/[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 223 by ph10, Mon Aug 20 11:07:53 2007 UTC revision 723 by ph10, Sat Oct 8 15:55:23 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185    searched linearly. Put all the names into a single string, in order to reduce
186    the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    const char *name;    int   len;                 /* Length of verb name */
192    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static verbitem verbs[] = {  static const char verbnames[] =
197    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
198    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
199    { "F",      1, OP_FAIL },    STRING_ACCEPT0
200    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
201    { "PRUNE",  5, OP_PRUNE },    STRING_F0
202    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
203    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
220    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
221    
222  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
223    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
225    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const 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 200  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const 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 207  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 is 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",    /** DEAD **/    "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"  /** DEAD **/
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized",    "(*VERB) not recognized\0"
403    "number is too big"    "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 303  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 339  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 414  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, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      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    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 458  ptr--;                            /* Set Line 649  ptr--;                            /* Set
649    
650  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
651    
652  /* 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
653  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.
654  Otherwise further processing may be required. */  Otherwise further processing may be required. */
655    
656  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
657  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
658  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
659    
660  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
661  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
662  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
663  #endif  #endif
664    
# Line 483  else Line 674  else
674      /* 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
675      error. */      error. */
676    
677      case 'l':      case CHAR_l:
678      case 'L':      case CHAR_L:
679      case 'N':      case CHAR_u:
680      case 'u':      case CHAR_U:
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by a number, either plain or braced. If positive, it      /* In a character class, \g is just a literal "g". Outside a character
685      is an absolute backreference. If negative, it is a relative backreference.      class, \g must be followed by one of a number of specific things:
686      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a  
687      reference to a named group. This is part of Perl's movement towards a      (1) A number, either plain or braced. If positive, it is an absolute
688      unified syntax for back references. As this is synonymous with \k{name}, we      backreference. If negative, it is a relative backreference. This is a Perl
689      fudge it up by pretending it really was \k. */      5.10 feature.
690    
691        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
692        is part of Perl's movement towards a unified syntax for back references. As
693        this is synonymous with \k{name}, we fudge it up by pretending it really
694        was \k.
695    
696        (3) For Oniguruma compatibility we also support \g followed by a name or a
697        number either in angle brackets or in single quotes. However, these are
698        (possibly recursive) subroutine calls, _not_ backreferences. Just return
699        the -ESC_g code (cf \k). */
700    
701        case CHAR_g:
702        if (isclass) break;
703        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704          {
705          c = -ESC_g;
706          break;
707          }
708    
709        /* Handle the Perl-compatible cases */
710    
711      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
712        {        {
713        const uschar *p;        const uschar *p;
714        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
715          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
716        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
717          {          {
718          c = -ESC_k;          c = -ESC_k;
719          break;          break;
# Line 514  else Line 723  else
723        }        }
724      else braced = FALSE;      else braced = FALSE;
725    
726      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
727        {        {
728        negated = TRUE;        negated = TRUE;
729        ptr++;        ptr++;
# Line 523  else Line 732  else
732    
733      c = 0;      c = 0;
734      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
735        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
736    
737      if (c < 0)      if (c < 0)   /* Integer overflow */
738        {        {
739        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
740        break;        break;
741        }        }
742    
743      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
744        {        {
745        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
746        break;        break;
747        }        }
748    
749        if (c == 0)
750          {
751          *errorcodeptr = ERR58;
752          break;
753          }
754    
755      if (negated)      if (negated)
756        {        {
757        if (c > bracount)        if (c > bracount)
# Line 562  else Line 777  else
777      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
778      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
779    
780      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:
781      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
782    
783      if (!isclass)      if (!isclass)
784        {        {
785        oldptr = ptr;        oldptr = ptr;
786        c -= '0';        c -= CHAR_0;
787        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
788          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
789        if (c < 0)        if (c < 0)    /* Integer overflow */
790          {          {
791          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
792          break;          break;
# Line 588  else Line 803  else
803      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.
804      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
805    
806      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
807        {        {
808        ptr--;        ptr--;
809        c = 0;        c = 0;
# Line 601  else Line 816  else
816      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
817      than 3 octal digits. */      than 3 octal digits. */
818    
819      case '0':      case CHAR_0:
820      c -= '0';      c -= CHAR_0;
821      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
822          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
823      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
824      break;      break;
825    
# Line 612  else Line 827  else
827      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
828      treated as a data character. */      treated as a data character. */
829    
830      case 'x':      case CHAR_x:
831      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
832        {        {
833        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
834        int count = 0;        int count = 0;
# Line 622  else Line 837  else
837        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
838          {          {
839          register int cc = *pt++;          register int cc = *pt++;
840          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
841          count++;          count++;
842    
843  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
844          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
845          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
846  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
847          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
848          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
849  #endif  #endif
850          }          }
851    
852        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
853          {          {
854          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
855          ptr = pt;          ptr = pt;
# Line 650  else Line 865  else
865      c = 0;      c = 0;
866      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
867        {        {
868        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
869        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
870  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
872        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
874        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
875        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876  #endif  #endif
877        }        }
878      break;      break;
879    
880      /* 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.
881      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
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case 'c':      case CHAR_c:
886      c = *(++ptr);      c = *(++ptr);
887      if (c == 0)      if (c == 0)
888        {        {
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894      if (c >= 'a' && c <= 'z') c -= 32;        {
895          *errorcodeptr = ERR68;
896          break;
897          }
898        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
904      break;      break;
905    
906      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
907      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
908      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
909      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
910      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
911    
912      default:      default:
913      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 920  else
920      }      }
921    }    }
922    
923    /* Perl supports \N{name} for character names, as well as plain \N for "not
924    newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929      *errorcodeptr = ERR37;
930    
931    /* If PCRE_UCP is set, we change the values for \d etc. */
932    
933    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
934      c -= (ESC_DU - ESC_D);
935    
936    /* Set the pointer to the final character before returning. */
937    
938  *ptrptr = ptr;  *ptrptr = ptr;
939  return c;  return c;
940  }  }
# Line 740  if (c == 0) goto ERROR_RETURN; Line 975  if (c == 0) goto ERROR_RETURN;
975  /* \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
976  negation. */  negation. */
977    
978  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
979    {    {
980    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
981      {      {
982      *negptr = TRUE;      *negptr = TRUE;
983      ptr++;      ptr++;
# Line 751  if (c == '{') Line 986  if (c == '{')
986      {      {
987      c = *(++ptr);      c = *(++ptr);
988      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
989      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
990      name[i] = c;      name[i] = c;
991      }      }
992    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
993    name[i] = 0;    name[i] = 0;
994    }    }
995    
# Line 776  top = _pcre_utt_size; Line 1011  top = _pcre_utt_size;
1011  while (bot < top)  while (bot < top)
1012    {    {
1013    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1014    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1015    if (c == 0)    if (c == 0)
1016      {      {
1017      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 800  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 860  int max = -1; Line 1062  int max = -1;
1062  /* 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
1063  an integer overflow. */  an integer overflow. */
1064    
1065  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1066  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1067    {    {
1068    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1072  if (min < 0 || min > 65535)
1072  /* 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.
1073  Also, max must not be less than min. */  Also, max must not be less than min. */
1074    
1075  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1076    {    {
1077    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1078      {      {
1079      max = 0;      max = 0;
1080      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1081      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1082        {        {
1083        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1102  return p;
1102    
1103    
1104  /*************************************************  /*************************************************
1105  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1106  *************************************************/  *************************************************/
1107    
1108  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1109    top-level call starts at the beginning of the pattern. All other calls must
1110    start at a parenthesis. It scans along a pattern's text looking for capturing
1111  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
1112  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
1113  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
1114  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  be terminated by '>' because that is checked in the first pass.  
1116    This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1131    count        current count of capturing parens so far encountered    cd           compile background data
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136      count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1139  */  */
1140    
1141  static int  static int
1142  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,
1143    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  const uschar *thisname;  uschar *ptr = *ptrptr;
1146    int start_count = *count;
1147    int hwm_count = start_count;
1148    BOOL dup_parens = FALSE;
1149    
1150    /* If the first character is a parenthesis, check on the type of group we are
1151    dealing with. The very first call may not start with a parenthesis. */
1152    
1153  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156    
1157      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159      /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161      else if (ptr[1] != CHAR_QUESTION_MARK)
1162        {
1163        *count += 1;
1164        if (name == NULL && *count == lorn) return *count;
1165        ptr++;
1166        }
1167    
1168      /* All cases now have (? at the start. Remember when we are in a group
1169      where the parenthesis numbers are duplicated. */
1170    
1171      else if (ptr[2] == CHAR_VERTICAL_LINE)
1172        {
1173        ptr += 3;
1174        dup_parens = TRUE;
1175        }
1176    
1177      /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179      else if (ptr[2] == CHAR_NUMBER_SIGN)
1180        {
1181        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182        goto FAIL_EXIT;
1183        }
1184    
1185      /* Handle a condition. If it is an assertion, just carry on so that it
1186      is processed as normal. If not, skip to the closing parenthesis of the
1187      condition (there can't be any nested parens). */
1188    
1189      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1190        {
1191        ptr += 2;
1192        if (ptr[1] != CHAR_QUESTION_MARK)
1193          {
1194          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1195          if (*ptr != 0) ptr++;
1196          }
1197        }
1198    
1199      /* Start with (? but not a condition. */
1200    
1201      else
1202        {
1203        ptr += 2;
1204        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1205    
1206        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1207    
1208        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1209            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1210          {
1211          int term;
1212          const uschar *thisname;
1213          *count += 1;
1214          if (name == NULL && *count == lorn) return *count;
1215          term = *ptr++;
1216          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1217          thisname = ptr;
1218          while (*ptr != term) ptr++;
1219          if (name != NULL && lorn == ptr - thisname &&
1220              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1221            return *count;
1222          term++;
1223          }
1224        }
1225      }
1226    
1227    /* Past any initial parenthesis handling, scan for parentheses or vertical
1228    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232    for (; ptr < cd->end_pattern; ptr++)
1233      {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
1236    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1237      {      {
1238      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1239      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1240        {        {
1241        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1242        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1243        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1244        }        }
1245      continue;      continue;
1246      }      }
1247    
1248    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1249      are handled for real. If the first character is '^', skip it. Also, if the
1250      first few characters (either before or after ^) are \Q\E or \E we skip them
1251      too. This makes for compatibility with Perl. Note the use of STR macros to
1252      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1253    
1254    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1255      {      {
1256      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1257        for (;;)
1258          {
1259          if (ptr[1] == CHAR_BACKSLASH)
1260            {
1261            if (ptr[2] == CHAR_E)
1262              ptr+= 2;
1263            else if (strncmp((const char *)ptr+2,
1264                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1265              ptr += 4;
1266            else
1267              break;
1268            }
1269          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1270            {
1271            negate_class = TRUE;
1272            ptr++;
1273            }
1274          else break;
1275          }
1276    
1277        /* If the next character is ']', it is a data character that must be
1278        skipped, except in JavaScript compatibility mode. */
1279    
1280        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1281            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1282          ptr++;
1283    
1284        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1285        {        {
1286        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1287        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1288          {          {
1289          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1290          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1291            {            {
1292            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1293            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1294            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1295            }            }
1296          continue;          continue;
1297          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1301  for (; *ptr != 0; ptr++)
1301    
1302    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1303    
1304    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1307      if (*ptr == 0) return -1;      while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315        if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
1318    
1319    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1320    
1321    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1322      {      {
1323      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1325      continue;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
1327    
1328    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1330        if (dup_parens && *count < hwm_count) *count = hwm_count;
1331        goto FAIL_EXIT;
1332        }
1333    
1334    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1335        {
1336        if (*count > hwm_count) hwm_count = *count;
1337        *count = start_count;
1338        }
1339      }
1340    
1341    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1342         *ptr != '\'')  *ptrptr = ptr;
1343      continue;  return -1;
1344    }
1345    
1346    
1347    
1348    
1349    /*************************************************
1350    *       Find forward referenced subpattern       *
1351    *************************************************/
1352    
1353    /* This function scans along a pattern's text looking for capturing
1354    subpatterns, and counting them. If it finds a named pattern that matches the
1355    name it is given, it returns its number. Alternatively, if the name is NULL, it
1356    returns when it reaches a given numbered subpattern. This is used for forward
1357    references to subpatterns. We used to be able to start this scan from the
1358    current compiling point, using the current count value from cd->bracount, and
1359    do it all in a single loop, but the addition of the possibility of duplicate
1360    subpattern numbers means that we have to scan from the very start, in order to
1361    take account of such duplicates, and to use a recursive function to keep track
1362    of the different types of group.
1363    
1364    Arguments:
1365      cd           compile background data
1366      name         name to seek, or NULL if seeking a numbered subpattern
1367      lorn         name length, or subpattern number if name is NULL
1368      xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371    count++;  Returns:       the number of the found subpattern, or -1 if not found
1372    */
1373    
1374    if (name == NULL && count == lorn) return count;  static int
1375    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376    if (term == '<') term = '>';    BOOL utf8)
1377    thisname = ptr;  {
1378    while (*ptr != term) ptr++;  uschar *ptr = (uschar *)cd->start_pattern;
1379    if (name != NULL && lorn == ptr - thisname &&  int count = 0;
1380        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  int rc;
1381      return count;  
1382    /* If the pattern does not start with an opening parenthesis, the first call
1383    to find_parens_sub() will scan right to the end (if necessary). However, if it
1384    does start with a parenthesis, find_parens_sub() will return when it hits the
1385    matching closing parens. That is why we have to have a loop. */
1386    
1387    for (;;)
1388      {
1389      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390      if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
1393  return -1;  return rc;
1394  }  }
1395    
1396    
1397    
1398    
1399  /*************************************************  /*************************************************
1400  *      Find first significant op code            *  *      Find first significant op code            *
1401  *************************************************/  *************************************************/
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  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
1405  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
1406  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
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    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  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1060  for (;;) Line 1435  for (;;)
1435    
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1443      break;      break;
# Line 1076  for (;;) Line 1453  for (;;)
1453    
1454    
1455  /*************************************************  /*************************************************
1456  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1457  *************************************************/  *************************************************/
1458    
1459  /* 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,
1460  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.
1461  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
1462    temporarily terminated with OP_END when this function is called.
1463    
1464    This function is called when a backward assertion is encountered, so that if it
1465    fails, the error message can point to the correct place in the pattern.
1466    However, we cannot do this when the assertion contains subroutine calls,
1467    because they can be forward references. We solve this by remembering this case
1468    and doing the check at the end; a flag specifies which mode we are running in.
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473      atend    TRUE if called when the pattern is complete
1474      cd       the "compile data" structure
1475    
1476  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1477                 or -1 if there is no fixed length,
1478               or -2 if \C was encountered               or -2 if \C was encountered
1479                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1105  branch, check the length against that of Line 1493  branch, check the length against that of
1493  for (;;)  for (;;)
1494    {    {
1495    int d;    int d;
1496      uschar *ce, *cs;
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509        case OP_ONCE_NC:
1510      case OP_COND:      case OP_COND:
1511      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1512      if (d < 0) return d;      if (d < 0) return d;
1513      branchlength += d;      branchlength += d;
1514      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1121  for (;;) Line 1517  for (;;)
1517    
1518      /* 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
1519      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
1520      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.
1521        Note that we must not include the OP_KETRxxx opcodes here, because they
1522        all imply an unlimited repeat. */
1523    
1524      case OP_ALT:      case OP_ALT:
1525      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1526      case OP_END:      case OP_END:
1527      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1528        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1135  for (;;) Line 1531  for (;;)
1531      branchlength = 0;      branchlength = 0;
1532      break;      break;
1533    
1534        /* A true recursion implies not fixed length, but a subroutine call may
1535        be OK. If the subroutine is a forward reference, we can't deal with
1536        it until the end of the pattern, so return -3. */
1537    
1538        case OP_RECURSE:
1539        if (!atend) return -3;
1540        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1541        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1542        if (cc > cs && cc < ce) return -1;                /* Recursion */
1543        d = find_fixedlength(cs + 2, utf8, atend, cd);
1544        if (d < 0) return d;
1545        branchlength += d;
1546        cc += 1 + LINK_SIZE;
1547        break;
1548    
1549      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1550    
1551      case OP_ASSERT:      case OP_ASSERT:
# Line 1148  for (;;) Line 1559  for (;;)
1559    
1560      case OP_REVERSE:      case OP_REVERSE:
1561      case OP_CREF:      case OP_CREF:
1562        case OP_NCREF:
1563      case OP_RREF:      case OP_RREF:
1564        case OP_NRREF:
1565      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1566      case OP_CALLOUT:      case OP_CALLOUT:
1567      case OP_SOD:      case OP_SOD:
1568      case OP_SOM:      case OP_SOM:
1569        case OP_SET_SOM:
1570      case OP_EOD:      case OP_EOD:
1571      case OP_EODN:      case OP_EODN:
1572      case OP_CIRC:      case OP_CIRC:
1573        case OP_CIRCM:
1574      case OP_DOLL:      case OP_DOLL:
1575        case OP_DOLLM:
1576      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1577      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1578      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1166  for (;;) Line 1581  for (;;)
1581      /* Handle literal characters */      /* Handle literal characters */
1582    
1583      case OP_CHAR:      case OP_CHAR:
1584      case OP_CHARNC:      case OP_CHARI:
1585      case OP_NOT:      case OP_NOT:
1586        case OP_NOTI:
1587      branchlength++;      branchlength++;
1588      cc += 2;      cc += 2;
1589  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1590      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1591  #endif  #endif
1592      break;      break;
1593    
# Line 1185  for (;;) Line 1598  for (;;)
1598      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1599      cc += 4;      cc += 4;
1600  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1601      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1602  #endif  #endif
1603      break;      break;
1604    
# Line 1212  for (;;) Line 1622  for (;;)
1622      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1623      case OP_WORDCHAR:      case OP_WORDCHAR:
1624      case OP_ANY:      case OP_ANY:
1625        case OP_ALLANY:
1626      branchlength++;      branchlength++;
1627      cc++;      cc++;
1628      break;      break;
# Line 1266  for (;;) Line 1677  for (;;)
1677    
1678    
1679  /*************************************************  /*************************************************
1680  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1681  *************************************************/  *************************************************/
1682    
1683  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1684  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1685    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1686    so that it can be called from pcre_study() when finding the minimum matching
1687    length.
1688    
1689  Arguments:  Arguments:
1690    code        points to start of expression    code        points to start of expression
1691    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1692    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1693    
1694  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
1695  */  */
1696    
1697  static const uschar *  const uschar *
1698  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1699  {  {
1700  for (;;)  for (;;)
1701    {    {
1702    register int c = *code;    register int c = *code;
1703    
1704    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1705    
1706    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1294  for (;;) Line 1709  for (;;)
1709    
1710    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1711    
1712      /* Handle recursion */
1713    
1714      else if (c == OP_REVERSE)
1715        {
1716        if (number < 0) return (uschar *)code;
1717        code += _pcre_OP_lengths[c];
1718        }
1719    
1720    /* Handle capturing bracket */    /* Handle capturing bracket */
1721    
1722    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1723               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1724      {      {
1725      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1726      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1305  for (;;) Line 1729  for (;;)
1729    
1730    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1731    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1732    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1733      must add in its length. */
1734    
1735    else    else
1736      {      {
# Line 1329  for (;;) Line 1754  for (;;)
1754        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1755        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1756        break;        break;
1757    
1758          case OP_MARK:
1759          case OP_PRUNE_ARG:
1760          case OP_SKIP_ARG:
1761          code += code[1];
1762          break;
1763    
1764          case OP_THEN_ARG:
1765          code += code[1];
1766          break;
1767        }        }
1768    
1769      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1343  for (;;) Line 1778  for (;;)
1778      if (utf8) switch(c)      if (utf8) switch(c)
1779        {        {
1780        case OP_CHAR:        case OP_CHAR:
1781        case OP_CHARNC:        case OP_CHARI:
1782        case OP_EXACT:        case OP_EXACT:
1783          case OP_EXACTI:
1784        case OP_UPTO:        case OP_UPTO:
1785          case OP_UPTOI:
1786        case OP_MINUPTO:        case OP_MINUPTO:
1787          case OP_MINUPTOI:
1788        case OP_POSUPTO:        case OP_POSUPTO:
1789          case OP_POSUPTOI:
1790        case OP_STAR:        case OP_STAR:
1791          case OP_STARI:
1792        case OP_MINSTAR:        case OP_MINSTAR:
1793          case OP_MINSTARI:
1794        case OP_POSSTAR:        case OP_POSSTAR:
1795          case OP_POSSTARI:
1796        case OP_PLUS:        case OP_PLUS:
1797          case OP_PLUSI:
1798        case OP_MINPLUS:        case OP_MINPLUS:
1799          case OP_MINPLUSI:
1800        case OP_POSPLUS:        case OP_POSPLUS:
1801          case OP_POSPLUSI:
1802        case OP_QUERY:        case OP_QUERY:
1803          case OP_QUERYI:
1804        case OP_MINQUERY:        case OP_MINQUERY:
1805          case OP_MINQUERYI:
1806        case OP_POSQUERY:        case OP_POSQUERY:
1807          case OP_POSQUERYI:
1808        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1809        break;        break;
1810        }        }
1811    #else
1812        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1813  #endif  #endif
1814      }      }
1815    }    }
# Line 1398  for (;;) Line 1848  for (;;)
1848    
1849    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1850    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1851    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1852      must add in its length. */
1853    
1854    else    else
1855      {      {
# Line 1422  for (;;) Line 1873  for (;;)
1873        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1874        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1875        break;        break;
1876    
1877          case OP_MARK:
1878          case OP_PRUNE_ARG:
1879          case OP_SKIP_ARG:
1880          code += code[1];
1881          break;
1882    
1883          case OP_THEN_ARG:
1884          code += code[1];
1885          break;
1886        }        }
1887    
1888      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1436  for (;;) Line 1897  for (;;)
1897      if (utf8) switch(c)      if (utf8) switch(c)
1898        {        {
1899        case OP_CHAR:        case OP_CHAR:
1900        case OP_CHARNC:        case OP_CHARI:
1901        case OP_EXACT:        case OP_EXACT:
1902          case OP_EXACTI:
1903        case OP_UPTO:        case OP_UPTO:
1904          case OP_UPTOI:
1905        case OP_MINUPTO:        case OP_MINUPTO:
1906          case OP_MINUPTOI:
1907        case OP_POSUPTO:        case OP_POSUPTO:
1908          case OP_POSUPTOI:
1909        case OP_STAR:        case OP_STAR:
1910          case OP_STARI:
1911        case OP_MINSTAR:        case OP_MINSTAR:
1912          case OP_MINSTARI:
1913        case OP_POSSTAR:        case OP_POSSTAR:
1914          case OP_POSSTARI:
1915        case OP_PLUS:        case OP_PLUS:
1916          case OP_PLUSI:
1917        case OP_MINPLUS:        case OP_MINPLUS:
1918          case OP_MINPLUSI:
1919        case OP_POSPLUS:        case OP_POSPLUS:
1920          case OP_POSPLUSI:
1921        case OP_QUERY:        case OP_QUERY:
1922          case OP_QUERYI:
1923        case OP_MINQUERY:        case OP_MINQUERY:
1924          case OP_MINQUERYI:
1925        case OP_POSQUERY:        case OP_POSQUERY:
1926          case OP_POSQUERYI:
1927        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1928        break;        break;
1929        }        }
1930    #else
1931        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1932  #endif  #endif
1933      }      }
1934    }    }
# Line 1468  for (;;) Line 1944  for (;;)
1944  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()
1945  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
1946  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1947  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
1948  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
1949    bracket whose current branch will already have been scanned.
1950    
1951  Arguments:  Arguments:
1952    code        points to start of search    code        points to start of search
1953    endcode     points to where to stop    endcode     points to where to stop
1954    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1955      cd          contains pointers to tables etc.
1956    
1957  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1958  */  */
1959    
1960  static BOOL  static BOOL
1961  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1962      compile_data *cd)
1963  {  {
1964  register int c;  register int c;
1965  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1966       code < endcode;       code < endcode;
1967       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1968    {    {
1969    const uschar *ccode;    const uschar *ccode;
1970    
1971    c = *code;    c = *code;
1972    
1973    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
1974      first_significant_code() with a TRUE final argument. */
1975    
1976    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
1977      {      {
     code += _pcre_OP_lengths[c];  
1978      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1979      c = *code;      c = *code;
1980      continue;      continue;
1981      }      }
1982    
1983    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
1984      implies a backward reference subroutine call, we can scan it. If it's a
1985      forward reference subroutine call, we can't. To detect forward reference
1986      we have to scan up the list that is kept in the workspace. This function is
1987      called only when doing the real compile, not during the pre-compile that
1988      measures the size of the compiled pattern. */
1989    
1990    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
1991      {      {
1992        const uschar *scode;
1993      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
1994    
1995      /* Scan a closed bracket */      /* Test for forward reference */
1996    
1997        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1998          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1999    
2000        /* Not a forward reference, test for completed backward reference */
2001    
2002      empty_branch = FALSE;      empty_branch = FALSE;
2003        scode = cd->start_code + GET(code, 1);
2004        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2005    
2006        /* Completed backwards reference */
2007    
2008      do      do
2009        {        {
2010        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2011            {
2012          empty_branch = TRUE;          empty_branch = TRUE;
2013            break;
2014            }
2015          scode += GET(scode, 1);
2016          }
2017        while (*scode == OP_ALT);
2018    
2019        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2020        continue;
2021        }
2022    
2023      /* Groups with zero repeats can of course be empty; skip them. */
2024    
2025      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2026          c == OP_BRAPOSZERO)
2027        {
2028        code += _pcre_OP_lengths[c];
2029        do code += GET(code, 1); while (*code == OP_ALT);
2030        c = *code;
2031        continue;
2032        }
2033    
2034      /* A nested group that is already marked as "could be empty" can just be
2035      skipped. */
2036    
2037      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2038          c == OP_SCBRA || c == OP_SCBRAPOS)
2039        {
2040        do code += GET(code, 1); while (*code == OP_ALT);
2041        c = *code;
2042        continue;
2043        }
2044    
2045      /* For other groups, scan the branches. */
2046    
2047      if (c == OP_BRA  || c == OP_BRAPOS ||
2048          c == OP_CBRA || c == OP_CBRAPOS ||
2049          c == OP_ONCE || c == OP_ONCE_NC ||
2050          c == OP_COND)
2051        {
2052        BOOL empty_branch;
2053        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2054    
2055        /* If a conditional group has only one branch, there is a second, implied,
2056        empty branch, so just skip over the conditional, because it could be empty.
2057        Otherwise, scan the individual branches of the group. */
2058    
2059        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2060        code += GET(code, 1);        code += GET(code, 1);
2061        else
2062          {
2063          empty_branch = FALSE;
2064          do
2065            {
2066            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2067              empty_branch = TRUE;
2068            code += GET(code, 1);
2069            }
2070          while (*code == OP_ALT);
2071          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2072        }        }
2073      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2074      c = *code;      c = *code;
2075      continue;      continue;
2076      }      }
# Line 1579  for (code = first_significant_code(code Line 2131  for (code = first_significant_code(code
2131      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2132      case OP_WORDCHAR:      case OP_WORDCHAR:
2133      case OP_ANY:      case OP_ANY:
2134        case OP_ALLANY:
2135      case OP_ANYBYTE:      case OP_ANYBYTE:
2136      case OP_CHAR:      case OP_CHAR:
2137      case OP_CHARNC:      case OP_CHARI:
2138      case OP_NOT:      case OP_NOT:
2139        case OP_NOTI:
2140      case OP_PLUS:      case OP_PLUS:
2141      case OP_MINPLUS:      case OP_MINPLUS:
2142      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1597  for (code = first_significant_code(code Line 2151  for (code = first_significant_code(code
2151      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2152      return FALSE;      return FALSE;
2153    
2154        /* These are going to continue, as they may be empty, but we have to
2155        fudge the length for the \p and \P cases. */
2156    
2157        case OP_TYPESTAR:
2158        case OP_TYPEMINSTAR:
2159        case OP_TYPEPOSSTAR:
2160        case OP_TYPEQUERY:
2161        case OP_TYPEMINQUERY:
2162        case OP_TYPEPOSQUERY:
2163        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2164        break;
2165    
2166        /* Same for these */
2167    
2168        case OP_TYPEUPTO:
2169        case OP_TYPEMINUPTO:
2170        case OP_TYPEPOSUPTO:
2171        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2172        break;
2173    
2174      /* End of branch */      /* End of branch */
2175    
2176      case OP_KET:      case OP_KET:
2177      case OP_KETRMAX:      case OP_KETRMAX:
2178      case OP_KETRMIN:      case OP_KETRMIN:
2179        case OP_KETRPOS:
2180      case OP_ALT:      case OP_ALT:
2181      return TRUE;      return TRUE;
2182    
# Line 1610  for (code = first_significant_code(code Line 2185  for (code = first_significant_code(code
2185    
2186  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2187      case OP_STAR:      case OP_STAR:
2188        case OP_STARI:
2189      case OP_MINSTAR:      case OP_MINSTAR:
2190        case OP_MINSTARI:
2191      case OP_POSSTAR:      case OP_POSSTAR:
2192        case OP_POSSTARI:
2193      case OP_QUERY:      case OP_QUERY:
2194        case OP_QUERYI:
2195      case OP_MINQUERY:      case OP_MINQUERY:
2196        case OP_MINQUERYI:
2197      case OP_POSQUERY:      case OP_POSQUERY:
2198        case OP_POSQUERYI:
2199        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2200        break;
2201    
2202      case OP_UPTO:      case OP_UPTO:
2203        case OP_UPTOI:
2204      case OP_MINUPTO:      case OP_MINUPTO:
2205        case OP_MINUPTOI:
2206      case OP_POSUPTO:      case OP_POSUPTO:
2207      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2208        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2209      break;      break;
2210  #endif  #endif
2211    
2212        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2213        string. */
2214    
2215        case OP_MARK:
2216        case OP_PRUNE_ARG:
2217        case OP_SKIP_ARG:
2218        code += code[1];
2219        break;
2220    
2221        case OP_THEN_ARG:
2222        code += code[1];
2223        break;
2224    
2225        /* None of the remaining opcodes are required to match a character. */
2226    
2227        default:
2228        break;
2229      }      }
2230    }    }
2231    
# Line 1637  return TRUE; Line 2242  return TRUE;
2242  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
2243  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,
2244  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.
2245    This function is called only during the real compile, not during the
2246    pre-compile.
2247    
2248  Arguments:  Arguments:
2249    code        points to start of the recursion    code        points to start of the recursion
2250    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2251    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2252    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2253      cd          pointers to tables etc
2254    
2255  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2256  */  */
2257    
2258  static BOOL  static BOOL
2259  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2260    BOOL utf8)    BOOL utf8, compile_data *cd)
2261  {  {
2262  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2263    {    {
2264    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2265        return FALSE;
2266    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2267    }    }
2268  return TRUE;  return TRUE;
# Line 1666  return TRUE; Line 2275  return TRUE;
2275  *************************************************/  *************************************************/
2276    
2277  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2278  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
2279  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2280  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2281    
2282    Originally, this function only recognized a sequence of letters between the
2283    terminators, but it seems that Perl recognizes any sequence of characters,
2284    though of course unknown POSIX names are subsequently rejected. Perl gives an
2285    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2286    didn't consider this to be a POSIX class. Likewise for [:1234:].
2287    
2288    The problem in trying to be exactly like Perl is in the handling of escapes. We
2289    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2290    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2291    below handles the special case of \], but does not try to do any other escape
2292    processing. This makes it different from Perl for cases such as [:l\ower:]
2293    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2294    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2295    I think.
2296    
2297    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2298    It seems that the appearance of a nested POSIX class supersedes an apparent
2299    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2300    a digit.
2301    
2302    In Perl, unescaped square brackets may also appear as part of class names. For
2303    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2304    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2305    seem right at all. PCRE does not allow closing square brackets in POSIX class
2306    names.
2307    
2308  Argument:  Arguments:
2309    ptr      pointer to the initial [    ptr      pointer to the initial [
2310    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2311    
2312  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2313  */  */
2314    
2315  static BOOL  static BOOL
2316  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2317  {  {
2318  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2319  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2320  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2321    {    {
2322    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2323    return TRUE;      ptr++;
2324      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2325      else
2326        {
2327        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2328          {
2329          *endptr = ptr;
2330          return TRUE;
2331          }
2332        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2333             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2334              ptr[1] == CHAR_EQUALS_SIGN) &&
2335            check_posix_syntax(ptr, endptr))
2336          return FALSE;
2337        }
2338    }    }
2339  return FALSE;  return FALSE;
2340  }  }
# Line 1713  Returns:     a value representing the na Line 2359  Returns:     a value representing the na
2359  static int  static int
2360  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2361  {  {
2362    const char *pn = posix_names;
2363  register int yield = 0;  register int yield = 0;
2364  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2365    {    {
2366    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2367      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2368      pn += posix_name_lengths[yield] + 1;
2369    yield++;    yield++;
2370    }    }
2371  return -1;  return -1;
# Line 1732  return -1; Line 2380  return -1;
2380  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2381  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2382  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
2383  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
2384  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
2385  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
2386  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
2387  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2388    OP_END.
2389    
2390  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2391  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 1759  adjust_recurse(uschar *group, int adjust Line 2408  adjust_recurse(uschar *group, int adjust
2408    uschar *save_hwm)    uschar *save_hwm)
2409  {  {
2410  uschar *ptr = group;  uschar *ptr = group;
2411    
2412  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2413    {    {
2414    int offset;    int offset;
# Line 1812  auto_callout(uschar *code, const uschar Line 2462  auto_callout(uschar *code, const uschar
2462  {  {
2463  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2464  *code++ = 255;  *code++ = 255;
2465  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2466  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2467  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2468  }  }
2469    
# Line 1838  Returns:             nothing Line 2488  Returns:             nothing
2488  static void  static void
2489  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2490  {  {
2491  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2492  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2493  }  }
2494    
# Line 1870  get_othercase_range(unsigned int *cptr, Line 2520  get_othercase_range(unsigned int *cptr,
2520  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2521    
2522  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2523    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2524    
2525  if (c > d) return FALSE;  if (c > d) return FALSE;
2526    
# Line 1879  next = othercase + 1; Line 2529  next = othercase + 1;
2529    
2530  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2531    {    {
2532    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2533    next++;    next++;
2534    }    }
2535    
# Line 1888  for (++c; c <= d; c++) Line 2538  for (++c; c <= d; c++)
2538    
2539  return TRUE;  return TRUE;
2540  }  }
2541    
2542    
2543    
2544    /*************************************************
2545    *        Check a character and a property        *
2546    *************************************************/
2547    
2548    /* This function is called by check_auto_possessive() when a property item
2549    is adjacent to a fixed character.
2550    
2551    Arguments:
2552      c            the character
2553      ptype        the property type
2554      pdata        the data for the type
2555      negated      TRUE if it's a negated property (\P or \p{^)
2556    
2557    Returns:       TRUE if auto-possessifying is OK
2558    */
2559    
2560    static BOOL
2561    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2562    {
2563    const ucd_record *prop = GET_UCD(c);
2564    switch(ptype)
2565      {
2566      case PT_LAMP:
2567      return (prop->chartype == ucp_Lu ||
2568              prop->chartype == ucp_Ll ||
2569              prop->chartype == ucp_Lt) == negated;
2570    
2571      case PT_GC:
2572      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2573    
2574      case PT_PC:
2575      return (pdata == prop->chartype) == negated;
2576    
2577      case PT_SC:
2578      return (pdata == prop->script) == negated;
2579    
2580      /* These are specials */
2581    
2582      case PT_ALNUM:
2583      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2584              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2585    
2586      case PT_SPACE:    /* Perl space */
2587      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2588              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2589              == negated;
2590    
2591      case PT_PXSPACE:  /* POSIX space */
2592      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2593              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2594              c == CHAR_FF || c == CHAR_CR)
2595              == negated;
2596    
2597      case PT_WORD:
2598      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2599              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2600              c == CHAR_UNDERSCORE) == negated;
2601      }
2602    return FALSE;
2603    }
2604  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2605    
2606    
# Line 1901  whether the next thing could possibly ma Line 2614  whether the next thing could possibly ma
2614  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2615    
2616  Arguments:  Arguments:
2617    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2618    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2619    ptr           next character in pattern    ptr           next character in pattern
2620    options       options bits    options       options bits
2621    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1913  Returns:        TRUE if possessifying is Line 2624  Returns:        TRUE if possessifying is
2624  */  */
2625    
2626  static BOOL  static BOOL
2627  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2628    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2629  {  {
2630  int next;  int c, next;
2631    int op_code = *previous++;
2632    
2633  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2634    
# Line 1925  if ((options & PCRE_EXTENDED) != 0) Line 2637  if ((options & PCRE_EXTENDED) != 0)
2637    for (;;)    for (;;)
2638      {      {
2639      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2640      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2641        {        {
2642        while (*(++ptr) != 0)        ptr++;
2643          while (*ptr != 0)
2644            {
2645          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2646            ptr++;
2647    #ifdef SUPPORT_UTF8
2648            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2649    #endif
2650            }
2651        }        }
2652      else break;      else break;
2653      }      }
# Line 1937  if ((options & PCRE_EXTENDED) != 0) Line 2656  if ((options & PCRE_EXTENDED) != 0)
2656  /* 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
2657  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2658    
2659  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2660    {    {
2661    int temperrorcode = 0;    int temperrorcode = 0;
2662    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1962  if ((options & PCRE_EXTENDED) != 0) Line 2681  if ((options & PCRE_EXTENDED) != 0)
2681    for (;;)    for (;;)
2682      {      {
2683      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2684      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2685        {        {
2686        while (*(++ptr) != 0)        ptr++;
2687          while (*ptr != 0)
2688            {
2689          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2690            ptr++;
2691    #ifdef SUPPORT_UTF8
2692            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2693    #endif
2694            }
2695        }        }
2696      else break;      else break;
2697      }      }
# Line 1973  if ((options & PCRE_EXTENDED) != 0) Line 2699  if ((options & PCRE_EXTENDED) != 0)
2699    
2700  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2701    
2702  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2703    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2704        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. */  
   
2705    
2706  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2707    the next item is a character. */
2708    
2709  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2710    {    {
2711    case OP_CHAR:    case OP_CHAR:
2712  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2713    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2714    #else
2715      c = *previous;
2716  #endif  #endif
2717    return item != next;    return c != next;
2718    
2719    /* 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
2720    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
2721    high-valued characters. */    high-valued characters. */
2722    
2723    case OP_CHARNC:    case OP_CHARI:
2724  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2725    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2726    #else
2727      c = *previous;
2728  #endif  #endif
2729    if (item == next) return FALSE;    if (c == next) return FALSE;
2730  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2731    if (utf8)    if (utf8)
2732      {      {
2733      unsigned int othercase;      unsigned int othercase;
2734      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2735  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2736      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2737  #else  #else
2738      othercase = NOTACHAR;      othercase = NOTACHAR;
2739  #endif  #endif
2740      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2741      }      }
2742    else    else
2743  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2744    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2745    
2746    /* 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
2747      opcodes are not used for multi-byte characters, because they are coded using
2748      an XCLASS instead. */
2749    
2750    case OP_NOT:    case OP_NOT:
2751    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2752    if (item == next) return TRUE;  
2753    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2754      if ((c = *previous) == next) return TRUE;
2755  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2756    if (utf8)    if (utf8)
2757      {      {
2758      unsigned int othercase;      unsigned int othercase;
2759      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2760  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2761      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2762  #else  #else
2763      othercase = NOTACHAR;      othercase = NOTACHAR;
2764  #endif  #endif
2765      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2766      }      }
2767    else    else
2768  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2769    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2770    
2771      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2772      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2773    
2774    case OP_DIGIT:    case OP_DIGIT:
2775    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2080  if (next >= 0) switch(op_code) Line 2812  if (next >= 0) switch(op_code)
2812      case 0x202f:      case 0x202f:
2813      case 0x205f:      case 0x205f:
2814      case 0x3000:      case 0x3000:
2815      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2816      default:      default:
2817      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2818      }      }
2819    
2820      case OP_ANYNL:
2821    case OP_VSPACE:    case OP_VSPACE:
2822    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2823    switch(next)    switch(next)
# Line 2096  if (next >= 0) switch(op_code) Line 2829  if (next >= 0) switch(op_code)
2829      case 0x85:      case 0x85:
2830      case 0x2028:      case 0x2028:
2831      case 0x2029:      case 0x2029:
2832      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2833      default:      default:
2834      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2835      }      }
2836    
2837    #ifdef SUPPORT_UCP
2838      case OP_PROP:
2839      return check_char_prop(next, previous[0], previous[1], FALSE);
2840    
2841      case OP_NOTPROP:
2842      return check_char_prop(next, previous[0], previous[1], TRUE);
2843    #endif
2844    
2845    default:    default:
2846    return FALSE;    return FALSE;
2847    }    }
2848    
2849    
2850  /* 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
2851    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2852    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2853    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2854    replaced by OP_PROP codes when PCRE_UCP is set. */
2855    
2856  switch(op_code)  switch(op_code)
2857    {    {
2858    case OP_CHAR:    case OP_CHAR:
2859    case OP_CHARNC:    case OP_CHARI:
2860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2861    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2862    #else
2863      c = *previous;
2864  #endif  #endif
2865    switch(-next)    switch(-next)
2866      {      {
2867      case ESC_d:      case ESC_d:
2868      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2869    
2870      case ESC_D:      case ESC_D:
2871      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2872    
2873      case ESC_s:      case ESC_s:
2874      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2875    
2876      case ESC_S:      case ESC_S:
2877      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2878    
2879      case ESC_w:      case ESC_w:
2880      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2881    
2882      case ESC_W:      case ESC_W:
2883      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2884    
2885      case ESC_h:      case ESC_h:
2886      case ESC_H:      case ESC_H:
2887      switch(item)      switch(c)
2888        {        {
2889        case 0x09:        case 0x09:
2890        case 0x20:        case 0x20:
# Line 2165  switch(op_code) Line 2912  switch(op_code)
2912    
2913      case ESC_v:      case ESC_v:
2914      case ESC_V:      case ESC_V:
2915      switch(item)      switch(c)
2916        {        {
2917        case 0x0a:        case 0x0a:
2918        case 0x0b:        case 0x0b:
# Line 2179  switch(op_code) Line 2926  switch(op_code)
2926        return -next == ESC_v;        return -next == ESC_v;
2927        }        }
2928    
2929        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2930        their substitutions and process them. The result will always be either
2931        -ESC_p or -ESC_P. Then fall through to process those values. */
2932    
2933    #ifdef SUPPORT_UCP
2934        case ESC_du:
2935        case ESC_DU:
2936        case ESC_wu:
2937        case ESC_WU:
2938        case ESC_su:
2939        case ESC_SU:
2940          {
2941          int temperrorcode = 0;
2942          ptr = substitutes[-next - ESC_DU];
2943          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2944          if (temperrorcode != 0) return FALSE;
2945          ptr++;    /* For compatibility */
2946          }
2947        /* Fall through */
2948    
2949        case ESC_p:
2950        case ESC_P:
2951          {
2952          int ptype, pdata, errorcodeptr;
2953          BOOL negated;
2954    
2955          ptr--;      /* Make ptr point at the p or P */
2956          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2957          if (ptype < 0) return FALSE;
2958          ptr++;      /* Point past the final curly ket */
2959    
2960          /* If the property item is optional, we have to give up. (When generated
2961          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2962          to the original \d etc. At this point, ptr will point to a zero byte. */
2963    
2964          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2965            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2966              return FALSE;
2967    
2968          /* Do the property check. */
2969    
2970          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2971          }
2972    #endif
2973    
2974      default:      default:
2975      return FALSE;      return FALSE;
2976      }      }
2977    
2978      /* In principle, support for Unicode properties should be integrated here as
2979      well. It means re-organizing the above code so as to get hold of the property
2980      values before switching on the op-code. However, I wonder how many patterns
2981      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2982      these op-codes are never generated.) */
2983    
2984    case OP_DIGIT:    case OP_DIGIT:
2985    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2986           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2989    return next == -ESC_d;    return next == -ESC_d;
2990    
2991    case OP_WHITESPACE:    case OP_WHITESPACE:
2992    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2993    
2994    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2995    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2996    
2997    case OP_HSPACE:    case OP_HSPACE:
2998    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2999             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3000    
3001    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3002    return next == -ESC_h;    return next == -ESC_h;
3003    
3004    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3005      case OP_ANYNL:
3006    case OP_VSPACE:    case OP_VSPACE:
3007    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3008    
3009    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3010    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3011    
3012    case OP_WORDCHAR:    case OP_WORDCHAR:
3013    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3014             next == -ESC_v || next == -ESC_R;
3015    
3016    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3017    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2242  Arguments: Line 3043  Arguments:
3043    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3044    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3045    bcptr          points to current branch chain    bcptr          points to current branch chain
3046      cond_depth     conditional nesting depth
3047    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3048    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3049                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2253  Returns:         TRUE on success Line 3055  Returns:         TRUE on success
3055  static BOOL  static BOOL
3056  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3057    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3058    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3059  {  {
3060  int repeat_type, op_type;  int repeat_type, op_type;
3061  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2262  int greedy_default, greedy_non_default; Line 3064  int greedy_default, greedy_non_default;
3064  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3065  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3066  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3067  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3068  int after_manual_callout = 0;  int after_manual_callout = 0;
3069  int length_prevgroup = 0;  int length_prevgroup = 0;
3070  register int c;  register int c;
# Line 2274  BOOL inescq = FALSE; Line 3076  BOOL inescq = FALSE;
3076  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3077  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3078  const uschar *tempptr;  const uschar *tempptr;
3079    const uschar *nestptr = NULL;
3080  uschar *previous = NULL;  uschar *previous = NULL;
3081  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3082  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3083  uschar classbits[32];  uschar classbits[32];
3084    
3085    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3086    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3087    dynamically as we process the pattern. */
3088    
3089  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3090  BOOL class_utf8;  BOOL class_utf8;
3091  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3092  uschar *class_utf8data;  uschar *class_utf8data;
3093    uschar *class_utf8data_base;
3094  uschar utf8_char[6];  uschar utf8_char[6];
3095  #else  #else
3096  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3097  #endif  #endif
3098    
3099  #ifdef DEBUG  #ifdef PCRE_DEBUG
3100  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3101  #endif  #endif
3102    
# Line 2322  req_caseopt = ((options & PCRE_CASELESS) Line 3129  req_caseopt = ((options & PCRE_CASELESS)
3129  for (;; ptr++)  for (;; ptr++)
3130    {    {
3131    BOOL negate_class;    BOOL negate_class;
3132      BOOL should_flip_negation;
3133    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3134    BOOL is_quantifier;    BOOL is_quantifier;
3135    BOOL is_recurse;    BOOL is_recurse;
# Line 2336  for (;; ptr++) Line 3144  for (;; ptr++)
3144    int subfirstbyte;    int subfirstbyte;
3145    int terminator;    int terminator;
3146    int mclength;    int mclength;
3147      int tempbracount;
3148    uschar mcbuffer[8];    uschar mcbuffer[8];
3149    
3150    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3151    
3152    c = *ptr;    c = *ptr;
3153    
3154      /* If we are at the end of a nested substitution, revert to the outer level
3155      string. Nesting only happens one level deep. */
3156    
3157      if (c == 0 && nestptr != NULL)
3158        {
3159        ptr = nestptr;
3160        nestptr = NULL;
3161        c = *ptr;
3162        }
3163    
3164    /* 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
3165    previous cycle of this loop. */    previous cycle of this loop. */
3166    
3167    if (lengthptr != NULL)    if (lengthptr != NULL)
3168      {      {
3169  #ifdef DEBUG  #ifdef PCRE_DEBUG
3170      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3171  #endif  #endif
3172      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3173        {        {
3174        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3175        goto FAILED;        goto FAILED;
# Line 2372  for (;; ptr++) Line 3191  for (;; ptr++)
3191        goto FAILED;        goto FAILED;
3192        }        }
3193    
3194      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3195      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));
3196    
3197      /* 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 2399  for (;; ptr++) Line 3218  for (;; ptr++)
3218    /* 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
3219    reference list. */    reference list. */
3220    
3221    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3222      {      {
3223      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3224      goto FAILED;      goto FAILED;
# Line 2409  for (;; ptr++) Line 3228  for (;; ptr++)
3228    
3229    if (inescq && c != 0)    if (inescq && c != 0)
3230      {      {
3231      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3232        {        {
3233        inescq = FALSE;        inescq = FALSE;
3234        ptr++;        ptr++;
# Line 2435  for (;; ptr++) Line 3254  for (;; ptr++)
3254    /* 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
3255    a quantifier. */    a quantifier. */
3256    
3257    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3258      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3259        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3260    
3261    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3262         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2446  for (;; ptr++) Line 3266  for (;; ptr++)
3266      previous_callout = NULL;      previous_callout = NULL;
3267      }      }
3268    
3269    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3270    
3271    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3272      {      {
3273      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3274      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3275        {        {
3276        while (*(++ptr) != 0)        ptr++;
3277          while (*ptr != 0)
3278          {          {
3279          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3280            ptr++;
3281    #ifdef SUPPORT_UTF8
3282            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3283    #endif
3284          }          }
3285        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3286    
# Line 2476  for (;; ptr++) Line 3301  for (;; ptr++)
3301      {      {
3302      /* ===================================================================*/      /* ===================================================================*/
3303      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3304      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3305      case ')':      case CHAR_RIGHT_PARENTHESIS:
3306      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3307      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3308      *codeptr = code;      *codeptr = code;
# Line 2489  for (;; ptr++) Line 3314  for (;; ptr++)
3314          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3315          goto FAILED;          goto FAILED;
3316          }          }
3317        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3318        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3319        }        }
3320      return TRUE;      return TRUE;
# Line 2499  for (;; ptr++) Line 3324  for (;; ptr++)
3324      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3325      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3326    
3327      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3328        previous = NULL;
3329      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3330        {        {
3331        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3332          *code++ = OP_CIRCM;
3333        }        }
3334      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3335      break;      break;
3336    
3337      case '$':      case CHAR_DOLLAR_SIGN:
3338      previous = NULL;      previous = NULL;
3339      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3340      break;      break;
3341    
3342      /* 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
3343      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3344    
3345      case '.':      case CHAR_DOT:
3346      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3347      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3348      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3349      previous = code;      previous = code;
3350      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3351      break;      break;
3352    
3353    
# Line 2536  for (;; ptr++) Line 3362  for (;; ptr++)
3362      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,
3363      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3364      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.
     */  
3365    
3366      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3367        default (Perl) mode, it is treated as a data character. */
3368    
3369        case CHAR_RIGHT_SQUARE_BRACKET:
3370        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3371          {
3372          *errorcodeptr = ERR64;
3373          goto FAILED;
3374          }
3375        goto NORMAL_CHAR;
3376    
3377        case CHAR_LEFT_SQUARE_BRACKET:
3378      previous = code;      previous = code;
3379    
3380      /* 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
3381      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. */
3382    
3383      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3384          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3385            check_posix_syntax(ptr, &tempptr))
3386        {        {
3387        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3388        goto FAILED;        goto FAILED;
3389        }        }
3390    
# Line 2559  for (;; ptr++) Line 3396  for (;; ptr++)
3396      for (;;)      for (;;)
3397        {        {
3398        c = *(++ptr);        c = *(++ptr);
3399        if (c == '\\')        if (c == CHAR_BACKSLASH)
3400          {          {
3401          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3402            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3403              else break;          else if (strncmp((const char *)ptr+1,
3404                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3405              ptr += 3;
3406            else
3407              break;
3408          }          }
3409        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3410          negate_class = TRUE;          negate_class = TRUE;
3411        else break;        else break;
3412        }        }
3413    
3414        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3415        an initial ']' is taken as a data character -- the code below handles
3416        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3417        [^] must match any character, so generate OP_ALLANY. */
3418    
3419        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3420            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3421          {
3422          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3423          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3424          zerofirstbyte = firstbyte;
3425          break;
3426          }
3427    
3428        /* If a class contains a negative special such as \S, we need to flip the
3429        negation flag at the end, so that support for characters > 255 works
3430        correctly (they are all included in the class). */
3431    
3432        should_flip_negation = FALSE;
3433    
3434      /* 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
3435      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
3436      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2587  for (;; ptr++) Line 3448  for (;; ptr++)
3448  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3449      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3450      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3451        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3452  #endif  #endif
3453    
3454      /* 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 2602  for (;; ptr++) Line 3464  for (;; ptr++)
3464          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3465          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3466          }          }
3467    
3468          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3469          data and reset the pointer. This is so that very large classes that
3470          contain a zillion UTF-8 characters no longer overwrite the work space
3471          (which is on the stack). */
3472    
3473          if (lengthptr != NULL)
3474            {
3475            *lengthptr += class_utf8data - class_utf8data_base;
3476            class_utf8data = class_utf8data_base;
3477            }
3478    
3479  #endif  #endif
3480    
3481        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3482    
3483        if (inescq)        if (inescq)
3484          {          {
3485          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3486            {            {
3487            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3488            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2623  for (;; ptr++) Line 3497  for (;; ptr++)
3497        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3498        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3499    
3500        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3501            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3502            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3503          {          {
3504          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3505          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3506          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3507          uschar pbits[32];          uschar pbits[32];
3508    
3509          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3510            {            {
3511            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3512            goto FAILED;            goto FAILED;
3513            }            }
3514    
3515          ptr += 2;          ptr += 2;
3516          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3517            {            {
3518            local_negate = TRUE;            local_negate = TRUE;
3519              should_flip_negation = TRUE;  /* Note negative special */
3520            ptr++;            ptr++;
3521            }            }
3522    
3523          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3524          if (posix_class < 0)          if (posix_class < 0)
3525            {            {
3526            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2659  for (;; ptr++) Line 3534  for (;; ptr++)
3534          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3535            posix_class = 0;            posix_class = 0;
3536    
3537          /* 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
3538          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3539          subtract bits that may be in the main map already. At the end we or the  
3540          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3541            if ((options & PCRE_UCP) != 0)
3542              {
3543              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3544              if (posix_substitutes[pc] != NULL)
3545                {
3546                nestptr = tempptr + 1;
3547                ptr = posix_substitutes[pc] - 1;
3548                continue;
3549                }
3550              }
3551    #endif
3552            /* In the non-UCP case, we build the bit map for the POSIX class in a
3553            chunk of local store because we may be adding and subtracting from it,
3554            and we don't want to subtract bits that may be in the main map already.
3555            At the end we or the result into the bit map that is being built. */
3556    
3557          posix_class *= 3;          posix_class *= 3;
3558    
# Line 2706  for (;; ptr++) Line 3596  for (;; ptr++)
3596    
3597        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3598        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
3599        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
3600        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
3601        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
3602        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3603          PCRE_EXTRA is set. */
3604    
3605        if (c == '\\')        if (c == CHAR_BACKSLASH)
3606          {          {
3607          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3608          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3609    
3610          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 */  
3611          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3612            {            {
3613            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3614              {              {
3615              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3616              }              }
# Line 2735  for (;; ptr++) Line 3624  for (;; ptr++)
3624            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3625            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3626    
3627            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3628              {              {
3629    #ifdef SUPPORT_UCP
3630                case ESC_du:     /* These are the values given for \d etc */
3631                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3632                case ESC_wu:     /* escape sequence with an appropriate \p */
3633                case ESC_WU:     /* or \P to test Unicode properties instead */
3634                case ESC_su:     /* of the default ASCII testing. */
3635                case ESC_SU:
3636                nestptr = ptr;
3637                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3638                class_charcount -= 2;                /* Undo! */
3639                continue;
3640    #endif
3641              case ESC_d:              case ESC_d:
3642              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3643              continue;              continue;
3644    
3645              case ESC_D:              case ESC_D:
3646                should_flip_negation = TRUE;
3647              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3648              continue;              continue;
3649    
# Line 2752  for (;; ptr++) Line 3652  for (;; ptr++)
3652              continue;              continue;
3653    
3654              case ESC_W:              case ESC_W:
3655                should_flip_negation = TRUE;
3656              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3657              continue;              continue;
3658    
3659                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3660                if it was previously set by something earlier in the character
3661                class. */
3662    
3663              case ESC_s:              case ESC_s:
3664              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3665              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3666                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3667              continue;              continue;
3668    
3669              case ESC_S:              case ESC_S:
3670                should_flip_negation = TRUE;
3671              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3672              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3673              continue;              continue;
3674    
3675              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3676              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3677              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3678              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2805  for (;; ptr++) Line 3696  for (;; ptr++)
3696                }                }
3697  #endif  #endif
3698              continue;              continue;
             }  
3699    
3700            if (-c == ESC_H)              case ESC_H:
             {  
3701              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3702                {                {
3703                int x = 0xff;                int x = 0xff;
# Line 2850  for (;; ptr++) Line 3739  for (;; ptr++)
3739                }                }
3740  #endif  #endif
3741              continue;              continue;
             }  
3742    
3743            if (-c == ESC_v)              case ESC_v:
             {  
3744              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3745              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3746              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2869  for (;; ptr++) Line 3756  for (;; ptr++)
3756                }                }
3757  #endif  #endif
3758              continue;              continue;
             }  
3759    
3760            if (-c == ESC_V)              case ESC_V:
             {  
3761              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3762                {                {
3763                int x = 0xff;                int x = 0xff;
# Line 2902  for (;; ptr++) Line 3787  for (;; ptr++)
3787                }                }
3788  #endif  #endif
3789              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3790    
3791  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3792            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3793              {              case ESC_P:
3794              BOOL negated;                {
3795              int pdata;                BOOL negated;
3796              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3797              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3798              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3799              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3800                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3801              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3802              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3803              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3804              continue;                class_charcount -= 2;   /* Not a < 256 character */
3805              }                continue;
3806                  }
3807  #endif  #endif
3808            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3809            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3810            treated as literals. */              treated as literals. */
3811    
3812            if ((options & PCRE_EXTRA) != 0)              default:
3813              {              if ((options & PCRE_EXTRA) != 0)
3814              *errorcodeptr = ERR7;                {
3815              goto FAILED;                *errorcodeptr = ERR7;
3816                  goto FAILED;
3817                  }
3818                class_charcount -= 2;  /* Undo the default count from above */
3819                c = *ptr;              /* Get the final character and fall through */
3820                break;
3821              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3822            }            }
3823    
3824          /* 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 2947  for (;; ptr++) Line 3832  for (;; ptr++)
3832        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3833    
3834        CHECK_RANGE:        CHECK_RANGE:
3835        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3836          {          {
3837          inescq = FALSE;          inescq = FALSE;
3838          ptr += 2;          ptr += 2;
# Line 2955  for (;; ptr++) Line 3840  for (;; ptr++)
3840    
3841        oldptr = ptr;        oldptr = ptr;
3842    
3843        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3844    
3845          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3846    
3847          /* Check for range */
3848    
3849          if (!inescq && ptr[1] == CHAR_MINUS)
3850          {          {
3851          int d;          int d;
3852          ptr += 2;          ptr += 2;
3853          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3854    
3855          /* 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
3856          mode. */          mode. */
3857    
3858          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3859            {            {
3860            ptr += 2;            ptr += 2;
3861            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3862                { ptr += 2; continue; }
3863            inescq = TRUE;            inescq = TRUE;
3864            break;            break;
3865            }            }
3866    
3867          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3868            {            {
3869            ptr = oldptr;            ptr = oldptr;
3870            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2991  for (;; ptr++) Line 3883  for (;; ptr++)
3883          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
3884          in such circumstances. */          in such circumstances. */
3885    
3886          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3887            {            {
3888            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3889            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3890    
3891            /* \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 */  
3892    
3893            if (d < 0)            if (d < 0)
3894              {              {
3895              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  
3896                {                {
3897                ptr = oldptr;                ptr = oldptr;
3898                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3022  for (;; ptr++) Line 3911  for (;; ptr++)
3911    
3912          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3913    
3914            /* Remember \r or \n */
3915    
3916            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3917    
3918          /* 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
3919          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3920          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3140  for (;; ptr++) Line 4033  for (;; ptr++)
4033          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4034            {            {
4035            unsigned int othercase;            unsigned int othercase;
4036            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4037              {              {
4038              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4039              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3165  for (;; ptr++) Line 4058  for (;; ptr++)
4058          }          }
4059        }        }
4060    
4061      /* 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.
4062        If we are at the end of an internal nested string, revert to the outer
4063        string. */
4064    
4065        while (((c = *(++ptr)) != 0 ||
4066               (nestptr != NULL &&
4067                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4068               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4069    
4070      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4071    
4072      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4073        {        {
4074        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4075        goto FAILED;        goto FAILED;
4076        }        }
4077    
4078      /* 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
4079      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4080      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
4081      optimize.      optimize.
4082    
4083      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4084      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4085      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4086      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4087    
4088      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
4089      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.
4090      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
4091      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
4092      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
4093      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4094    
4095  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4096      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
4097        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
4098  #else  #else
4099      if (class_charcount == 1)      if (class_charcount == 1)
# Line 3201  for (;; ptr++) Line 4101  for (;; ptr++)
4101        {        {
4102        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4103    
4104        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4105    
4106        if (negate_class)        if (negate_class)
4107          {          {
4108          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4109          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4110          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4111          *code++ = class_lastchar;          *code++ = class_lastchar;
4112          break;          break;
4113          }          }
# Line 3237  for (;; ptr++) Line 4137  for (;; ptr++)
4137      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4138    
4139      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4140      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4141      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
4142        characters > 255 are in the class, so any that were explicitly given as
4143        well can be ignored. If (when there are explicit characters > 255 that must
4144        be listed) there are no characters < 256, we can omit the bitmap in the
4145        actual compiled code. */
4146    
4147  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4148      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4149        {        {
4150        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4151        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3267  for (;; ptr++) Line 4171  for (;; ptr++)
4171        }        }
4172  #endif  #endif
4173    
4174      /* 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
4175      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
4176      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
4177      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4178        negating it if necessary. */
4179    
4180        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4181      if (negate_class)      if (negate_class)
4182        {        {
       *code++ = OP_NCLASS;  
4183        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4184          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4185        }        }
4186      else      else
4187        {        {
       *code++ = OP_CLASS;  
4188        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4189        }        }
4190      code += 32;      code += 32;
# Line 3291  for (;; ptr++) Line 4195  for (;; ptr++)
4195      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4196      has been tested above. */      has been tested above. */
4197    
4198      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4199      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4200      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4201      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4202      goto REPEAT;      goto REPEAT;
4203    
4204      case '*':      case CHAR_ASTERISK:
4205      repeat_min = 0;      repeat_min = 0;
4206      repeat_max = -1;      repeat_max = -1;
4207      goto REPEAT;      goto REPEAT;
4208    
4209      case '+':      case CHAR_PLUS:
4210      repeat_min = 1;      repeat_min = 1;
4211      repeat_max = -1;      repeat_max = -1;
4212      goto REPEAT;      goto REPEAT;
4213    
4214      case '?':      case CHAR_QUESTION_MARK:
4215      repeat_min = 0;      repeat_min = 0;
4216      repeat_max = 1;      repeat_max = 1;
4217    
# Line 3331  for (;; ptr++) Line 4235  for (;; ptr++)
4235      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4236      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4237    
4238      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4239      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4240    
4241      tempcode = previous;      tempcode = previous;
4242    
# Line 3342  for (;; ptr++) Line 4246  for (;; ptr++)
4246      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
4247      repeat type to the non-default. */      repeat type to the non-default. */
4248    
4249      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4250        {        {
4251        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4252        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4253        ptr++;        ptr++;
4254        }        }
4255      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4256        {        {
4257        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4258        ptr++;        ptr++;
4259        }        }
4260      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4261    
4262        /* If previous was a recursion call, wrap it in atomic brackets so that
4263        previous becomes the atomic group. All recursions were so wrapped in the
4264        past, but it no longer happens for non-repeated recursions. In fact, the
4265        repeated ones could be re-implemented independently so as not to need this,
4266        but for the moment we rely on the code for repeating groups. */
4267    
4268        if (*previous == OP_RECURSE)
4269          {
4270          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4271          *previous = OP_ONCE;
4272          PUT(previous, 1, 2 + 2*LINK_SIZE);
4273          previous[2 + 2*LINK_SIZE] = OP_KET;
4274          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4275          code += 2 + 2 * LINK_SIZE;
4276          length_prevgroup = 3 + 3*LINK_SIZE;
4277    
4278          /* When actually compiling, we need to check whether this was a forward
4279          reference, and if so, adjust the offset. */
4280    
4281          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4282            {
4283            int offset = GET(cd->hwm, -LINK_SIZE);
4284            if (offset == previous + 1 - cd->start_code)
4285              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4286            }
4287          }
4288    
4289        /* Now handle repetition for the different types of item. */
4290    
4291      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4292      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
4293      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4294      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
4295      instead.  */      instead.  */
4296    
4297      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4298        {        {
4299          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4300    
4301        /* 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
4302        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
4303        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 3395  for (;; ptr++) Line 4330  for (;; ptr++)
4330    
4331        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4332            repeat_max < 0 &&            repeat_max < 0 &&
4333            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4334          {          {
4335          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4336          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3408  for (;; ptr++) Line 4342  for (;; ptr++)
4342      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4343      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-
4344      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4345      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
4346      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4347    
4348      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4349        {        {
4350        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4351        c = previous[1];        c = previous[1];
4352        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4353            repeat_max < 0 &&            repeat_max < 0 &&
4354            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4355          {          {
4356          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4357          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3441  for (;; ptr++) Line 4375  for (;; ptr++)
4375    
4376        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4377            repeat_max < 0 &&            repeat_max < 0 &&
4378            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4379          {          {
4380          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4381          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3463  for (;; ptr++) Line 4397  for (;; ptr++)
4397    
4398        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4399    
4400          /*--------------------------------------------------------------------*/
4401          /* This code is obsolete from release 8.00; the restriction was finally
4402          removed: */
4403    
4404        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4405        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4406    
4407        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4408          /*--------------------------------------------------------------------*/
4409    
4410        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4411    
# Line 3605  for (;; ptr++) Line 4544  for (;; ptr++)
4544  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4545               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4546  #endif  #endif
4547               *previous == OP_REF)               *previous == OP_REF ||
4548                 *previous == OP_REFI)
4549        {        {
4550        if (repeat_max == 0)        if (repeat_max == 0)
4551          {          {
# Line 3613  for (;; ptr++) Line 4553  for (;; ptr++)
4553          goto END_REPEAT;          goto END_REPEAT;
4554          }          }
4555    
4556          /*--------------------------------------------------------------------*/
4557          /* This code is obsolete from release 8.00; the restriction was finally
4558          removed: */
4559    
4560        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4561        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4562    
4563        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4564          /*--------------------------------------------------------------------*/
4565    
4566        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4567          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3634  for (;; ptr++) Line 4579  for (;; ptr++)
4579        }        }
4580    
4581      /* 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
4582      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4583        opcodes such as BRA and CBRA, as this is the place where they get converted
4584        into the more special varieties such as BRAPOS and SBRA. A test for >=
4585        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4586        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4587        repetition of assertions, but now it does, for Perl compatibility. */
4588    
4589      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4590        {        {
4591        register int i;        register int i;
4592        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4593        uschar *bralink = NULL;        uschar *bralink = NULL;
4594          uschar *brazeroptr = NULL;
4595    
4596        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4597          we just ignore the repeat. */
4598    
4599        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4600          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4601    
4602        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4603        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4604        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4605        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4606        pointer. */  
4607          if (*previous < OP_ONCE)    /* Assertion */
4608        if (repeat_max == -1)          {
4609          {          if (repeat_min > 0) goto END_REPEAT;
4610          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4611          }          }
4612    
4613        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3672  for (;; ptr++) Line 4617  for (;; ptr++)
4617        the code gets far too messy. There are several special subcases when the        the code gets far too messy. There are several special subcases when the
4618        minimum is zero. */        minimum is zero. */
4619    
4620        if (repeat_min == 0)        if (repeat_min == 0)
4621          {          {
4622          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4623          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4624    
4625          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4626          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4627          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4628          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4629          the start of the whole regex. Temporarily terminate the pattern while          **   }
4630          doing this. */  
4631            However, that fails when a group or a subgroup within it is referenced
4632            as a subroutine from elsewhere in the pattern, so now we stick in
4633            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4634            don't have a list of which groups are referenced, we cannot do this
4635            selectively.
4636    
4637            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4638            and do no more at this point. However, we do need to adjust any
4639            OP_RECURSE calls inside the group that refer to the group itself or any
4640            internal or forward referenced group, because the offset is from the
4641            start of the whole regex. Temporarily terminate the pattern while doing
4642            this. */
4643    
4644          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4645            {            {
4646            *code = OP_END;            *code = OP_END;
4647            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4648            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4649            code++;            code++;
4650              if (repeat_max == 0)
4651                {
4652                *previous++ = OP_SKIPZERO;
4653                goto END_REPEAT;
4654                }
4655              brazeroptr = previous;    /* Save for possessive optimizing */
4656            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4657            }            }
4658    
# Line 3720  for (;; ptr++) Line 4677  for (;; ptr++)
4677            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4678            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4679    
4680            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4681            bralink = previous;            bralink = previous;
4682            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4683            }            }
# Line 3741  for (;; ptr++) Line 4698  for (;; ptr++)
4698            {            {
4699            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4700            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4701            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4702              integer type when available, otherwise double. */
4703    
4704            if (lengthptr != NULL)            if (lengthptr != NULL)
4705              {              {
4706              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4707              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4708                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4709                        (INT64_OR_DOUBLE)INT_MAX ||
4710                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4711                {                {
4712                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3793  for (;; ptr++) Line 4752  for (;; ptr++)
4752          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4753          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4754          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4755          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4756            a 64-bit integer type when available, otherwise double. */
4757    
4758          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4759            {            {
4760            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4761                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4762            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4763                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4764                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4765                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4766              {              {
4767              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3826  for (;; ptr++) Line 4786  for (;; ptr++)
4786              {              {
4787              int offset;              int offset;
4788              *code++ = OP_BRA;              *code++ = OP_BRA;
4789              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4790              bralink = code;              bralink = code;
4791              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4792              }              }
# Line 3847  for (;; ptr++) Line 4807  for (;; ptr++)
4807          while (bralink != NULL)          while (bralink != NULL)
4808            {            {
4809            int oldlinkoffset;            int oldlinkoffset;
4810            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4811            uschar *bra = code - offset;            uschar *bra = code - offset;
4812            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4813            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3857  for (;; ptr++) Line 4817  for (;; ptr++)
4817            }            }
4818          }          }
4819    
4820        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4821        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4822        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4823        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4824          deal with possessive ONCEs specially.
4825    
4826          Otherwise, if the quantifier was possessive, we convert the BRA code to
4827          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4828          at runtime to detect this kind of subpattern at both the start and at the
4829          end.) The use of special opcodes makes it possible to reduce greatly the
4830          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4831          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4832          the default action below, of wrapping everything inside atomic brackets,
4833          does not happen.
4834    
4835        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4836        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4837        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4838        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4839        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4840    
4841        else        else
4842          {          {
4843          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4844          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4845          *ketcode = OP_KETRMAX + repeat_type;  
4846          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4847                possessive_quantifier) *bracode = OP_BRA;
4848    
4849            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4850              *ketcode = OP_KETRMAX + repeat_type;
4851            else
4852            {            {
4853            uschar *scode = bracode;            if (possessive_quantifier)
4854            do              {
4855                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4856                *ketcode = OP_KETRPOS;
4857                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4858                possessive_quantifier = FALSE;
4859                }
4860              else *ketcode = OP_KETRMAX + repeat_type;
4861    
4862              if (lengthptr == NULL)
4863              {              {
4864              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4865                do
4866                {                {
4867                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4868                break;                  {
4869                    *bracode += OP_SBRA - OP_BRA;
4870                    break;
4871                    }
4872                  scode += GET(scode, 1);
4873                }                }
4874              scode += GET(scode, 1);              while (*scode == OP_ALT);
4875              }              }
           while (*scode == OP_ALT);  
4876            }            }
4877          }          }
4878        }        }
4879    
4880        /* If previous is OP_FAIL, it was generated by an empty class [] in
4881        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4882        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4883        error above. We can just ignore the repeat in JS case. */
4884    
4885        else if (*previous == OP_FAIL) goto END_REPEAT;
4886    
4887      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4888    
4889      else      else
# Line 3899  for (;; ptr++) Line 4893  for (;; ptr++)
4893        }        }
4894    
4895      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4896      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4897      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4898      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4899      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4900      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4901      tempcode, not at previous, which might be the first part of a string whose  
4902      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4903        just above, so possessive_quantifier is always FALSE for them at this
4904        stage.
4905    
4906        Note that the repeated item starts at tempcode, not at previous, which
4907        might be the first part of a string whose (former) last char we repeated.
4908    
4909      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4910      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 3914  for (;; ptr++) Line 4913  for (;; ptr++)
4913      if (possessive_quantifier)      if (possessive_quantifier)
4914        {        {
4915        int len;        int len;
4916        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4917            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4918            tempcode += _pcre_OP_lengths[*tempcode] +
4919              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4920    
4921          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4922            {
4923          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4924        len = code - tempcode;  #ifdef SUPPORT_UTF8
4925            if (utf8 && tempcode[-1] >= 0xc0)
4926              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4927    #endif
4928            }
4929    
4930          len = (int)(code - tempcode);
4931        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4932          {          {
4933          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3925  for (;; ptr++) Line 4935  for (;; ptr++)
4935          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4936          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4937    
4938          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4939          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4940          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4941          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4942    
4943          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4944          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4945          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4946          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4947    
4948            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4949            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4950            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4951            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4952    
4953            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4954            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4955            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4956            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4957    
4958            /* Because we are moving code along, we must ensure that any
4959            pending recursive references are updated. */
4960    
4961          default:          default:
4962            *code = OP_END;
4963            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4964          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4965          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4966          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3962  for (;; ptr++) Line 4987  for (;; ptr++)
4987      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4988      parenthesis forms.  */      parenthesis forms.  */
4989    
4990      case '(':      case CHAR_LEFT_PARENTHESIS:
4991      newoptions = options;      newoptions = options;
4992      skipbytes = 0;      skipbytes = 0;
4993      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 3971  for (;; ptr++) Line 4996  for (;; ptr++)
4996    
4997      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4998    
4999      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5000             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5001        {        {
5002        int i, namelen;        int i, namelen;
5003        const uschar *name = ++ptr;        int arglen = 0;
5004          const char *vn = verbnames;
5005          const uschar *name = ptr + 1;
5006          const uschar *arg = NULL;
5007        previous = NULL;        previous = NULL;
5008        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5009        if (*ptr == ':')        namelen = (int)(ptr - name);
5010    
5011          /* It appears that Perl allows any characters whatsoever, other than
5012          a closing parenthesis, to appear in arguments, so we no longer insist on
5013          letters, digits, and underscores. */
5014    
5015          if (*ptr == CHAR_COLON)
5016          {          {
5017          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5018          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5019            arglen = (int)(ptr - arg);
5020          }          }
5021        if (*ptr != ')')  
5022          if (*ptr != CHAR_RIGHT_PARENTHESIS)
5023          {          {
5024          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5025          goto FAILED;          goto FAILED;
5026          }          }
5027        namelen = ptr - name;  
5028          /* Scan the table of verb names */
5029    
5030        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5031          {          {
5032          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5033              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5034            {            {
5035            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5036            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5037            break;  
5038              if (verbs[i].op == OP_ACCEPT)
5039                {
5040                open_capitem *oc;
5041                if (arglen != 0)
5042                  {
5043                  *errorcodeptr = ERR59;
5044                  goto FAILED;
5045                  }
5046                cd->had_accept = TRUE;
5047                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5048                  {
5049                  *code++ = OP_CLOSE;
5050                  PUT2INC(code, 0, oc->number);
5051                  }
5052