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

Diff of /code/trunk/pcre_compile.c

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

revision 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 629 by ph10, Fri Jul 22 09:18:11 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 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_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    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
# Line 76  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       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -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       0,      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,      0, -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 106  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
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 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       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,     0, -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,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  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  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      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 const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    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 uschar 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 };
# Line 164  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 uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  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
317  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
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number or name after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
387    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by an (optionally braced) non-zero number"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399  };    "a numbered reference must not be zero\0"
400      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 262  For convenience, we use the same bit def Line 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 298  static const unsigned char digitab[] = Line 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 373  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
554  /*************************************************  /*************************************************
555    *            Find an error text                  *
556    *************************************************/
557    
558    /* The error texts are now all in one long string, to save on relocations. As
559    some of the text is of unknown length, we can't use a table of offsets.
560    Instead, just count through the strings. This is not a performance issue
561    because it happens only when there has been a compilation error.
562    
563    Argument:   the error number
564    Returns:    pointer to the error string
565    */
566    
567    static const char *
568    find_error_text(int n)
569    {
570    const char *s = error_texts;
571    for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576    return s;
577    }
578    
579    
580    /*************************************************
581  *            Handle escapes                      *  *            Handle escapes                      *
582  *************************************************/  *************************************************/
583    
# Line 399  Arguments: Line 598  Arguments:
598    
599  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
600                   negative => a special escape sequence                   negative => a special escape sequence
601                   on error, errorptr is set                   on error, errorcodeptr is set
602  */  */
603    
604  static int  static int
# Line 417  ptr--;                            /* Set Line 616  ptr--;                            /* Set
616    
617  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
618    
619  /* 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
620  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.
621  Otherwise further processing may be required. */  Otherwise further processing may be required. */
622    
623  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
624  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
625  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
626    
627  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
628  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
629  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
630  #endif  #endif
631    
# Line 442  else Line 641  else
641      /* 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
642      error. */      error. */
643    
644      case 'l':      case CHAR_l:
645      case 'L':      case CHAR_L:
646      case 'N':      case CHAR_u:
647      case 'u':      case CHAR_U:
     case 'U':  
648      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
649      break;      break;
650    
651      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
652      is an absolute backreference. If negative, it is a relative backreference.  
653      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
654        backreference. If negative, it is a relative backreference. This is a Perl
655        5.10 feature.
656    
657        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
658        is part of Perl's movement towards a unified syntax for back references. As
659        this is synonymous with \k{name}, we fudge it up by pretending it really
660        was \k.
661    
662        (3) For Oniguruma compatibility we also support \g followed by a name or a
663        number either in angle brackets or in single quotes. However, these are
664        (possibly recursive) subroutine calls, _not_ backreferences. Just return
665        the -ESC_g code (cf \k). */
666    
667        case CHAR_g:
668        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
669          {
670          c = -ESC_g;
671          break;
672          }
673    
674        /* Handle the Perl-compatible cases */
675    
676      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
677        {        {
678          const uschar *p;
679          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
680            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
681          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
682            {
683            c = -ESC_k;
684            break;
685            }
686        braced = TRUE;        braced = TRUE;
687        ptr++;        ptr++;
688        }        }
689      else braced = FALSE;      else braced = FALSE;
690    
691      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
692        {        {
693        negated = TRUE;        negated = TRUE;
694        ptr++;        ptr++;
# Line 471  else Line 697  else
697    
698      c = 0;      c = 0;
699      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
700        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
701    
702        if (c < 0)   /* Integer overflow */
703          {
704          *errorcodeptr = ERR61;
705          break;
706          }
707    
708      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
709        {        {
710        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
711        return 0;        break;
712          }
713    
714        if (c == 0)
715          {
716          *errorcodeptr = ERR58;
717          break;
718        }        }
719    
720      if (negated)      if (negated)
# Line 484  else Line 722  else
722        if (c > bracount)        if (c > bracount)
723          {          {
724          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
725          return 0;          break;
726          }          }
727        c = bracount - (c - 1);        c = bracount - (c - 1);
728        }        }
# Line 504  else Line 742  else
742      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
743      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
744    
745      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:
746      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
747    
748      if (!isclass)      if (!isclass)
749        {        {
750        oldptr = ptr;        oldptr = ptr;
751        c -= '0';        c -= CHAR_0;
752        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
753          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
754          if (c < 0)    /* Integer overflow */
755            {
756            *errorcodeptr = ERR61;
757            break;
758            }
759        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
760          {          {
761          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 768  else
768      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.
769      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
770    
771      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
772        {        {
773        ptr--;        ptr--;
774        c = 0;        c = 0;
# Line 538  else Line 781  else
781      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
782      than 3 octal digits. */      than 3 octal digits. */
783    
784      case '0':      case CHAR_0:
785      c -= '0';      c -= CHAR_0;
786      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
787          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
788      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
789      break;      break;
790    
# Line 549  else Line 792  else
792      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
793      treated as a data character. */      treated as a data character. */
794    
795      case 'x':      case CHAR_x:
796      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
797        {        {
798        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
799        int count = 0;        int count = 0;
# Line 559  else Line 802  else
802        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
803          {          {
804          register int cc = *pt++;          register int cc = *pt++;
805          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
806          count++;          count++;
807    
808  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
809          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
810          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
811  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
812          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
813          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
814  #endif  #endif
815          }          }
816    
817        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
818          {          {
819          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
820          ptr = pt;          ptr = pt;
# Line 587  else Line 830  else
830      c = 0;      c = 0;
831      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
832        {        {
833        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
834        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
835  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
836        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
837        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
838  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
839        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
840        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
841  #endif  #endif
842        }        }
843      break;      break;
844    
845      /* 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.
846      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
847        coding is ASCII-specific, but then the whole concept of \cx is
848      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
849    
850      case 'c':      case CHAR_c:
851      c = *(++ptr);      c = *(++ptr);
852      if (c == 0)      if (c == 0)
853        {        {
854        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
855        return 0;        break;
856        }        }
857    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
858  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
859      if (c >= 'a' && c <= 'z') c -= 32;        {
860          *errorcodeptr = ERR68;
861          break;
862          }
863        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
864      c ^= 0x40;      c ^= 0x40;
865  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
866      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
867      c ^= 0xC0;      c ^= 0xC0;
868  #endif  #endif
869      break;      break;
870    
871      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
872      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
873      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
874      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
875      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
876    
877      default:      default:
878      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 885  else
885      }      }
886    }    }
887    
888    /* Perl supports \N{name} for character names, as well as plain \N for "not
889    newline". PCRE does not support \N{name}. */
890    
891    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
892      *errorcodeptr = ERR37;
893    
894    /* If PCRE_UCP is set, we change the values for \d etc. */
895    
896    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
897      c -= (ESC_DU - ESC_D);
898    
899    /* Set the pointer to the final character before returning. */
900    
901  *ptrptr = ptr;  *ptrptr = ptr;
902  return c;  return c;
903  }  }
# Line 677  if (c == 0) goto ERROR_RETURN; Line 938  if (c == 0) goto ERROR_RETURN;
938  /* \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
939  negation. */  negation. */
940    
941  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
942    {    {
943    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
944      {      {
945      *negptr = TRUE;      *negptr = TRUE;
946      ptr++;      ptr++;
947      }      }
948    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
949      {      {
950      c = *(++ptr);      c = *(++ptr);
951      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
952      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
953      name[i] = c;      name[i] = c;
954      }      }
955    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
956    name[i] = 0;    name[i] = 0;
957    }    }
958    
# Line 713  top = _pcre_utt_size; Line 974  top = _pcre_utt_size;
974  while (bot < top)  while (bot < top)
975    {    {
976    i = (bot + top) >> 1;    i = (bot + top) >> 1;
977    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
978    if (c == 0)    if (c == 0)
979      {      {
980      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 756  is_counted_repeat(const uschar *p) Line 1017  is_counted_repeat(const uschar *p)
1017  {  {
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1021    
1022  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1023  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1024    
1025  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1026  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1027    
1028  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1029  }  }
1030    
1031    
# Line 797  int max = -1; Line 1058  int max = -1;
1058  /* 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
1059  an integer overflow. */  an integer overflow. */
1060    
1061  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1062  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1063    {    {
1064    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1068  if (min < 0 || min > 65535)
1068  /* 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.
1069  Also, max must not be less than min. */  Also, max must not be less than min. */
1070    
1071  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1072    {    {
1073    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1074      {      {
1075      max = 0;      max = 0;
1076      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1077      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1078        {        {
1079        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1098  return p;
1098    
1099    
1100  /*************************************************  /*************************************************
1101  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1102  *************************************************/  *************************************************/
1103    
1104  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1105    top-level call starts at the beginning of the pattern. All other calls must
1106    start at a parenthesis. It scans along a pattern's text looking for capturing
1107  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
1108  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
1109  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
1110  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1111  be terminated by '>' because that is checked in the first pass.  
1112    This function was originally called only from the second pass, in which we know
1113    that if (?< or (?' or (?P< is encountered, the name will be correctly
1114    terminated because that is checked in the first pass. There is now one call to
1115    this function in the first pass, to check for a recursive back reference by
1116    name (so that we can make the whole group atomic). In this case, we need check
1117    only up to the current position in the pattern, and that is still OK because
1118    and previous occurrences will have been checked. To make this work, the test
1119    for "end of pattern" is a check against cd->end_pattern in the main loop,
1120    instead of looking for a binary zero. This means that the special first-pass
1121    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1122    processing items within the loop are OK, because afterwards the main loop will
1123    terminate.)
1124    
1125  Arguments:  Arguments:
1126    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1127    count        current count of capturing parens so far encountered    cd           compile background data
1128    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1129    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1130    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1131      utf8         TRUE if we are in UTF-8 mode
1132      count        pointer to the current capturing subpattern number (updated)
1133    
1134  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1135  */  */
1136    
1137  static int  static int
1138  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1139    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1140  {  {
1141  const uschar *thisname;  uschar *ptr = *ptrptr;
1142    int start_count = *count;
1143    int hwm_count = start_count;
1144    BOOL dup_parens = FALSE;
1145    
1146    /* If the first character is a parenthesis, check on the type of group we are
1147    dealing with. The very first call may not start with a parenthesis. */
1148    
1149  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1150    {    {
1151    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1152    
1153      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1154    
1155      /* Handle a normal, unnamed capturing parenthesis. */
1156    
1157      else if (ptr[1] != CHAR_QUESTION_MARK)
1158        {
1159        *count += 1;
1160        if (name == NULL && *count == lorn) return *count;
1161        ptr++;
1162        }
1163    
1164      /* All cases now have (? at the start. Remember when we are in a group
1165      where the parenthesis numbers are duplicated. */
1166    
1167      else if (ptr[2] == CHAR_VERTICAL_LINE)
1168        {
1169        ptr += 3;
1170        dup_parens = TRUE;
1171        }
1172    
1173      /* Handle comments; all characters are allowed until a ket is reached. */
1174    
1175      else if (ptr[2] == CHAR_NUMBER_SIGN)
1176        {
1177        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1178        goto FAIL_EXIT;
1179        }
1180    
1181      /* Handle a condition. If it is an assertion, just carry on so that it
1182      is processed as normal. If not, skip to the closing parenthesis of the
1183      condition (there can't be any nested parens). */
1184    
1185      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1186        {
1187        ptr += 2;
1188        if (ptr[1] != CHAR_QUESTION_MARK)
1189          {
1190          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1191          if (*ptr != 0) ptr++;
1192          }
1193        }
1194    
1195      /* Start with (? but not a condition. */
1196    
1197      else
1198        {
1199        ptr += 2;
1200        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1201    
1202        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1203    
1204        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1205            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1206          {
1207          int term;
1208          const uschar *thisname;
1209          *count += 1;
1210          if (name == NULL && *count == lorn) return *count;
1211          term = *ptr++;
1212          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1213          thisname = ptr;
1214          while (*ptr != term) ptr++;
1215          if (name != NULL && lorn == ptr - thisname &&
1216              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1217            return *count;
1218          term++;
1219          }
1220        }
1221      }
1222    
1223    /* Past any initial parenthesis handling, scan for parentheses or vertical
1224    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1225    first-pass call when this value is temporarily adjusted to stop at the current
1226    position. So DO NOT change this to a test for binary zero. */
1227    
1228    for (; ptr < cd->end_pattern; ptr++)
1229      {
1230    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1231    
1232    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1233      {      {
1234      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1235      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1236        {        {
1237        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1238        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1239        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1240        }        }
1241      continue;      continue;
1242      }      }
1243    
1244    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1245      are handled for real. If the first character is '^', skip it. Also, if the
1246      first few characters (either before or after ^) are \Q\E or \E we skip them
1247      too. This makes for compatibility with Perl. Note the use of STR macros to
1248      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1249    
1250    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1251      {      {
1252      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1253        for (;;)
1254          {
1255          if (ptr[1] == CHAR_BACKSLASH)
1256            {
1257            if (ptr[2] == CHAR_E)
1258              ptr+= 2;
1259            else if (strncmp((const char *)ptr+2,
1260                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1261              ptr += 4;
1262            else
1263              break;
1264            }
1265          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1266            {
1267            negate_class = TRUE;
1268            ptr++;
1269            }
1270          else break;
1271          }
1272    
1273        /* If the next character is ']', it is a data character that must be
1274        skipped, except in JavaScript compatibility mode. */
1275    
1276        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1277            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1278          ptr++;
1279    
1280        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1281        {        {
1282        if (*ptr == '\\')        if (*ptr == 0) return -1;
1283          if (*ptr == CHAR_BACKSLASH)
1284          {          {
1285          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1286          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1287            {            {
1288            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1289            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1290            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1291            }            }
1292          continue;          continue;
1293          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1297  for (; *ptr != 0; ptr++)
1297    
1298    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1299    
1300    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1301      {      {
1302      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1303      if (*ptr == 0) return -1;      while (*ptr != 0)
1304          {
1305          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1306          ptr++;
1307    #ifdef SUPPORT_UTF8
1308          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1309    #endif
1310          }
1311        if (*ptr == 0) goto FAIL_EXIT;
1312      continue;      continue;
1313      }      }
1314    
1315    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1316    
1317    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1318      {      {
1319      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1320      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1321      continue;      if (*ptr == 0) goto FAIL_EXIT;
1322        }
1323    
1324      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1325        {
1326        if (dup_parens && *count < hwm_count) *count = hwm_count;
1327        goto FAIL_EXIT;
1328      }      }
1329    
1330    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1331    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1332        if (*count > hwm_count) hwm_count = *count;
1333        *count = start_count;
1334        }
1335      }
1336    
1337    FAIL_EXIT:
1338    *ptrptr = ptr;
1339    return -1;
1340    }
1341    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1342    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1343    
   count++;  
1344    
1345    if (name == NULL && count == lorn) return count;  /*************************************************
1346    term = *ptr++;  *       Find forward referenced subpattern       *
1347    if (term == '<') term = '>';  *************************************************/
1348    thisname = ptr;  
1349    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1350    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1351        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1352      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1353    references to subpatterns. We used to be able to start this scan from the
1354    current compiling point, using the current count value from cd->bracount, and
1355    do it all in a single loop, but the addition of the possibility of duplicate
1356    subpattern numbers means that we have to scan from the very start, in order to
1357    take account of such duplicates, and to use a recursive function to keep track
1358    of the different types of group.
1359    
1360    Arguments:
1361      cd           compile background data
1362      name         name to seek, or NULL if seeking a numbered subpattern
1363      lorn         name length, or subpattern number if name is NULL
1364      xmode        TRUE if we are in /x mode
1365      utf8         TRUE if we are in UTF-8 mode
1366    
1367    Returns:       the number of the found subpattern, or -1 if not found
1368    */
1369    
1370    static int
1371    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1372      BOOL utf8)
1373    {
1374    uschar *ptr = (uschar *)cd->start_pattern;
1375    int count = 0;
1376    int rc;
1377    
1378    /* If the pattern does not start with an opening parenthesis, the first call
1379    to find_parens_sub() will scan right to the end (if necessary). However, if it
1380    does start with a parenthesis, find_parens_sub() will return when it hits the
1381    matching closing parens. That is why we have to have a loop. */
1382    
1383    for (;;)
1384      {
1385      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1386      if (rc > 0 || *ptr++ == 0) break;
1387    }    }
1388    
1389  return -1;  return rc;
1390  }  }
1391    
1392    
1393    
1394    
1395  /*************************************************  /*************************************************
1396  *      Find first significant op code            *  *      Find first significant op code            *
1397  *************************************************/  *************************************************/
1398    
1399  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1400  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
1401  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
1402  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
1403  assertions, and also the \b assertion; for others it does not.  does not.
1404    
1405  Arguments:  Arguments:
1406    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  
1407    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1408    
1409  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1410  */  */
1411    
1412  static const uschar*  static const uschar*
1413  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1414  {  {
1415  for (;;)  for (;;)
1416    {    {
1417    switch ((int)*code)    switch ((int)*code)
1418      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1419      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1420      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1421      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 996  for (;;) Line 1431  for (;;)
1431    
1432      case OP_CALLOUT:      case OP_CALLOUT:
1433      case OP_CREF:      case OP_CREF:
1434        case OP_NCREF:
1435      case OP_RREF:      case OP_RREF:
1436        case OP_NRREF:
1437      case OP_DEF:      case OP_DEF:
1438      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1439      break;      break;
# Line 1012  for (;;) Line 1449  for (;;)
1449    
1450    
1451  /*************************************************  /*************************************************
1452  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1453  *************************************************/  *************************************************/
1454    
1455  /* 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,
1456  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.
1457  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
1458    temporarily terminated with OP_END when this function is called.
1459    
1460    This function is called when a backward assertion is encountered, so that if it
1461    fails, the error message can point to the correct place in the pattern.
1462    However, we cannot do this when the assertion contains subroutine calls,
1463    because they can be forward references. We solve this by remembering this case
1464    and doing the check at the end; a flag specifies which mode we are running in.
1465    
1466  Arguments:  Arguments:
1467    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1468    options  the compiling options    utf8     TRUE in UTF-8 mode
1469      atend    TRUE if called when the pattern is complete
1470      cd       the "compile data" structure
1471    
1472  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1473                 or -1 if there is no fixed length,
1474               or -2 if \C was encountered               or -2 if \C was encountered
1475                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1476  */  */
1477    
1478  static int  static int
1479  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1480  {  {
1481  int length = -1;  int length = -1;
1482    
# Line 1041  branch, check the length against that of Line 1489  branch, check the length against that of
1489  for (;;)  for (;;)
1490    {    {
1491    int d;    int d;
1492      uschar *ce, *cs;
1493    register int op = *cc;    register int op = *cc;
   
1494    switch (op)    switch (op)
1495      {      {
1496        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1497        OP_BRA (normal non-capturing bracket) because the other variants of these
1498        opcodes are all concerned with unlimited repeated groups, which of course
1499        are not of fixed length. They will cause a -1 response from the default
1500        case of this switch. */
1501    
1502      case OP_CBRA:      case OP_CBRA:
1503      case OP_BRA:      case OP_BRA:
1504      case OP_ONCE:      case OP_ONCE:
1505      case OP_COND:      case OP_COND:
1506      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1507      if (d < 0) return d;      if (d < 0) return d;
1508      branchlength += d;      branchlength += d;
1509      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1058  for (;;) Line 1512  for (;;)
1512    
1513      /* 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
1514      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1515      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1516        Note that we must not include the OP_KETRxxx opcodes here, because they
1517        all imply an unlimited repeat. */
1518    
1519      case OP_ALT:      case OP_ALT:
1520      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1521      case OP_END:      case OP_END:
1522      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1523        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1072  for (;;) Line 1526  for (;;)
1526      branchlength = 0;      branchlength = 0;
1527      break;      break;
1528    
1529        /* A true recursion implies not fixed length, but a subroutine call may
1530        be OK. If the subroutine is a forward reference, we can't deal with
1531        it until the end of the pattern, so return -3. */
1532    
1533        case OP_RECURSE:
1534        if (!atend) return -3;
1535        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1536        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1537        if (cc > cs && cc < ce) return -1;                /* Recursion */
1538        d = find_fixedlength(cs + 2, utf8, atend, cd);
1539        if (d < 0) return d;
1540        branchlength += d;
1541        cc += 1 + LINK_SIZE;
1542        break;
1543    
1544      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1545    
1546      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1554  for (;;)
1554    
1555      case OP_REVERSE:      case OP_REVERSE:
1556      case OP_CREF:      case OP_CREF:
1557        case OP_NCREF:
1558      case OP_RREF:      case OP_RREF:
1559        case OP_NRREF:
1560      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1561      case OP_CALLOUT:      case OP_CALLOUT:
1562      case OP_SOD:      case OP_SOD:
1563      case OP_SOM:      case OP_SOM:
1564        case OP_SET_SOM:
1565      case OP_EOD:      case OP_EOD:
1566      case OP_EODN:      case OP_EODN:
1567      case OP_CIRC:      case OP_CIRC:
1568        case OP_CIRCM:
1569      case OP_DOLL:      case OP_DOLL:
1570        case OP_DOLLM:
1571      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1572      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1573      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1103  for (;;) Line 1576  for (;;)
1576      /* Handle literal characters */      /* Handle literal characters */
1577    
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARNC:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581        case OP_NOTI:
1582      branchlength++;      branchlength++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1586  #endif  #endif
1587      break;      break;
1588    
# Line 1122  for (;;) Line 1593  for (;;)
1593      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1594      cc += 4;      cc += 4;
1595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1596      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1597  #endif  #endif
1598      break;      break;
1599    
1600      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1601      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1602        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1603      cc += 4;      cc += 4;
1604      break;      break;
1605    
# Line 1148  for (;;) Line 1617  for (;;)
1617      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1618      case OP_WORDCHAR:      case OP_WORDCHAR:
1619      case OP_ANY:      case OP_ANY:
1620        case OP_ALLANY:
1621      branchlength++;      branchlength++;
1622      cc++;      cc++;
1623      break;      break;
# Line 1202  for (;;) Line 1672  for (;;)
1672    
1673    
1674  /*************************************************  /*************************************************
1675  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1676  *************************************************/  *************************************************/
1677    
1678  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1679  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1680    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1681    so that it can be called from pcre_study() when finding the minimum matching
1682    length.
1683    
1684  Arguments:  Arguments:
1685    code        points to start of expression    code        points to start of expression
1686    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1687    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1688    
1689  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
1690  */  */
1691    
1692  static const uschar *  const uschar *
1693  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1694  {  {
1695  for (;;)  for (;;)
1696    {    {
1697    register int c = *code;    register int c = *code;
1698    
1699    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1700    
1701    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1230  for (;;) Line 1704  for (;;)
1704    
1705    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1706    
1707      /* Handle recursion */
1708    
1709      else if (c == OP_REVERSE)
1710        {
1711        if (number < 0) return (uschar *)code;
1712        code += _pcre_OP_lengths[c];
1713        }
1714    
1715    /* Handle capturing bracket */    /* Handle capturing bracket */
1716    
1717    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1718               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1719      {      {
1720      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1721      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1722      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1723      }      }
1724    
1725    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1726    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1727    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1728      must add in its length. */
1729    
1730    else    else
1731      {      {
1732        switch(c)
1733          {
1734          case OP_TYPESTAR:
1735          case OP_TYPEMINSTAR:
1736          case OP_TYPEPLUS:
1737          case OP_TYPEMINPLUS:
1738          case OP_TYPEQUERY:
1739          case OP_TYPEMINQUERY:
1740          case OP_TYPEPOSSTAR:
1741          case OP_TYPEPOSPLUS:
1742          case OP_TYPEPOSQUERY:
1743          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1744          break;
1745    
1746          case OP_TYPEUPTO:
1747          case OP_TYPEMINUPTO:
1748          case OP_TYPEEXACT:
1749          case OP_TYPEPOSUPTO:
1750          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1751          break;
1752    
1753          case OP_MARK:
1754          case OP_PRUNE_ARG:
1755          case OP_SKIP_ARG:
1756          code += code[1];
1757          break;
1758    
1759          case OP_THEN_ARG:
1760          code += code[1+LINK_SIZE];
1761          break;
1762          }
1763    
1764        /* Add in the fixed length from the table */
1765    
1766      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1767    
1768      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1769      a multi-byte character. The length in the table is a minimum, so we have to
1770      arrange to skip the extra bytes. */
1771    
1772  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1773      if (utf8) switch(c)      if (utf8) switch(c)
1774        {        {
1775        case OP_CHAR:        case OP_CHAR:
1776        case OP_CHARNC:        case OP_CHARI:
1777        case OP_EXACT:        case OP_EXACT:
1778          case OP_EXACTI:
1779        case OP_UPTO:        case OP_UPTO:
1780          case OP_UPTOI:
1781        case OP_MINUPTO:        case OP_MINUPTO:
1782          case OP_MINUPTOI:
1783        case OP_POSUPTO:        case OP_POSUPTO:
1784          case OP_POSUPTOI:
1785        case OP_STAR:        case OP_STAR:
1786          case OP_STARI:
1787        case OP_MINSTAR:        case OP_MINSTAR:
1788          case OP_MINSTARI:
1789        case OP_POSSTAR:        case OP_POSSTAR:
1790          case OP_POSSTARI:
1791        case OP_PLUS:        case OP_PLUS:
1792          case OP_PLUSI:
1793        case OP_MINPLUS:        case OP_MINPLUS:
1794          case OP_MINPLUSI:
1795        case OP_POSPLUS:        case OP_POSPLUS:
1796          case OP_POSPLUSI:
1797        case OP_QUERY:        case OP_QUERY:
1798          case OP_QUERYI:
1799        case OP_MINQUERY:        case OP_MINQUERY:
1800          case OP_MINQUERYI:
1801        case OP_POSQUERY:        case OP_POSQUERY:
1802          case OP_POSQUERYI:
1803        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1804        break;        break;
1805        }        }
1806    #else
1807        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1808  #endif  #endif
1809      }      }
1810    }    }
# Line 1303  for (;;) Line 1841  for (;;)
1841    
1842    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1843    
1844    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1845    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1846    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1847    bytes. */    must add in its length. */
1848    
1849    else    else
1850      {      {
1851        switch(c)
1852          {
1853          case OP_TYPESTAR:
1854          case OP_TYPEMINSTAR:
1855          case OP_TYPEPLUS:
1856          case OP_TYPEMINPLUS:
1857          case OP_TYPEQUERY:
1858          case OP_TYPEMINQUERY:
1859          case OP_TYPEPOSSTAR:
1860          case OP_TYPEPOSPLUS:
1861          case OP_TYPEPOSQUERY:
1862          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1863          break;
1864    
1865          case OP_TYPEPOSUPTO:
1866          case OP_TYPEUPTO:
1867          case OP_TYPEMINUPTO:
1868          case OP_TYPEEXACT:
1869          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1870          break;
1871    
1872          case OP_MARK:
1873          case OP_PRUNE_ARG:
1874          case OP_SKIP_ARG:
1875          code += code[1];
1876          break;
1877    
1878          case OP_THEN_ARG:
1879          code += code[1+LINK_SIZE];
1880          break;
1881          }
1882    
1883        /* Add in the fixed length from the table */
1884    
1885      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1886    
1887        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1888        by a multi-byte character. The length in the table is a minimum, so we have
1889        to arrange to skip the extra bytes. */
1890    
1891  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1892      if (utf8) switch(c)      if (utf8) switch(c)
1893        {        {
1894        case OP_CHAR:        case OP_CHAR:
1895        case OP_CHARNC:        case OP_CHARI:
1896        case OP_EXACT:        case OP_EXACT:
1897          case OP_EXACTI:
1898        case OP_UPTO:        case OP_UPTO:
1899          case OP_UPTOI:
1900        case OP_MINUPTO:        case OP_MINUPTO:
1901          case OP_MINUPTOI:
1902        case OP_POSUPTO:        case OP_POSUPTO:
1903          case OP_POSUPTOI:
1904        case OP_STAR:        case OP_STAR:
1905          case OP_STARI:
1906        case OP_MINSTAR:        case OP_MINSTAR:
1907          case OP_MINSTARI:
1908        case OP_POSSTAR:        case OP_POSSTAR:
1909          case OP_POSSTARI:
1910        case OP_PLUS:        case OP_PLUS:
1911          case OP_PLUSI:
1912        case OP_MINPLUS:        case OP_MINPLUS:
1913          case OP_MINPLUSI:
1914        case OP_POSPLUS:        case OP_POSPLUS:
1915          case OP_POSPLUSI:
1916        case OP_QUERY:        case OP_QUERY:
1917          case OP_QUERYI:
1918        case OP_MINQUERY:        case OP_MINQUERY:
1919          case OP_MINQUERYI:
1920        case OP_POSQUERY:        case OP_POSQUERY:
1921          case OP_POSQUERYI:
1922        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1923        break;        break;
1924        }        }
1925    #else
1926        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1927  #endif  #endif
1928      }      }
1929    }    }
# Line 1347  for (;;) Line 1939  for (;;)
1939  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()
1940  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
1941  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1942  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
1943  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
1944    bracket whose current branch will already have been scanned.
1945    
1946  Arguments:  Arguments:
1947    code        points to start of search    code        points to start of search
1948    endcode     points to where to stop    endcode     points to where to stop
1949    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1950      cd          contains pointers to tables etc.
1951    
1952  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1953  */  */
1954    
1955  static BOOL  static BOOL
1956  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1957      compile_data *cd)
1958  {  {
1959  register int c;  register int c;
1960  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1961       code < endcode;       code < endcode;
1962       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1963    {    {
1964    const uschar *ccode;    const uschar *ccode;
1965    
1966    c = *code;    c = *code;
1967    
1968    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1969      first_significant_code() with a TRUE final argument. */
1970    
1971      if (c == OP_ASSERT)
1972      {      {
1973      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1974      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1975        continue;
1976        }
1977    
1978      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1979      implies a backward reference subroutine call, we can scan it. If it's a
1980      forward reference subroutine call, we can't. To detect forward reference
1981      we have to scan up the list that is kept in the workspace. This function is
1982      called only when doing the real compile, not during the pre-compile that
1983      measures the size of the compiled pattern. */
1984    
1985      if (c == OP_RECURSE)
1986        {
1987        const uschar *scode;
1988        BOOL empty_branch;
1989    
1990        /* Test for forward reference */
1991    
1992        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1993          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1994    
1995        /* Not a forward reference, test for completed backward reference */
1996    
1997      empty_branch = FALSE;      empty_branch = FALSE;
1998        scode = cd->start_code + GET(code, 1);
1999        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2000    
2001        /* Completed backwards reference */
2002    
2003      do      do
2004        {        {
2005        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2006            {
2007          empty_branch = TRUE;          empty_branch = TRUE;
2008        code += GET(code, 1);          break;
2009            }
2010          scode += GET(scode, 1);
2011        }        }
2012      while (*code == OP_ALT);      while (*scode == OP_ALT);
2013      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2014        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2015        continue;
2016        }
2017    
2018      /* Groups with zero repeats can of course be empty; skip them. */
2019    
2020      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2021          c == OP_BRAPOSZERO)
2022        {
2023        code += _pcre_OP_lengths[c];
2024        do code += GET(code, 1); while (*code == OP_ALT);
2025        c = *code;
2026        continue;
2027        }
2028    
2029      /* A nested group that is already marked as "could be empty" can just be
2030      skipped. */
2031    
2032      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2033          c == OP_SCBRA || c == OP_SCBRAPOS)
2034        {
2035        do code += GET(code, 1); while (*code == OP_ALT);
2036        c = *code;
2037        continue;
2038        }
2039    
2040      /* For other groups, scan the branches. */
2041    
2042      if (c == OP_BRA  || c == OP_BRAPOS ||
2043          c == OP_CBRA || c == OP_CBRAPOS ||
2044          c == OP_ONCE || c == OP_COND)
2045        {
2046        BOOL empty_branch;
2047        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2048    
2049        /* If a conditional group has only one branch, there is a second, implied,
2050        empty branch, so just skip over the conditional, because it could be empty.
2051        Otherwise, scan the individual branches of the group. */
2052    
2053      /* Move past the KET and fudge things so that the increment in the "for"      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2054      above has no effect. */        code += GET(code, 1);
2055        else
2056          {
2057          empty_branch = FALSE;
2058          do
2059            {
2060            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2061              empty_branch = TRUE;
2062            code += GET(code, 1);
2063            }
2064          while (*code == OP_ALT);
2065          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2066          }
2067    
2068      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2069      continue;      continue;
2070      }      }
2071    
# Line 1399  for (code = first_significant_code(code Line 2073  for (code = first_significant_code(code
2073    
2074    switch (c)    switch (c)
2075      {      {
2076      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2077        cannot be represented just by a bit map. This includes negated single
2078        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2079        actual length is stored in the compiled code, so we must update "code"
2080        here. */
2081    
2082  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2083      case OP_XCLASS:      case OP_XCLASS:
2084      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2085      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2086  #endif  #endif
2087    
# Line 1447  for (code = first_significant_code(code Line 2125  for (code = first_significant_code(code
2125      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2126      case OP_WORDCHAR:      case OP_WORDCHAR:
2127      case OP_ANY:      case OP_ANY:
2128        case OP_ALLANY:
2129      case OP_ANYBYTE:      case OP_ANYBYTE:
2130      case OP_CHAR:      case OP_CHAR:
2131      case OP_CHARNC:      case OP_CHARI:
2132      case OP_NOT:      case OP_NOT:
2133        case OP_NOTI:
2134      case OP_PLUS:      case OP_PLUS:
2135      case OP_MINPLUS:      case OP_MINPLUS:
2136      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1465  for (code = first_significant_code(code Line 2145  for (code = first_significant_code(code
2145      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2146      return FALSE;      return FALSE;
2147    
2148        /* These are going to continue, as they may be empty, but we have to
2149        fudge the length for the \p and \P cases. */
2150    
2151        case OP_TYPESTAR:
2152        case OP_TYPEMINSTAR:
2153        case OP_TYPEPOSSTAR:
2154        case OP_TYPEQUERY:
2155        case OP_TYPEMINQUERY:
2156        case OP_TYPEPOSQUERY:
2157        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2158        break;
2159    
2160        /* Same for these */
2161    
2162        case OP_TYPEUPTO:
2163        case OP_TYPEMINUPTO:
2164        case OP_TYPEPOSUPTO:
2165        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2166        break;
2167    
2168      /* End of branch */      /* End of branch */
2169    
2170      case OP_KET:      case OP_KET:
2171      case OP_KETRMAX:      case OP_KETRMAX:
2172      case OP_KETRMIN:      case OP_KETRMIN:
2173        case OP_KETRPOS:
2174      case OP_ALT:      case OP_ALT:
2175      return TRUE;      return TRUE;
2176    
# Line 1478  for (code = first_significant_code(code Line 2179  for (code = first_significant_code(code
2179    
2180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2181      case OP_STAR:      case OP_STAR:
2182        case OP_STARI:
2183      case OP_MINSTAR:      case OP_MINSTAR:
2184        case OP_MINSTARI:
2185      case OP_POSSTAR:      case OP_POSSTAR:
2186        case OP_POSSTARI:
2187      case OP_QUERY:      case OP_QUERY:
2188        case OP_QUERYI:
2189      case OP_MINQUERY:      case OP_MINQUERY:
2190        case OP_MINQUERYI:
2191      case OP_POSQUERY:      case OP_POSQUERY:
2192        case OP_POSQUERYI:
2193        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2194        break;
2195    
2196      case OP_UPTO:      case OP_UPTO:
2197        case OP_UPTOI:
2198      case OP_MINUPTO:      case OP_MINUPTO:
2199        case OP_MINUPTOI:
2200      case OP_POSUPTO:      case OP_POSUPTO:
2201      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2202        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2203      break;      break;
2204  #endif  #endif
2205    
2206        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2207        string. */
2208    
2209        case OP_MARK:
2210        case OP_PRUNE_ARG:
2211        case OP_SKIP_ARG:
2212        code += code[1];
2213        break;
2214    
2215        case OP_THEN_ARG:
2216        code += code[1+LINK_SIZE];
2217        break;
2218    
2219        /* None of the remaining opcodes are required to match a character. */
2220    
2221        default:
2222        break;
2223      }      }
2224    }    }
2225    
# Line 1505  return TRUE; Line 2236  return TRUE;
2236  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
2237  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,
2238  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.
2239    This function is called only during the real compile, not during the
2240    pre-compile.
2241    
2242  Arguments:  Arguments:
2243    code        points to start of the recursion    code        points to start of the recursion
2244    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2245    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2246    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2247      cd          pointers to tables etc
2248    
2249  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2250  */  */
2251    
2252  static BOOL  static BOOL
2253  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2254    BOOL utf8)    BOOL utf8, compile_data *cd)
2255  {  {
2256  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2257    {    {
2258    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2259        return FALSE;
2260    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2261    }    }
2262  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2269  return TRUE;
2269  *************************************************/  *************************************************/
2270    
2271  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2272  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
2273  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2274  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2275    
2276    Originally, this function only recognized a sequence of letters between the
2277    terminators, but it seems that Perl recognizes any sequence of characters,
2278    though of course unknown POSIX names are subsequently rejected. Perl gives an
2279    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2280    didn't consider this to be a POSIX class. Likewise for [:1234:].
2281    
2282    The problem in trying to be exactly like Perl is in the handling of escapes. We
2283    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2284    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2285    below handles the special case of \], but does not try to do any other escape
2286    processing. This makes it different from Perl for cases such as [:l\ower:]
2287    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2288    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2289    I think.
2290    
2291  Argument:  Arguments:
2292    ptr      pointer to the initial [    ptr      pointer to the initial [
2293    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2294    
2295  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2296  */  */
2297    
2298  static BOOL  static BOOL
2299  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2300  {  {
2301  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2302  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2303  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2304    {    {
2305    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2306    return TRUE;      {
2307        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2308        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2309          {
2310          *endptr = ptr;
2311          return TRUE;
2312          }
2313        }
2314    }    }
2315  return FALSE;  return FALSE;
2316  }  }
# Line 1581  Returns:     a value representing the na Line 2335  Returns:     a value representing the na
2335  static int  static int
2336  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2337  {  {
2338    const char *pn = posix_names;
2339  register int yield = 0;  register int yield = 0;
2340  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2341    {    {
2342    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2343      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2344      pn += posix_name_lengths[yield] + 1;
2345    yield++;    yield++;
2346    }    }
2347  return -1;  return -1;
# Line 1600  return -1; Line 2356  return -1;
2356  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2357  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2358  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
2359  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
2360  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
2361  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
2362  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
2363  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2364    OP_END.
2365    
2366  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2367  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 1627  adjust_recurse(uschar *group, int adjust Line 2384  adjust_recurse(uschar *group, int adjust
2384    uschar *save_hwm)    uschar *save_hwm)
2385  {  {
2386  uschar *ptr = group;  uschar *ptr = group;
2387    
2388  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2389    {    {
2390    int offset;    int offset;
# Line 1680  auto_callout(uschar *code, const uschar Line 2438  auto_callout(uschar *code, const uschar
2438  {  {
2439  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2440  *code++ = 255;  *code++ = 255;
2441  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2442  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2443  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2444  }  }
2445    
# Line 1706  Returns:             nothing Line 2464  Returns:             nothing
2464  static void  static void
2465  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2466  {  {
2467  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2468  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2469  }  }
2470    
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2496  get_othercase_range(unsigned int *cptr,
2496  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2497    
2498  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2499    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2500    
2501  if (c > d) return FALSE;  if (c > d) return FALSE;
2502    
# Line 1747  next = othercase + 1; Line 2505  next = othercase + 1;
2505    
2506  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2507    {    {
2508    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2509    next++;    next++;
2510    }    }
2511    
# Line 1756  for (++c; c <= d; c++) Line 2514  for (++c; c <= d; c++)
2514    
2515  return TRUE;  return TRUE;
2516  }  }
2517    
2518    
2519    
2520    /*************************************************
2521    *        Check a character and a property        *
2522    *************************************************/
2523    
2524    /* This function is called by check_auto_possessive() when a property item
2525    is adjacent to a fixed character.
2526    
2527    Arguments:
2528      c            the character
2529      ptype        the property type
2530      pdata        the data for the type
2531      negated      TRUE if it's a negated property (\P or \p{^)
2532    
2533    Returns:       TRUE if auto-possessifying is OK
2534    */
2535    
2536    static BOOL
2537    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2538    {
2539    const ucd_record *prop = GET_UCD(c);
2540    switch(ptype)
2541      {
2542      case PT_LAMP:
2543      return (prop->chartype == ucp_Lu ||
2544              prop->chartype == ucp_Ll ||
2545              prop->chartype == ucp_Lt) == negated;
2546    
2547      case PT_GC:
2548      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2549    
2550      case PT_PC:
2551      return (pdata == prop->chartype) == negated;
2552    
2553      case PT_SC:
2554      return (pdata == prop->script) == negated;
2555    
2556      /* These are specials */
2557    
2558      case PT_ALNUM:
2559      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2560              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2561    
2562      case PT_SPACE:    /* Perl space */
2563      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2564              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2565              == negated;
2566    
2567      case PT_PXSPACE:  /* POSIX space */
2568      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2569              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2570              c == CHAR_FF || c == CHAR_CR)
2571              == negated;
2572    
2573      case PT_WORD:
2574      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2575              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2576              c == CHAR_UNDERSCORE) == negated;
2577      }
2578    return FALSE;
2579    }
2580  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2581    
2582    
# Line 1769  whether the next thing could possibly ma Line 2590  whether the next thing could possibly ma
2590  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2591    
2592  Arguments:  Arguments:
2593    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2594    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2595    ptr           next character in pattern    ptr           next character in pattern
2596    options       options bits    options       options bits
2597    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns:        TRUE if possessifying is Line 2600  Returns:        TRUE if possessifying is
2600  */  */
2601    
2602  static BOOL  static BOOL
2603  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2604    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2605  {  {
2606  int next;  int c, next;
2607    int op_code = *previous++;
2608    
2609  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2610    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2613  if ((options & PCRE_EXTENDED) != 0)
2613    for (;;)    for (;;)
2614      {      {
2615      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2616      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2617        {        {
2618        while (*(++ptr) != 0)        ptr++;
2619          while (*ptr != 0)
2620            {
2621          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2622            ptr++;
2623    #ifdef SUPPORT_UTF8
2624            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2625    #endif
2626            }
2627        }        }
2628      else break;      else break;
2629      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2632  if ((options & PCRE_EXTENDED) != 0)
2632  /* 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
2633  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2634    
2635  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2636    {    {
2637    int temperrorcode = 0;    int temperrorcode = 0;
2638    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2657  if ((options & PCRE_EXTENDED) != 0)
2657    for (;;)    for (;;)
2658      {      {
2659      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2660      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2661        {        {
2662        while (*(++ptr) != 0)        ptr++;
2663          while (*ptr != 0)
2664            {
2665          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2666            ptr++;
2667    #ifdef SUPPORT_UTF8
2668            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2669    #endif
2670            }
2671        }        }
2672      else break;      else break;
2673      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2675  if ((options & PCRE_EXTENDED) != 0)
2675    
2676  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2677    
2678  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2679    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2680        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. */  
   
2681    
2682  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2683    the next item is a character. */
2684    
2685  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2686    {    {
2687    case OP_CHAR:    case OP_CHAR:
2688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2689    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2690    #else
2691      c = *previous;
2692  #endif  #endif
2693    return item != next;    return c != next;
2694    
2695    /* 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
2696    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
2697    high-valued characters. */    high-valued characters. */
2698    
2699    case OP_CHARNC:    case OP_CHARI:
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2702    #else
2703      c = *previous;
2704  #endif  #endif
2705    if (item == next) return FALSE;    if (c == next) return FALSE;
2706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2707    if (utf8)    if (utf8)
2708      {      {
2709      unsigned int othercase;      unsigned int othercase;
2710      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2711  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2712      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2713  #else  #else
2714      othercase = NOTACHAR;      othercase = NOTACHAR;
2715  #endif  #endif
2716      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2717      }      }
2718    else    else
2719  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2720    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* 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
2723      opcodes are not used for multi-byte characters, because they are coded using
2724      an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2728    if (item == next) return TRUE;  
2729    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2730      if ((c = *previous) == next) return TRUE;
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
2733      {      {
2734      unsigned int othercase;      unsigned int othercase;
2735      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2736  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2737      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2738  #else  #else
2739      othercase = NOTACHAR;      othercase = NOTACHAR;
2740  #endif  #endif
2741      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2742      }      }
2743    else    else
2744  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2745    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2746    
2747      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2748      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2749    
2750    case OP_DIGIT:    case OP_DIGIT:
2751    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1925  if (next >= 0) switch(op_code) Line 2765  if (next >= 0) switch(op_code)
2765    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2766    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2767    
2768      case OP_HSPACE:
2769      case OP_NOT_HSPACE:
2770      switch(next)
2771        {
2772        case 0x09:
2773        case 0x20:
2774        case 0xa0:
2775        case 0x1680:
2776        case 0x180e:
2777        case 0x2000:
2778        case 0x2001:
2779        case 0x2002:
2780        case 0x2003:
2781        case 0x2004:
2782        case 0x2005:
2783        case 0x2006:
2784        case 0x2007:
2785        case 0x2008:
2786        case 0x2009:
2787        case 0x200A:
2788        case 0x202f:
2789        case 0x205f:
2790        case 0x3000:
2791        return op_code == OP_NOT_HSPACE;
2792        default:
2793        return op_code != OP_NOT_HSPACE;
2794        }
2795    
2796      case OP_ANYNL:
2797      case OP_VSPACE:
2798      case OP_NOT_VSPACE:
2799      switch(next)
2800        {
2801        case 0x0a:
2802        case 0x0b:
2803        case 0x0c:
2804        case 0x0d:
2805        case 0x85:
2806        case 0x2028:
2807        case 0x2029:
2808        return op_code == OP_NOT_VSPACE;
2809        default:
2810        return op_code != OP_NOT_VSPACE;
2811        }
2812    
2813    #ifdef SUPPORT_UCP
2814      case OP_PROP:
2815      return check_char_prop(next, previous[0], previous[1], FALSE);
2816    
2817      case OP_NOTPROP:
2818      return check_char_prop(next, previous[0], previous[1], TRUE);
2819    #endif
2820    
2821    default:    default:
2822    return FALSE;    return FALSE;
2823    }    }
2824    
2825    
2826  /* 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
2827    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2828    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2829    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2830    replaced by OP_PROP codes when PCRE_UCP is set. */
2831    
2832  switch(op_code)  switch(op_code)
2833    {    {
2834    case OP_CHAR:    case OP_CHAR:
2835    case OP_CHARNC:    case OP_CHARI:
2836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2837    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2838    #else
2839      c = *previous;
2840  #endif  #endif
2841    switch(-next)    switch(-next)
2842      {      {
2843      case ESC_d:      case ESC_d:
2844      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2845    
2846      case ESC_D:      case ESC_D:
2847      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2848    
2849      case ESC_s:      case ESC_s:
2850      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2851    
2852      case ESC_S:      case ESC_S:
2853      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2854    
2855      case ESC_w:      case ESC_w:
2856      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2857    
2858      case ESC_W:      case ESC_W:
2859      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2860    
2861        case ESC_h:
2862        case ESC_H:
2863        switch(c)
2864          {
2865          case 0x09:
2866          case 0x20:
2867          case 0xa0:
2868          case 0x1680:
2869          case 0x180e:
2870          case 0x2000:
2871          case 0x2001:
2872          case 0x2002:
2873          case 0x2003:
2874          case 0x2004:
2875          case 0x2005:
2876          case 0x2006:
2877          case 0x2007:
2878          case 0x2008:
2879          case 0x2009:
2880          case 0x200A:
2881          case 0x202f:
2882          case 0x205f:
2883          case 0x3000:
2884          return -next != ESC_h;
2885          default:
2886          return -next == ESC_h;
2887          }
2888    
2889        case ESC_v:
2890        case ESC_V:
2891        switch(c)
2892          {
2893          case 0x0a:
2894          case 0x0b:
2895          case 0x0c:
2896          case 0x0d:
2897          case 0x85:
2898          case 0x2028:
2899          case 0x2029:
2900          return -next != ESC_v;
2901          default:
2902          return -next == ESC_v;
2903          }
2904    
2905        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2906        their substitutions and process them. The result will always be either
2907        -ESC_p or -ESC_P. Then fall through to process those values. */
2908    
2909    #ifdef SUPPORT_UCP
2910        case ESC_du:
2911        case ESC_DU:
2912        case ESC_wu:
2913        case ESC_WU:
2914        case ESC_su:
2915        case ESC_SU:
2916          {
2917          int temperrorcode = 0;
2918          ptr = substitutes[-next - ESC_DU];
2919          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2920          if (temperrorcode != 0) return FALSE;
2921          ptr++;    /* For compatibility */
2922          }
2923        /* Fall through */
2924    
2925        case ESC_p:
2926        case ESC_P:
2927          {
2928          int ptype, pdata, errorcodeptr;
2929          BOOL negated;
2930    
2931          ptr--;      /* Make ptr point at the p or P */
2932          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2933          if (ptype < 0) return FALSE;
2934          ptr++;      /* Point past the final curly ket */
2935    
2936          /* If the property item is optional, we have to give up. (When generated
2937          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2938          to the original \d etc. At this point, ptr will point to a zero byte. */
2939    
2940          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2941            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2942              return FALSE;
2943    
2944          /* Do the property check. */
2945    
2946          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2947          }
2948    #endif
2949    
2950      default:      default:
2951      return FALSE;      return FALSE;
2952      }      }
2953    
2954      /* In principle, support for Unicode properties should be integrated here as
2955      well. It means re-organizing the above code so as to get hold of the property
2956      values before switching on the op-code. However, I wonder how many patterns
2957      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2958      these op-codes are never generated.) */
2959    
2960    case OP_DIGIT:    case OP_DIGIT:
2961    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2962             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2963    
2964    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2965    return next == -ESC_d;    return next == -ESC_d;
2966    
2967    case OP_WHITESPACE:    case OP_WHITESPACE:
2968    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2969    
2970    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2971    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2972    
2973      case OP_HSPACE:
2974      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2975             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2976    
2977      case OP_NOT_HSPACE:
2978      return next == -ESC_h;
2979    
2980      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2981      case OP_ANYNL:
2982      case OP_VSPACE:
2983      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2984    
2985      case OP_NOT_VSPACE:
2986      return next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_WORDCHAR:    case OP_WORDCHAR:
2989    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2990             next == -ESC_v || next == -ESC_R;
2991    
2992    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2993    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2040  BOOL inescq = FALSE; Line 3051  BOOL inescq = FALSE;
3051  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3052  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3053  const uschar *tempptr;  const uschar *tempptr;
3054    const uschar *nestptr = NULL;
3055  uschar *previous = NULL;  uschar *previous = NULL;
3056  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3057  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2049  uschar classbits[32]; Line 3061  uschar classbits[32];
3061  BOOL class_utf8;  BOOL class_utf8;
3062  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3063  uschar *class_utf8data;  uschar *class_utf8data;
3064    uschar *class_utf8data_base;
3065  uschar utf8_char[6];  uschar utf8_char[6];
3066  #else  #else
3067  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
3068  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3069  #endif  #endif
3070    
3071  #ifdef DEBUG  #ifdef PCRE_DEBUG
3072  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3073  #endif  #endif
3074    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 3101  req_caseopt = ((options & PCRE_CASELESS)
3101  for (;; ptr++)  for (;; ptr++)
3102    {    {
3103    BOOL negate_class;    BOOL negate_class;
3104      BOOL should_flip_negation;
3105    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3106    BOOL is_quantifier;    BOOL is_quantifier;
3107    BOOL is_recurse;    BOOL is_recurse;
3108      BOOL reset_bracount;
3109    int class_charcount;    int class_charcount;
3110    int class_lastchar;    int class_lastchar;
3111    int newoptions;    int newoptions;
3112    int recno;    int recno;
3113      int refsign;
3114    int skipbytes;    int skipbytes;
3115    int subreqbyte;    int subreqbyte;
3116    int subfirstbyte;    int subfirstbyte;
# Line 2106  for (;; ptr++) Line 3122  for (;; ptr++)
3122    
3123    c = *ptr;    c = *ptr;
3124    
3125      /* If we are at the end of a nested substitution, revert to the outer level
3126      string. Nesting only happens one level deep. */
3127    
3128      if (c == 0 && nestptr != NULL)
3129        {
3130        ptr = nestptr;
3131        nestptr = NULL;
3132        c = *ptr;
3133        }
3134    
3135    /* 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
3136    previous cycle of this loop. */    previous cycle of this loop. */
3137    
3138    if (lengthptr != NULL)    if (lengthptr != NULL)
3139      {      {
3140  #ifdef DEBUG  #ifdef PCRE_DEBUG
3141      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3142  #endif  #endif
3143      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3144        {        {
3145        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3146        goto FAILED;        goto FAILED;
# Line 2127  for (;; ptr++) Line 3153  for (;; ptr++)
3153      */      */
3154    
3155      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3156      *lengthptr += code - last_code;  
3157        /* Paranoid check for integer overflow */
3158    
3159        if (OFLOW_MAX - *lengthptr < code - last_code)
3160          {
3161          *errorcodeptr = ERR20;
3162          goto FAILED;
3163          }
3164    
3165        *lengthptr += (int)(code - last_code);
3166      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3167    
3168      /* 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
# Line 2154  for (;; ptr++) Line 3189  for (;; ptr++)
3189    /* 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
3190    reference list. */    reference list. */
3191    
3192    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3193      {      {
3194      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3195      goto FAILED;      goto FAILED;
# Line 2164  for (;; ptr++) Line 3199  for (;; ptr++)
3199    
3200    if (inescq && c != 0)    if (inescq && c != 0)
3201      {      {
3202      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3203        {        {
3204        inescq = FALSE;        inescq = FALSE;
3205        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 3225  for (;; ptr++)
3225    /* 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
3226    a quantifier. */    a quantifier. */
3227    
3228    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3229      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3230        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3231    
3232    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3233         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 3242  for (;; ptr++)
3242    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3243      {      {
3244      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3245      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3246        {        {
3247        while (*(++ptr) != 0)        ptr++;
3248          while (*ptr != 0)
3249          {          {
3250          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3251            ptr++;
3252    #ifdef SUPPORT_UTF8
3253            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3254    #endif
3255          }          }
3256        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3257    
# Line 2231  for (;; ptr++) Line 3272  for (;; ptr++)
3272      {      {
3273      /* ===================================================================*/      /* ===================================================================*/
3274      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3275      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3276      case ')':      case CHAR_RIGHT_PARENTHESIS:
3277      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3278      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3279      *codeptr = code;      *codeptr = code;
3280      *ptrptr = ptr;      *ptrptr = ptr;
3281      if (lengthptr != NULL)      if (lengthptr != NULL)
3282        {        {
3283        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3284            {
3285            *errorcodeptr = ERR20;
3286            goto FAILED;
3287            }
3288          *lengthptr += (int)(code - last_code);   /* To include callout length */
3289        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3290        }        }
3291      return TRUE;      return TRUE;
# Line 2249  for (;; ptr++) Line 3295  for (;; ptr++)
3295      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3296      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3297    
3298      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3299        previous = NULL;
3300      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3301        {        {
3302        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3303          *code++ = OP_CIRCM;
3304        }        }
3305      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3306      break;      break;
3307    
3308      case '$':      case CHAR_DOLLAR_SIGN:
3309      previous = NULL;      previous = NULL;
3310      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3311      break;      break;
3312    
3313      /* 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
3314      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3315    
3316      case '.':      case CHAR_DOT:
3317      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3318      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3319      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3320      previous = code;      previous = code;
3321      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3322      break;      break;
3323    
3324    
# Line 2286  for (;; ptr++) Line 3333  for (;; ptr++)
3333      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,
3334      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3335      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.
     */  
3336    
3337      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3338        default (Perl) mode, it is treated as a data character. */
3339    
3340        case CHAR_RIGHT_SQUARE_BRACKET:
3341        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3342          {
3343          *errorcodeptr = ERR64;
3344          goto FAILED;
3345          }
3346        goto NORMAL_CHAR;
3347    
3348        case CHAR_LEFT_SQUARE_BRACKET:
3349      previous = code;      previous = code;
3350    
3351      /* 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
3352      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. */
3353    
3354      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3355          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3356            check_posix_syntax(ptr, &tempptr))
3357        {        {
3358        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3359        goto FAILED;        goto FAILED;
3360        }        }
3361    
3362      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3363        if the first few characters (either before or after ^) are \Q\E or \E we
3364        skip them too. This makes for compatibility with Perl. */
3365    
3366      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3367        for (;;)
3368        {        {
       negate_class = TRUE;  
3369        c = *(++ptr);        c = *(++ptr);
3370          if (c == CHAR_BACKSLASH)
3371            {
3372            if (ptr[1] == CHAR_E)
3373              ptr++;
3374            else if (strncmp((const char *)ptr+1,
3375                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3376              ptr += 3;
3377            else
3378              break;
3379            }
3380          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3381            negate_class = TRUE;
3382          else break;
3383        }        }
3384      else  
3385        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3386        an initial ']' is taken as a data character -- the code below handles
3387        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3388        [^] must match any character, so generate OP_ALLANY. */
3389    
3390        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3391            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3392        {        {
3393        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3394          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3395          zerofirstbyte = firstbyte;
3396          break;
3397        }        }
3398    
3399        /* If a class contains a negative special such as \S, we need to flip the
3400        negation flag at the end, so that support for characters > 255 works
3401        correctly (they are all included in the class). */
3402    
3403        should_flip_negation = FALSE;
3404    
3405      /* 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
3406      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
3407      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 3419  for (;; ptr++)
3419  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3420      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3421      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3422        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3423  #endif  #endif
3424    
3425      /* 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 2345  for (;; ptr++) Line 3435  for (;; ptr++)
3435          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3436          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3437          }          }
3438    
3439          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3440          data and reset the pointer. This is so that very large classes that
3441          contain a zillion UTF-8 characters no longer overwrite the work space
3442          (which is on the stack). */
3443    
3444          if (lengthptr != NULL)
3445            {
3446            *lengthptr += class_utf8data - class_utf8data_base;
3447            class_utf8data = class_utf8data_base;
3448            }
3449    
3450  #endif  #endif
3451    
3452        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3453    
3454        if (inescq)        if (inescq)
3455          {          {
3456          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3457            {            {
3458            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3459            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3468  for (;; ptr++)
3468        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3469        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3470    
3471        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3472            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3473            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3474          {          {
3475          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3476          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3477          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3478          uschar pbits[32];          uschar pbits[32];
3479    
3480          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3481            {            {
3482            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3483            goto FAILED;            goto FAILED;
3484            }            }
3485    
3486          ptr += 2;          ptr += 2;
3487          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3488            {            {
3489            local_negate = TRUE;            local_negate = TRUE;
3490              should_flip_negation = TRUE;  /* Note negative special */
3491            ptr++;            ptr++;
3492            }            }
3493    
3494          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3495          if (posix_class < 0)          if (posix_class < 0)
3496            {            {
3497            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2402  for (;; ptr++) Line 3505  for (;; ptr++)
3505          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3506            posix_class = 0;            posix_class = 0;
3507    
3508          /* 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
3509          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3510          subtract bits that may be in the main map already. At the end we or the  
3511          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3512            if ((options & PCRE_UCP) != 0)
3513              {
3514              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3515              if (posix_substitutes[pc] != NULL)
3516                {
3517                nestptr = tempptr + 1;
3518                ptr = posix_substitutes[pc] - 1;
3519                continue;
3520                }
3521              }
3522    #endif
3523            /* In the non-UCP case, we build the bit map for the POSIX class in a
3524            chunk of local store because we may be adding and subtracting from it,
3525            and we don't want to subtract bits that may be in the main map already.
3526            At the end we or the result into the bit map that is being built. */
3527    
3528          posix_class *= 3;          posix_class *= 3;
3529    
# Line 2449  for (;; ptr++) Line 3567  for (;; ptr++)
3567    
3568        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3569        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
3570        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
3571        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
3572        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
3573        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3574          PCRE_EXTRA is set. */
3575    
3576        if (c == '\\')        if (c == CHAR_BACKSLASH)
3577          {          {
3578          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3579          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3580    
3581          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 */  
3582          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3583            {            {
3584            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3585              {              {
3586              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3587              }              }
3588            else inescq = TRUE;            else inescq = TRUE;
3589            continue;            continue;
3590            }            }
3591            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3592    
3593          if (c < 0)          if (c < 0)
3594            {            {
3595            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3596            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3597    
3598            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3599              {              {
3600    #ifdef SUPPORT_UCP
3601                case ESC_du:     /* These are the values given for \d etc */
3602                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3603                case ESC_wu:     /* escape sequence with an appropriate \p */
3604                case ESC_WU:     /* or \P to test Unicode properties instead */
3605                case ESC_su:     /* of the default ASCII testing. */
3606                case ESC_SU:
3607                nestptr = ptr;
3608                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3609                class_charcount -= 2;                /* Undo! */
3610                continue;
3611    #endif
3612              case ESC_d:              case ESC_d:
3613              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3614              continue;              continue;
3615    
3616              case ESC_D:              case ESC_D:
3617                should_flip_negation = TRUE;
3618              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3619              continue;              continue;
3620    
# Line 2494  for (;; ptr++) Line 3623  for (;; ptr++)
3623              continue;              continue;
3624    
3625              case ESC_W:              case ESC_W:
3626                should_flip_negation = TRUE;
3627              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3628              continue;              continue;
3629    
3630                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3631                if it was previously set by something earlier in the character
3632                class. */
3633    
3634              case ESC_s:              case ESC_s:
3635              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3636              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3637                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3638              continue;              continue;
3639    
3640              case ESC_S:              case ESC_S:
3641                should_flip_negation = TRUE;
3642              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3643              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3644              continue;              continue;
3645    
3646              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3647                SETBIT(classbits, 0x09); /* VT */
3648                SETBIT(classbits, 0x20); /* SPACE */
3649                SETBIT(classbits, 0xa0); /* NSBP */
3650    #ifdef SUPPORT_UTF8
3651                if (utf8)
3652                  {
3653                  class_utf8 = TRUE;
3654                  *class_utf8data++ = XCL_SINGLE;
3655                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3656                  *class_utf8data++ = XCL_SINGLE;
3657                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3658                  *class_utf8data++ = XCL_RANGE;
3659                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3660                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3661                  *class_utf8data++ = XCL_SINGLE;
3662                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3663                  *class_utf8data++ = XCL_SINGLE;
3664                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3665                  *class_utf8data++ = XCL_SINGLE;
3666                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3667                  }
3668    #endif
3669              continue;              continue;
3670    
3671              default:    /* Not recognized; fall through */              case ESC_H:
3672              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3673              }                {
3674                  int x = 0xff;
3675                  switch (c)
3676                    {
3677                    case 0x09/8: x ^= 1 << (0x09%8); break;
3678                    case 0x20/8: x ^= 1 << (0x20%8); break;
3679                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3680                    default: break;
3681                    }
3682                  classbits[c] |= x;
3683                  }
3684    
3685            /* In the pre-compile phase, just do the recognition. */  #ifdef SUPPORT_UTF8
3686                if (utf8)
3687                  {
3688                  class_utf8 = TRUE;
3689                  *class_utf8data++ = XCL_RANGE;
3690                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3691                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3692                  *class_utf8data++ = XCL_RANGE;
3693                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3694                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3695                  *class_utf8data++ = XCL_RANGE;
3696                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3697                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3698                  *class_utf8data++ = XCL_RANGE;
3699                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3700                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3701                  *class_utf8data++ = XCL_RANGE;
3702                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3703                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3704                  *class_utf8data++ = XCL_RANGE;
3705                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3706                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3707                  *class_utf8data++ = XCL_RANGE;
3708                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3709                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3710                  }
3711    #endif
3712                continue;
3713    
3714            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_v:
3715                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              SETBIT(classbits, 0x0a); /* LF */
3716                SETBIT(classbits, 0x0b); /* VT */
3717                SETBIT(classbits, 0x0c); /* FF */
3718                SETBIT(classbits, 0x0d); /* CR */
3719                SETBIT(classbits, 0x85); /* NEL */
3720    #ifdef SUPPORT_UTF8
3721                if (utf8)
3722                  {
3723                  class_utf8 = TRUE;
3724                  *class_utf8data++ = XCL_RANGE;
3725                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3726                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3727                  }
3728    #endif
3729                continue;
3730    
3731            /* We need to deal with \P and \p in both phases. */              case ESC_V:
3732                for (c = 0; c < 32; c++)
3733                  {
3734                  int x = 0xff;
3735                  switch (c)
3736                    {
3737                    case 0x0a/8: x ^= 1 << (0x0a%8);
3738                                 x ^= 1 << (0x0b%8);
3739                                 x ^= 1 << (0x0c%8);
3740                                 x ^= 1 << (0x0d%8);
3741                                 break;
3742                    case 0x85/8: x ^= 1 << (0x85%8); break;
3743                    default: break;
3744                    }
3745                  classbits[c] |= x;
3746                  }
3747    
3748  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UTF8
3749            if (-c == ESC_p || -c == ESC_P)              if (utf8)
3750              {                {
3751              BOOL negated;                class_utf8 = TRUE;
3752              int pdata;                *class_utf8data++ = XCL_RANGE;
3753              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3754              if (ptype < 0) goto FAILED;                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3755              class_utf8 = TRUE;                *class_utf8data++ = XCL_RANGE;
3756              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3757                XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3758              *class_utf8data++ = ptype;                }
3759              *class_utf8data++ = pdata;  #endif
             class_charcount -= 2;   /* Not a < 256 character */  
3760              continue;              continue;
3761              }  
3762    #ifdef SUPPORT_UCP
3763                case ESC_p:
3764                case ESC_P:
3765                  {
3766                  BOOL negated;
3767                  int pdata;
3768                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3769                  if (ptype < 0) goto FAILED;
3770                  class_utf8 = TRUE;
3771                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3772                    XCL_PROP : XCL_NOTPROP;
3773                  *class_utf8data++ = ptype;
3774                  *class_utf8data++ = pdata;
3775                  class_charcount -= 2;   /* Not a < 256 character */
3776                  continue;
3777                  }
3778  #endif  #endif
3779            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3780            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3781            treated as literals. */              treated as literals. */
3782    
3783            if ((options & PCRE_EXTRA) != 0)              default:
3784              {              if ((options & PCRE_EXTRA) != 0)
3785              *errorcodeptr = ERR7;                {
3786              goto FAILED;                *errorcodeptr = ERR7;
3787                  goto FAILED;
3788                  }
3789                class_charcount -= 2;  /* Undo the default count from above */
3790                c = *ptr;              /* Get the final character and fall through */
3791                break;
3792              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3793            }            }
3794    
3795          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 2562  for (;; ptr++) Line 3803  for (;; ptr++)
3803        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3804    
3805        CHECK_RANGE:        CHECK_RANGE:
3806        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3807          {          {
3808          inescq = FALSE;          inescq = FALSE;
3809          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3811  for (;; ptr++)
3811    
3812        oldptr = ptr;        oldptr = ptr;
3813    
3814        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3815    
3816          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3817    
3818          /* Check for range */
3819    
3820          if (!inescq && ptr[1] == CHAR_MINUS)
3821          {          {
3822          int d;          int d;
3823          ptr += 2;          ptr += 2;
3824          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3825    
3826          /* If we hit \Q (not followed by \E) at this point, go into escaped          /* If we hit \Q (not followed by \E) at this point, go into escaped
3827          mode. */          mode. */
3828    
3829          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3830            {            {
3831            ptr += 2;            ptr += 2;
3832            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3833                { ptr += 2; continue; }
3834            inescq = TRUE;            inescq = TRUE;
3835            break;            break;
3836            }            }
3837    
3838          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3839            {            {
3840            ptr = oldptr;            ptr = oldptr;
3841            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3854  for (;; ptr++)
3854          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3855          in such circumstances. */          in such circumstances. */
3856    
3857          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3858            {            {
3859            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3860            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3861    
3862            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3863    
3864            if (d < 0)            if (d < 0)
3865              {              {
3866              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X';  
             else if (d == -ESC_R) d = 'R'; else  
3867                {                {
3868                ptr = oldptr;                ptr = oldptr;
3869                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3882  for (;; ptr++)
3882    
3883          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3884    
3885            /* Remember \r or \n */
3886    
3887            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3888    
3889          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3890          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3891          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3908  for (;; ptr++)
3908              unsigned int origd = d;              unsigned int origd = d;
3909              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3910                {                {
3911                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3912                      ocd <= (unsigned int)d)
3913                    continue;                          /* Skip embedded ranges */
3914    
3915                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3916                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3917                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3918                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3919                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3920                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3921                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3922                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3923                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3924                  d = ocd;                  d = ocd;
3925                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 4004  for (;; ptr++)
4004          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4005            {            {
4006            unsigned int othercase;            unsigned int othercase;
4007            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4008              {              {
4009              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4010              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2776  for (;; ptr++) Line 4029  for (;; ptr++)
4029          }          }
4030        }        }
4031    
4032      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4033        If we are at the end of an internal nested string, revert to the outer
4034        string. */
4035    
4036        while (((c = *(++ptr)) != 0 ||
4037               (nestptr != NULL &&
4038                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4039               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4040    
4041      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4042    
4043      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4044        {        {
4045        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4046        goto FAILED;        goto FAILED;
4047        }        }
4048    
4049      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4050      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
4051      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
4052      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4053      single-bytes only. This is an historical hangover. Maybe one day we can  
4054      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4055        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4056        operate on single-bytes characters only. This is an historical hangover.
4057        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4058    
4059      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4060      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4061      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4062      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4063      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4064      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4065    
4066  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4067      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4068            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4069  #else  #else
4070      if (class_charcount == 1)      if (class_charcount == 1)
4071  #endif  #endif
4072        {        {
4073        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4074    
4075        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4076    
4077        if (negate_class)        if (negate_class)
4078          {          {
4079          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4080          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4081          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4082          *code++ = class_lastchar;          *code++ = class_lastchar;
4083          break;          break;
4084          }          }
# Line 2847  for (;; ptr++) Line 4108  for (;; ptr++)
4108      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4109    
4110      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4111      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4112      we can omit the bitmap in the actual compiled code. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
4113        characters > 255 are in the class, so any that were explicitly given as
4114        well can be ignored. If (when there are explicit characters > 255 that must
4115        be listed) there are no characters < 256, we can omit the bitmap in the
4116        actual compiled code. */
4117    
4118  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4119      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4120        {        {
4121        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4122        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 4142  for (;; ptr++)
4142        }        }
4143  #endif  #endif
4144    
4145      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
4146      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4147      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
4148      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4149        negating it if necessary. */
4150    
4151        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4152      if (negate_class)      if (negate_class)
4153        {        {
       *code++ = OP_NCLASS;  
4154        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4155          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4156        }        }
4157      else      else
4158        {        {
       *code++ = OP_CLASS;  
4159        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4160        }        }
4161      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 4166  for (;; ptr++)
4166      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4167      has been tested above. */      has been tested above. */
4168    
4169      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4170      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4171      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4172      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4173      goto REPEAT;      goto REPEAT;
4174    
4175      case '*':      case CHAR_ASTERISK:
4176      repeat_min = 0;      repeat_min = 0;
4177      repeat_max = -1;      repeat_max = -1;
4178      goto REPEAT;      goto REPEAT;
4179    
4180      case '+':      case CHAR_PLUS:
4181      repeat_min = 1;      repeat_min = 1;
4182      repeat_max = -1;      repeat_max = -1;
4183      goto REPEAT;      goto REPEAT;
4184    
4185      case '?':      case CHAR_QUESTION_MARK:
4186      repeat_min = 0;      repeat_min = 0;
4187      repeat_max = 1;      repeat_max = 1;
4188    
# Line 2952  for (;; ptr++) Line 4217  for (;; ptr++)
4217      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
4218      repeat type to the non-default. */      repeat type to the non-default. */
4219    
4220      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4221        {        {
4222        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4223        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4224        ptr++;        ptr++;
4225        }        }
4226      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4227        {        {
4228        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4229        ptr++;        ptr++;
4230        }        }
4231      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4232    
4233        /* If previous was a recursion call, wrap it in atomic brackets so that
4234        previous becomes the atomic group. All recursions were so wrapped in the
4235        past, but it no longer happens for non-repeated recursions. In fact, the
4236        repeated ones could be re-implemented independently so as not to need this,
4237        but for the moment we rely on the code for repeating groups. */
4238    
4239        if (*previous == OP_RECURSE)
4240          {
4241          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4242          *previous = OP_ONCE;
4243          PUT(previous, 1, 2 + 2*LINK_SIZE);
4244          previous[2 + 2*LINK_SIZE] = OP_KET;
4245          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4246          code += 2 + 2 * LINK_SIZE;
4247          length_prevgroup = 3 + 3*LINK_SIZE;
4248    
4249          /* When actually compiling, we need to check whether this was a forward
4250          reference, and if so, adjust the offset. */
4251    
4252          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4253            {
4254            int offset = GET(cd->hwm, -LINK_SIZE);
4255            if (offset == previous + 1 - cd->start_code)
4256              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4257            }
4258          }
4259    
4260        /* Now handle repetition for the different types of item. */
4261    
4262      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4263      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 2971  for (;; ptr++) Line 4265  for (;; ptr++)
4265      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4266      instead.  */      instead.  */
4267    
4268      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4269        {        {
4270          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4271    
4272        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4273        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4274        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3005  for (;; ptr++) Line 4301  for (;; ptr++)
4301    
4302        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4303            repeat_max < 0 &&            repeat_max < 0 &&
4304            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4305          {          {
4306          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4307          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3018  for (;; ptr++) Line 4313  for (;; ptr++)
4313      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4314      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4315      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4316      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4317      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4318    
4319      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4320        {        {
4321        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4322        c = previous[1];        c = previous[1];
4323        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4324            repeat_max < 0 &&            repeat_max < 0 &&
4325            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4326          {          {
4327          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4328          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3051  for (;; ptr++) Line 4346  for (;; ptr++)
4346    
4347        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4348            repeat_max < 0 &&            repeat_max < 0 &&
4349            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4350          {          {
4351          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4352          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3073  for (;; ptr++) Line 4368  for (;; ptr++)
4368    
4369        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4370    
4371          /*--------------------------------------------------------------------*/
4372          /* This code is obsolete from release 8.00; the restriction was finally
4373          removed: */
4374    
4375        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4376        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4377    
4378        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4379          /*--------------------------------------------------------------------*/
4380    
4381        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4382    
# Line 3215  for (;; ptr++) Line 4515  for (;; ptr++)
4515  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4516               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4517  #endif  #endif
4518               *previous == OP_REF)               *previous == OP_REF ||
4519                 *previous == OP_REFI)
4520        {        {
4521        if (repeat_max == 0)        if (repeat_max == 0)
4522          {          {
# Line 3223  for (;; ptr++) Line 4524  for (;; ptr++)
4524          goto END_REPEAT;          goto END_REPEAT;
4525          }          }
4526    
4527          /*--------------------------------------------------------------------*/
4528          /* This code is obsolete from release 8.00; the restriction was finally
4529          removed: */
4530    
4531        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4532        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4533    
4534        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4535          /*--------------------------------------------------------------------*/
4536    
4537        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4538          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3244  for (;; ptr++) Line 4550  for (;; ptr++)
4550        }        }
4551    
4552      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4553      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4554        KET opcodes, as this is the place where they get converted into the more
4555        special varieties. */
4556    
4557      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4558               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4559        {