recurses; r != NULL; r = r->prev)
2403            if (r->group == scode) break;
2404          if (r != NULL) continue;   /* Mutual recursion */
2405          }
2406    
2407        /* Completed reference; scan the referenced group, remembering it on the
2408        stack chain to detect mutual recursions. */
2409    
2410        empty_branch = FALSE;
2411        this_recurse.prev = recurses;
2412        this_recurse.group = scode;
2413    
2414        do
2415          {
2416          if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2417            {
2418            empty_branch = TRUE;
2419            break;
2420            }
2421          scode += GET(scode, 1);
2422          }
2423        while (*scode == OP_ALT);
2424    
2425        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2426        continue;
2427        }
2428    
2429    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2430    
2431    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2432          c == OP_BRAPOSZERO)
2433        {
2434        code += PRIV(OP_lengths)[c];
2435        do code += GET(code, 1); while (*code == OP_ALT);
2436        c = *code;
2437        continue;
2438        }
2439    
2440      /* A nested group that is already marked as "could be empty" can just be
2441      skipped. */
2442    
2443      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2444          c == OP_SCBRA || c == OP_SCBRAPOS)
2445      {      {
     code += _pcre_OP_lengths[c];  
2446      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2447      c = *code;      c = *code;
2448      continue;      continue;
# Line 1577  for (code = first_significant_code(code Line 2450  for (code = first_significant_code(code
2450    
2451    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2452    
2453    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2454          c == OP_CBRA || c == OP_CBRAPOS ||
2455          c == OP_ONCE || c == OP_ONCE_NC ||
2456          c == OP_COND)
2457      {      {
2458      BOOL empty_branch;      BOOL empty_branch;
2459      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2460    
2461      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2462        empty branch, so just skip over the conditional, because it could be empty.
2463        Otherwise, scan the individual branches of the group. */
2464    
2465      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
2466        code += GET(code, 1);        code += GET(code, 1);
2467        else
2468          {
2469          empty_branch = FALSE;
2470          do
2471            {
2472            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2473              empty_branch = TRUE;
2474            code += GET(code, 1);
2475            }
2476          while (*code == OP_ALT);
2477          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2478        }        }
2479      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2480      c = *code;      c = *code;
2481      continue;      continue;
2482      }      }
# Line 1603  for (code = first_significant_code(code Line 2487  for (code = first_significant_code(code
2487      {      {
2488      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2489      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2490      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2491      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2492      here. */      here. */
2493    
2494  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2495      case OP_XCLASS:      case OP_XCLASS:
2496      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2497      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1615  for (code = first_significant_code(code Line 2499  for (code = first_significant_code(code
2499    
2500      case OP_CLASS:      case OP_CLASS:
2501      case OP_NCLASS:      case OP_NCLASS:
2502      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2503    
2504  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2505      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2506  #endif  #endif
2507    
# Line 1627  for (code = first_significant_code(code Line 2511  for (code = first_significant_code(code
2511        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2512        case OP_CRQUERY:        case OP_CRQUERY:
2513        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2514          case OP_CRPOSSTAR:
2515          case OP_CRPOSQUERY:
2516        break;        break;
2517    
2518        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2519        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2520        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2521          case OP_CRPOSPLUS:
2522        return FALSE;        return FALSE;
2523    
2524        case OP_CRRANGE:        case OP_CRRANGE:
2525        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2526          case OP_CRPOSRANGE:
2527        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2528        break;        break;
2529        }        }
# Line 1643  for (code = first_significant_code(code Line 2531  for (code = first_significant_code(code
2531    
2532      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2533    
2534        case OP_ANY:
2535        case OP_ALLANY:
2536        case OP_ANYBYTE:
2537    
2538      case OP_PROP:      case OP_PROP:
2539      case OP_NOTPROP:      case OP_NOTPROP:
2540        case OP_ANYNL:
2541    
2542        case OP_NOT_HSPACE:
2543        case OP_HSPACE:
2544        case OP_NOT_VSPACE:
2545        case OP_VSPACE:
2546      case OP_EXTUNI:      case OP_EXTUNI:
2547    
2548      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2549      case OP_DIGIT:      case OP_DIGIT:
2550      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2551      case OP_WHITESPACE:      case OP_WHITESPACE:
2552      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2553      case OP_WORDCHAR:      case OP_WORDCHAR:
2554      case OP_ANY:  
     case OP_ANYBYTE:  
2555      case OP_CHAR:      case OP_CHAR:
2556      case OP_CHARNC:      case OP_CHARI:
2557      case OP_NOT:      case OP_NOT:
2558        case OP_NOTI:
2559    
2560      case OP_PLUS:      case OP_PLUS:
2561        case OP_PLUSI:
2562      case OP_MINPLUS:      case OP_MINPLUS:
2563      case OP_POSPLUS:      case OP_MINPLUSI:
2564      case OP_EXACT:  
2565      case OP_NOTPLUS:      case OP_NOTPLUS:
2566        case OP_NOTPLUSI:
2567      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2568        case OP_NOTMINPLUSI:
2569    
2570        case OP_POSPLUS:
2571        case OP_POSPLUSI:
2572      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2573        case OP_NOTPOSPLUSI:
2574    
2575        case OP_EXACT:
2576        case OP_EXACTI:
2577      case OP_NOTEXACT:      case OP_NOTEXACT:
2578        case OP_NOTEXACTI:
2579    
2580      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2581      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2582      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2583      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2584    
2585      return FALSE;      return FALSE;
2586    
2587      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 1688  for (code = first_significant_code(code Line 2601  for (code = first_significant_code(code
2601      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2602      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2603      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2604      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2605          code += 2;
2606      break;      break;
2607    
2608      /* End of branch */      /* End of branch */
# Line 1696  for (code = first_significant_code(code Line 2610  for (code = first_significant_code(code
2610      case OP_KET:      case OP_KET:
2611      case OP_KETRMAX:      case OP_KETRMAX:
2612      case OP_KETRMIN:      case OP_KETRMIN:
2613        case OP_KETRPOS:
2614      case OP_ALT:      case OP_ALT:
2615      return TRUE;      return TRUE;
2616    
2617      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2618      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2619        followed by a multibyte character. */
2620    
2621  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2622      case OP_STAR:      case OP_STAR:
2623        case OP_STARI:
2624        case OP_NOTSTAR:
2625        case OP_NOTSTARI:
2626    
2627      case OP_MINSTAR:      case OP_MINSTAR:
2628        case OP_MINSTARI:
2629        case OP_NOTMINSTAR:
2630        case OP_NOTMINSTARI:
2631    
2632      case OP_POSSTAR:      case OP_POSSTAR:
2633        case OP_POSSTARI:
2634        case OP_NOTPOSSTAR:
2635        case OP_NOTPOSSTARI:
2636    
2637      case OP_QUERY:      case OP_QUERY:
2638        case OP_QUERYI:
2639        case OP_NOTQUERY:
2640        case OP_NOTQUERYI:
2641    
2642      case OP_MINQUERY:      case OP_MINQUERY:
2643        case OP_MINQUERYI:
2644        case OP_NOTMINQUERY:
2645        case OP_NOTMINQUERYI:
2646    
2647      case OP_POSQUERY:      case OP_POSQUERY:
2648        case OP_POSQUERYI:
2649        case OP_NOTPOSQUERY:
2650        case OP_NOTPOSQUERYI:
2651    
2652        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2653        break;
2654    
2655      case OP_UPTO:      case OP_UPTO:
2656        case OP_UPTOI:
2657        case OP_NOTUPTO:
2658        case OP_NOTUPTOI:
2659    
2660      case OP_MINUPTO:      case OP_MINUPTO:
2661        case OP_MINUPTOI:
2662        case OP_NOTMINUPTO:
2663        case OP_NOTMINUPTOI:
2664    
2665      case OP_POSUPTO:      case OP_POSUPTO:
2666      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2667        case OP_NOTPOSUPTO:
2668        case OP_NOTPOSUPTOI:
2669    
2670        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2671      break;      break;
2672  #endif  #endif
2673    
2674        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2675        string. */
2676    
2677        case OP_MARK:
2678        case OP_PRUNE_ARG:
2679        case OP_SKIP_ARG:
2680        case OP_THEN_ARG:
2681        code += code[1];
2682        break;
2683    
2684        /* None of the remaining opcodes are required to match a character. */
2685    
2686        default:
2687        break;
2688      }      }
2689    }    }
2690    
# Line 1731  return TRUE; Line 2701  return TRUE;
2701  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
2702  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,
2703  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.
2704    This function is called only during the real compile, not during the
2705    pre-compile.
2706    
2707  Arguments:  Arguments:
2708    code        points to start of the recursion    code        points to start of the recursion
2709    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2710    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2711    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2712      cd          pointers to tables etc
2713    
2714  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2715  */  */
2716    
2717  static BOOL  static BOOL
2718  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2719    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2720  {  {
2721  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2722    {    {
2723    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2724        return FALSE;
2725    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2726    }    }
2727  return TRUE;  return TRUE;
2728    }
2729    
2730    
2731    
2732    /*************************************************
2733    *        Base opcode of repeated opcodes         *
2734    *************************************************/
2735    
2736    /* Returns the base opcode for repeated single character type opcodes. If the
2737    opcode is not a repeated character type, it returns with the original value.
2738    
2739    Arguments:  c opcode
2740    Returns:    base opcode for the type
2741    */
2742    
2743    static pcre_uchar
2744    get_repeat_base(pcre_uchar c)
2745    {
2746    return (c > OP_TYPEPOSUPTO)? c :
2747           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2748           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2749           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2750           (c >= OP_STARI)?      OP_STARI :
2751                                 OP_STAR;
2752    }
2753    
2754    
2755    
2756    #ifdef SUPPORT_UCP
2757    /*************************************************
2758    *        Check a character and a property        *
2759    *************************************************/
2760    
2761    /* This function is called by check_auto_possessive() when a property item
2762    is adjacent to a fixed character.
2763    
2764    Arguments:
2765      c            the character
2766      ptype        the property type
2767      pdata        the data for the type
2768      negated      TRUE if it's a negated property (\P or \p{^)
2769    
2770    Returns:       TRUE if auto-possessifying is OK
2771    */
2772    
2773    static BOOL
2774    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2775      BOOL negated)
2776    {
2777    const pcre_uint32 *p;
2778    const ucd_record *prop = GET_UCD(c);
2779    
2780    switch(ptype)
2781      {
2782      case PT_LAMP:
2783      return (prop->chartype == ucp_Lu ||
2784              prop->chartype == ucp_Ll ||
2785              prop->chartype == ucp_Lt) == negated;
2786    
2787      case PT_GC:
2788      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2789    
2790      case PT_PC:
2791      return (pdata == prop->chartype) == negated;
2792    
2793      case PT_SC:
2794      return (pdata == prop->script) == negated;
2795    
2796      /* These are specials */
2797    
2798      case PT_ALNUM:
2799      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2800              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2801    
2802      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2803      means that Perl space and POSIX space are now identical. PCRE was changed
2804      at release 8.34. */
2805    
2806      case PT_SPACE:    /* Perl space */
2807      case PT_PXSPACE:  /* POSIX space */
2808      switch(c)
2809        {
2810        HSPACE_CASES:
2811        VSPACE_CASES:
2812        return negated;
2813    
2814        default:
2815        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2816        }
2817      break;  /* Control never reaches here */
2818    
2819      case PT_WORD:
2820      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2821              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2822              c == CHAR_UNDERSCORE) == negated;
2823    
2824      case PT_CLIST:
2825      p = PRIV(ucd_caseless_sets) + prop->caseset;
2826      for (;;)
2827        {
2828        if (c < *p) return !negated;
2829        if (c == *p++) return negated;
2830        }
2831      break;  /* Control never reaches here */
2832      }
2833    
2834    return FALSE;
2835    }
2836    #endif  /* SUPPORT_UCP */
2837    
2838    
2839    
2840    /*************************************************
2841    *        Fill the character property list        *
2842    *************************************************/
2843    
2844    /* Checks whether the code points to an opcode that can take part in auto-
2845    possessification, and if so, fills a list with its properties.
2846    
2847    Arguments:
2848      code        points to start of expression
2849      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2850      fcc         points to case-flipping table
2851      list        points to output list
2852                  list[0] will be filled with the opcode
2853                  list[1] will be non-zero if this opcode
2854                    can match an empty character string
2855                  list[2..7] depends on the opcode
2856    
2857    Returns:      points to the start of the next opcode if *code is accepted
2858                  NULL if *code is not accepted
2859    */
2860    
2861    static const pcre_uchar *
2862    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2863      const pcre_uint8 *fcc, pcre_uint32 *list)
2864    {
2865    pcre_uchar c = *code;
2866    pcre_uchar base;
2867    const pcre_uchar *end;
2868    pcre_uint32 chr;
2869    
2870    #ifdef SUPPORT_UCP
2871    pcre_uint32 *clist_dest;
2872    const pcre_uint32 *clist_src;
2873    #else
2874    utf = utf;  /* Suppress "unused parameter" compiler warning */
2875    #endif
2876    
2877    list[0] = c;
2878    list[1] = FALSE;
2879    code++;
2880    
2881    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2882      {
2883      base = get_repeat_base(c);
2884      c -= (base - OP_STAR);
2885    
2886      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2887        code += IMM2_SIZE;
2888    
2889      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2890    
2891      switch(base)
2892        {
2893        case OP_STAR:
2894        list[0] = OP_CHAR;
2895        break;
2896    
2897        case OP_STARI:
2898        list[0] = OP_CHARI;
2899        break;
2900    
2901        case OP_NOTSTAR:
2902        list[0] = OP_NOT;
2903        break;
2904    
2905        case OP_NOTSTARI:
2906        list[0] = OP_NOTI;
2907        break;
2908    
2909        case OP_TYPESTAR:
2910        list[0] = *code;
2911        code++;
2912        break;
2913        }
2914      c = list[0];
2915      }
2916    
2917    switch(c)
2918      {
2919      case OP_NOT_DIGIT:
2920      case OP_DIGIT:
2921      case OP_NOT_WHITESPACE:
2922      case OP_WHITESPACE:
2923      case OP_NOT_WORDCHAR:
2924      case OP_WORDCHAR:
2925      case OP_ANY:
2926      case OP_ALLANY:
2927      case OP_ANYNL:
2928      case OP_NOT_HSPACE:
2929      case OP_HSPACE:
2930      case OP_NOT_VSPACE:
2931      case OP_VSPACE:
2932      case OP_EXTUNI:
2933      case OP_EODN:
2934      case OP_EOD:
2935      case OP_DOLL:
2936      case OP_DOLLM:
2937      return code;
2938    
2939      case OP_CHAR:
2940      case OP_NOT:
2941      GETCHARINCTEST(chr, code);
2942      list[2] = chr;
2943      list[3] = NOTACHAR;
2944      return code;
2945    
2946      case OP_CHARI:
2947      case OP_NOTI:
2948      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2949      GETCHARINCTEST(chr, code);
2950      list[2] = chr;
2951    
2952    #ifdef SUPPORT_UCP
2953      if (chr < 128 || (chr < 256 && !utf))
2954        list[3] = fcc[chr];
2955      else
2956        list[3] = UCD_OTHERCASE(chr);
2957    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2958      list[3] = (chr < 256) ? fcc[chr] : chr;
2959    #else
2960      list[3] = fcc[chr];
2961    #endif
2962    
2963      /* The othercase might be the same value. */
2964    
2965      if (chr == list[3])
2966        list[3] = NOTACHAR;
2967      else
2968        list[4] = NOTACHAR;
2969      return code;
2970    
2971    #ifdef SUPPORT_UCP
2972      case OP_PROP:
2973      case OP_NOTPROP:
2974      if (code[0] != PT_CLIST)
2975        {
2976        list[2] = code[0];
2977        list[3] = code[1];
2978        return code + 2;
2979        }
2980    
2981      /* Convert only if we have enough space. */
2982    
2983      clist_src = PRIV(ucd_caseless_sets) + code[1];
2984      clist_dest = list + 2;
2985      code += 2;
2986    
2987      do {
2988         if (clist_dest >= list + 8)
2989           {
2990           /* Early return if there is not enough space. This should never
2991           happen, since all clists are shorter than 5 character now. */
2992           list[2] = code[0];
2993           list[3] = code[1];
2994           return code;
2995           }
2996         *clist_dest++ = *clist_src;
2997         }
2998      while(*clist_src++ != NOTACHAR);
2999    
3000      /* All characters are stored. The terminating NOTACHAR
3001      is copied form the clist itself. */
3002    
3003      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3004      return code;
3005    #endif
3006    
3007      case OP_NCLASS:
3008      case OP_CLASS:
3009    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3010      case OP_XCLASS:
3011      if (c == OP_XCLASS)
3012        end = code + GET(code, 0) - 1;
3013      else
3014    #endif
3015        end = code + 32 / sizeof(pcre_uchar);
3016    
3017      switch(*end)
3018        {
3019        case OP_CRSTAR:
3020        case OP_CRMINSTAR:
3021        case OP_CRQUERY:
3022        case OP_CRMINQUERY:
3023        case OP_CRPOSSTAR:
3024        case OP_CRPOSQUERY:
3025        list[1] = TRUE;
3026        end++;
3027        break;
3028    
3029        case OP_CRPLUS:
3030        case OP_CRMINPLUS:
3031        case OP_CRPOSPLUS:
3032        end++;
3033        break;
3034    
3035        case OP_CRRANGE:
3036        case OP_CRMINRANGE:
3037        case OP_CRPOSRANGE:
3038        list[1] = (GET2(end, 1) == 0);
3039        end += 1 + 2 * IMM2_SIZE;
3040        break;
3041        }
3042      list[2] = (pcre_uint32)(end - code);
3043      return end;
3044      }
3045    return NULL;    /* Opcode not accepted */
3046    }
3047    
3048    
3049    
3050    /*************************************************
3051    *    Scan further character sets for match       *
3052    *************************************************/
3053    
3054    /* Checks whether the base and the current opcode have a common character, in
3055    which case the base cannot be possessified.
3056    
3057    Arguments:
3058      code        points to the byte code
3059      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3060      cd          static compile data
3061      base_list   the data list of the base opcode
3062    
3063    Returns:      TRUE if the auto-possessification is possible
3064    */
3065    
3066    static BOOL
3067    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3068      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3069    {
3070    pcre_uchar c;
3071    pcre_uint32 list[8];
3072    const pcre_uint32 *chr_ptr;
3073    const pcre_uint32 *ochr_ptr;
3074    const pcre_uint32 *list_ptr;
3075    const pcre_uchar *next_code;
3076    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3077    const pcre_uchar *xclass_flags;
3078    #endif
3079    const pcre_uint8 *class_bitset;
3080    const pcre_uint8 *set1, *set2, *set_end;
3081    pcre_uint32 chr;
3082    BOOL accepted, invert_bits;
3083    BOOL entered_a_group = FALSE;
3084    
3085    /* Note: the base_list[1] contains whether the current opcode has greedy
3086    (represented by a non-zero value) quantifier. This is a different from
3087    other character type lists, which stores here that the character iterator
3088    matches to an empty string (also represented by a non-zero value). */
3089    
3090    for(;;)
3091      {
3092      /* All operations move the code pointer forward.
3093      Therefore infinite recursions are not possible. */
3094    
3095      c = *code;
3096    
3097      /* Skip over callouts */
3098    
3099      if (c == OP_CALLOUT)
3100        {
3101        code += PRIV(OP_lengths)[c];
3102        continue;
3103        }
3104    
3105      if (c == OP_ALT)
3106        {
3107        do code += GET(code, 1); while (*code == OP_ALT);
3108        c = *code;
3109        }
3110    
3111      switch(c)
3112        {
3113        case OP_END:
3114        case OP_KETRPOS:
3115        /* TRUE only in greedy case. The non-greedy case could be replaced by
3116        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3117        uses more memory, which we cannot get at this stage.) */
3118    
3119        return base_list[1] != 0;
3120    
3121        case OP_KET:
3122        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3123        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3124        cannot be converted to a possessive form. */
3125    
3126        if (base_list[1] == 0) return FALSE;
3127    
3128        switch(*(code - GET(code, 1)))
3129          {
3130          case OP_ASSERT:
3131          case OP_ASSERT_NOT:
3132          case OP_ASSERTBACK:
3133          case OP_ASSERTBACK_NOT:
3134          case OP_ONCE:
3135          case OP_ONCE_NC:
3136          /* Atomic sub-patterns and assertions can always auto-possessify their
3137          last iterator. However, if the group was entered as a result of checking
3138          a previous iterator, this is not possible. */
3139    
3140          return !entered_a_group;
3141          }
3142    
3143        code += PRIV(OP_lengths)[c];
3144        continue;
3145    
3146        case OP_ONCE:
3147        case OP_ONCE_NC:
3148        case OP_BRA:
3149        case OP_CBRA:
3150        next_code = code + GET(code, 1);
3151        code += PRIV(OP_lengths)[c];
3152    
3153        while (*next_code == OP_ALT)
3154          {
3155          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3156          code = next_code + 1 + LINK_SIZE;
3157          next_code += GET(next_code, 1);
3158          }
3159    
3160        entered_a_group = TRUE;
3161        continue;
3162    
3163        case OP_BRAZERO:
3164        case OP_BRAMINZERO:
3165    
3166        next_code = code + 1;
3167        if (*next_code != OP_BRA && *next_code != OP_CBRA
3168            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3169    
3170        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3171    
3172        /* The bracket content will be checked by the
3173        OP_BRA/OP_CBRA case above. */
3174        next_code += 1 + LINK_SIZE;
3175        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3176          return FALSE;
3177    
3178        code += PRIV(OP_lengths)[c];
3179        continue;
3180    
3181        default:
3182        break;
3183        }
3184    
3185      /* Check for a supported opcode, and load its properties. */
3186    
3187      code = get_chr_property_list(code, utf, cd->fcc, list);
3188      if (code == NULL) return FALSE;    /* Unsupported */
3189    
3190      /* If either opcode is a small character list, set pointers for comparing
3191      characters from that list with another list, or with a property. */
3192    
3193      if (base_list[0] == OP_CHAR)
3194        {
3195        chr_ptr = base_list + 2;
3196        list_ptr = list;
3197        }
3198      else if (list[0] == OP_CHAR)
3199        {
3200        chr_ptr = list + 2;
3201        list_ptr = base_list;
3202        }
3203    
3204      /* Character bitsets can also be compared to certain opcodes. */
3205    
3206      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3207    #ifdef COMPILE_PCRE8
3208          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3209          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3210    #endif
3211          )
3212        {
3213    #ifdef COMPILE_PCRE8
3214        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3215    #else
3216        if (base_list[0] == OP_CLASS)
3217    #endif
3218          {
3219          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3220          list_ptr = list;
3221          }
3222        else
3223          {
3224          set1 = (pcre_uint8 *)(code - list[2]);
3225          list_ptr = base_list;
3226          }
3227    
3228        invert_bits = FALSE;
3229        switch(list_ptr[0])
3230          {
3231          case OP_CLASS:
3232          case OP_NCLASS:
3233          set2 = (pcre_uint8 *)
3234            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3235          break;
3236    
3237    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3238          case OP_XCLASS:
3239          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3240          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3241          if ((*xclass_flags & XCL_MAP) == 0)
3242            {
3243            /* No bits are set for characters < 256. */
3244            if (list[1] == 0) return TRUE;
3245            /* Might be an empty repeat. */
3246            continue;
3247            }
3248          set2 = (pcre_uint8 *)(xclass_flags + 1);
3249          break;
3250    #endif
3251    
3252          case OP_NOT_DIGIT:
3253          invert_bits = TRUE;
3254          /* Fall through */
3255          case OP_DIGIT:
3256          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3257          break;
3258    
3259          case OP_NOT_WHITESPACE:
3260          invert_bits = TRUE;
3261          /* Fall through */
3262          case OP_WHITESPACE:
3263          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3264          break;
3265    
3266          case OP_NOT_WORDCHAR:
3267          invert_bits = TRUE;
3268          /* Fall through */
3269          case OP_WORDCHAR:
3270          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3271          break;
3272    
3273          default:
3274          return FALSE;
3275          }
3276    
3277        /* Because the sets are unaligned, we need
3278        to perform byte comparison here. */
3279        set_end = set1 + 32;
3280        if (invert_bits)
3281          {
3282          do
3283            {
3284            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3285            }
3286          while (set1 < set_end);
3287          }
3288        else
3289          {
3290          do
3291            {
3292            if ((*set1++ & *set2++) != 0) return FALSE;
3293            }
3294          while (set1 < set_end);
3295          }
3296    
3297        if (list[1] == 0) return TRUE;
3298        /* Might be an empty repeat. */
3299        continue;
3300        }
3301    
3302      /* Some property combinations also acceptable. Unicode property opcodes are
3303      processed specially; the rest can be handled with a lookup table. */
3304    
3305      else
3306        {
3307        pcre_uint32 leftop, rightop;
3308    
3309        leftop = base_list[0];
3310        rightop = list[0];
3311    
3312    #ifdef SUPPORT_UCP
3313        accepted = FALSE; /* Always set in non-unicode case. */
3314        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3315          {
3316          if (rightop == OP_EOD)
3317            accepted = TRUE;
3318          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3319            {
3320            int n;
3321            const pcre_uint8 *p;
3322            BOOL same = leftop == rightop;
3323            BOOL lisprop = leftop == OP_PROP;
3324            BOOL risprop = rightop == OP_PROP;
3325            BOOL bothprop = lisprop && risprop;
3326    
3327            /* There's a table that specifies how each combination is to be
3328            processed:
3329              0   Always return FALSE (never auto-possessify)
3330              1   Character groups are distinct (possessify if both are OP_PROP)
3331              2   Check character categories in the same group (general or particular)
3332              3   Return TRUE if the two opcodes are not the same
3333              ... see comments below
3334            */
3335    
3336            n = propposstab[base_list[2]][list[2]];
3337            switch(n)
3338              {
3339              case 0: break;
3340              case 1: accepted = bothprop; break;
3341              case 2: accepted = (base_list[3] == list[3]) != same; break;
3342              case 3: accepted = !same; break;
3343    
3344              case 4:  /* Left general category, right particular category */
3345              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3346              break;
3347    
3348              case 5:  /* Right general category, left particular category */
3349              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3350              break;
3351    
3352              /* This code is logically tricky. Think hard before fiddling with it.
3353              The posspropstab table has four entries per row. Each row relates to
3354              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3355              Only WORD actually needs all four entries, but using repeats for the
3356              others means they can all use the same code below.
3357    
3358              The first two entries in each row are Unicode general categories, and
3359              apply always, because all the characters they include are part of the
3360              PCRE character set. The third and fourth entries are a general and a
3361              particular category, respectively, that include one or more relevant
3362              characters. One or the other is used, depending on whether the check
3363              is for a general or a particular category. However, in both cases the
3364              category contains more characters than the specials that are defined
3365              for the property being tested against. Therefore, it cannot be used
3366              in a NOTPROP case.
3367    
3368              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3369              Underscore is covered by ucp_P or ucp_Po. */
3370    
3371              case 6:  /* Left alphanum vs right general category */
3372              case 7:  /* Left space vs right general category */
3373              case 8:  /* Left word vs right general category */
3374              p = posspropstab[n-6];
3375              accepted = risprop && lisprop ==
3376                (list[3] != p[0] &&
3377                 list[3] != p[1] &&
3378                (list[3] != p[2] || !lisprop));
3379              break;
3380    
3381              case 9:   /* Right alphanum vs left general category */
3382              case 10:  /* Right space vs left general category */
3383              case 11:  /* Right word vs left general category */
3384              p = posspropstab[n-9];
3385              accepted = lisprop && risprop ==
3386                (base_list[3] != p[0] &&
3387                 base_list[3] != p[1] &&
3388                (base_list[3] != p[2] || !risprop));
3389              break;
3390    
3391              case 12:  /* Left alphanum vs right particular category */
3392              case 13:  /* Left space vs right particular category */
3393              case 14:  /* Left word vs right particular category */
3394              p = posspropstab[n-12];
3395              accepted = risprop && lisprop ==
3396                (catposstab[p[0]][list[3]] &&
3397                 catposstab[p[1]][list[3]] &&
3398                (list[3] != p[3] || !lisprop));
3399              break;
3400    
3401              case 15:  /* Right alphanum vs left particular category */
3402              case 16:  /* Right space vs left particular category */
3403              case 17:  /* Right word vs left particular category */
3404              p = posspropstab[n-15];
3405              accepted = lisprop && risprop ==
3406                (catposstab[p[0]][base_list[3]] &&
3407                 catposstab[p[1]][base_list[3]] &&
3408                (base_list[3] != p[3] || !risprop));
3409              break;
3410              }
3411            }
3412          }
3413    
3414        else
3415    #endif  /* SUPPORT_UCP */
3416    
3417        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3418               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3419               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3420    
3421        if (!accepted) return FALSE;
3422    
3423        if (list[1] == 0) return TRUE;
3424        /* Might be an empty repeat. */
3425        continue;
3426        }
3427    
3428      /* Control reaches here only if one of the items is a small character list.
3429      All characters are checked against the other side. */
3430    
3431      do
3432        {
3433        chr = *chr_ptr;
3434    
3435        switch(list_ptr[0])
3436          {
3437          case OP_CHAR:
3438          ochr_ptr = list_ptr + 2;
3439          do
3440            {
3441            if (chr == *ochr_ptr) return FALSE;
3442            ochr_ptr++;
3443            }
3444          while(*ochr_ptr != NOTACHAR);
3445          break;
3446    
3447          case OP_NOT:
3448          ochr_ptr = list_ptr + 2;
3449          do
3450            {
3451            if (chr == *ochr_ptr)
3452              break;
3453            ochr_ptr++;
3454            }
3455          while(*ochr_ptr != NOTACHAR);
3456          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3457          break;
3458    
3459          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3460          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3461    
3462          case OP_DIGIT:
3463          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3464          break;
3465    
3466          case OP_NOT_DIGIT:
3467          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3468          break;
3469    
3470          case OP_WHITESPACE:
3471          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3472          break;
3473    
3474          case OP_NOT_WHITESPACE:
3475          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3476          break;
3477    
3478          case OP_WORDCHAR:
3479          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3480          break;
3481    
3482          case OP_NOT_WORDCHAR:
3483          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3484          break;
3485    
3486          case OP_HSPACE:
3487          switch(chr)
3488            {
3489            HSPACE_CASES: return FALSE;
3490            default: break;
3491            }
3492          break;
3493    
3494          case OP_NOT_HSPACE:
3495          switch(chr)
3496            {
3497            HSPACE_CASES: break;
3498            default: return FALSE;
3499            }
3500          break;
3501    
3502          case OP_ANYNL:
3503          case OP_VSPACE:
3504          switch(chr)
3505            {
3506            VSPACE_CASES: return FALSE;
3507            default: break;
3508            }
3509          break;
3510    
3511          case OP_NOT_VSPACE:
3512          switch(chr)
3513            {
3514            VSPACE_CASES: break;
3515            default: return FALSE;
3516            }
3517          break;
3518    
3519          case OP_DOLL:
3520          case OP_EODN:
3521          switch (chr)
3522            {
3523            case CHAR_CR:
3524            case CHAR_LF:
3525            case CHAR_VT:
3526            case CHAR_FF:
3527            case CHAR_NEL:
3528    #ifndef EBCDIC
3529            case 0x2028:
3530            case 0x2029:
3531    #endif  /* Not EBCDIC */
3532            return FALSE;
3533            }
3534          break;
3535    
3536          case OP_EOD:    /* Can always possessify before \z */
3537          break;
3538    
3539    #ifdef SUPPORT_UCP
3540          case OP_PROP:
3541          case OP_NOTPROP:
3542          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3543                list_ptr[0] == OP_NOTPROP))
3544            return FALSE;
3545          break;
3546    #endif
3547    
3548          case OP_NCLASS:
3549          if (chr > 255) return FALSE;
3550          /* Fall through */
3551    
3552          case OP_CLASS:
3553          if (chr > 255) break;
3554          class_bitset = (pcre_uint8 *)
3555            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3556          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3557          break;
3558    
3559    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3560          case OP_XCLASS:
3561          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3562              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3563          break;
3564    #endif
3565    
3566          default:
3567          return FALSE;
3568          }
3569    
3570        chr_ptr++;
3571        }
3572      while(*chr_ptr != NOTACHAR);
3573    
3574      /* At least one character must be matched from this opcode. */
3575    
3576      if (list[1] == 0) return TRUE;
3577      }
3578    
3579    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3580    but some compilers complain about an unreachable statement. */
3581    
3582    }
3583    
3584    
3585    
3586    /*************************************************
3587    *    Scan compiled regex for auto-possession     *
3588    *************************************************/
3589    
3590    /* Replaces single character iterations with their possessive alternatives
3591    if appropriate. This function modifies the compiled opcode!
3592    
3593    Arguments:
3594      code        points to start of the byte code
3595      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3596      cd          static compile data
3597    
3598    Returns:      nothing
3599    */
3600    
3601    static void
3602    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3603    {
3604    register pcre_uchar c;
3605    const pcre_uchar *end;
3606    pcre_uchar *repeat_opcode;
3607    pcre_uint32 list[8];
3608    
3609    for (;;)
3610      {
3611      c = *code;
3612    
3613      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3614        {
3615        c -= get_repeat_base(c) - OP_STAR;
3616        end = (c <= OP_MINUPTO) ?
3617          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3618        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3619    
3620        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3621          {
3622          switch(c)
3623            {
3624            case OP_STAR:
3625            *code += OP_POSSTAR - OP_STAR;
3626            break;