/[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 178 by ph10, Wed Jun 13 08:44:34 2007 UTC revision 518 by ph10, Tue May 18 15:47:01 2010 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) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 62  used by pcretest. DEBUG is not defined w Line 66  used by pcretest. DEBUG is not defined w
66    
67  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 81  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 120  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,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-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 130  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 169  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 176  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
317  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number or name after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
387    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by 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  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  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 268  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 304  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 379  static const unsigned char ebcdic_charta Line 544  static const unsigned char ebcdic_charta
544  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
545    
546  static BOOL  static BOOL
547    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
548      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
549    
550    
551    
552  /*************************************************  /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578    /*************************************************
579  *            Handle escapes                      *  *            Handle escapes                      *
580  *************************************************/  *************************************************/
581    
# Line 405  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
# Line 423  ptr--;                            /* Set Line 614  ptr--;                            /* Set
614    
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* 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
618  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.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 448  else Line 639  else
639      /* 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
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
650      is an absolute backreference. If negative, it is a relative backreference.  
651      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
652      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
653      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
654      fudge it up by pretending it really was \k. */  
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665      case 'g':      case CHAR_g:
666      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675        {        {
676        const uschar *p;        const uschar *p;
677        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680          {          {
681          c = -ESC_k;          c = -ESC_k;
682          break;          break;
# Line 479  else Line 686  else
686        }        }
687      else braced = FALSE;      else braced = FALSE;
688    
689      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
690        {        {
691        negated = TRUE;        negated = TRUE;
692        ptr++;        ptr++;
# Line 488  else Line 695  else
695    
696      c = 0;      c = 0;
697      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
698        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707        {        {
708        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
709        return 0;        break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716        }        }
717    
718      if (negated)      if (negated)
# Line 501  else Line 720  else
720        if (c > bracount)        if (c > bracount)
721          {          {
722          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
723          return 0;          break;
724          }          }
725        c = bracount - (c - 1);        c = bracount - (c - 1);
726        }        }
# Line 521  else Line 740  else
740      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
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      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:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 542  else Line 766  else
766      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.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 555  else Line 779  else
779      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
780      than 3 octal digits. */      than 3 octal digits. */
781    
782      case '0':      case CHAR_0:
783      c -= '0';      c -= CHAR_0;
784      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
786      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
# Line 566  else Line 790  else
790      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
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 576  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 604  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
# Line 620  else Line 844  else
844      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
845      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
857      c ^= 0x40;      c ^= 0x40;
858  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      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
867      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
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 654  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 694  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \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
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
940      }      }
941    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 730  top = _pcre_utt_size; Line 967  top = _pcre_utt_size;
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top) >> 1;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0)    if (c == 0)
972      {      {
973      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 773  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 814  int max = -1; Line 1051  int max = -1;
1051  /* 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
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 824  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* 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.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 854  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1095  *************************************************/  *************************************************/
1096    
1097  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100  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
1101  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
1102  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1103  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1104  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107  Arguments:  Arguments:
1108    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1109    count        current count of capturing parens so far encountered    cd           compile background data
1110    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1111    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1112    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  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,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1133          ptr[2] == CHAR_VERTICAL_LINE)
1134        {
1135        ptr += 3;
1136        dup_parens = TRUE;
1137        }
1138    
1139      /* Handle a normal, unnamed capturing parenthesis */
1140    
1141      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1142        {
1143        *count += 1;
1144        if (name == NULL && *count == lorn) return *count;
1145        ptr++;
1146        }
1147    
1148      /* Handle a condition. If it is an assertion, just carry on so that it
1149      is processed as normal. If not, skip to the closing parenthesis of the
1150      condition (there can't be any nested parens. */
1151    
1152      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1153        {
1154        ptr += 2;
1155        if (ptr[1] != CHAR_QUESTION_MARK)
1156          {
1157          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1158          if (*ptr != 0) ptr++;
1159          }
1160        }
1161    
1162      /* We have either (? or (* and not a condition */
1163    
1164      else
1165        {
1166        ptr += 2;
1167        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1168    
1169        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1170    
1171        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1172            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1173          {
1174          int term;
1175          const uschar *thisname;
1176          *count += 1;
1177          if (name == NULL && *count == lorn) return *count;
1178          term = *ptr++;
1179          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1180          thisname = ptr;
1181          while (*ptr != term) ptr++;
1182          if (name != NULL && lorn == ptr - thisname &&
1183              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184            return *count;
1185          term++;
1186          }
1187        }
1188      }
1189    
1190    /* Past any initial parenthesis handling, scan for parentheses or vertical
1191    bars. */
1192    
1193    for (; *ptr != 0; ptr++)
1194      {
1195    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1196    
1197    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1198      {      {
1199      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1200      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1201        {        {
1202        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1203        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1204        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1205        }        }
1206      continue;      continue;
1207      }      }
1208    
1209    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1210      are handled for real. If the first character is '^', skip it. Also, if the
1211      first few characters (either before or after ^) are \Q\E or \E we skip them
1212      too. This makes for compatibility with Perl. Note the use of STR macros to
1213      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1214    
1215      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1216        {
1217        BOOL negate_class = FALSE;
1218        for (;;)
1219          {
1220          if (ptr[1] == CHAR_BACKSLASH)
1221            {
1222            if (ptr[2] == CHAR_E)
1223              ptr+= 2;
1224            else if (strncmp((const char *)ptr+2,
1225                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1226              ptr += 4;
1227            else
1228              break;
1229            }
1230          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231            {
1232            negate_class = TRUE;
1233            ptr++;
1234            }
1235          else break;
1236          }
1237    
1238    if (*ptr == '[')      /* If the next character is ']', it is a data character that must be
1239      {      skipped, except in JavaScript compatibility mode. */
1240      while (*(++ptr) != ']')  
1241        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1242            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1243          ptr++;
1244    
1245        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1246        {        {
1247        if (*ptr == '\\')        if (*ptr == 0) return -1;
1248          if (*ptr == CHAR_BACKSLASH)
1249          {          {
1250          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1251          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1252            {            {
1253            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1254            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1255            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1256            }            }
1257          continue;          continue;
1258          }          }
# Line 921  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262    
1263    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1264    
1265    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1266      {      {
1267      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1268      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1269      continue;      continue;
1270      }      }
1271    
1272    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1273    
1274    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1275      {      {
1276      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1277      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1278      continue;      if (*ptr == 0) goto FAIL_EXIT;
1279      }      }
1280    
1281    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1282    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1283        if (dup_parens && *count < hwm_count) *count = hwm_count;
1284        *ptrptr = ptr;
1285        return -1;
1286        }
1287    
1288    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1289        {
1290        if (*count > hwm_count) hwm_count = *count;
1291        *count = start_count;
1292        }
1293      }
1294    
1295    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1296         *ptr != '\'')  *ptrptr = ptr;
1297      continue;  return -1;
1298    }
1299    
1300    
1301    
1302    
1303    /*************************************************
1304    *       Find forward referenced subpattern       *
1305    *************************************************/
1306    
1307    /* This function scans along a pattern's text looking for capturing
1308    subpatterns, and counting them. If it finds a named pattern that matches the
1309    name it is given, it returns its number. Alternatively, if the name is NULL, it
1310    returns when it reaches a given numbered subpattern. This is used for forward
1311    references to subpatterns. We used to be able to start this scan from the
1312    current compiling point, using the current count value from cd->bracount, and
1313    do it all in a single loop, but the addition of the possibility of duplicate
1314    subpattern numbers means that we have to scan from the very start, in order to
1315    take account of such duplicates, and to use a recursive function to keep track
1316    of the different types of group.
1317    
1318    Arguments:
1319      cd           compile background data
1320      name         name to seek, or NULL if seeking a numbered subpattern
1321      lorn         name length, or subpattern number if name is NULL
1322      xmode        TRUE if we are in /x mode
1323    
1324    Returns:       the number of the found subpattern, or -1 if not found
1325    */
1326    
1327    count++;  static int
1328    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1329    {
1330    uschar *ptr = (uschar *)cd->start_pattern;
1331    int count = 0;
1332    int rc;
1333    
1334    /* If the pattern does not start with an opening parenthesis, the first call
1335    to find_parens_sub() will scan right to the end (if necessary). However, if it
1336    does start with a parenthesis, find_parens_sub() will return when it hits the
1337    matching closing parens. That is why we have to have a loop. */
1338    
1339    if (name == NULL && count == lorn) return count;  for (;;)
1340    term = *ptr++;    {
1341    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1343    }    }
1344    
1345  return -1;  return rc;
1346  }  }
1347    
1348    
1349    
1350    
1351  /*************************************************  /*************************************************
1352  *      Find first significant op code            *  *      Find first significant op code            *
1353  *************************************************/  *************************************************/
# Line 1013  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400        case OP_NCREF:
1401      case OP_RREF:      case OP_RREF:
1402        case OP_NRREF:
1403      case OP_DEF:      case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
# Line 1029  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* 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,
1422  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.
1423  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
1424    temporarily terminated with OP_END when this function is called.
1425    
1426    This function is called when a backward assertion is encountered, so that if it
1427    fails, the error message can point to the correct place in the pattern.
1428    However, we cannot do this when the assertion contains subroutine calls,
1429    because they can be forward references. We solve this by remembering this case
1430    and doing the check at the end; a flag specifies which mode we are running in.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435      atend    TRUE if called when the pattern is complete
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 1058  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
   
1460    switch (op)    switch (op)
1461      {      {
1462      case OP_CBRA:      case OP_CBRA:
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1089  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* A true recursion implies not fixed length, but a subroutine call may
1490        be OK. If the subroutine is a forward reference, we can't deal with
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 1102  for (;;) Line 1514  for (;;)
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518      case OP_RREF:      case OP_RREF:
1519        case OP_NRREF:
1520      case OP_DEF:      case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 1125  for (;;) Line 1540  for (;;)
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1545  #endif  #endif
1546      break;      break;
1547    
# Line 1139  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1557  #endif  #endif
1558      break;      break;
1559    
1560      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1561      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1562        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1563      cc += 4;      cc += 4;
1564      break;      break;
1565    
# Line 1165  for (;;) Line 1577  for (;;)
1577      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1578      case OP_WORDCHAR:      case OP_WORDCHAR:
1579      case OP_ANY:      case OP_ANY:
1580        case OP_ALLANY:
1581      branchlength++;      branchlength++;
1582      cc++;      cc++;
1583      break;      break;
# Line 1219  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1640    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  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
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
1655  for (;;)  for (;;)
1656    {    {
# Line 1247  for (;;) Line 1663  for (;;)
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674    /* Handle capturing bracket */    /* Handle capturing bracket */
1675    
1676    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1256  for (;;) Line 1680  for (;;)
1680      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1681      }      }
1682    
1683    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1684    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1685    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
1690        switch(c)
1691          {
1692          case OP_TYPESTAR:
1693          case OP_TYPEMINSTAR:
1694          case OP_TYPEPLUS:
1695          case OP_TYPEMINPLUS:
1696          case OP_TYPEQUERY:
1697          case OP_TYPEMINQUERY:
1698          case OP_TYPEPOSSTAR:
1699          case OP_TYPEPOSPLUS:
1700          case OP_TYPEPOSQUERY:
1701          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1702          break;
1703    
1704          case OP_TYPEUPTO:
1705          case OP_TYPEMINUPTO:
1706          case OP_TYPEEXACT:
1707          case OP_TYPEPOSUPTO:
1708          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709          break;
1710    
1711          case OP_MARK:
1712          case OP_PRUNE_ARG:
1713          case OP_SKIP_ARG:
1714          case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717          }
1718    
1719        /* Add in the fixed length from the table */
1720    
1721      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1722    
1723      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1724      a multi-byte character. The length in the table is a minimum, so we have to
1725      arrange to skip the extra bytes. */
1726    
1727  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1728      if (utf8) switch(c)      if (utf8) switch(c)
1729        {        {
# Line 1284  for (;;) Line 1745  for (;;)
1745        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1746        break;        break;
1747        }        }
1748    #else
1749        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1750  #endif  #endif
1751      }      }
1752    }    }
# Line 1320  for (;;) Line 1783  for (;;)
1783    
1784    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1785    
1786    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1787    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1788    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1789    bytes. */    must add in its length. */
1790    
1791    else    else
1792      {      {
1793        switch(c)
1794          {
1795          case OP_TYPESTAR:
1796          case OP_TYPEMINSTAR:
1797          case OP_TYPEPLUS:
1798          case OP_TYPEMINPLUS:
1799          case OP_TYPEQUERY:
1800          case OP_TYPEMINQUERY:
1801          case OP_TYPEPOSSTAR:
1802          case OP_TYPEPOSPLUS:
1803          case OP_TYPEPOSQUERY:
1804          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1805          break;
1806    
1807          case OP_TYPEPOSUPTO:
1808          case OP_TYPEUPTO:
1809          case OP_TYPEMINUPTO:
1810          case OP_TYPEEXACT:
1811          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812          break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820          }
1821    
1822        /* Add in the fixed length from the table */
1823    
1824      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1825    
1826        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1827        by a multi-byte character. The length in the table is a minimum, so we have
1828        to arrange to skip the extra bytes. */
1829    
1830  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1831      if (utf8) switch(c)      if (utf8) switch(c)
1832        {        {
# Line 1349  for (;;) Line 1848  for (;;)
1848        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1849        break;        break;
1850        }        }
1851    #else
1852        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1853  #endif  #endif
1854      }      }
1855    }    }
# Line 1364  for (;;) Line 1865  for (;;)
1865  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()
1866  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
1867  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1868  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
1869  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
1870    bracket whose current branch will already have been scanned.
1871    
1872  Arguments:  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1387  for (code = first_significant_code(code Line 1891  for (code = first_significant_code(code
1891    
1892    c = *code;    c = *code;
1893    
1894      /* Skip over forward assertions; the other assertions are skipped by
1895      first_significant_code() with a TRUE final argument. */
1896    
1897      if (c == OP_ASSERT)
1898        {
1899        do code += GET(code, 1); while (*code == OP_ALT);
1900        c = *code;
1901        continue;
1902        }
1903    
1904    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1905    
1906    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1907      {      {
1908      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1909      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1397  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      continue;      continue;
1912      }      }
1913    
1914      /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922        do
1923          {
1924          if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926            empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1937    
1938    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1939      {      {
1940      BOOL empty_branch;      BOOL empty_branch;
1941      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1942    
1943      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1944        empty branch, so just skip over the conditional, because it could be empty.
1945        Otherwise, scan the individual branches of the group. */
1946    
1947      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1948        code += GET(code, 1);        code += GET(code, 1);
1949        else
1950          {
1951          empty_branch = FALSE;
1952          do
1953            {
1954            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955              empty_branch = TRUE;
1956            code += GET(code, 1);
1957            }
1958          while (*code == OP_ALT);
1959          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1960        }        }
1961      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1962      c = *code;      c = *code;
1963      continue;      continue;
1964      }      }
# Line 1423  for (code = first_significant_code(code Line 1967  for (code = first_significant_code(code
1967    
1968    switch (c)    switch (c)
1969      {      {
1970      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1971        cannot be represented just by a bit map. This includes negated single
1972        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1973        actual length is stored in the compiled code, so we must update "code"
1974        here. */
1975    
1976  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1977      case OP_XCLASS:      case OP_XCLASS:
1978      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1979      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1980  #endif  #endif
1981    
# Line 1471  for (code = first_significant_code(code Line 2019  for (code = first_significant_code(code
2019      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2020      case OP_WORDCHAR:      case OP_WORDCHAR:
2021      case OP_ANY:      case OP_ANY:
2022        case OP_ALLANY:
2023      case OP_ANYBYTE:      case OP_ANYBYTE:
2024      case OP_CHAR:      case OP_CHAR:
2025      case OP_CHARNC:      case OP_CHARNC:
# Line 1489  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2039      return FALSE;      return FALSE;
2040    
2041        /* These are going to continue, as they may be empty, but we have to
2042        fudge the length for the \p and \P cases. */
2043    
2044        case OP_TYPESTAR:
2045        case OP_TYPEMINSTAR:
2046        case OP_TYPEPOSSTAR:
2047        case OP_TYPEQUERY:
2048        case OP_TYPEMINQUERY:
2049        case OP_TYPEPOSQUERY:
2050        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2051        break;
2052    
2053        /* Same for these */
2054    
2055        case OP_TYPEUPTO:
2056        case OP_TYPEMINUPTO:
2057        case OP_TYPEPOSUPTO:
2058        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2059        break;
2060    
2061      /* End of branch */      /* End of branch */
2062    
2063      case OP_KET:      case OP_KET:
# Line 1507  for (code = first_significant_code(code Line 2076  for (code = first_significant_code(code
2076      case OP_QUERY:      case OP_QUERY:
2077      case OP_MINQUERY:      case OP_MINQUERY:
2078      case OP_POSQUERY:      case OP_POSQUERY:
2079        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080        break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      case OP_POSUPTO:      case OP_POSUPTO:
2085      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1535  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 1558  return TRUE; Line 2147  return TRUE;
2147  *************************************************/  *************************************************/
2148    
2149  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2150  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
2151  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2152  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2153    
2154    Originally, this function only recognized a sequence of letters between the
2155    terminators, but it seems that Perl recognizes any sequence of characters,
2156    though of course unknown POSIX names are subsequently rejected. Perl gives an
2157    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2158    didn't consider this to be a POSIX class. Likewise for [:1234:].
2159    
2160    The problem in trying to be exactly like Perl is in the handling of escapes. We
2161    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2162    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2163    below handles the special case of \], but does not try to do any other escape
2164    processing. This makes it different from Perl for cases such as [:l\ower:]
2165    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2166    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2167    I think.
2168    
2169  Argument:  Arguments:
2170    ptr      pointer to the initial [    ptr      pointer to the initial [
2171    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2172    
2173  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2174  */  */
2175    
2176  static BOOL  static BOOL
2177  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2178  {  {
2179  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2180  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2181  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2182    {    {
2183    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2184    return TRUE;      {
2185        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2186        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2187          {
2188          *endptr = ptr;
2189          return TRUE;
2190          }
2191        }
2192    }    }
2193  return FALSE;  return FALSE;
2194  }  }
# Line 1605  Returns:     a value representing the na Line 2213  Returns:     a value representing the na
2213  static int  static int
2214  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2215  {  {
2216    const char *pn = posix_names;
2217  register int yield = 0;  register int yield = 0;
2218  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2219    {    {
2220    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2221      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2222      pn += posix_name_lengths[yield] + 1;
2223    yield++;    yield++;
2224    }    }
2225  return -1;  return -1;
# Line 1624  return -1; Line 2234  return -1;
2234  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2235  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2236  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
2237  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
2238  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
2239  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
2240  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
2241  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2242    OP_END.
2243    
2244  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2245  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 1651  adjust_recurse(uschar *group, int adjust Line 2262  adjust_recurse(uschar *group, int adjust
2262    uschar *save_hwm)    uschar *save_hwm)
2263  {  {
2264  uschar *ptr = group;  uschar *ptr = group;
2265    
2266  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2267    {    {
2268    int offset;    int offset;
# Line 1762  get_othercase_range(unsigned int *cptr, Line 2374  get_othercase_range(unsigned int *cptr,
2374  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2375    
2376  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2377    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2378    
2379  if (c > d) return FALSE;  if (c > d) return FALSE;
2380    
# Line 1771  next = othercase + 1; Line 2383  next = othercase + 1;
2383    
2384  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2385    {    {
2386    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2387    next++;    next++;
2388    }    }
2389    
# Line 1817  if ((options & PCRE_EXTENDED) != 0) Line 2429  if ((options & PCRE_EXTENDED) != 0)
2429    for (;;)    for (;;)
2430      {      {
2431      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2432      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2433        {        {
2434        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2435          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 2441  if ((options & PCRE_EXTENDED) != 0)
2441  /* 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
2442  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2443    
2444  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2445    {    {
2446    int temperrorcode = 0;    int temperrorcode = 0;
2447    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1854  if ((options & PCRE_EXTENDED) != 0) Line 2466  if ((options & PCRE_EXTENDED) != 0)
2466    for (;;)    for (;;)
2467      {      {
2468      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2469      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2470        {        {
2471        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2472          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1865  if ((options & PCRE_EXTENDED) != 0) Line 2477  if ((options & PCRE_EXTENDED) != 0)
2477    
2478  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2479    
2480  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2481    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2482        return FALSE;
2483    
2484  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2485  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1881  if (next >= 0) switch(op_code) Line 2494  if (next >= 0) switch(op_code)
2494    case OP_CHAR:    case OP_CHAR:
2495  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2496    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2497    #else
2498      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2499  #endif  #endif
2500    return item != next;    return item != next;
2501    
# Line 1899  if (next >= 0) switch(op_code) Line 2514  if (next >= 0) switch(op_code)
2514      unsigned int othercase;      unsigned int othercase;
2515      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2516  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2517      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2518  #else  #else
2519      othercase = NOTACHAR;      othercase = NOTACHAR;
2520  #endif  #endif
# Line 1912  if (next >= 0) switch(op_code) Line 2527  if (next >= 0) switch(op_code)
2527    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2528    
2529    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2530    if (item == next) return TRUE;    if (item == next) return TRUE;
2531    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2532  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1921  if (next >= 0) switch(op_code) Line 2535  if (next >= 0) switch(op_code)
2535      unsigned int othercase;      unsigned int othercase;
2536      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2537  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2538      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2539  #else  #else
2540      othercase = NOTACHAR;      othercase = NOTACHAR;
2541  #endif  #endif
# Line 1949  if (next >= 0) switch(op_code) Line 2563  if (next >= 0) switch(op_code)
2563    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2564    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2565    
2566      case OP_HSPACE:
2567      case OP_NOT_HSPACE:
2568      switch(next)
2569        {
2570        case 0x09:
2571        case 0x20:
2572        case 0xa0:
2573        case 0x1680:
2574        case 0x180e:
2575        case 0x2000:
2576        case 0x2001:
2577        case 0x2002:
2578        case 0x2003:
2579        case 0x2004:
2580        case 0x2005:
2581        case 0x2006:
2582        case 0x2007:
2583        case 0x2008:
2584        case 0x2009:
2585        case 0x200A:
2586        case 0x202f:
2587        case 0x205f:
2588        case 0x3000:
2589        return op_code != OP_HSPACE;
2590        default:
2591        return op_code == OP_HSPACE;
2592        }
2593    
2594      case OP_VSPACE:
2595      case OP_NOT_VSPACE:
2596      switch(next)
2597        {
2598        case 0x0a:
2599        case 0x0b:
2600        case 0x0c:
2601        case 0x0d:
2602        case 0x85:
2603        case 0x2028:
2604        case 0x2029:
2605        return op_code != OP_VSPACE;
2606        default:
2607        return op_code == OP_VSPACE;
2608        }
2609    
2610    default:    default:
2611    return FALSE;    return FALSE;
2612    }    }
# Line 1983  switch(op_code) Line 2641  switch(op_code)
2641      case ESC_W:      case ESC_W:
2642      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2643    
2644        case ESC_h:
2645        case ESC_H:
2646        switch(item)
2647          {
2648          case 0x09:
2649          case 0x20:
2650          case 0xa0:
2651          case 0x1680:
2652          case 0x180e:
2653          case 0x2000:
2654          case 0x2001:
2655          case 0x2002:
2656          case 0x2003:
2657          case 0x2004:
2658          case 0x2005:
2659          case 0x2006:
2660          case 0x2007:
2661          case 0x2008:
2662          case 0x2009:
2663          case 0x200A:
2664          case 0x202f:
2665          case 0x205f:
2666          case 0x3000:
2667          return -next != ESC_h;
2668          default:
2669          return -next == ESC_h;
2670          }
2671    
2672        case ESC_v:
2673        case ESC_V:
2674        switch(item)
2675          {
2676          case 0x0a:
2677          case 0x0b:
2678          case 0x0c:
2679          case 0x0d:
2680          case 0x85:
2681          case 0x2028:
2682          case 0x2029:
2683          return -next != ESC_v;
2684          default:
2685          return -next == ESC_v;
2686          }
2687    
2688      default:      default:
2689      return FALSE;      return FALSE;
2690      }      }
2691    
2692    case OP_DIGIT:    case OP_DIGIT:
2693    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2694             next == -ESC_h || next == -ESC_v;
2695    
2696    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2697    return next == -ESC_d;    return next == -ESC_d;
# Line 1997  switch(op_code) Line 2700  switch(op_code)
2700    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2701    
2702    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2703    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2704    
2705      case OP_HSPACE:
2706      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2707    
2708      case OP_NOT_HSPACE:
2709      return next == -ESC_h;
2710    
2711      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2712      case OP_VSPACE:
2713      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2714    
2715      case OP_NOT_VSPACE:
2716      return next == -ESC_v;
2717    
2718    case OP_WORDCHAR:    case OP_WORDCHAR:
2719    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2720    
2721    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2722    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2064  BOOL inescq = FALSE; Line 2780  BOOL inescq = FALSE;
2780  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2781  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2782  const uschar *tempptr;  const uschar *tempptr;
2783    const uschar *nestptr = NULL;
2784  uschar *previous = NULL;  uschar *previous = NULL;
2785  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2786  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2073  uschar classbits[32]; Line 2790  uschar classbits[32];
2790  BOOL class_utf8;  BOOL class_utf8;
2791  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2792  uschar *class_utf8data;  uschar *class_utf8data;
2793    uschar *class_utf8data_base;
2794  uschar utf8_char[6];  uschar utf8_char[6];
2795  #else  #else
2796  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2797  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2798  #endif  #endif
2799    
2800  #ifdef DEBUG  #ifdef PCRE_DEBUG
2801  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2802  #endif  #endif
2803    
# Line 2112  req_caseopt = ((options & PCRE_CASELESS) Line 2830  req_caseopt = ((options & PCRE_CASELESS)
2830  for (;; ptr++)  for (;; ptr++)
2831    {    {
2832    BOOL negate_class;    BOOL negate_class;
2833      BOOL should_flip_negation;
2834    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2835    BOOL is_quantifier;    BOOL is_quantifier;
2836    BOOL is_recurse;    BOOL is_recurse;
2837    BOOL reset_bracount;    BOOL reset_bracount;
2838    int class_charcount;    int class_charcount;
2839    int class_lastchar;    int class_lastchar;
2840    int newoptions;    int newoptions;
# Line 2132  for (;; ptr++) Line 2851  for (;; ptr++)
2851    
2852    c = *ptr;    c = *ptr;
2853    
2854      /* If we are at the end of a nested substitution, revert to the outer level
2855      string. Nesting only happens one level deep. */
2856    
2857      if (c == 0 && nestptr != NULL)
2858        {
2859        ptr = nestptr;
2860        nestptr = NULL;
2861        c = *ptr;
2862        }
2863    
2864    /* 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
2865    previous cycle of this loop. */    previous cycle of this loop. */
2866    
2867    if (lengthptr != NULL)    if (lengthptr != NULL)
2868      {      {
2869  #ifdef DEBUG  #ifdef PCRE_DEBUG
2870      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2871  #endif  #endif
2872      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2873        {        {
2874        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2875        goto FAILED;        goto FAILED;
# Line 2153  for (;; ptr++) Line 2882  for (;; ptr++)
2882      */      */
2883    
2884      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2885    
2886        /* Paranoid check for integer overflow */
2887    
2888        if (OFLOW_MAX - *lengthptr < code - last_code)
2889          {
2890          *errorcodeptr = ERR20;
2891          goto FAILED;
2892          }
2893    
2894      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2895      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));
2896    
# Line 2180  for (;; ptr++) Line 2918  for (;; ptr++)
2918    /* 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
2919    reference list. */    reference list. */
2920    
2921    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2922      {      {
2923      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2924      goto FAILED;      goto FAILED;
# Line 2190  for (;; ptr++) Line 2928  for (;; ptr++)
2928    
2929    if (inescq && c != 0)    if (inescq && c != 0)
2930      {      {
2931      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2932        {        {
2933        inescq = FALSE;        inescq = FALSE;
2934        ptr++;        ptr++;
# Line 2216  for (;; ptr++) Line 2954  for (;; ptr++)
2954    /* 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
2955    a quantifier. */    a quantifier. */
2956    
2957    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2958      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2959        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2960    
2961    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2962         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2232  for (;; ptr++) Line 2971  for (;; ptr++)
2971    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2972      {      {
2973      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2974      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2975        {        {
2976        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2977          {          {
# Line 2257  for (;; ptr++) Line 2996  for (;; ptr++)
2996      {      {
2997      /* ===================================================================*/      /* ===================================================================*/
2998      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2999      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3000      case ')':      case CHAR_RIGHT_PARENTHESIS:
3001      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3002      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3003      *codeptr = code;      *codeptr = code;
3004      *ptrptr = ptr;      *ptrptr = ptr;
3005      if (lengthptr != NULL)      if (lengthptr != NULL)
3006        {        {
3007          if (OFLOW_MAX - *lengthptr < code - last_code)
3008            {
3009            *errorcodeptr = ERR20;
3010            goto FAILED;
3011            }
3012        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
3013        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3014        }        }
# Line 2275  for (;; ptr++) Line 3019  for (;; ptr++)
3019      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3020      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3021    
3022      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3023      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3024        {        {
3025        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2284  for (;; ptr++) Line 3028  for (;; ptr++)
3028      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3029      break;      break;
3030    
3031      case '$':      case CHAR_DOLLAR_SIGN:
3032      previous = NULL;      previous = NULL;
3033      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3034      break;      break;
# Line 2292  for (;; ptr++) Line 3036  for (;; ptr++)
3036      /* 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
3037      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3038    
3039      case '.':      case CHAR_DOT:
3040      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3041      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3042      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3043      previous = code;      previous = code;
3044      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3045      break;      break;
3046    
3047    
# Line 2312  for (;; ptr++) Line 3056  for (;; ptr++)
3056      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,
3057      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3058      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.
     */  
3059    
3060      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3061        default (Perl) mode, it is treated as a data character. */
3062    
3063        case CHAR_RIGHT_SQUARE_BRACKET:
3064        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3065          {
3066          *errorcodeptr = ERR64;
3067          goto FAILED;
3068          }
3069        goto NORMAL_CHAR;
3070    
3071        case CHAR_LEFT_SQUARE_BRACKET:
3072      previous = code;      previous = code;
3073    
3074      /* 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
3075      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. */
3076    
3077      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3078          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3079            check_posix_syntax(ptr, &tempptr))
3080        {        {
3081        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3082        goto FAILED;        goto FAILED;
3083        }        }
3084    
3085      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3086        if the first few characters (either before or after ^) are \Q\E or \E we
3087        skip them too. This makes for compatibility with Perl. */
3088    
3089      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3090        for (;;)
3091        {        {
       negate_class = TRUE;  
3092        c = *(++ptr);        c = *(++ptr);
3093          if (c == CHAR_BACKSLASH)
3094            {
3095            if (ptr[1] == CHAR_E)
3096              ptr++;
3097            else if (strncmp((const char *)ptr+1,
3098                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3099              ptr += 3;
3100            else
3101              break;
3102            }
3103          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3104            negate_class = TRUE;
3105          else break;
3106        }        }
3107      else  
3108        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3109        an initial ']' is taken as a data character -- the code below handles
3110        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3111        [^] must match any character, so generate OP_ALLANY. */
3112    
3113        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3114            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3115        {        {
3116        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3117          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3118          zerofirstbyte = firstbyte;
3119          break;
3120        }        }
3121    
3122        /* If a class contains a negative special such as \S, we need to flip the
3123        negation flag at the end, so that support for characters > 255 works
3124        correctly (they are all included in the class). */
3125    
3126        should_flip_negation = FALSE;
3127    
3128      /* 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
3129      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
3130      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2356  for (;; ptr++) Line 3142  for (;; ptr++)
3142  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3143      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3144      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3145        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3146  #endif  #endif
3147    
3148      /* 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 2371  for (;; ptr++) Line 3158  for (;; ptr++)
3158          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3159          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3160          }          }
3161    
3162          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3163          data and reset the pointer. This is so that very large classes that
3164          contain a zillion UTF-8 characters no longer overwrite the work space
3165          (which is on the stack). */
3166    
3167          if (lengthptr != NULL)
3168            {
3169            *lengthptr += class_utf8data - class_utf8data_base;
3170            class_utf8data = class_utf8data_base;
3171            }
3172    
3173  #endif  #endif
3174    
3175        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3176    
3177        if (inescq)        if (inescq)
3178          {          {
3179          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3180            {            {
3181            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3182            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2392  for (;; ptr++) Line 3191  for (;; ptr++)
3191        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3192        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3193    
3194        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3195            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3196            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3197          {          {
3198          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3199          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3200          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3201          uschar pbits[32];          uschar pbits[32];
3202    
3203          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3204            {            {
3205            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3206            goto FAILED;            goto FAILED;
3207            }            }
3208    
3209          ptr += 2;          ptr += 2;
3210          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3211            {            {
3212            local_negate = TRUE;            local_negate = TRUE;
3213              should_flip_negation = TRUE;  /* Note negative special */
3214            ptr++;            ptr++;
3215            }            }
3216    
# Line 2427  for (;; ptr++) Line 3227  for (;; ptr++)
3227    
3228          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3229            posix_class = 0;            posix_class = 0;
3230    
3231          /* 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
3232          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3233          subtract bits that may be in the main map already. At the end we or the  
3234          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3235            if ((options & PCRE_UCP) != 0)
3236              {
3237              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3238              if (posix_substitutes[pc] != NULL)
3239                {
3240                nestptr = tempptr + 1;
3241                ptr = posix_substitutes[pc] - 1;
3242                continue;
3243                }
3244              }
3245    #endif
3246            /* In the non-UCP case, we build the bit map for the POSIX class in a
3247            chunk of local store because we may be adding and subtracting from it,
3248            and we don't want to subtract bits that may be in the main map already.
3249            At the end we or the result into the bit map that is being built. */
3250    
3251          posix_class *= 3;          posix_class *= 3;
3252    
# Line 2475  for (;; ptr++) Line 3290  for (;; ptr++)
3290    
3291        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3292        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
3293        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
3294        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
3295        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
3296        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3297          PCRE_EXTRA is set. */
3298    
3299        if (c == '\\')        if (c == CHAR_BACKSLASH)
3300          {          {
3301          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3302          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3303    
3304          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 */  
3305          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3306            {            {
3307            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3308              {              {
3309              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3310              }              }
3311            else inescq = TRUE;            else inescq = TRUE;
3312            continue;            continue;
3313            }            }
3314            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3315    
3316          if (c < 0)          if (c < 0)
3317            {            {
3318            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3319            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3320    
3321            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3322              {              {
3323    #ifdef SUPPORT_UCP
3324                case ESC_du:     /* These are the values given for \d etc */
3325                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3326                case ESC_wu:     /* escape sequence with an appropriate \p */
3327                case ESC_WU:     /* or \P to test Unicode properties instead */
3328                case ESC_su:     /* of the default ASCII testing. */
3329                case ESC_SU:
3330                nestptr = ptr;
3331                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3332                class_charcount -= 2;                /* Undo! */
3333                continue;
3334    #endif
3335              case ESC_d:              case ESC_d:
3336              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3337              continue;              continue;
3338    
3339              case ESC_D:              case ESC_D:
3340                should_flip_negation = TRUE;
3341              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3342              continue;              continue;
3343    
# Line 2520  for (;; ptr++) Line 3346  for (;; ptr++)
3346              continue;              continue;
3347    
3348              case ESC_W:              case ESC_W:
3349                should_flip_negation = TRUE;
3350              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3351              continue;              continue;
3352    
# Line 2529  for (;; ptr++) Line 3356  for (;; ptr++)
3356              continue;              continue;
3357    
3358              case ESC_S:              case ESC_S:
3359                should_flip_negation = TRUE;
3360              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3361              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3362              continue;              continue;
3363    
3364              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)  
             {  
3365              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3366              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3367              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
3368  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3369              if (utf8)              if (utf8)
3370                {                {
3371                class_utf8 = TRUE;                class_utf8 = TRUE;
3372                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3373                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3374                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3375                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3376                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3377                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3378                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3379                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3380                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3381                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3382                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3383                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3384                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3385                }                }
3386  #endif  #endif
3387              continue;              continue;
             }  
3388    
3389            if (-c == ESC_H)              case ESC_H:
             {  
3390              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3391                {                {
3392                int x = 0xff;                int x = 0xff;
3393                switch (c)                switch (c)
3394                  {                  {
3395                  case 0x09/8: x ^= 1 << (0x09%8); break;                  case 0x09/8: x ^= 1 << (0x09%8); break;
3396                  case 0x20/8: x ^= 1 << (0x20%8); break;                  case 0x20/8: x ^= 1 << (0x20%8); break;
3397                  case 0xa0/8: x ^= 1 << (0xa0%8); break;                  case 0xa0/8: x ^= 1 << (0xa0%8); break;
3398                  default: break;                  default: break;
3399                  }                  }
3400                classbits[c] |= x;                classbits[c] |= x;
3401                }                }
3402    
3403  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3404              if (utf8)              if (utf8)
3405                {                {
3406                class_utf8 = TRUE;                class_utf8 = TRUE;
3407                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3408                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3409                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3410                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3411                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3412                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3413                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3414                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3415                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3416                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3417                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3418                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3419                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3420                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3421                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3422                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3423                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3424                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3425                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3426                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3427                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3428                }                }
3429  #endif  #endif
3430              continue;              continue;
             }  
3431    
3432            if (-c == ESC_v)              case ESC_v:
             {  
3433              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3434              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3435              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
3436              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
3437              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
3438  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3439              if (utf8)              if (utf8)
3440                {                {
3441                class_utf8 = TRUE;                class_utf8 = TRUE;
3442                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3443                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3444                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3445                }                }
3446  #endif  #endif
3447              continue;              continue;
             }  
3448    
3449            if (-c == ESC_V)              case ESC_V:
             {  
3450              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3451                {                {
3452                int x = 0xff;                int x = 0xff;
3453                switch (c)                switch (c)
3454                  {                  {
3455                  case 0x0a/8: x ^= 1 << (0x0a%8);                  case 0x0a/8: x ^= 1 << (0x0a%8);
3456                               x ^= 1 << (0x0b%8);                               x ^= 1 << (0x0b%8);
3457                               x ^= 1 << (0x0c%8);                               x ^= 1 << (0x0c%8);
3458                               x ^= 1 << (0x0d%8);                               x ^= 1 << (0x0d%8);
3459                               break;                               break;
3460                  case 0x85/8: x ^= 1 << (0x85%8); break;                  case 0x85/8: x ^= 1 << (0x85%8); break;
3461                  default: break;                  default: break;
3462                  }                  }
3463                classbits[c] |= x;                classbits[c] |= x;
3464                }                }
3465    
3466  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3467              if (utf8)              if (utf8)
3468                {                {
3469                class_utf8 = TRUE;                class_utf8 = TRUE;
3470                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3471                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3472                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3473                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3474                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3475                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3476                }                }
3477  #endif  #endif
3478              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3479    
3480  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3481            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3482              {              case ESC_P:
3483              BOOL negated;                {
3484              int pdata;                BOOL negated;
3485              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3486              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3487              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3488              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3489                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3490              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3491              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3492              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3493              continue;                class_charcount -= 2;   /* Not a < 256 character */
3494              }                continue;
3495                  }
3496  #endif  #endif
3497            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3498            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3499            treated as literals. */              treated as literals. */
3500    
3501            if ((options & PCRE_EXTRA) != 0)              default:
3502              {              if ((options & PCRE_EXTRA) != 0)
3503              *errorcodeptr = ERR7;                {
3504              goto FAILED;                *errorcodeptr = ERR7;
3505                  goto FAILED;
3506                  }
3507                class_charcount -= 2;  /* Undo the default count from above */
3508                c = *ptr;              /* Get the final character and fall through */
3509                break;
3510              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3511            }            }
3512    
3513          /* 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 2715  for (;; ptr++) Line 3521  for (;; ptr++)
3521        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3522    
3523        CHECK_RANGE:        CHECK_RANGE:
3524        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3525          {          {
3526          inescq = FALSE;          inescq = FALSE;
3527          ptr += 2;          ptr += 2;
# Line 2723  for (;; ptr++) Line 3529  for (;; ptr++)
3529    
3530        oldptr = ptr;        oldptr = ptr;
3531    
3532        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3533    
3534          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3535    
3536          /* Check for range */
3537    
3538          if (!inescq && ptr[1] == CHAR_MINUS)
3539          {          {
3540          int d;          int d;
3541          ptr += 2;          ptr += 2;
3542          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3543    
3544          /* 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
3545          mode. */          mode. */
3546    
3547          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3548            {            {
3549            ptr += 2;            ptr += 2;
3550            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3551                { ptr += 2; continue; }
3552            inescq = TRUE;            inescq = TRUE;
3553            break;            break;
3554            }            }
3555    
3556          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3557            {            {
3558            ptr = oldptr;            ptr = oldptr;
3559            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2759  for (;; ptr++) Line 3572  for (;; ptr++)
3572          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
3573          in such circumstances. */          in such circumstances. */
3574    
3575          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3576            {            {
3577            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3578            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3579    
3580            /* \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 */  
3581    
3582            if (d < 0)            if (d < 0)
3583              {              {
3584              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  
3585                {                {
3586                ptr = oldptr;                ptr = oldptr;
3587                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2790  for (;; ptr++) Line 3600  for (;; ptr++)
3600    
3601          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3602    
3603            /* Remember \r or \n */
3604    
3605            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3606    
3607          /* 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
3608          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3609          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2812  for (;; ptr++) Line 3626  for (;; ptr++)
3626              unsigned int origd = d;              unsigned int origd = d;
3627              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3628                {                {
3629                if (occ >= (unsigned int)c &&                if (occ >= (unsigned int)c &&
3630                    ocd <= (unsigned int)d)                    ocd <= (unsigned int)d)
3631                  continue;                          /* Skip embedded ranges */                  continue;                          /* Skip embedded ranges */
3632    
3633                if (occ < (unsigned int)c  &&                if (occ < (unsigned int)c  &&
3634                    ocd >= (unsigned int)c - 1)      /* Extend the basic range */                    ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3635                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3636                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3637                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3638                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3639                if (ocd > (unsigned int)d &&                if (ocd > (unsigned int)d &&
3640                    occ <= (unsigned int)d + 1)      /* always shorter than    */                    occ <= (unsigned int)d + 1)      /* always shorter than    */
3641                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3642                  d = ocd;                  d = ocd;
# Line 2908  for (;; ptr++) Line 3722  for (;; ptr++)
3722          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3723            {            {
3724            unsigned int othercase;            unsigned int othercase;
3725            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3726              {              {
3727              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3728              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2933  for (;; ptr++) Line 3747  for (;; ptr++)
3747          }          }
3748        }        }
3749    
3750      /* 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.
3751        If we are at the end of an internal nested string, revert to the outer
3752        string. */
3753    
3754        while (((c = *(++ptr)) != 0 ||
3755               (nestptr != NULL &&
3756                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3757               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3758    
3759      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3760    
3761      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3762        {        {
3763        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3764        goto FAILED;        goto FAILED;
3765        }        }
3766    
3767      /* 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
3768      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3769      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3770      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3771      single-bytes only. This is an historical hangover. Maybe one day we can  
3772      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3773        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3774        operate on single-bytes only. This is an historical hangover. Maybe one day
3775        we can tidy these opcodes to handle multi-byte characters.
3776    
3777      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
3778      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2956  for (;; ptr++) Line 3780  for (;; ptr++)
3780      can cause firstbyte to be set. Otherwise, there can be no first char if      can cause firstbyte to be set. Otherwise, there can be no first char if
3781      this item is first, whatever repeat count may follow. In the case of      this item is first, whatever repeat count may follow. In the case of
3782      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3783    
3784  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3785      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3786            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3787  #else  #else
3788      if (class_charcount == 1)      if (class_charcount == 1)
3789  #endif  #endif
# Line 3004  for (;; ptr++) Line 3826  for (;; ptr++)
3826      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3827    
3828      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3829      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3830      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
3831        characters > 255 are in the class, so any that were explicitly given as
3832        well can be ignored. If (when there are explicit characters > 255 that must
3833        be listed) there are no characters < 256, we can omit the bitmap in the
3834        actual compiled code. */
3835    
3836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3837      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3838        {        {
3839        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3840        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3034  for (;; ptr++) Line 3860  for (;; ptr++)
3860        }        }
3861  #endif  #endif
3862    
3863      /* 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
3864      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
3865      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
3866      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3867        negating it if necessary. */
3868    
3869        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3870      if (negate_class)      if (negate_class)
3871        {        {
       *code++ = OP_NCLASS;  
3872        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3873          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3874        }        }
3875      else      else
3876        {        {
       *code++ = OP_CLASS;  
3877        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3878        }        }
3879      code += 32;      code += 32;
# Line 3058  for (;; ptr++) Line 3884  for (;; ptr++)
3884      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3885      has been tested above. */      has been tested above. */
3886    
3887      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3888      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3889      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3890      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3891      goto REPEAT;      goto REPEAT;
3892    
3893      case '*':      case CHAR_ASTERISK:
3894      repeat_min = 0;      repeat_min = 0;
3895      repeat_max = -1;      repeat_max = -1;
3896      goto REPEAT;      goto REPEAT;
3897    
3898      case '+':      case CHAR_PLUS:
3899      repeat_min = 1;      repeat_min = 1;
3900      repeat_max = -1;      repeat_max = -1;
3901      goto REPEAT;      goto REPEAT;
3902    
3903      case '?':      case CHAR_QUESTION_MARK:
3904      repeat_min = 0;      repeat_min = 0;
3905      repeat_max = 1;      repeat_max = 1;
3906    
# Line 3109  for (;; ptr++) Line 3935  for (;; ptr++)
3935      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
3936      repeat type to the non-default. */      repeat type to the non-default. */
3937    
3938      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3939        {        {
3940        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3941        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3942        ptr++;        ptr++;
3943        }        }
3944      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3945        {        {
3946        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3947        ptr++;        ptr++;
# Line 3230  for (;; ptr++) Line 4056  for (;; ptr++)
4056    
4057        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4058    
4059          /*--------------------------------------------------------------------*/
4060          /* This code is obsolete from release 8.00; the restriction was finally
4061          removed: */
4062    
4063        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4064        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4065    
4066        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4067          /*--------------------------------------------------------------------*/
4068    
4069        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4070    
# Line 3380  for (;; ptr++) Line 4211  for (;; ptr++)
4211          goto END_REPEAT;          goto END_REPEAT;
4212          }          }
4213    
4214          /*--------------------------------------------------------------------*/
4215          /* This code is obsolete from release 8.00; the restriction was finally
4216          removed: */
4217    
4218        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4219        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4220    
4221        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4222          /*--------------------------------------------------------------------*/
4223    
4224        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4225          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3415  for (;; ptr++) Line 4251  for (;; ptr++)
4251    
4252        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4253          {          {
4254          *errorcodeptr = ERR55;          *errorcodeptr = ERR55;
         goto FAILED;  
         }  
   
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
4255          goto FAILED;          goto FAILED;
4256          }          }
4257    
# Line 3449  for (;; ptr++) Line 4277  for (;; ptr++)
4277    
4278        if (repeat_min == 0)        if (repeat_min == 0)
4279          {          {
4280          /* 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
4281          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4282    
4283          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4284          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4285          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4286          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4287          the start of the whole regex. Temporarily terminate the pattern while          **   }
4288          doing this. */  
4289            However, that fails when a group is referenced as a subroutine from
4290            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4291            so that it is skipped on execution. As we don't have a list of which
4292            groups are referenced, we cannot do this selectively.
4293    
4294            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4295            and do no more at this point. However, we do need to adjust any
4296            OP_RECURSE calls inside the group that refer to the group itself or any
4297            internal or forward referenced group, because the offset is from the
4298            start of the whole regex. Temporarily terminate the pattern while doing
4299            this. */
4300    
4301          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4302            {            {
4303            *code = OP_END;            *code = OP_END;
4304            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4305            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4306            code++;            code++;
4307              if (repeat_max == 0)
4308                {
4309                *previous++ = OP_SKIPZERO;
4310                goto END_REPEAT;
4311                }
4312            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4313            }            }
4314    
# Line 3515  for (;; ptr++) Line 4353  for (;; ptr++)
4353          if (repeat_min > 1)          if (repeat_min > 1)
4354            {            {
4355            /* 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
4356            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4357              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4358              integer type when available, otherwise double. */
4359    
4360            if (lengthptr != NULL)            if (lengthptr != NULL)
4361              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4362                int delta = (repeat_min - 1)*length_prevgroup;
4363                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4364                      (INT64_OR_DOUBLE)length_prevgroup >
4365                        (INT64_OR_DOUBLE)INT_MAX ||
4366                    OFLOW_MAX - *lengthptr < delta)
4367                  {
4368                  *errorcodeptr = ERR20;
4369                  goto FAILED;
4370                  }
4371                *lengthptr += delta;
4372                }
4373    
4374            /* This is compiling for real */            /* This is compiling for real */
4375    
# Line 3556  for (;; ptr++) Line 4407  for (;; ptr++)
4407          /* 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
4408          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
4409          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
4410          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4411            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4412            a 64-bit integer type when available, otherwise double. */
4413    
4414          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4415            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4416              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4417                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4418              if ((INT64_OR_DOUBLE)repeat_max *
4419                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4420                      > (INT64_OR_DOUBLE)INT_MAX ||
4421                  OFLOW_MAX - *lengthptr < delta)
4422                {
4423                *errorcodeptr = ERR20;
4424                goto FAILED;
4425                }
4426              *lengthptr += delta;
4427              }
4428    
4429          /* This is compiling for real */          /* This is compiling for real */
4430    
# Line 3630  for (;; ptr++) Line 4494  for (;; ptr++)
4494            uschar *scode = bracode;            uschar *scode = bracode;
4495            do            do
4496              {              {
4497              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4498                {                {
4499                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4500                break;                break;
# Line 3642  for (;; ptr++) Line 4506  for (;; ptr++)
4506          }          }
4507        }        }
4508    
4509        /* If previous is OP_FAIL, it was generated by an empty class [] in
4510        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4511        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4512        error above. We can just ignore the repeat in JS case. */
4513    
4514        else if (*previous == OP_FAIL) goto END_REPEAT;
4515    
4516      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4517    
4518      else      else
# Line 3666  for (;; ptr++) Line 4537  for (;; ptr++)
4537      if (possessive_quantifier)      if (possessive_quantifier)
4538        {        {
4539        int len;        int len;
4540        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4541            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4542            tempcode += _pcre_OP_lengths[*tempcode] +
4543              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4544    
4545          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4546            {
4547          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4548    #ifdef SUPPORT_UTF8
4549            if (utf8 && tempcode[-1] >= 0xc0)
4550              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4551    #endif
4552            }
4553    
4554        len = code - tempcode;        len = code - tempcode;
4555        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4556          {          {
# Line 3687  for (;; ptr++) Line 4569  for (;; ptr++)
4569          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4570          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4571    
4572            /* Because we are moving code along, we must ensure that any
4573            pending recursive references are updated. */
4574    
4575          default:          default:
4576            *code = OP_END;
4577            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4578          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4579          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4580          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3712  for (;; ptr++) Line 4599  for (;; ptr++)
4599      /* ===================================================================*/      /* ===================================================================*/
4600      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4601      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4602      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4603    
4604      case '(':      case CHAR_LEFT_PARENTHESIS:
4605      newoptions = options;      newoptions = options;
4606      skipbytes = 0;      skipbytes = 0;
4607      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4608      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4609      reset_bracount = FALSE;      reset_bracount = FALSE;
4610    
4611        /* First deal with various "verbs" that can be introduced by '*'. */
4612    
4613        if (*(++ptr) == CHAR_ASTERISK &&
4614             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4615          {
4616          int i, namelen;
4617          int arglen = 0;
4618          const char *vn = verbnames;
4619          const uschar *name = ptr + 1;
4620          const uschar *arg = NULL;
4621          previous = NULL;
4622          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4623          namelen = ptr - name;
4624    
4625          if (*ptr == CHAR_COLON)
4626            {
4627            arg = ++ptr;
4628            while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4629              || *ptr == '_') ptr++;
4630            arglen = ptr - arg;
4631            }
4632    
4633          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4634            {
4635            *errorcodeptr = ERR60;
4636            goto FAILED;
4637            }
4638    
4639          /* Scan the table of verb names */
4640    
4641          for (i = 0; i < verbcount; i++)
4642            {
4643            if (namelen == verbs[i].len &&
4644                strncmp((char *)name, vn, namelen) == 0)
4645              {
4646              /* Check for open captures before ACCEPT */
4647    
4648              if (verbs[i].op == OP_ACCEPT)
4649                {
4650                open_capitem *oc;
4651                cd->had_accept = TRUE;
4652                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4653                  {
4654                  *code++ = OP_CLOSE;
4655                  PUT2INC(code, 0, oc->number);
4656                  }
4657                }
4658    
4659              /* Handle the cases with/without an argument */
4660    
4661              if (arglen == 0)
4662                {
4663                if (verbs[i].op < 0)   /* Argument is mandatory */
4664                  {
4665                  *errorcodeptr = ERR66;
4666                  goto FAILED;
4667                  }
4668                *code++ = verbs[i].op;
4669                }
4670    
4671              else
4672                {
4673                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4674                  {
4675                  *errorcodeptr = ERR59;
4676                  goto FAILED;
4677                  }
4678                *code++ = verbs[i].op_arg;
4679                *code++ = arglen;
4680                memcpy(code, arg, arglen);
4681                code += arglen;
4682                *code++ = 0;
4683                }
4684    
4685              break;  /* Found verb, exit loop */
4686              }
4687    
4688            vn += verbs[i].len + 1;
4689            }
4690    
4691          if (i < verbcount) continue;    /* Successfully handled a verb */
4692          *errorcodeptr = ERR60;          /* Verb not recognized */
4693          goto FAILED;
4694          }
4695    
4696      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4697        appearance of any of them means that this is not a capturing group. */
4698    
4699        else if (*ptr == CHAR_QUESTION_MARK)
4700        {        {
4701        int i, set, unset, namelen;        int i, set, unset, namelen;
4702        int *optset;        int *optset;
# Line 3732  for (;; ptr++) Line 4705  for (;; ptr++)
4705    
4706        switch (*(++ptr))        switch (*(++ptr))
4707          {          {
4708          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4709          ptr++;          ptr++;
4710          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4711          if (*ptr == 0)          if (*ptr == 0)
4712            {            {
4713            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3744  for (;; ptr++) Line 4717  for (;; ptr++)
4717    
4718    
4719          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4720          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4721          reset_bracount = TRUE;          reset_bracount = TRUE;
4722          /* Fall through */          /* Fall through */
4723    
4724          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4725          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4726          bravalue = OP_BRA;          bravalue = OP_BRA;
4727          ptr++;          ptr++;
4728          break;          break;
4729    
4730    
4731          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4732          case '(':          case CHAR_LEFT_PARENTHESIS:
4733          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4734    
4735          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 3776  for (;; ptr++) Line 4749  for (;; ptr++)
4749          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4750          including assertions, are processed. */          including assertions, are processed. */
4751    
4752          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4753                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4754            break;            break;
4755    
4756          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 3788  for (;; ptr++) Line 4762  for (;; ptr++)
4762    
4763          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4764    
4765          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4766            {            {
4767            terminator = -1;            terminator = -1;
4768            ptr += 2;            ptr += 2;
# Line 3798  for (;; ptr++) Line 4772  for (;; ptr++)
4772          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4773          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4774    
4775          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4776            {            {
4777            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4778            ptr++;            ptr++;
4779            }            }
4780          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4781            {            {
4782            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4783            ptr++;            ptr++;
4784            }            }
4785          else          else
4786            {            {
4787            terminator = 0;            terminator = 0;
4788            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4789            }            }
4790    
4791          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 3831  for (;; ptr++) Line 4805  for (;; ptr++)
4805            {            {
4806            if (recno >= 0)            if (recno >= 0)
4807              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4808                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4809            ptr++;            ptr++;
4810            }            }
4811          namelen = ptr - name;          namelen = ptr - name;
4812    
4813          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4814                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4815            {            {
4816            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4817            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3858  for (;; ptr++) Line 4833  for (;; ptr++)
4833              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4834              goto FAILED;              goto FAILED;
4835              }              }
4836            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4837                cd->bracount - recno + 1 : recno +cd->bracount;
4838              if (recno <= 0 || recno > cd->final_bracount)
4839              {              {
4840              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4841              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4842              }              }
           else recno += cd->bracount;  
4843            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4844            break;            break;
4845            }            }
4846    
4847          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4848          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4849            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4850            except they record that the reference was originally to a name. The
4851            information is used to check duplicate names. */
4852    
4853          slot = cd->name_table;          slot = cd->name_table;
4854          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3888  for (;; ptr++) Line 4863  for (;; ptr++)
4863            {            {
4864            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4865            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4866              code[1+LINK_SIZE]++;
4867            }            }
4868    
4869          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4870    
4871          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4872                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4873            {            {
4874            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4875              code[1+LINK_SIZE]++;
4876            }            }
4877    
4878          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 3913  for (;; ptr++) Line 4890  for (;; ptr++)
4890          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4891          specific group number. */          specific group number. */
4892    
4893          else if (*name == 'R')          else if (*name == CHAR_R)
4894            {            {
4895            recno = 0;            recno = 0;
4896            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3923  for (;; ptr++) Line 4900  for (;; ptr++)
4900                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4901                goto FAILED;                goto FAILED;
4902                }                }
4903              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4904              }              }
4905            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4906            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3933  for (;; ptr++) Line 4910  for (;; ptr++)
4910          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4911          false. */          false. */
4912    
4913          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4914            {            {
4915            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4916            skipbytes = 1;            skipbytes = 1;
4917            }            }
4918    
4919          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4920            in the second pass here, so final_bracount is set. */
4921    
4922          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4923            {            {
4924            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4925            }            }
# Line 3957  for (;; ptr++) Line 4935  for (;; ptr++)
4935    
4936    
4937          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4938          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4939          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4940          ptr++;          ptr++;
4941          break;          break;
4942    
4943    
4944          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4945          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4946          ptr++;          ptr++;
4947            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4948              {
4949              *code++ = OP_FAIL;
4950              previous = NULL;
4951              continue;
4952              }
4953            bravalue = OP_ASSERT_NOT;
4954          break;          break;
4955    
4956    
4957          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4958          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4959          switch (ptr[1])          switch (ptr[1])
4960            {            {
4961            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4962            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4963            ptr += 2;            ptr += 2;
4964            break;            break;
4965    
4966            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4967            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4968            ptr += 2;            ptr += 2;
4969            break;            break;
# Line 3994  for (;; ptr++) Line 4978  for (;; ptr++)
4978    
4979    
4980          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4981          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4982          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4983          ptr++;          ptr++;
4984          break;          break;
4985    
4986    
4987          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4988          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4989          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4990          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4991          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4992            {            {
4993            int n = 0;            int n = 0;
4994            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4995              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4996            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4997              {              {
4998              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4999              goto FAILED;              goto FAILED;
# Line 4029  for (;; ptr++) Line 5013  for (;; ptr++)
5013    
5014    
5015          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5016          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5017          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5018                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5019            {            {
5020            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5021            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5022            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5023            }            }
5024          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5025            {            {
5026            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5027            goto FAILED;            goto FAILED;
# Line 4046  for (;; ptr++) Line 5031  for (;; ptr++)
5031    
5032          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5033          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5034          case '\'':          case CHAR_APOSTROPHE:
5035            {            {
5036            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5037                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5038            name = ++ptr;            name = ++ptr;
5039    
5040            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4079  for (;; ptr++) Line 5065  for (;; ptr++)
5065                }                }
5066              }              }
5067    
5068            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5069              alphabetical order. Duplicate names for different numbers are
5070              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5071              number are always OK. (An existing number can be re-used if (?|
5072              appears in the pattern.) In either event, a duplicate name results in
5073              a duplicate entry in the table, even if the number is the same. This
5074              is because the number of names, and hence the table size, is computed
5075              in the pre-compile, and it affects various numbers and pointers which
5076              would all have to be modified, and the compiled code moved down, if
5077              duplicates with the same number were omitted from the table. This
5078              doesn't seem worth the hassle. However, *different* names for the
5079              same number are not permitted. */
5080    
5081            else            else
5082              {              {
5083                BOOL dupname = FALSE;
5084              slot = cd->name_table;              slot = cd->name_table;
5085    
5086              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5087                {                {
5088                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4091  for (;; ptr++) Line 5090  for (;; ptr++)
5090                  {                  {
5091                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5092                    {                    {
5093                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5094                          (options & PCRE_DUPNAMES) == 0)
5095                      {                      {
5096                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5097                      goto FAILED;                      goto FAILED;
5098                      }                      }
5099                      else dupname = TRUE;
5100                    }                    }
5101                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5102                  }                  }
5103    
5104                  /* Make space in the table and break the loop for an earlier
5105                  name. For a duplicate or later name, carry on. We do this for
5106                  duplicates so that in the simple case (when ?(| is not used) they
5107                  are in order of their numbers. */
5108    
5109                if (crc < 0)                if (crc < 0)
5110                  {                  {
5111                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5112                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5113                  break;                  break;
5114                  }                  }
5115    
5116                  /* Continue the loop for a later or duplicate name */
5117    
5118                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5119                }                }
5120    
5121                /* For non-duplicate names, check for a duplicate number before
5122                adding the new name. */
5123    
5124                if (!dupname)
5125                  {
5126                  uschar *cslot = cd->name_table;
5127                  for (i = 0; i < cd->names_found; i++)
5128                    {
5129                    if (cslot != slot)
5130                      {
5131                      if (GET2(cslot, 0) == cd->bracount + 1)
5132                        {
5133                        *errorcodeptr = ERR65;
5134                        goto FAILED;
5135                        }
5136                      }
5137                    else i--;
5138                    cslot += cd->name_entry_size;
5139                    }
5140                  }
5141    
5142              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5143              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5144              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5145              }              }
5146            }            }
5147    
5148          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5149            encountered. */
5150    
         ptr++;                    /* Move past > or ' */  
5151          cd->names_found++;          cd->names_found++;
5152            ptr++;                    /* Move past > or ' */
5153          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5154    
5155    
5156          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5157          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5158          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5159          is_recurse = TRUE;          is_recurse = TRUE;
5160          /* Fall through */          /* Fall through */
5161    
5162          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
5163          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5164          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
5165            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5166            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5167    
5168          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5169          name = ++ptr;          name = ++ptr;
# Line 4141  for (;; ptr++) Line 5175  for (;; ptr++)
5175    
5176          if (lengthptr != NULL)          if (lengthptr != NULL)
5177            {            {
5178              if (namelen == 0)
5179                {
5180                *errorcodeptr = ERR62;
5181                goto FAILED;
5182                }
5183            if (*ptr != terminator)            if (*ptr != terminator)
5184              {              {
5185              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4154  for (;; ptr++) Line 5193  for (;; ptr++)
5193            recno = 0;            recno = 0;
5194            }            }
5195    
5196          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5197            first, and then check that we have reached the end of the name in the
5198            table. That way, if the name that is longer than any in the table,
5199            the comparison will fail without reading beyond the table entry. */
5200    
5201          else          else
5202            {            {
5203            slot = cd->name_table;            slot = cd->name_table;
5204            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5205              {              {
5206              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5207                    slot[2+namelen] == 0)
5208                  break;
5209              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5210              }              }
5211    
# Line 4170  for (;; ptr++) Line 5214  for (;; ptr++)
5214              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5215              }              }
5216            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5217                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5218                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5219              {              {
5220              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4186  for (;; ptr++) Line 5230  for (;; ptr++)
5230    
5231    
5232          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5233          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5234          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5235          /* Fall through */          /* Fall through */
5236    
5237    
5238          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5239          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5240          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5241          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5242            {            {
5243            const uschar *called;            const uschar *called;
5244              terminator = CHAR_RIGHT_PARENTHESIS;
5245    
5246              /* Come here from the \g<...> and \g'...' code (Oniguruma
5247              compatibility). However, the syntax has been checked to ensure that
5248              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5249              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5250              ever be taken. */
5251    
5252            if ((refsign = *ptr) == '+') ptr++;            HANDLE_NUMERICAL_RECURSION:
5253            else if (refsign == '-')  
5254              if ((refsign = *ptr) == CHAR_PLUS)
5255                {
5256                ptr++;
5257                if ((digitab[*ptr] & ctype_digit) == 0)
5258                  {
5259                  *errorcodeptr = ERR63;
5260                  goto FAILED;
5261                  }
5262                }
5263              else if (refsign == CHAR_MINUS)
5264              {              {
5265              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5266                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4208  for (;; ptr++) Line 5269  for (;; ptr++)
5269    
5270            recno = 0;            recno = 0;
5271            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5272              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5273    
5274            if (*ptr != ')')            if (*ptr != terminator)
5275              {              {
5276              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5277              goto FAILED;              goto FAILED;
5278              }              }
5279    
5280            if (refsign == '-')            if (refsign == CHAR_MINUS)
5281              {              {
5282              if (recno == 0)              if (recno == 0)
5283                {                {
# Line 4230  for (;; ptr++) Line 5291  for (;; ptr++)
5291                goto FAILED;                goto FAILED;
5292                }                }
5293              }              }
5294            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5295              {              {
5296              if (recno == 0)              if (recno == 0)
5297                {                {
# Line 4257  for (;; ptr++) Line 5318  for (;; ptr++)
5318            if (lengthptr == NULL)            if (lengthptr == NULL)
5319              {              {
5320              *code = OP_END;              *code = OP_END;
5321              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5322                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5323    
5324              /* Forward reference */              /* Forward reference */
5325    
5326              if (called == NULL)              if (called == NULL)
5327                {                {
5328                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5329                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5330                  {                  {
5331                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5332                  goto FAILED;                  goto FAILED;
5333                  }                  }
5334    
5335                  /* Fudge the value of "called" so that when it is inserted as an
5336                  offset below, what it actually inserted is the reference number
5337                  of the group. */
5338    
5339                called = cd->start_code + recno;                called = cd->start_code + recno;
5340                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5341                }                }
# Line 4278  for (;; ptr++) Line 5345  for (;; ptr++)
5345              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */