/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

code/trunk/pcre_compile.c revision 218 by ph10, Thu Aug 16 10:13:23 2007 UTC code/branches/pcre16/pcre_compile.c revision 764 by zherczeg, Wed Nov 23 17:23:20 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 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 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
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_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  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
103  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
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  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,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185    searched linearly. Put all the names into a single string, in order to reduce
186    the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    const char *name;    int   len;                 /* Length of verb name */
192    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static verbitem verbs[] = {  static const char verbnames[] =
197    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
198    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
199    { "F",      1, OP_FAIL },    STRING_ACCEPT0
200    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
201    { "PRUNE",  5, OP_PRUNE },    STRING_F0
202    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
203    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
220    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
221    
222  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
223    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
225    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
235    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 };
236    
237  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 200  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const pcre_uchar string_PNd[]  = {
270      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
271      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
272    static const pcre_uchar string_pNd[]  = {
273      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
274      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
275    static const pcre_uchar string_PXsp[] = {
276      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
277      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
278    static const pcre_uchar string_pXsp[] = {
279      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
280      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
281    static const pcre_uchar string_PXwd[] = {
282      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
283      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
284    static const pcre_uchar string_pXwd[] = {
285      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
286      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
287    
288    static const pcre_uchar *substitutes[] = {
289      string_PNd,           /* \D */
290      string_pNd,           /* \d */
291      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
292      string_pXsp,          /* \s */
293      string_PXwd,          /* \W */
294      string_pXwd           /* \w */
295    };
296    
297    static const pcre_uchar string_pL[] =   {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_pLl[] =  {
301      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pLu[] =  {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_pXan[] = {
307      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_h[] =    {
310      CHAR_BACKSLASH, CHAR_h, '\0' };
311    static const pcre_uchar string_pXps[] = {
312      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_PL[] =   {
315      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    static const pcre_uchar string_PLl[] =  {
318      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
319      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
320    static const pcre_uchar string_PLu[] =  {
321      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
322      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
323    static const pcre_uchar string_PXan[] = {
324      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
325      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
326    static const pcre_uchar string_H[] =    {
327      CHAR_BACKSLASH, CHAR_H, '\0' };
328    static const pcre_uchar string_PXps[] = {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    
332    static const pcre_uchar *posix_substitutes[] = {
333      string_pL,            /* alpha */
334      string_pLl,           /* lower */
335      string_pLu,           /* upper */
336      string_pXan,          /* alnum */
337      NULL,                 /* ascii */
338      string_h,             /* blank */
339      NULL,                 /* cntrl */
340      string_pNd,           /* digit */
341      NULL,                 /* graph */
342      NULL,                 /* print */
343      NULL,                 /* punct */
344      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
345      string_pXwd,          /* word */
346      NULL,                 /* xdigit */
347      /* Negated cases */
348      string_PL,            /* ^alpha */
349      string_PLl,           /* ^lower */
350      string_PLu,           /* ^upper */
351      string_PXan,          /* ^alnum */
352      NULL,                 /* ^ascii */
353      string_H,             /* ^blank */
354      NULL,                 /* ^cntrl */
355      string_PNd,           /* ^digit */
356      NULL,                 /* ^graph */
357      NULL,                 /* ^print */
358      NULL,                 /* ^punct */
359      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
360      string_PXwd,          /* ^word */
361      NULL                  /* ^xdigit */
362    };
363    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
364    #endif
365    
366  #define STRING(a)  # a  #define STRING(a)  # a
367  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 207  static const int posix_class_maps[] = { Line 369  static const int posix_class_maps[] = {
369  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
370  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
371  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
372  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
373    the number of relocations needed when a shared library is loaded dynamically,
374  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
375    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
376    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
377    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
378    "unrecognized character follows \\",  
379    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
380    substring, so that the whole string ends with \0\0, which can be detected when
381    counting through. */
382    
383    static const char error_texts[] =
384      "no error\0"
385      "\\ at end of pattern\0"
386      "\\c at end of pattern\0"
387      "unrecognized character follows \\\0"
388      "numbers out of order in {} quantifier\0"
389    /* 5 */    /* 5 */
390    "number too big in {} quantifier",    "number too big in {} quantifier\0"
391    "missing terminating ] for character class",    "missing terminating ] for character class\0"
392    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
393    "range out of order in character class",    "range out of order in character class\0"
394    "nothing to repeat",    "nothing to repeat\0"
395    /* 10 */    /* 10 */
396    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
397    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
398    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
399    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
400    "missing )",    "missing )\0"
401    /* 15 */    /* 15 */
402    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
403    "erroffset passed as NULL",    "erroffset passed as NULL\0"
404    "unknown option bit(s) set",    "unknown option bit(s) set\0"
405    "missing ) after comment",    "missing ) after comment\0"
406    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
407    /* 20 */    /* 20 */
408    "regular expression is too large",    "regular expression is too large\0"
409    "failed to get memory",    "failed to get memory\0"
410    "unmatched parentheses",    "unmatched parentheses\0"
411    "internal error: code overflow",    "internal error: code overflow\0"
412    "unrecognized character after (?<",    "unrecognized character after (?<\0"
413    /* 25 */    /* 25 */
414    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
415    "malformed number or name after (?(",    "malformed number or name after (?(\0"
416    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
417    "assertion expected after (?(",    "assertion expected after (?(\0"
418    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
419    /* 30 */    /* 30 */
420    "unknown POSIX class name",    "unknown POSIX class name\0"
421    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
422    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
423    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
424    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
425    /* 35 */    /* 35 */
426    "invalid condition (?(0)",    "invalid condition (?(0)\0"
427    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
428    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
429    "number after (?C is > 255",    "number after (?C is > 255\0"
430    "closing ) for (?C expected",    "closing ) for (?C expected\0"
431    /* 40 */    /* 40 */
432    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
433    "unrecognized character after (?P",    "unrecognized character after (?P\0"
434    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
435    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
436    "invalid UTF-8 string",    "invalid UTF-8 string\0"
437    /* 45 */    /* 45 */
438    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
439    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
440    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
441    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
442    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
443    /* 50 */    /* 50 */
444    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
445    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
446    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
447    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
448    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
449    /* 55 */    /* 55 */
450    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
451    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
452    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
453    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
454    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
455    /* 60 */    /* 60 */
456    "(*VERB) not recognized",    "(*VERB) not recognized\0"
457    "number is too big"    "number is too big\0"
458  };    "subpattern name expected\0"
459      "digit expected after (?+\0"
460      "] is an invalid data character in JavaScript compatibility mode\0"
461      /* 65 */
462      "different names for subpatterns of the same number are not allowed\0"
463      "(*MARK) must have an argument\0"
464      "this version of PCRE is not compiled with PCRE_UCP support\0"
465      "\\c must be followed by an ASCII character\0"
466      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
467      /* 70 */
468      "internal error: unknown opcode in find_fixedlength()\0"
469      ;
470    
471  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
472  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 303  For convenience, we use the same bit def Line 484  For convenience, we use the same bit def
484    
485  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
486    
487  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
488    
489    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
490    UTF-8 mode. */
491    
492  static const unsigned char digitab[] =  static const unsigned char digitab[] =
493    {    {
494    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 339  static const unsigned char digitab[] = Line 524  static const unsigned char digitab[] =
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
525    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
526    
527  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
528    
529    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
530    
531  static const unsigned char digitab[] =  static const unsigned char digitab[] =
532    {    {
533    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 414  static const unsigned char ebcdic_charta Line 602  static const unsigned char ebcdic_charta
602  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
603    
604  static BOOL  static BOOL
605    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
606      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
607    
608    
609    
610  /*************************************************  /*************************************************
611    *            Find an error text                  *
612    *************************************************/
613    
614    /* The error texts are now all in one long string, to save on relocations. As
615    some of the text is of unknown length, we can't use a table of offsets.
616    Instead, just count through the strings. This is not a performance issue
617    because it happens only when there has been a compilation error.
618    
619    Argument:   the error number
620    Returns:    pointer to the error string
621    */
622    
623    static const char *
624    find_error_text(int n)
625    {
626    const char *s = error_texts;
627    for (; n > 0; n--)
628      {
629      while (*s++ != 0) {};
630      if (*s == 0) return "Error text not found (please report)";
631      }
632    return s;
633    }
634    
635    
636    /*************************************************
637    *            Check for counted repeat            *
638    *************************************************/
639    
640    /* This function is called when a '{' is encountered in a place where it might
641    start a quantifier. It looks ahead to see if it really is a quantifier or not.
642    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
643    where the ddds are digits.
644    
645    Arguments:
646      p         pointer to the first char after '{'
647    
648    Returns:    TRUE or FALSE
649    */
650    
651    static BOOL
652    is_counted_repeat(const pcre_uchar *p)
653    {
654    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
655    while ((digitab[*p] & ctype_digit) != 0) p++;
656    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
657    
658    if (*p++ != CHAR_COMMA) return FALSE;
659    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
660    
661    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
662    while ((digitab[*p] & ctype_digit) != 0) p++;
663    
664    return (*p == CHAR_RIGHT_CURLY_BRACKET);
665    }
666    
667    
668    
669    /*************************************************
670  *            Handle escapes                      *  *            Handle escapes                      *
671  *************************************************/  *************************************************/
672    
# Line 444  Returns:         zero or positive => a d Line 691  Returns:         zero or positive => a d
691  */  */
692    
693  static int  static int
694  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
695    int options, BOOL isclass)    int options, BOOL isclass)
696  {  {
697  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
698  const uschar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
699  int c, i;  int c, i;
700    
701  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 458  ptr--;                            /* Set Line 705  ptr--;                            /* Set
705    
706  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
707    
708  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
709  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.
710  Otherwise further processing may be required. */  Otherwise further processing may be required. */
711    
712  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
713  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
714  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
715    
716  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
717  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
718  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
719  #endif  #endif
720    
# Line 475  else if ((i = escapes[c - 0x48]) != 0) Line 722  else if ((i = escapes[c - 0x48]) != 0)
722    
723  else  else
724    {    {
725    const uschar *oldptr;    const pcre_uchar *oldptr;
726    BOOL braced, negated;    BOOL braced, negated;
727    
728    switch (c)    switch (c)
# Line 483  else Line 730  else
730      /* 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
731      error. */      error. */
732    
733      case 'l':      case CHAR_l:
734      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
735      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
736      break;      break;
737    
738      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
739      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
740      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
741      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
742      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
743      fudge it up by pretending it really was \k. */        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
744               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
745      case 'g':          {
746      if (ptr[1] == '{')          c = 0;
747        {          for (i = 0; i < 4; ++i)
748        const uschar *p;            {
749        for (p = ptr+2; *p != 0 && *p != '}'; p++)            register int cc = *(++ptr);
750          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
751        if (*p != 0 && *p != '}')            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
752              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
753    #else           /* EBCDIC coding */
754              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
755              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
756    #endif
757              }
758            }
759          }
760        else
761          *errorcodeptr = ERR37;
762        break;
763    
764        case CHAR_U:
765        /* In JavaScript, \U is an uppercase U letter. */
766        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
767        break;
768    
769        /* In a character class, \g is just a literal "g". Outside a character
770        class, \g must be followed by one of a number of specific things:
771    
772        (1) A number, either plain or braced. If positive, it is an absolute
773        backreference. If negative, it is a relative backreference. This is a Perl
774        5.10 feature.
775    
776        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
777        is part of Perl's movement towards a unified syntax for back references. As
778        this is synonymous with \k{name}, we fudge it up by pretending it really
779        was \k.
780    
781        (3) For Oniguruma compatibility we also support \g followed by a name or a
782        number either in angle brackets or in single quotes. However, these are
783        (possibly recursive) subroutine calls, _not_ backreferences. Just return
784        the -ESC_g code (cf \k). */
785    
786        case CHAR_g:
787        if (isclass) break;
788        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
789          {
790          c = -ESC_g;
791          break;
792          }
793    
794        /* Handle the Perl-compatible cases */
795    
796        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
797          {
798          const pcre_uchar *p;
799          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
800            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
801          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
802          {          {
803          c = -ESC_k;          c = -ESC_k;
804          break;          break;
# Line 514  else Line 808  else
808        }        }
809      else braced = FALSE;      else braced = FALSE;
810    
811      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
812        {        {
813        negated = TRUE;        negated = TRUE;
814        ptr++;        ptr++;
# Line 523  else Line 817  else
817    
818      c = 0;      c = 0;
819      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
820        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
821    
822      if (c < 0)      if (c < 0)   /* Integer overflow */
823        {        {
824        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
825        break;        break;
826        }        }
827    
828      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
829        {        {
830        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
831        break;        break;
832        }        }
833    
834        if (c == 0)
835          {
836          *errorcodeptr = ERR58;
837          break;
838          }
839    
840      if (negated)      if (negated)
841        {        {
842        if (c > bracount)        if (c > bracount)
# Line 562  else Line 862  else
862      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
863      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
864    
865      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:
866      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
867    
868      if (!isclass)      if (!isclass)
869        {        {
870        oldptr = ptr;        oldptr = ptr;
871        c -= '0';        c -= CHAR_0;
872        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
873          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
874        if (c < 0)        if (c < 0)    /* Integer overflow */
875          {          {
876          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
877          break;          break;
878          }          }
879        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
880          {          {
881          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 588  else Line 888  else
888      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
889      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
890    
891      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
892        {        {
893        ptr--;        ptr--;
894        c = 0;        c = 0;
# Line 601  else Line 901  else
901      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, but no more
902      than 3 octal digits. */      than 3 octal digits. */
903    
904      case '0':      case CHAR_0:
905      c -= '0';      c -= CHAR_0;
906      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
907          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
908      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
909      break;      break;
910    
# Line 612  else Line 912  else
912      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
913      treated as a data character. */      treated as a data character. */
914    
915      case 'x':      case CHAR_x:
916      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
917          {
918          /* In JavaScript, \x must be followed by two hexadecimal numbers.
919          Otherwise it is a lowercase x letter. */
920          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
921            {
922            c = 0;
923            for (i = 0; i < 2; ++i)
924              {
925              register int cc = *(++ptr);
926    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
927              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
928              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
929    #else           /* EBCDIC coding */
930              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
931              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
932    #endif
933              }
934            }
935          break;
936          }
937    
938        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
939        {        {
940        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
941        int count = 0;        int count = 0;
942    
943        c = 0;        c = 0;
944        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
945          {          {
946          register int cc = *pt++;          register int cc = *pt++;
947          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
948          count++;          count++;
949    
950  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
951          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
952          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
953  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
954          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
955          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
956  #endif  #endif
957          }          }
958    
959        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
960          {          {
961          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
962          ptr = pt;          ptr = pt;
# Line 650  else Line 972  else
972      c = 0;      c = 0;
973      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
974        {        {
975        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
976        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
977  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
978        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
979        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
980  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
981        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
982        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
983  #endif  #endif
984        }        }
985      break;      break;
986    
987      /* 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.
988      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
989        coding is ASCII-specific, but then the whole concept of \cx is
990      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
991    
992      case 'c':      case CHAR_c:
993      c = *(++ptr);      c = *(++ptr);
994      if (c == 0)      if (c == 0)
995        {        {
996        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
997        break;        break;
998        }        }
999    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1000  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1001      if (c >= 'a' && c <= 'z') c -= 32;        {
1002          *errorcodeptr = ERR68;
1003          break;
1004          }
1005        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1006      c ^= 0x40;      c ^= 0x40;
1007  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1008      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1009      c ^= 0xC0;      c ^= 0xC0;
1010  #endif  #endif
1011      break;      break;
1012    
1013      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1014      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1015      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
1016      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
1017      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1018    
1019      default:      default:
1020      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 1027  else
1027      }      }
1028    }    }
1029    
1030    /* Perl supports \N{name} for character names, as well as plain \N for "not
1031    newline". PCRE does not support \N{name}. However, it does support
1032    quantification such as \N{2,3}. */
1033    
1034    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1035         !is_counted_repeat(ptr+2))
1036      *errorcodeptr = ERR37;
1037    
1038    /* If PCRE_UCP is set, we change the values for \d etc. */
1039    
1040    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1041      c -= (ESC_DU - ESC_D);
1042    
1043    /* Set the pointer to the final character before returning. */
1044    
1045  *ptrptr = ptr;  *ptrptr = ptr;
1046  return c;  return c;
1047  }  }
# Line 726  Returns:         type value from ucp_typ Line 1068  Returns:         type value from ucp_typ
1068  */  */
1069    
1070  static int  static int
1071  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1072  {  {
1073  int c, i, bot, top;  int c, i, bot, top;
1074  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1075  char name[32];  pcre_uchar name[32];
1076    
1077  c = *(++ptr);  c = *(++ptr);
1078  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 740  if (c == 0) goto ERROR_RETURN; Line 1082  if (c == 0) goto ERROR_RETURN;
1082  /* \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
1083  negation. */  negation. */
1084    
1085  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1086    {    {
1087    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1088      {      {
1089      *negptr = TRUE;      *negptr = TRUE;
1090      ptr++;      ptr++;
# Line 751  if (c == '{') Line 1093  if (c == '{')
1093      {      {
1094      c = *(++ptr);      c = *(++ptr);
1095      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1096      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1097      name[i] = c;      name[i] = c;
1098      }      }
1099    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1100    name[i] = 0;    name[i] = 0;
1101    }    }
1102    
# Line 771  else Line 1113  else
1113  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1114    
1115  bot = 0;  bot = 0;
1116  top = _pcre_utt_size;  top = PRIV(utt_size);
1117    
1118  while (bot < top)  while (bot < top)
1119    {    {
1120    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1121    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1122    if (c == 0)    if (c == 0)
1123      {      {
1124      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1125      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1126      }      }
1127    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1128    }    }
# Line 800  return -1; Line 1142  return -1;
1142    
1143    
1144  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1145  *         Read repeat counts                     *  *         Read repeat counts                     *
1146  *************************************************/  *************************************************/
1147    
# Line 851  Returns:         pointer to '}' on succe Line 1160  Returns:         pointer to '}' on succe
1160                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1161  */  */
1162    
1163  static const uschar *  static const pcre_uchar *
1164  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)
1165  {  {
1166  int min = 0;  int min = 0;
1167  int max = -1;  int max = -1;
# Line 860  int max = -1; Line 1169  int max = -1;
1169  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1170  an integer overflow. */  an integer overflow. */
1171    
1172  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1173  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1174    {    {
1175    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1179  if (min < 0 || min > 65535)
1179  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1180  Also, max must not be less than min. */  Also, max must not be less than min. */
1181    
1182  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1183    {    {
1184    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1185      {      {
1186      max = 0;      max = 0;
1187      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1188      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1189        {        {
1190        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1209  return p;
1209    
1210    
1211  /*************************************************  /*************************************************
1212  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1213  *************************************************/  *************************************************/
1214    
1215  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1216    top-level call starts at the beginning of the pattern. All other calls must
1217    start at a parenthesis. It scans along a pattern's text looking for capturing
1218  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1219  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1220  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1221  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1222  be terminated by '>' because that is checked in the first pass.  
1223    This function was originally called only from the second pass, in which we know
1224    that if (?< or (?' or (?P< is encountered, the name will be correctly
1225    terminated because that is checked in the first pass. There is now one call to
1226    this function in the first pass, to check for a recursive back reference by
1227    name (so that we can make the whole group atomic). In this case, we need check
1228    only up to the current position in the pattern, and that is still OK because
1229    and previous occurrences will have been checked. To make this work, the test
1230    for "end of pattern" is a check against cd->end_pattern in the main loop,
1231    instead of looking for a binary zero. This means that the special first-pass
1232    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1233    processing items within the loop are OK, because afterwards the main loop will
1234    terminate.)
1235    
1236  Arguments:  Arguments:
1237    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1238    count        current count of capturing parens so far encountered    cd           compile background data
1239    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1240    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1241    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1242      utf8         TRUE if we are in UTF-8 mode
1243      count        pointer to the current capturing subpattern number (updated)
1244    
1245  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1246  */  */
1247    
1248  static int  static int
1249  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1250    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1251  {  {
1252  const uschar *thisname;  pcre_uchar *ptr = *ptrptr;
1253    int start_count = *count;
1254    int hwm_count = start_count;
1255    BOOL dup_parens = FALSE;
1256    
1257    /* If the first character is a parenthesis, check on the type of group we are
1258    dealing with. The very first call may not start with a parenthesis. */
1259    
1260  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1261    {    {
1262    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1263    
1264      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1265    
1266      /* Handle a normal, unnamed capturing parenthesis. */
1267    
1268      else if (ptr[1] != CHAR_QUESTION_MARK)
1269        {
1270        *count += 1;
1271        if (name == NULL && *count == lorn) return *count;
1272        ptr++;
1273        }
1274    
1275      /* All cases now have (? at the start. Remember when we are in a group
1276      where the parenthesis numbers are duplicated. */
1277    
1278      else if (ptr[2] == CHAR_VERTICAL_LINE)
1279        {
1280        ptr += 3;
1281        dup_parens = TRUE;
1282        }
1283    
1284      /* Handle comments; all characters are allowed until a ket is reached. */
1285    
1286      else if (ptr[2] == CHAR_NUMBER_SIGN)
1287        {
1288        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1289        goto FAIL_EXIT;
1290        }
1291    
1292      /* Handle a condition. If it is an assertion, just carry on so that it
1293      is processed as normal. If not, skip to the closing parenthesis of the
1294      condition (there can't be any nested parens). */
1295    
1296      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1297        {
1298        ptr += 2;
1299        if (ptr[1] != CHAR_QUESTION_MARK)
1300          {
1301          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1302          if (*ptr != 0) ptr++;
1303          }
1304        }
1305    
1306      /* Start with (? but not a condition. */
1307    
1308      else
1309        {
1310        ptr += 2;
1311        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1312    
1313        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1314    
1315        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1316            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1317          {
1318          int term;
1319          const pcre_uchar *thisname;
1320          *count += 1;
1321          if (name == NULL && *count == lorn) return *count;
1322          term = *ptr++;
1323          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1324          thisname = ptr;
1325          while (*ptr != term) ptr++;
1326          if (name != NULL && lorn == ptr - thisname &&
1327              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1328            return *count;
1329          term++;
1330          }
1331        }
1332      }
1333    
1334    /* Past any initial parenthesis handling, scan for parentheses or vertical
1335    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1336    first-pass call when this value is temporarily adjusted to stop at the current
1337    position. So DO NOT change this to a test for binary zero. */
1338    
1339    for (; ptr < cd->end_pattern; ptr++)
1340      {
1341    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1342    
1343    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1344      {      {
1345      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1346      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1347        {        {
1348        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1349        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1350        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1351        }        }
1352      continue;      continue;
1353      }      }
1354    
1355    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1356      are handled for real. If the first character is '^', skip it. Also, if the
1357      first few characters (either before or after ^) are \Q\E or \E we skip them
1358      too. This makes for compatibility with Perl. Note the use of STR macros to
1359      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1360    
1361    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1362      {      {
1363      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1364        for (;;)
1365          {
1366          if (ptr[1] == CHAR_BACKSLASH)
1367            {
1368            if (ptr[2] == CHAR_E)
1369              ptr+= 2;
1370            else if (STRNCMP_UC_C8(ptr + 2,
1371                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1372              ptr += 4;
1373            else
1374              break;
1375            }
1376          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1377            {
1378            negate_class = TRUE;
1379            ptr++;
1380            }
1381          else break;
1382          }
1383    
1384        /* If the next character is ']', it is a data character that must be
1385        skipped, except in JavaScript compatibility mode. */
1386    
1387        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1388            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1389          ptr++;
1390    
1391        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1392        {        {
1393        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1394        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1395          {          {
1396          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1397          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1398            {            {
1399            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1400            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1401            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1402            }            }
1403          continue;          continue;
1404          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1408  for (; *ptr != 0; ptr++)
1408    
1409    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1410    
1411    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1412      {      {
1413      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1414      if (*ptr == 0) return -1;      while (*ptr != 0)
1415          {
1416          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1417          ptr++;
1418    #ifdef SUPPORT_UTF8
1419          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1420    #endif
1421          }
1422        if (*ptr == 0) goto FAIL_EXIT;
1423      continue;      continue;
1424      }      }
1425    
1426    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1427    
1428    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1429      {      {
1430      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1431      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1432      continue;      if (*ptr == 0) goto FAIL_EXIT;
1433      }      }
1434    
1435    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1436    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1437        if (dup_parens && *count < hwm_count) *count = hwm_count;
1438        goto FAIL_EXIT;
1439        }
1440    
1441      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1442        {
1443        if (*count > hwm_count) hwm_count = *count;
1444        *count = start_count;
1445        }
1446      }
1447    
1448    FAIL_EXIT:
1449    *ptrptr = ptr;
1450    return -1;
1451    }
1452    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1453    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1454    
   count++;  
1455    
1456    if (name == NULL && count == lorn) return count;  /*************************************************
1457    term = *ptr++;  *       Find forward referenced subpattern       *
1458    if (term == '<') term = '>';  *************************************************/
1459    thisname = ptr;  
1460    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1461    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1462        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1463      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1464    references to subpatterns. We used to be able to start this scan from the
1465    current compiling point, using the current count value from cd->bracount, and
1466    do it all in a single loop, but the addition of the possibility of duplicate
1467    subpattern numbers means that we have to scan from the very start, in order to
1468    take account of such duplicates, and to use a recursive function to keep track
1469    of the different types of group.
1470    
1471    Arguments:
1472      cd           compile background data
1473      name         name to seek, or NULL if seeking a numbered subpattern
1474      lorn         name length, or subpattern number if name is NULL
1475      xmode        TRUE if we are in /x mode
1476      utf8         TRUE if we are in UTF-8 mode
1477    
1478    Returns:       the number of the found subpattern, or -1 if not found
1479    */
1480    
1481    static int
1482    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1483      BOOL utf8)
1484    {
1485    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1486    int count = 0;
1487    int rc;
1488    
1489    /* If the pattern does not start with an opening parenthesis, the first call
1490    to find_parens_sub() will scan right to the end (if necessary). However, if it
1491    does start with a parenthesis, find_parens_sub() will return when it hits the
1492    matching closing parens. That is why we have to have a loop. */
1493    
1494    for (;;)
1495      {
1496      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1497      if (rc > 0 || *ptr++ == 0) break;
1498    }    }
1499    
1500  return -1;  return rc;
1501  }  }
1502    
1503    
1504    
1505    
1506  /*************************************************  /*************************************************
1507  *      Find first significant op code            *  *      Find first significant op code            *
1508  *************************************************/  *************************************************/
1509    
1510  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1511  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1512  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1513  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1514  assertions, and also the \b assertion; for others it does not.  does not.
1515    
1516  Arguments:  Arguments:
1517    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1518    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1519    
1520  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1521  */  */
1522    
1523  static const uschar*  static const pcre_uchar*
1524  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1525  {  {
1526  for (;;)  for (;;)
1527    {    {
1528    switch ((int)*code)    switch ((int)*code)
1529      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1530      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1531      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1532      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1533      if (!skipassert) return code;      if (!skipassert) return code;
1534      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1535      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1536      break;      break;
1537    
1538      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1060  for (;;) Line 1542  for (;;)
1542    
1543      case OP_CALLOUT:      case OP_CALLOUT:
1544      case OP_CREF:      case OP_CREF:
1545        case OP_NCREF:
1546      case OP_RREF:      case OP_RREF:
1547        case OP_NRREF:
1548      case OP_DEF:      case OP_DEF:
1549      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1550      break;      break;
1551    
1552      default:      default:
# Line 1076  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1564  *************************************************/  *************************************************/
1565    
1566  /* 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,
1567  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.
1568  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
1569    temporarily terminated with OP_END when this function is called.
1570    
1571    This function is called when a backward assertion is encountered, so that if it
1572    fails, the error message can point to the correct place in the pattern.
1573    However, we cannot do this when the assertion contains subroutine calls,
1574    because they can be forward references. We solve this by remembering this case
1575    and doing the check at the end; a flag specifies which mode we are running in.
1576    
1577  Arguments:  Arguments:
1578    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1579    options  the compiling options    utf8     TRUE in UTF-8 mode
1580      atend    TRUE if called when the pattern is complete
1581  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1582               or -2 if \C was encountered  
1583    Returns:   the fixed length,
1584                 or -1 if there is no fixed length,
1585                 or -2 if \C was encountered (in UTF-8 mode only)
1586                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1587                 or -4 if an unknown opcode was encountered (internal error)
1588  */  */
1589    
1590  static int  static int
1591  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf8, BOOL atend, compile_data *cd)
1592  {  {
1593  int length = -1;  int length = -1;
1594    
1595  register int branchlength = 0;  register int branchlength = 0;
1596  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1597    
1598  /* 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
1599  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1105  branch, check the length against that of Line 1601  branch, check the length against that of
1601  for (;;)  for (;;)
1602    {    {
1603    int d;    int d;
1604      pcre_uchar *ce, *cs;
1605    register int op = *cc;    register int op = *cc;
1606    switch (op)    switch (op)
1607      {      {
1608        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1609        OP_BRA (normal non-capturing bracket) because the other variants of these
1610        opcodes are all concerned with unlimited repeated groups, which of course
1611        are not of fixed length. */
1612    
1613      case OP_CBRA:      case OP_CBRA:
1614      case OP_BRA:      case OP_BRA:
1615      case OP_ONCE:      case OP_ONCE:
1616        case OP_ONCE_NC:
1617      case OP_COND:      case OP_COND:
1618      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1619      if (d < 0) return d;      if (d < 0) return d;
1620      branchlength += d;      branchlength += d;
1621      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1622      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1623      break;      break;
1624    
1625      /* 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.
1626      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
1627      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
1628        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1629        because they all imply an unlimited repeat. */
1630    
1631      case OP_ALT:      case OP_ALT:
1632      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1633      case OP_END:      case OP_END:
1634        case OP_ACCEPT:
1635        case OP_ASSERT_ACCEPT:
1636      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1637        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1638      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1135  for (;;) Line 1640  for (;;)
1640      branchlength = 0;      branchlength = 0;
1641      break;      break;
1642    
1643        /* A true recursion implies not fixed length, but a subroutine call may
1644        be OK. If the subroutine is a forward reference, we can't deal with
1645        it until the end of the pattern, so return -3. */
1646    
1647        case OP_RECURSE:
1648        if (!atend) return -3;
1649        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1650        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1651        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1652        d = find_fixedlength(cs + 2, utf8, atend, cd);
1653        if (d < 0) return d;
1654        branchlength += d;
1655        cc += 1 + LINK_SIZE;
1656        break;
1657    
1658      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1659    
1660      case OP_ASSERT:      case OP_ASSERT:
# Line 1146  for (;;) Line 1666  for (;;)
1666    
1667      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1668    
1669      case OP_REVERSE:      case OP_MARK:
1670        case OP_PRUNE_ARG:
1671        case OP_SKIP_ARG:
1672        case OP_THEN_ARG:
1673        cc += cc[1] + PRIV(OP_lengths)[*cc];
1674        break;
1675    
1676        case OP_CALLOUT:
1677        case OP_CIRC:
1678        case OP_CIRCM:
1679        case OP_CLOSE:
1680        case OP_COMMIT:
1681      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1682      case OP_DEF:      case OP_DEF:
1683      case OP_OPT:      case OP_DOLL:
1684      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1685      case OP_EOD:      case OP_EOD:
1686      case OP_EODN:      case OP_EODN:
1687      case OP_CIRC:      case OP_FAIL:
1688      case OP_DOLL:      case OP_NCREF:
1689        case OP_NRREF:
1690      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1691        case OP_PRUNE:
1692        case OP_REVERSE:
1693        case OP_RREF:
1694        case OP_SET_SOM:
1695        case OP_SKIP:
1696        case OP_SOD:
1697        case OP_SOM:
1698        case OP_THEN:
1699      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1700      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1701      break;      break;
1702    
1703      /* Handle literal characters */      /* Handle literal characters */
1704    
1705      case OP_CHAR:      case OP_CHAR:
1706      case OP_CHARNC:      case OP_CHARI:
1707      case OP_NOT:      case OP_NOT:
1708        case OP_NOTI:
1709      branchlength++;      branchlength++;
1710      cc += 2;      cc += 2;
1711  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1712      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1713  #endif  #endif
1714      break;      break;
1715    
# Line 1182  for (;;) Line 1717  for (;;)
1717      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1718    
1719      case OP_EXACT:      case OP_EXACT:
1720        case OP_EXACTI:
1721        case OP_NOTEXACT:
1722        case OP_NOTEXACTI:
1723      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1724      cc += 4;      cc += 4;
1725  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1726      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1727  #endif  #endif
1728      break;      break;
1729    
1730      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1731      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1732      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1733      cc += 4;      cc += 4;
1734      break;      break;
1735    
# Line 1205  for (;;) Line 1740  for (;;)
1740      cc += 2;      cc += 2;
1741      /* Fall through */      /* Fall through */
1742    
1743        case OP_HSPACE:
1744        case OP_VSPACE:
1745        case OP_NOT_HSPACE:
1746        case OP_NOT_VSPACE:
1747      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1748      case OP_DIGIT:      case OP_DIGIT:
1749      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1212  for (;;) Line 1751  for (;;)
1751      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1752      case OP_WORDCHAR:      case OP_WORDCHAR:
1753      case OP_ANY:      case OP_ANY:
1754        case OP_ALLANY:
1755      branchlength++;      branchlength++;
1756      cc++;      cc++;
1757      break;      break;
1758    
1759      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1760        otherwise \C is coded as OP_ALLANY. */
1761    
1762      case OP_ANYBYTE:      case OP_ANYBYTE:
1763      return -2;      return -2;
# Line 1235  for (;;) Line 1776  for (;;)
1776    
1777      switch (*cc)      switch (*cc)
1778        {        {
1779          case OP_CRPLUS:
1780          case OP_CRMINPLUS:
1781        case OP_CRSTAR:        case OP_CRSTAR:
1782        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1783        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1255  for (;;) Line 1798  for (;;)
1798    
1799      /* Anything else is variable length */      /* Anything else is variable length */
1800    
1801      default:      case OP_ANYNL:
1802        case OP_BRAMINZERO:
1803        case OP_BRAPOS:
1804        case OP_BRAPOSZERO:
1805        case OP_BRAZERO:
1806        case OP_CBRAPOS:
1807        case OP_EXTUNI:
1808        case OP_KETRMAX:
1809        case OP_KETRMIN:
1810        case OP_KETRPOS:
1811        case OP_MINPLUS:
1812        case OP_MINPLUSI:
1813        case OP_MINQUERY:
1814        case OP_MINQUERYI:
1815        case OP_MINSTAR:
1816        case OP_MINSTARI:
1817        case OP_MINUPTO:
1818        case OP_MINUPTOI:
1819        case OP_NOTMINPLUS:
1820        case OP_NOTMINPLUSI:
1821        case OP_NOTMINQUERY:
1822        case OP_NOTMINQUERYI:
1823        case OP_NOTMINSTAR:
1824        case OP_NOTMINSTARI:
1825        case OP_NOTMINUPTO:
1826        case OP_NOTMINUPTOI:
1827        case OP_NOTPLUS:
1828        case OP_NOTPLUSI:
1829        case OP_NOTPOSPLUS:
1830        case OP_NOTPOSPLUSI:
1831        case OP_NOTPOSQUERY:
1832        case OP_NOTPOSQUERYI:
1833        case OP_NOTPOSSTAR:
1834        case OP_NOTPOSSTARI:
1835        case OP_NOTPOSUPTO:
1836        case OP_NOTPOSUPTOI:
1837        case OP_NOTQUERY:
1838        case OP_NOTQUERYI:
1839        case OP_NOTSTAR:
1840        case OP_NOTSTARI:
1841        case OP_NOTUPTO:
1842        case OP_NOTUPTOI:
1843        case OP_PLUS:
1844        case OP_PLUSI:
1845        case OP_POSPLUS:
1846        case OP_POSPLUSI:
1847        case OP_POSQUERY:
1848        case OP_POSQUERYI:
1849        case OP_POSSTAR:
1850        case OP_POSSTARI:
1851        case OP_POSUPTO:
1852        case OP_POSUPTOI:
1853        case OP_QUERY:
1854        case OP_QUERYI:
1855        case OP_REF:
1856        case OP_REFI:
1857        case OP_SBRA:
1858        case OP_SBRAPOS:
1859        case OP_SCBRA:
1860        case OP_SCBRAPOS:
1861        case OP_SCOND:
1862        case OP_SKIPZERO:
1863        case OP_STAR:
1864        case OP_STARI:
1865        case OP_TYPEMINPLUS:
1866        case OP_TYPEMINQUERY:
1867        case OP_TYPEMINSTAR:
1868        case OP_TYPEMINUPTO:
1869        case OP_TYPEPLUS:
1870        case OP_TYPEPOSPLUS:
1871        case OP_TYPEPOSQUERY:
1872        case OP_TYPEPOSSTAR:
1873        case OP_TYPEPOSUPTO:
1874        case OP_TYPEQUERY:
1875        case OP_TYPESTAR:
1876        case OP_TYPEUPTO:
1877        case OP_UPTO:
1878        case OP_UPTOI:
1879      return -1;      return -1;
1880    
1881        /* Catch unrecognized opcodes so that when new ones are added they
1882        are not forgotten, as has happened in the past. */
1883    
1884        default:
1885        return -4;
1886      }      }
1887    }    }
1888  /* Control never gets here */  /* Control never gets here */
# Line 1266  for (;;) Line 1892  for (;;)
1892    
1893    
1894  /*************************************************  /*************************************************
1895  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1896  *************************************************/  *************************************************/
1897    
1898  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1899  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1900    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1901    so that it can be called from pcre_study() when finding the minimum matching
1902    length.
1903    
1904  Arguments:  Arguments:
1905    code        points to start of expression    code        points to start of expression
1906    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1907    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1908    
1909  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
1910  */  */
1911    
1912  static const uschar *  const pcre_uchar *
1913  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf8, int number)
1914  {  {
1915  for (;;)  for (;;)
1916    {    {
1917    register int c = *code;    register int c = *code;
1918    
1919    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1920    
1921    /* 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 1294  for (;;) Line 1924  for (;;)
1924    
1925    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1926    
1927      /* Handle recursion */
1928    
1929      else if (c == OP_REVERSE)
1930        {
1931        if (number < 0) return (pcre_uchar *)code;
1932        code += PRIV(OP_lengths)[c];
1933        }
1934    
1935    /* Handle capturing bracket */    /* Handle capturing bracket */
1936    
1937    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1938               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1939      {      {
1940      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1941      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
1942      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1943      }      }
1944    
1945    /* 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
1946    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
1947    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1948      must add in its length. */
1949    
1950    else    else
1951      {      {
# Line 1317  for (;;) Line 1957  for (;;)
1957        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1958        case OP_TYPEQUERY:        case OP_TYPEQUERY:
1959        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
1960        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
1961        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
1962        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
1963        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1964        break;        break;
1965        }  
1966          case OP_TYPEUPTO:
1967          case OP_TYPEMINUPTO:
1968          case OP_TYPEEXACT:
1969          case OP_TYPEPOSUPTO:
1970          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1971          break;
1972    
1973          case OP_MARK:
1974          case OP_PRUNE_ARG:
1975          case OP_SKIP_ARG:
1976          code += code[1];
1977          break;
1978    
1979          case OP_THEN_ARG:
1980          code += code[1];
1981          break;
1982          }
1983    
1984      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
1985    
1986      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1987    
1988    /* 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
1989    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
1990    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
1991    
1992  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1993      if (utf8) switch(c)      if (utf8) switch(c)
1994        {        {
1995        case OP_CHAR:        case OP_CHAR:
1996        case OP_CHARNC:        case OP_CHARI:
1997        case OP_EXACT:        case OP_EXACT:
1998          case OP_EXACTI:
1999        case OP_UPTO:        case OP_UPTO:
2000          case OP_UPTOI:
2001        case OP_MINUPTO:        case OP_MINUPTO:
2002          case OP_MINUPTOI:
2003        case OP_POSUPTO:        case OP_POSUPTO:
2004          case OP_POSUPTOI:
2005        case OP_STAR:        case OP_STAR:
2006          case OP_STARI:
2007        case OP_MINSTAR:        case OP_MINSTAR:
2008          case OP_MINSTARI:
2009        case OP_POSSTAR:        case OP_POSSTAR:
2010          case OP_POSSTARI:
2011        case OP_PLUS:        case OP_PLUS:
2012          case OP_PLUSI:
2013        case OP_MINPLUS:        case OP_MINPLUS:
2014          case OP_MINPLUSI:
2015        case OP_POSPLUS:        case OP_POSPLUS:
2016          case OP_POSPLUSI:
2017        case OP_QUERY:        case OP_QUERY:
2018          case OP_QUERYI:
2019        case OP_MINQUERY:        case OP_MINQUERY:
2020          case OP_MINQUERYI:
2021        case OP_POSQUERY:        case OP_POSQUERY:
2022        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2023          if (code[-1] >= 0xc0) code += PRIV(utf8_table4)[code[-1] & 0x3f];
2024        break;        break;
2025        }        }
2026    #else
2027        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2028  #endif  #endif
2029      }      }
2030    }    }
# Line 1378  Arguments: Line 2046  Arguments:
2046  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
2047  */  */
2048    
2049  static const uschar *  static const pcre_uchar *
2050  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf8)
2051  {  {
2052  for (;;)  for (;;)
2053    {    {
2054    register int c = *code;    register int c = *code;
2055    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2056    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
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
2059    map. This includes negated single high-valued characters. The length in    map. This includes negated single high-valued characters. The length in
2060    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
2061    
2062    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2063    
2064    /* 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
2065    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
2066    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2067      must add in its length. */
2068    
2069    else    else
2070      {      {
# Line 1407  for (;;) Line 2076  for (;;)
2076        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
2077        case OP_TYPEQUERY:        case OP_TYPEQUERY:
2078        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
2079        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
2080        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
2081        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
2082        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2083        break;        break;
2084        }  
2085          case OP_TYPEPOSUPTO:
2086          case OP_TYPEUPTO:
2087          case OP_TYPEMINUPTO:
2088          case OP_TYPEEXACT:
2089          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2090          break;
2091    
2092          case OP_MARK:
2093          case OP_PRUNE_ARG:
2094          case OP_SKIP_ARG:
2095          code += code[1];
2096          break;
2097    
2098          case OP_THEN_ARG:
2099          code += code[1];
2100          break;
2101          }
2102    
2103      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2104    
2105      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2106    
2107      /* 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
2108      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
2109      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2110    
2111  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2112      if (utf8) switch(c)      if (utf8) switch(c)
2113        {        {
2114        case OP_CHAR:        case OP_CHAR:
2115        case OP_CHARNC:        case OP_CHARI:
2116        case OP_EXACT:        case OP_EXACT:
2117          case OP_EXACTI:
2118        case OP_UPTO:        case OP_UPTO:
2119          case OP_UPTOI:
2120        case OP_MINUPTO:        case OP_MINUPTO:
2121          case OP_MINUPTOI:
2122        case OP_POSUPTO:        case OP_POSUPTO:
2123          case OP_POSUPTOI:
2124        case OP_STAR:        case OP_STAR:
2125          case OP_STARI:
2126        case OP_MINSTAR:        case OP_MINSTAR:
2127          case OP_MINSTARI:
2128        case OP_POSSTAR:        case OP_POSSTAR:
2129          case OP_POSSTARI:
2130        case OP_PLUS:        case OP_PLUS:
2131          case OP_PLUSI:
2132        case OP_MINPLUS:        case OP_MINPLUS:
2133          case OP_MINPLUSI:
2134        case OP_POSPLUS:        case OP_POSPLUS:
2135          case OP_POSPLUSI:
2136        case OP_QUERY:        case OP_QUERY:
2137          case OP_QUERYI:
2138        case OP_MINQUERY:        case OP_MINQUERY:
2139          case OP_MINQUERYI:
2140        case OP_POSQUERY:        case OP_POSQUERY:
2141        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2142          if (code[-1] >= 0xc0) code += PRIV(utf8_table4)[code[-1] & 0x3f];
2143        break;        break;
2144        }        }
2145    #else
2146        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2147  #endif  #endif
2148      }      }
2149    }    }
# Line 1462  for (;;) Line 2159  for (;;)
2159  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2160  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
2161  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2162  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2163  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2164    bracket whose current branch will already have been scanned.
2165    
2166  Arguments:  Arguments:
2167    code        points to start of search    code        points to start of search
2168    endcode     points to where to stop    endcode     points to where to stop
2169    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2170      cd          contains pointers to tables etc.
2171    
2172  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2173  */  */
2174    
2175  static BOOL  static BOOL
2176  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2177      BOOL utf8, compile_data *cd)
2178  {  {
2179  register int c;  register int c;
2180  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2181       code < endcode;       code < endcode;
2182       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2183    {    {
2184    const uschar *ccode;    const pcre_uchar *ccode;
2185    
2186    c = *code;    c = *code;
2187    
2188    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2189      first_significant_code() with a TRUE final argument. */
2190    
2191    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2192      {      {
     code += _pcre_OP_lengths[c];  
2193      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2194      c = *code;      c = *code;
2195      continue;      continue;
2196      }      }
2197    
2198    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2199      implies a backward reference subroutine call, we can scan it. If it's a
2200      forward reference subroutine call, we can't. To detect forward reference
2201      we have to scan up the list that is kept in the workspace. This function is
2202      called only when doing the real compile, not during the pre-compile that
2203      measures the size of the compiled pattern. */
2204    
2205    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2206      {      {
2207        const pcre_uchar *scode;
2208      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2209    
2210      /* Scan a closed bracket */      /* Test for forward reference */
2211    
2212        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2213          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2214    
2215        /* Not a forward reference, test for completed backward reference */
2216    
2217      empty_branch = FALSE;      empty_branch = FALSE;
2218        scode = cd->start_code + GET(code, 1);
2219        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2220    
2221        /* Completed backwards reference */
2222    
2223      do      do
2224        {        {
2225        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2226            {
2227          empty_branch = TRUE;          empty_branch = TRUE;
2228            break;
2229            }
2230          scode += GET(scode, 1);
2231          }
2232        while (*scode == OP_ALT);
2233    
2234        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2235        continue;
2236        }
2237    
2238      /* Groups with zero repeats can of course be empty; skip them. */
2239    
2240      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2241          c == OP_BRAPOSZERO)
2242        {
2243        code += PRIV(OP_lengths)[c];
2244        do code += GET(code, 1); while (*code == OP_ALT);
2245        c = *code;
2246        continue;
2247        }
2248    
2249      /* A nested group that is already marked as "could be empty" can just be
2250      skipped. */
2251    
2252      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2253          c == OP_SCBRA || c == OP_SCBRAPOS)
2254        {
2255        do code += GET(code, 1); while (*code == OP_ALT);
2256        c = *code;
2257        continue;
2258        }
2259    
2260      /* For other groups, scan the branches. */
2261    
2262      if (c == OP_BRA  || c == OP_BRAPOS ||
2263          c == OP_CBRA || c == OP_CBRAPOS ||
2264          c == OP_ONCE || c == OP_ONCE_NC ||
2265          c == OP_COND)
2266        {
2267        BOOL empty_branch;
2268        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2269    
2270        /* If a conditional group has only one branch, there is a second, implied,
2271        empty branch, so just skip over the conditional, because it could be empty.
2272        Otherwise, scan the individual branches of the group. */
2273    
2274        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2275        code += GET(code, 1);        code += GET(code, 1);
2276        else
2277          {
2278          empty_branch = FALSE;
2279          do
2280            {
2281            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2282              empty_branch = TRUE;
2283            code += GET(code, 1);
2284            }
2285          while (*code == OP_ALT);
2286          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2287        }        }
2288      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2289      c = *code;      c = *code;
2290      continue;      continue;
2291      }      }
# Line 1523  for (code = first_significant_code(code Line 2296  for (code = first_significant_code(code
2296      {      {
2297      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2298      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2299      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2300      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"
2301      here. */      here. */
2302    
2303  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1573  for (code = first_significant_code(code Line 2346  for (code = first_significant_code(code
2346      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2347      case OP_WORDCHAR:      case OP_WORDCHAR:
2348      case OP_ANY:      case OP_ANY:
2349        case OP_ALLANY:
2350      case OP_ANYBYTE:      case OP_ANYBYTE:
2351      case OP_CHAR:      case OP_CHAR:
2352      case OP_CHARNC:      case OP_CHARI:
2353      case OP_NOT:      case OP_NOT:
2354        case OP_NOTI:
2355      case OP_PLUS:      case OP_PLUS:
2356      case OP_MINPLUS:      case OP_MINPLUS:
2357      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1591  for (code = first_significant_code(code Line 2366  for (code = first_significant_code(code
2366      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2367      return FALSE;      return FALSE;
2368    
2369        /* These are going to continue, as they may be empty, but we have to
2370        fudge the length for the \p and \P cases. */
2371    
2372        case OP_TYPESTAR:
2373        case OP_TYPEMINSTAR:
2374        case OP_TYPEPOSSTAR:
2375        case OP_TYPEQUERY:
2376        case OP_TYPEMINQUERY:
2377        case OP_TYPEPOSQUERY:
2378        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2379        break;
2380    
2381        /* Same for these */
2382    
2383        case OP_TYPEUPTO:
2384        case OP_TYPEMINUPTO:
2385        case OP_TYPEPOSUPTO:
2386        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2387        break;
2388    
2389      /* End of branch */      /* End of branch */
2390    
2391      case OP_KET:      case OP_KET:
2392      case OP_KETRMAX:      case OP_KETRMAX:
2393      case OP_KETRMIN:      case OP_KETRMIN:
2394        case OP_KETRPOS:
2395      case OP_ALT:      case OP_ALT:
2396      return TRUE;      return TRUE;
2397    
# Line 1604  for (code = first_significant_code(code Line 2400  for (code = first_significant_code(code
2400    
2401  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2402      case OP_STAR:      case OP_STAR:
2403        case OP_STARI:
2404      case OP_MINSTAR:      case OP_MINSTAR:
2405        case OP_MINSTARI:
2406      case OP_POSSTAR:      case OP_POSSTAR:
2407        case OP_POSSTARI:
2408      case OP_QUERY:      case OP_QUERY:
2409        case OP_QUERYI:
2410      case OP_MINQUERY:      case OP_MINQUERY:
2411        case OP_MINQUERYI:
2412      case OP_POSQUERY:      case OP_POSQUERY:
2413        case OP_POSQUERYI:
2414        if (utf8 && code[1] >= 0xc0) code += PRIV(utf8_table4)[code[1] & 0x3f];
2415        break;
2416    
2417      case OP_UPTO:      case OP_UPTO:
2418        case OP_UPTOI:
2419      case OP_MINUPTO:      case OP_MINUPTO:
2420        case OP_MINUPTOI:
2421      case OP_POSUPTO:      case OP_POSUPTO:
2422      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2423        if (utf8 && code[3] >= 0xc0) code += PRIV(utf8_table4)[code[3] & 0x3f];
2424      break;      break;
2425  #endif  #endif
2426    
2427        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2428        string. */
2429    
2430        case OP_MARK:
2431        case OP_PRUNE_ARG:
2432        case OP_SKIP_ARG:
2433        code += code[1];
2434        break;
2435    
2436        case OP_THEN_ARG:
2437        code += code[1];
2438        break;
2439    
2440        /* None of the remaining opcodes are required to match a character. */
2441    
2442        default:
2443        break;
2444      }      }
2445    }    }
2446    
# Line 1631  return TRUE; Line 2457  return TRUE;
2457  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
2458  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,
2459  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.
2460    This function is called only during the real compile, not during the
2461    pre-compile.
2462    
2463  Arguments:  Arguments:
2464    code        points to start of the recursion    code        points to start of the recursion
2465    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2466    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2467    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2468      cd          pointers to tables etc
2469    
2470  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2471  */  */
2472    
2473  static BOOL  static BOOL
2474  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2475    BOOL utf8)    branch_chain *bcptr, BOOL utf8, compile_data *cd)
2476  {  {
2477  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2478    {    {
2479    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2480        return FALSE;
2481    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2482    }    }
2483  return TRUE;  return TRUE;
# Line 1660  return TRUE; Line 2490  return TRUE;
2490  *************************************************/  *************************************************/
2491    
2492  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2493  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2494  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2495  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2496    
2497    Originally, this function only recognized a sequence of letters between the
2498    terminators, but it seems that Perl recognizes any sequence of characters,
2499    though of course unknown POSIX names are subsequently rejected. Perl gives an
2500    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2501    didn't consider this to be a POSIX class. Likewise for [:1234:].
2502    
2503    The problem in trying to be exactly like Perl is in the handling of escapes. We
2504    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2505    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2506    below handles the special case of \], but does not try to do any other escape
2507    processing. This makes it different from Perl for cases such as [:l\ower:]
2508    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2509    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2510    I think.
2511    
2512    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2513    It seems that the appearance of a nested POSIX class supersedes an apparent
2514    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2515    a digit.
2516    
2517    In Perl, unescaped square brackets may also appear as part of class names. For
2518    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2519    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2520    seem right at all. PCRE does not allow closing square brackets in POSIX class
2521    names.
2522    
2523  Argument:  Arguments:
2524    ptr      pointer to the initial [    ptr      pointer to the initial [
2525    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2526    
2527  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2528  */  */
2529    
2530  static BOOL  static BOOL
2531  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2532  {  {
2533  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2534  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2535  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2536    {    {
2537    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2538    return TRUE;      ptr++;
2539      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2540      else
2541        {
2542        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2543          {
2544          *endptr = ptr;
2545          return TRUE;
2546          }
2547        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2548             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2549              ptr[1] == CHAR_EQUALS_SIGN) &&
2550            check_posix_syntax(ptr, endptr))
2551          return FALSE;
2552        }
2553    }    }
2554  return FALSE;  return FALSE;
2555  }  }
# Line 1705  Returns:     a value representing the na Line 2572  Returns:     a value representing the na
2572  */  */
2573    
2574  static int  static int
2575  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2576  {  {
2577    const char *pn = posix_names;
2578  register int yield = 0;  register int yield = 0;
2579  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2580    {    {
2581    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2582      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2583      pn += posix_name_lengths[yield] + 1;
2584    yield++;    yield++;
2585    }    }
2586  return -1;  return -1;
# Line 1726  return -1; Line 2595  return -1;
2595  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2596  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2597  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2598  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2599  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2600  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2601  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2602  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2603    OP_END.
2604    
2605  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2606  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1749  Returns:     nothing Line 2619  Returns:     nothing
2619  */  */
2620    
2621  static void  static void
2622  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf8, compile_data *cd,
2623    uschar *save_hwm)    pcre_uchar *save_hwm)
2624  {  {
2625  uschar *ptr = group;  pcre_uchar *ptr = group;
2626  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2627    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf8)) != NULL)
2628    {    {
2629    int offset;    int offset;
2630    uschar *hc;    pcre_uchar *hc;
2631    
2632    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2633    reference. */    reference. */
# Line 1801  Arguments: Line 2672  Arguments:
2672  Returns:         new code pointer  Returns:         new code pointer
2673  */  */
2674    
2675  static uschar *  static pcre_uchar *
2676  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2677  {  {
2678  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2679  *code++ = 255;  *code++ = 255;
2680  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2681  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2682  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2683  }  }
2684    
2685    
# Line 1830  Returns:             nothing Line 2701  Returns:             nothing
2701  */  */
2702    
2703  static void  static void
2704  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2705  {  {
2706  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2707  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2708  }  }
2709    
# Line 1864  get_othercase_range(unsigned int *cptr, Line 2735  get_othercase_range(unsigned int *cptr,
2735  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2736    
2737  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2738    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2739    
2740  if (c > d) return FALSE;  if (c > d) return FALSE;
2741    
# Line 1873  next = othercase + 1; Line 2744  next = othercase + 1;
2744    
2745  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2746    {    {
2747    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2748    next++;    next++;
2749    }    }
2750    
# Line 1882  for (++c; c <= d; c++) Line 2753  for (++c; c <= d; c++)
2753    
2754  return TRUE;  return TRUE;
2755  }  }
2756    
2757    
2758    
2759    /*************************************************
2760    *        Check a character and a property        *
2761    *************************************************/
2762    
2763    /* This function is called by check_auto_possessive() when a property item
2764    is adjacent to a fixed character.
2765    
2766    Arguments:
2767      c            the character
2768      ptype        the property type
2769      pdata        the data for the type
2770      negated      TRUE if it's a negated property (\P or \p{^)
2771    
2772    Returns:       TRUE if auto-possessifying is OK
2773    */
2774    
2775    static BOOL
2776    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2777    {
2778    const ucd_record *prop = GET_UCD(c);
2779    switch(ptype)
2780      {
2781      case PT_LAMP:
2782      return (prop->chartype == ucp_Lu ||
2783              prop->chartype == ucp_Ll ||
2784              prop->chartype == ucp_Lt) == negated;
2785    
2786      case PT_GC:
2787      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2788    
2789      case PT_PC:
2790      return (pdata == prop->chartype) == negated;
2791    
2792      case PT_SC:
2793      return (pdata == prop->script) == negated;
2794    
2795      /* These are specials */
2796    
2797      case PT_ALNUM:
2798      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2799              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2800    
2801      case PT_SPACE:    /* Perl space */
2802      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2803              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2804              == negated;
2805    
2806      case PT_PXSPACE:  /* POSIX space */
2807      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2808              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2809              c == CHAR_FF || c == CHAR_CR)
2810              == negated;
2811    
2812      case PT_WORD:
2813      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2814              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2815              c == CHAR_UNDERSCORE) == negated;
2816      }
2817    return FALSE;
2818    }
2819  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2820    
2821    
# Line 1895  whether the next thing could possibly ma Line 2829  whether the next thing could possibly ma
2829  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2830    
2831  Arguments:  Arguments:
2832    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2833    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2834    ptr           next character in pattern    ptr           next character in pattern
2835    options       options bits    options       options bits
2836    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1907  Returns:        TRUE if possessifying is Line 2839  Returns:        TRUE if possessifying is
2839  */  */
2840    
2841  static BOOL  static BOOL
2842  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf8,
2843    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2844  {  {
2845  int next;  int c, next;
2846    int op_code = *previous++;
2847    
2848  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2849    
# Line 1919  if ((options & PCRE_EXTENDED) != 0) Line 2852  if ((options & PCRE_EXTENDED) != 0)
2852    for (;;)    for (;;)
2853      {      {
2854      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2855      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2856        {        {
2857        while (*(++ptr) != 0)        ptr++;
2858          while (*ptr != 0)
2859            {
2860          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2861            ptr++;
2862    #ifdef SUPPORT_UTF8
2863            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2864    #endif
2865            }
2866        }        }
2867      else break;      else break;
2868      }      }
# Line 1931  if ((options & PCRE_EXTENDED) != 0) Line 2871  if ((options & PCRE_EXTENDED) != 0)
2871  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2872  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2873    
2874  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2875    {    {
2876    int temperrorcode = 0;    int temperrorcode = 0;
2877    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1956  if ((options & PCRE_EXTENDED) != 0) Line 2896  if ((options & PCRE_EXTENDED) != 0)
2896    for (;;)    for (;;)
2897      {      {
2898      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2899      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2900        {        {
2901        while (*(++ptr) != 0)        ptr++;
2902          while (*ptr != 0)
2903            {
2904          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2905            ptr++;
2906    #ifdef SUPPORT_UTF8
2907            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2908    #endif
2909            }
2910        }        }
2911      else break;      else break;
2912      }      }
# Line 1967  if ((options & PCRE_EXTENDED) != 0) Line 2914  if ((options & PCRE_EXTENDED) != 0)
2914    
2915  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2916    
2917  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2918    return FALSE;    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2919        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2920    
2921  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2922    the next item is a character. */
2923    
2924  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2925    {    {
2926    case OP_CHAR:    case OP_CHAR:
2927  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2928    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2929    #else
2930      c = *previous;
2931  #endif  #endif
2932    return item != next;    return c != next;
2933    
2934    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2935    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2936    high-valued characters. */    high-valued characters. */
2937    
2938    case OP_CHARNC:    case OP_CHARI:
2939  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2940    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2941    #else
2942      c = *previous;
2943  #endif  #endif
2944    if (item == next) return FALSE;    if (c == next) return FALSE;
2945  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2946    if (utf8)    if (utf8)
2947      {      {
2948      unsigned int othercase;      unsigned int othercase;
2949      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2950  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2951      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2952  #else  #else
2953      othercase = NOTACHAR;      othercase = NOTACHAR;
2954  #endif  #endif
2955      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2956      }      }
2957    else    else
2958  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2959    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2960    
2961    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2962      opcodes are not used for multi-byte characters, because they are coded using
2963      an XCLASS instead. */
2964    
2965    case OP_NOT:    case OP_NOT:
2966    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2967    if (item == next) return TRUE;  
2968    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2969      if ((c = *previous) == next) return TRUE;
2970  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2971    if (utf8)    if (utf8)
2972      {      {
2973      unsigned int othercase;      unsigned int othercase;
2974      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2975  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2976      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2977  #else  #else
2978      othercase = NOTACHAR;      othercase = NOTACHAR;
2979  #endif  #endif
2980      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2981      }      }
2982    else    else
2983  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2984    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2985    
2986      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2987      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2988    
2989    case OP_DIGIT:    case OP_DIGIT:
2990    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2074  if (next >= 0) switch(op_code) Line 3027  if (next >= 0) switch(op_code)
3027      case 0x202f:      case 0x202f:
3028      case 0x205f:      case 0x205f:
3029      case 0x3000:      case 0x3000:
3030      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3031      default:      default:
3032      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3033      }      }
3034    
3035      case OP_ANYNL:
3036    case OP_VSPACE:    case OP_VSPACE:
3037    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3038    switch(next)    switch(next)
# Line 2090  if (next >= 0) switch(op_code) Line 3044  if (next >= 0) switch(op_code)
3044      case 0x85:      case 0x85:
3045      case 0x2028:      case 0x2028:
3046      case 0x2029:      case 0x2029:
3047      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3048      default:      default:
3049      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3050      }      }
3051    
3052    #ifdef SUPPORT_UCP
3053      case OP_PROP:
3054      return check_char_prop(next, previous[0], previous[1], FALSE);
3055    
3056      case OP_NOTPROP:
3057      return check_char_prop(next, previous[0], previous[1], TRUE);
3058    #endif
3059    
3060    default:    default:
3061    return FALSE;    return FALSE;
3062    }    }
3063    
3064    
3065  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3066    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3067    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3068    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3069    replaced by OP_PROP codes when PCRE_UCP is set. */
3070    
3071  switch(op_code)  switch(op_code)
3072    {    {
3073    case OP_CHAR:    case OP_CHAR:
3074    case OP_CHARNC:    case OP_CHARI:
3075  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3076    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3077    #else
3078      c = *previous;
3079  #endif  #endif
3080    switch(-next)    switch(-next)
3081      {      {
3082      case ESC_d:      case ESC_d:
3083      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3084    
3085      case ESC_D:      case ESC_D:
3086      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3087    
3088      case ESC_s:      case ESC_s:
3089      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3090    
3091      case ESC_S:      case ESC_S:
3092      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3093    
3094      case ESC_w:      case ESC_w:
3095      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3096    
3097      case ESC_W:      case ESC_W:
3098      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3099    
3100      case ESC_h:      case ESC_h:
3101      case ESC_H:      case ESC_H:
3102      switch(item)      switch(c)
3103        {        {
3104        case 0x09:        case 0x09:
3105        case 0x20:        case 0x20:
# Line 2159  switch(op_code) Line 3127  switch(op_code)
3127    
3128      case ESC_v:      case ESC_v:
3129      case ESC_V:      case ESC_V:
3130      switch(item)      switch(c)
3131        {        {
3132        case 0x0a:        case 0x0a:
3133        case 0x0b:        case 0x0b:
# Line 2173  switch(op_code) Line 3141  switch(op_code)
3141        return -next == ESC_v;        return -next == ESC_v;
3142        }        }
3143    
3144        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3145        their substitutions and process them. The result will always be either
3146        -ESC_p or -ESC_P. Then fall through to process those values. */
3147    
3148    #ifdef SUPPORT_UCP
3149        case ESC_du:
3150        case ESC_DU:
3151        case ESC_wu:
3152        case ESC_WU:
3153        case ESC_su:
3154        case ESC_SU:
3155          {
3156          int temperrorcode = 0;
3157          ptr = substitutes[-next - ESC_DU];
3158          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3159          if (temperrorcode != 0) return FALSE;
3160          ptr++;    /* For compatibility */
3161          }
3162        /* Fall through */
3163    
3164        case ESC_p:
3165        case ESC_P:
3166          {
3167          int ptype, pdata, errorcodeptr;
3168          BOOL negated;
3169    
3170          ptr--;      /* Make ptr point at the p or P */
3171          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3172          if (ptype < 0) return FALSE;
3173          ptr++;      /* Point past the final curly ket */
3174    
3175          /* If the property item is optional, we have to give up. (When generated
3176          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3177          to the original \d etc. At this point, ptr will point to a zero byte. */
3178    
3179          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3180            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3181              return FALSE;
3182    
3183          /* Do the property check. */
3184    
3185          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3186          }
3187    #endif
3188    
3189      default:      default:
3190      return FALSE;      return FALSE;
3191      }      }
3192    
3193      /* In principle, support for Unicode properties should be integrated here as
3194      well. It means re-organizing the above code so as to get hold of the property
3195      values before switching on the op-code. However, I wonder how many patterns
3196      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3197      these op-codes are never generated.) */
3198    
3199    case OP_DIGIT:    case OP_DIGIT:
3200    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3201           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3202    
3203    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3204    return next == -ESC_d;    return next == -ESC_d;
3205    
3206    case OP_WHITESPACE:    case OP_WHITESPACE:
3207    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3208    
3209    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3210    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3211    
3212    case OP_HSPACE:    case OP_HSPACE:
3213    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3214             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3215    
3216    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3217    return next == -ESC_h;    return next == -ESC_h;
3218    
3219    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3220      case OP_ANYNL:
3221    case OP_VSPACE:    case OP_VSPACE:
3222    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3223    
3224    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3225    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3226    
3227    case OP_WORDCHAR:    case OP_WORDCHAR:
3228    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3229             next == -ESC_v || next == -ESC_R;
3230    
3231    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3232    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2236  Arguments: Line 3258  Arguments:
3258    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3259    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3260    bcptr          points to current branch chain    bcptr          points to current branch chain
3261      cond_depth     conditional nesting depth
3262    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3263    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3264                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2245  Returns:         TRUE on success Line 3268  Returns:         TRUE on success
3268  */  */
3269    
3270  static BOOL  static BOOL
3271  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3272    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, int *firstbyteptr,
3273    compile_data *cd, int *lengthptr)    int *reqbyteptr, branch_chain *bcptr, int cond_depth, compile_data *cd,
3274      int *lengthptr)
3275  {  {
3276  int repeat_type, op_type;  int repeat_type, op_type;
3277  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2256  int greedy_default, greedy_non_default; Line 3280  int greedy_default, greedy_non_default;
3280  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3281  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3282  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3283  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3284  int after_manual_callout = 0;  int after_manual_callout = 0;
3285  int length_prevgroup = 0;  int length_prevgroup = 0;
3286  register int c;  register int c;
3287  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3288  uschar *last_code = code;  pcre_uchar *last_code = code;
3289  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3290  uschar *tempcode;  pcre_uchar *tempcode;
3291  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3292  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3293  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3294  const uschar *tempptr;  const pcre_uchar *tempptr;
3295  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3296  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3297  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3298  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3299    pcre_uchar classbits[32];
3300    
3301    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3302    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3303    dynamically as we process the pattern. */
3304    
3305  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3306  BOOL class_utf8;  BOOL class_utf8;
3307  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3308  uschar *class_utf8data;  pcre_uint8 *class_utf8data;
3309  uschar utf8_char[6];  pcre_uint8 *class_utf8data_base;
3310    pcre_uint8 utf8_char[6];
3311  #else  #else
3312  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3313  #endif  #endif
3314    
3315  #ifdef DEBUG  #ifdef PCRE_DEBUG
3316  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3317  #endif  #endif
3318    
# Line 2316  req_caseopt = ((options & PCRE_CASELESS) Line 3345  req_caseopt = ((options & PCRE_CASELESS)
3345  for (;; ptr++)  for (;; ptr++)
3346    {    {
3347    BOOL negate_class;    BOOL negate_class;
3348      BOOL should_flip_negation;
3349    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3350    BOOL is_quantifier;    BOOL is_quantifier;
3351    BOOL is_recurse;    BOOL is_recurse;
# Line 2330  for (;; ptr++) Line 3360  for (;; ptr++)
3360    int subfirstbyte;    int subfirstbyte;
3361    int terminator;    int terminator;
3362    int mclength;    int mclength;
3363    uschar mcbuffer[8];    int tempbracount;
3364      pcre_uchar mcbuffer[8];
3365    
3366    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3367    
3368    c = *ptr;    c = *ptr;
3369    
3370      /* If we are at the end of a nested substitution, revert to the outer level
3371      string. Nesting only happens one level deep. */
3372    
3373      if (c == 0 && nestptr != NULL)
3374        {
3375        ptr = nestptr;
3376        nestptr = NULL;
3377        c = *ptr;
3378        }
3379    
3380    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3381    previous cycle of this loop. */    previous cycle of this loop. */
3382    
3383    if (lengthptr != NULL)    if (lengthptr != NULL)
3384      {      {
3385  #ifdef DEBUG  #ifdef PCRE_DEBUG
3386      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3387  #endif  #endif
3388      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3389        {        {
3390        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3391        goto FAILED;        goto FAILED;
# Line 2366  for (;; ptr++) Line 3407  for (;; ptr++)
3407        goto FAILED;        goto FAILED;
3408        }        }
3409    
3410      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3411      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3412          c));
3413    
3414      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3415      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2393  for (;; ptr++) Line 3435  for (;; ptr++)
3435    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3436    reference list. */    reference list. */
3437    
3438    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3439      {      {
3440      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3441      goto FAILED;      goto FAILED;
# Line 2403  for (;; ptr++) Line 3445  for (;; ptr++)
3445    
3446    if (inescq && c != 0)    if (inescq && c != 0)
3447      {      {
3448      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3449        {        {
3450        inescq = FALSE;        inescq = FALSE;
3451        ptr++;        ptr++;
# Line 2429  for (;; ptr++) Line 3471  for (;; ptr++)
3471    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3472    a quantifier. */    a quantifier. */
3473    
3474    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3475      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3476        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3477    
3478    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3479         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2440  for (;; ptr++) Line 3483  for (;; ptr++)
3483      previous_callout = NULL;      previous_callout = NULL;
3484      }      }
3485    
3486    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3487    
3488    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3489      {      {
3490      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3491      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3492        {        {
3493        while (*(++ptr) != 0)        ptr++;
3494          while (*ptr != 0)
3495          {          {
3496          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3497            ptr++;
3498    #ifdef SUPPORT_UTF8
3499            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3500    #endif
3501          }          }
3502        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3503    
# Line 2470  for (;; ptr++) Line 3518  for (;; ptr++)
3518      {      {
3519      /* ===================================================================*/      /* ===================================================================*/
3520      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3521      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3522      case ')':      case CHAR_RIGHT_PARENTHESIS:
3523      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3524      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3525      *codeptr = code;      *codeptr = code;
# Line 2483  for (;; ptr++) Line 3531  for (;; ptr++)
3531          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3532          goto FAILED;          goto FAILED;
3533          }          }
3534        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3535        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3536        }        }
3537      return TRUE;      return TRUE;
# Line 2493  for (;; ptr++) Line 3541  for (;; ptr++)
3541      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3542      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3543    
3544      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3545        previous = NULL;
3546      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3547        {        {
3548        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3549          *code++ = OP_CIRCM;
3550        }        }
3551      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3552      break;      break;
3553    
3554      case '$':      case CHAR_DOLLAR_SIGN:
3555      previous = NULL;      previous = NULL;
3556      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3557      break;      break;
3558    
3559      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3560      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3561    
3562      case '.':      case CHAR_DOT:
3563      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3564      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3565      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3566      previous = code;      previous = code;
3567      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3568      break;      break;
3569    
3570    
# Line 2530  for (;; ptr++) Line 3579  for (;; ptr++)
3579      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3580      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3581      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3582    
3583      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3584        default (Perl) mode, it is treated as a data character. */
3585    
3586        case CHAR_RIGHT_SQUARE_BRACKET:
3587        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3588          {
3589          *errorcodeptr = ERR64;
3590          goto FAILED;
3591          }
3592        goto NORMAL_CHAR;
3593    
3594        case CHAR_LEFT_SQUARE_BRACKET:
3595      previous = code;      previous = code;
3596    
3597      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3598      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3599    
3600      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3601          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3602            check_posix_syntax(ptr, &tempptr))
3603        {        {
3604        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3605        goto FAILED;        goto FAILED;
3606        }        }
3607    
# Line 2553  for (;; ptr++) Line 3613  for (;; ptr++)
3613      for (;;)      for (;;)
3614        {        {
3615        c = *(++ptr);        c = *(++ptr);
3616        if (c == '\\')        if (c == CHAR_BACKSLASH)
3617          {          {
3618          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3619            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3620              else break;          else if (STRNCMP_UC_C8(ptr + 1,
3621                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3622              ptr += 3;
3623            else
3624              break;
3625          }          }
3626        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3627          negate_class = TRUE;          negate_class = TRUE;
3628        else break;        else break;
3629        }        }
3630    
3631        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3632        an initial ']' is taken as a data character -- the code below handles
3633        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3634        [^] must match any character, so generate OP_ALLANY. */
3635    
3636        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3637            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3638          {
3639          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3640          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3641          zerofirstbyte = firstbyte;
3642          break;
3643          }
3644    
3645        /* If a class contains a negative special such as \S, we need to flip the
3646        negation flag at the end, so that support for characters > 255 works
3647        correctly (they are all included in the class). */
3648    
3649        should_flip_negation = FALSE;
3650    
3651      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
3652      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
3653      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2576  for (;; ptr++) Line 3660  for (;; ptr++)
3660      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3661      */      */
3662    
3663      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3664    
3665  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3666      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3667      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3668        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3669  #endif  #endif
3670    
3671      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2589  for (;; ptr++) Line 3674  for (;; ptr++)
3674    
3675      if (c != 0) do      if (c != 0) do
3676        {        {
3677        const uschar *oldptr;        const pcre_uchar *oldptr;
3678    
3679  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3680        if (utf8 && c > 127)        if (utf8 && c > 127)
3681          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3682          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3683          }          }
3684    
3685          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3686          data and reset the pointer. This is so that very large classes that
3687          contain a zillion UTF-8 characters no longer overwrite the work space
3688          (which is on the stack). */
3689    
3690          if (lengthptr != NULL)
3691            {
3692            *lengthptr += class_utf8data - class_utf8data_base;
3693            class_utf8data = class_utf8data_base;
3694            }
3695    
3696  #endif  #endif
3697    
3698        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3699    
3700        if (inescq)        if (inescq)
3701          {          {
3702          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3703            {            {
3704            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3705            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2617  for (;; ptr++) Line 3714  for (;; ptr++)
3714        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3715        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3716    
3717        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3718            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3719            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3720          {          {
3721          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3722          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3723          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3724          uschar pbits[32];          pcre_uint8 pbits[32];
3725    
3726          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3727            {            {
3728            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3729            goto FAILED;            goto FAILED;
3730            }            }
3731    
3732          ptr += 2;          ptr += 2;
3733          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3734            {            {
3735            local_negate = TRUE;            local_negate = TRUE;
3736              should_flip_negation = TRUE;  /* Note negative special */
3737            ptr++;            ptr++;
3738            }            }
3739    
3740          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3741          if (posix_class < 0)          if (posix_class < 0)
3742            {            {
3743            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2653  for (;; ptr++) Line 3751  for (;; ptr++)
3751          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3752            posix_class = 0;            posix_class = 0;
3753    
3754          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3755          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3756          subtract bits that may be in the main map already. At the end we or the  
3757          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3758            if ((options & PCRE_UCP) != 0)
3759              {
3760              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3761              if (posix_substitutes[pc] != NULL)
3762                {
3763                nestptr = tempptr + 1;
3764                ptr = posix_substitutes[pc] - 1;
3765                continue;
3766                }
3767              }
3768    #endif
3769            /* In the non-UCP case, we build the bit map for the POSIX class in a
3770            chunk of local store because we may be adding and subtracting from it,
3771            and we don't want to subtract bits that may be in the main map already.
3772            At the end we or the result into the bit map that is being built. */
3773    
3774          posix_class *= 3;          posix_class *= 3;
3775    
3776          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3777    
3778          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3779            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3780    
3781          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3782    
# Line 2700  for (;; ptr++) Line 3813  for (;; ptr++)
3813    
3814        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3815        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3816        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3817        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3818        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3819        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3820          PCRE_EXTRA is set. */
3821    
3822        if (c == '\\')        if (c == CHAR_BACKSLASH)
3823          {          {
3824          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3825          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3826    
3827          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3828          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3829            {            {
3830            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3831              {              {
3832              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3833              }              }
3834            else inescq = TRUE;            else inescq = TRUE;
3835            continue;            continue;
3836            }            }
3837          else if (-c == ESC_E) continue;  /* Ignore orphan \E */          else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3838    
3839          if (c < 0)          if (c < 0)
3840            {            {
3841            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3842            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3843    
3844            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3845              {              {
3846    #ifdef SUPPORT_UCP
3847                case ESC_du:     /* These are the values given for \d etc */
3848                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3849                case ESC_wu:     /* escape sequence with an appropriate \p */
3850                case ESC_WU:     /* or \P to test Unicode properties instead */
3851                case ESC_su:     /* of the default ASCII testing. */
3852                case ESC_SU:
3853                nestptr = ptr;
3854                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3855                class_charcount -= 2;                /* Undo! */
3856                continue;
3857    #endif
3858              case ESC_d:              case ESC_d:
3859              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3860              continue;              continue;
3861    
3862              case ESC_D:              case ESC_D:
3863                should_flip_negation = TRUE;
3864              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3865              continue;              continue;
3866    
# Line 2746  for (;; ptr++) Line 3869  for (;; ptr++)
3869              continue;              continue;
3870    
3871              case ESC_W:              case ESC_W:
3872                should_flip_negation = TRUE;
3873              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3874              continue;              continue;
3875    
3876                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3877                if it was previously set by something earlier in the character
3878                class. */
3879    
3880              case ESC_s:              case ESC_s:
3881              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3882              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3883                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3884              continue;              continue;
3885    
3886              case ESC_S:              case ESC_S:
3887                should_flip_negation = TRUE;
3888              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3889              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3890              continue;              continue;
3891    
3892              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3893              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3894              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3895              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2784  for (;; ptr++) Line 3898  for (;; ptr++)
3898                {                {
3899                class_utf8 = TRUE;                class_utf8 = TRUE;
3900                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3901                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1680, class_utf8data);
3902                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3903                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180e, class_utf8data);
3904                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3905                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2000, class_utf8data);
3906                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x200A, class_utf8data);
3907                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3908                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x202f, class_utf8data);
3909                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3910                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x205f, class_utf8data);
3911                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3912                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x3000, class_utf8data);
3913                }                }
3914  #endif  #endif
3915              continue;              continue;
             }  
3916    
3917            if (-c == ESC_H)              case ESC_H:
             {  
3918              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3919                {                {
3920                int x = 0xff;                int x = 0xff;
# Line 2821  for (;; ptr++) Line 3933  for (;; ptr++)
3933                {                {
3934                class_utf8 = TRUE;                class_utf8 = TRUE;
3935                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3936                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x0100, class_utf8data);
3937                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x167f, class_utf8data);
3938                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3939                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1681, class_utf8data);
3940                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180d, class_utf8data);
3941                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3942                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180f, class_utf8data);
3943                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1fff, class_utf8data);
3944                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3945                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x200B, class_utf8data);
3946                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x202e, class_utf8data);
3947                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3948                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2030, class_utf8data);
3949                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x205e, class_utf8data);
3950                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3951                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2060, class_utf8data);
3952                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2fff, class_utf8data);
3953                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3954                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x3001, class_utf8data);
3955                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x7fffffff, class_utf8data);
3956                }                }
3957  #endif  #endif
3958              continue;              continue;
             }  
3959    
3960            if (-c == ESC_v)              case ESC_v:
             {  
3961              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3962              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3963              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2858  for (;; ptr++) Line 3968  for (;; ptr++)
3968                {                {
3969                class_utf8 = TRUE;                class_utf8 = TRUE;
3970                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3971                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2028, class_utf8data);
3972                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2029, class_utf8data);
3973                }                }
3974  #endif  #endif
3975              continue;              continue;
             }  
3976    
3977            if (-c == ESC_V)              case ESC_V:
             {  
3978              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3979                {                {
3980                int x = 0xff;                int x = 0xff;
# Line 2888  for (;; ptr++) Line 3996  for (;; ptr++)
3996                {                {
3997                class_utf8 = TRUE;                class_utf8 = TRUE;
3998                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3999                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x0100, class_utf8data);
4000                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2027, class_utf8data);
4001                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
4002                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2029, class_utf8data);
4003                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x7fffffff, class_utf8data);
4004                }                }
4005  #endif  #endif
4006              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
4007    
4008  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4009            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
4010              {              case ESC_P:
4011              BOOL negated;                {
4012              int pdata;                BOOL negated;
4013              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
4014              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4015              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
4016              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
4