/[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 213 by ph10, Wed Aug 15 11:34:14 2007 UTC revision 758 by ph10, Mon Nov 21 12:05:36 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185    searched linearly. Put all the names into a single string, in order to reduce
186    the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    const char *name;    int   len;                 /* Length of verb name */
192    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static verbitem verbs[] = {  static const char verbnames[] =
197    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
198    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
199    { "F",      1, OP_FAIL },    STRING_ACCEPT0
200    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
201    { "PRUNE",  5, OP_PRUNE },    STRING_F0
202    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
203    { "THEN",   4, OP_THEN  }    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
220    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
221    
222  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
223    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
225    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 200  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 207  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
317  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression is too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number or name after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
387    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized",    "(*VERB) not recognized\0"
403    "number is too big"    "number is too big\0"
404  };    "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      "\\N is not supported in a class\0"
416      ;
417    
418  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
419  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 303  For convenience, we use the same bit def Line 431  For convenience, we use the same bit def
431    
432  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
433    
434  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
435    
436    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
437    UTF-8 mode. */
438    
439  static const unsigned char digitab[] =  static const unsigned char digitab[] =
440    {    {
441    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 339  static const unsigned char digitab[] = Line 471  static const unsigned char digitab[] =
471    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
472    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
473    
474  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
475    
476    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
477    
478  static const unsigned char digitab[] =  static const unsigned char digitab[] =
479    {    {
480    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 414  static const unsigned char ebcdic_charta Line 549  static const unsigned char ebcdic_charta
549  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
550    
551  static BOOL  static BOOL
552    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
553      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
554    
555    
556    
557  /*************************************************  /*************************************************
558    *            Find an error text                  *
559    *************************************************/
560    
561    /* The error texts are now all in one long string, to save on relocations. As
562    some of the text is of unknown length, we can't use a table of offsets.
563    Instead, just count through the strings. This is not a performance issue
564    because it happens only when there has been a compilation error.
565    
566    Argument:   the error number
567    Returns:    pointer to the error string
568    */
569    
570    static const char *
571    find_error_text(int n)
572    {
573    const char *s = error_texts;
574    for (; n > 0; n--)
575      {
576      while (*s++ != 0) {};
577      if (*s == 0) return "Error text not found (please report)";
578      }
579    return s;
580    }
581    
582    
583    /*************************************************
584    *            Check for counted repeat            *
585    *************************************************/
586    
587    /* This function is called when a '{' is encountered in a place where it might
588    start a quantifier. It looks ahead to see if it really is a quantifier or not.
589    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
590    where the ddds are digits.
591    
592    Arguments:
593      p         pointer to the first char after '{'
594    
595    Returns:    TRUE or FALSE
596    */
597    
598    static BOOL
599    is_counted_repeat(const uschar *p)
600    {
601    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
602    while ((digitab[*p] & ctype_digit) != 0) p++;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if (*p++ != CHAR_COMMA) return FALSE;
606    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
607    
608    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
609    while ((digitab[*p] & ctype_digit) != 0) p++;
610    
611    return (*p == CHAR_RIGHT_CURLY_BRACKET);
612    }
613    
614    
615    
616    /*************************************************
617  *            Handle escapes                      *  *            Handle escapes                      *
618  *************************************************/  *************************************************/
619    
# Line 458  ptr--;                            /* Set Line 652  ptr--;                            /* Set
652    
653  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
654    
655  /* 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
656  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.
657  Otherwise further processing may be required. */  Otherwise further processing may be required. */
658    
659  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
660  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
661  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
662    
663  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
664  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
665  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
666  #endif  #endif
667    
# Line 483  else Line 677  else
677      /* 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
678      error. */      error. */
679    
680      case 'l':      case CHAR_l:
681      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
682      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
683      break;      break;
684    
685      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
686      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
687      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
688      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
689      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
690      fudge it up by pretending it really was \k. */        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
691               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
692            {
693            c = 0;
694            for (i = 0; i < 4; ++i)
695              {
696              register int cc = *(++ptr);
697    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
698              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
699              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
700    #else           /* EBCDIC coding */
701              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
702              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
703    #endif
704              }
705            }
706          }
707        else
708          *errorcodeptr = ERR37;
709        break;
710    
711        case CHAR_U:
712        /* In JavaScript, \U is an uppercase U letter. */
713        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
714        break;
715    
716        /* In a character class, \g is just a literal "g". Outside a character
717        class, \g must be followed by one of a number of specific things:
718    
719        (1) A number, either plain or braced. If positive, it is an absolute
720        backreference. If negative, it is a relative backreference. This is a Perl
721        5.10 feature.
722    
723        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
724        is part of Perl's movement towards a unified syntax for back references. As
725        this is synonymous with \k{name}, we fudge it up by pretending it really
726        was \k.
727    
728        (3) For Oniguruma compatibility we also support \g followed by a name or a
729        number either in angle brackets or in single quotes. However, these are
730        (possibly recursive) subroutine calls, _not_ backreferences. Just return
731        the -ESC_g code (cf \k). */
732    
733        case CHAR_g:
734        if (isclass) break;
735        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
736          {
737          c = -ESC_g;
738          break;
739          }
740    
741      case 'g':      /* Handle the Perl-compatible cases */
742      if (ptr[1] == '{')  
743        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
744        {        {
745        const uschar *p;        const uschar *p;
746        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
747          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
748        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
749          {          {
750          c = -ESC_k;          c = -ESC_k;
751          break;          break;
# Line 514  else Line 755  else
755        }        }
756      else braced = FALSE;      else braced = FALSE;
757    
758      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
759        {        {
760        negated = TRUE;        negated = TRUE;
761        ptr++;        ptr++;
# Line 523  else Line 764  else
764    
765      c = 0;      c = 0;
766      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
767        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
768    
769      if (c < 0)      if (c < 0)   /* Integer overflow */
770        {        {
771        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
772        break;        break;
773        }        }
774    
775      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
776        {        {
777        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
778        break;        break;
779        }        }
780    
781        if (c == 0)
782          {
783          *errorcodeptr = ERR58;
784          break;
785          }
786    
787      if (negated)      if (negated)
788        {        {
789        if (c > bracount)        if (c > bracount)
# Line 562  else Line 809  else
809      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
810      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
811    
812      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:
813      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
814    
815      if (!isclass)      if (!isclass)
816        {        {
817        oldptr = ptr;        oldptr = ptr;
818        c -= '0';        c -= CHAR_0;
819        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
820          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
821        if (c < 0)        if (c < 0)    /* Integer overflow */
822          {          {
823          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
824          break;          break;
825          }          }
826        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
827          {          {
828          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 588  else Line 835  else
835      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.
836      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
837    
838      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
839        {        {
840        ptr--;        ptr--;
841        c = 0;        c = 0;
# Line 601  else Line 848  else
848      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
849      than 3 octal digits. */      than 3 octal digits. */
850    
851      case '0':      case CHAR_0:
852      c -= '0';      c -= CHAR_0;
853      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
854          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
855      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
856      break;      break;
857    
# Line 612  else Line 859  else
859      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
860      treated as a data character. */      treated as a data character. */
861    
862      case 'x':      case CHAR_x:
863      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
864          {
865          /* In JavaScript, \x must be followed by two hexadecimal numbers.
866          Otherwise it is a lowercase x letter. */
867          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
868            {
869            c = 0;
870            for (i = 0; i < 2; ++i)
871              {
872              register int cc = *(++ptr);
873    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
874              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
875              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
876    #else           /* EBCDIC coding */
877              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
878              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
879    #endif
880              }
881            }
882          break;
883          }
884    
885        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
886        {        {
887        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
888        int count = 0;        int count = 0;
# Line 622  else Line 891  else
891        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
892          {          {
893          register int cc = *pt++;          register int cc = *pt++;
894          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
895          count++;          count++;
896    
897  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
898          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
899          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
900  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
901          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
902          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
903  #endif  #endif
904          }          }
905    
906        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
907          {          {
908          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
909          ptr = pt;          ptr = pt;
# Line 650  else Line 919  else
919      c = 0;      c = 0;
920      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
921        {        {
922        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
923        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
924  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
925        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
926        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
927  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
928        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
929        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
930  #endif  #endif
931        }        }
932      break;      break;
933    
934      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
935      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
936        coding is ASCII-specific, but then the whole concept of \cx is
937      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
938    
939      case 'c':      case CHAR_c:
940      c = *(++ptr);      c = *(++ptr);
941      if (c == 0)      if (c == 0)
942        {        {
943        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
944        break;        break;
945        }        }
946    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
947  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
948      if (c >= 'a' && c <= 'z') c -= 32;        {
949          *errorcodeptr = ERR68;
950          break;
951          }
952        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
953      c ^= 0x40;      c ^= 0x40;
954  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
955      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
956      c ^= 0xC0;      c ^= 0xC0;
957  #endif  #endif
958      break;      break;
959    
960      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
961      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
962      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
963      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
964      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
965    
966      default:      default:
967      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 974  else
974      }      }
975    }    }
976    
977    /* Perl supports \N{name} for character names, as well as plain \N for "not
978    newline". PCRE does not support \N{name}. However, it does support
979    quantification such as \N{2,3}. */
980    
981    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
982         !is_counted_repeat(ptr+2))
983      *errorcodeptr = ERR37;
984    
985    /* If PCRE_UCP is set, we change the values for \d etc. */
986    
987    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
988      c -= (ESC_DU - ESC_D);
989    
990    /* Set the pointer to the final character before returning. */
991    
992  *ptrptr = ptr;  *ptrptr = ptr;
993  return c;  return c;
994  }  }
# Line 740  if (c == 0) goto ERROR_RETURN; Line 1029  if (c == 0) goto ERROR_RETURN;
1029  /* \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
1030  negation. */  negation. */
1031    
1032  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1033    {    {
1034    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1035      {      {
1036      *negptr = TRUE;      *negptr = TRUE;
1037      ptr++;      ptr++;
# Line 751  if (c == '{') Line 1040  if (c == '{')
1040      {      {
1041      c = *(++ptr);      c = *(++ptr);
1042      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1043      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1044      name[i] = c;      name[i] = c;
1045      }      }
1046    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1047    name[i] = 0;    name[i] = 0;
1048    }    }
1049    
# Line 776  top = _pcre_utt_size; Line 1065  top = _pcre_utt_size;
1065  while (bot < top)  while (bot < top)
1066    {    {
1067    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1068    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1069    if (c == 0)    if (c == 0)
1070      {      {
1071      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 800  return -1; Line 1089  return -1;
1089    
1090    
1091  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1092  *         Read repeat counts                     *  *         Read repeat counts                     *
1093  *************************************************/  *************************************************/
1094    
# Line 860  int max = -1; Line 1116  int max = -1;
1116  /* 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
1117  an integer overflow. */  an integer overflow. */
1118    
1119  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1120  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1121    {    {
1122    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1126  if (min < 0 || min > 65535)
1126  /* 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.
1127  Also, max must not be less than min. */  Also, max must not be less than min. */
1128    
1129  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1130    {    {
1131    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1132      {      {
1133      max = 0;      max = 0;
1134      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1135      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1136        {        {
1137        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1156  return p;
1156    
1157    
1158  /*************************************************  /*************************************************
1159  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1160  *************************************************/  *************************************************/
1161    
1162  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1163    top-level call starts at the beginning of the pattern. All other calls must
1164    start at a parenthesis. It scans along a pattern's text looking for capturing
1165  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
1166  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
1167  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1168  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1169  be terminated by '>' because that is checked in the first pass.  
1170    This function was originally called only from the second pass, in which we know
1171    that if (?< or (?' or (?P< is encountered, the name will be correctly
1172    terminated because that is checked in the first pass. There is now one call to
1173    this function in the first pass, to check for a recursive back reference by
1174    name (so that we can make the whole group atomic). In this case, we need check
1175    only up to the current position in the pattern, and that is still OK because
1176    and previous occurrences will have been checked. To make this work, the test
1177    for "end of pattern" is a check against cd->end_pattern in the main loop,
1178    instead of looking for a binary zero. This means that the special first-pass
1179    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1180    processing items within the loop are OK, because afterwards the main loop will
1181    terminate.)
1182    
1183  Arguments:  Arguments:
1184    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1185    count        current count of capturing parens so far encountered    cd           compile background data
1186    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1187    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1188    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1189      utf8         TRUE if we are in UTF-8 mode
1190      count        pointer to the current capturing subpattern number (updated)
1191    
1192  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1193  */  */
1194    
1195  static int  static int
1196  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,
1197    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1198  {  {
1199  const uschar *thisname;  uschar *ptr = *ptrptr;
1200    int start_count = *count;
1201    int hwm_count = start_count;
1202    BOOL dup_parens = FALSE;
1203    
1204    /* If the first character is a parenthesis, check on the type of group we are
1205    dealing with. The very first call may not start with a parenthesis. */
1206    
1207  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1208    {    {
1209    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1210    
1211      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1212    
1213      /* Handle a normal, unnamed capturing parenthesis. */
1214    
1215      else if (ptr[1] != CHAR_QUESTION_MARK)
1216        {
1217        *count += 1;
1218        if (name == NULL && *count == lorn) return *count;
1219        ptr++;
1220        }
1221    
1222      /* All cases now have (? at the start. Remember when we are in a group
1223      where the parenthesis numbers are duplicated. */
1224    
1225      else if (ptr[2] == CHAR_VERTICAL_LINE)
1226        {
1227        ptr += 3;
1228        dup_parens = TRUE;
1229        }
1230    
1231      /* Handle comments; all characters are allowed until a ket is reached. */
1232    
1233      else if (ptr[2] == CHAR_NUMBER_SIGN)
1234        {
1235        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1236        goto FAIL_EXIT;
1237        }
1238    
1239      /* Handle a condition. If it is an assertion, just carry on so that it
1240      is processed as normal. If not, skip to the closing parenthesis of the
1241      condition (there can't be any nested parens). */
1242    
1243      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1244        {
1245        ptr += 2;
1246        if (ptr[1] != CHAR_QUESTION_MARK)
1247          {
1248          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1249          if (*ptr != 0) ptr++;
1250          }
1251        }
1252    
1253      /* Start with (? but not a condition. */
1254    
1255      else
1256        {
1257        ptr += 2;
1258        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1259    
1260        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1261    
1262        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1263            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1264          {
1265          int term;
1266          const uschar *thisname;
1267          *count += 1;
1268          if (name == NULL && *count == lorn) return *count;
1269          term = *ptr++;
1270          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1271          thisname = ptr;
1272          while (*ptr != term) ptr++;
1273          if (name != NULL && lorn == ptr - thisname &&
1274              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1275            return *count;
1276          term++;
1277          }
1278        }
1279      }
1280    
1281    /* Past any initial parenthesis handling, scan for parentheses or vertical
1282    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1283    first-pass call when this value is temporarily adjusted to stop at the current
1284    position. So DO NOT change this to a test for binary zero. */
1285    
1286    for (; ptr < cd->end_pattern; ptr++)
1287      {
1288    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1289    
1290    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1291      {      {
1292      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1293      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1294        {        {
1295        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1296        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1297        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1298        }        }
1299      continue;      continue;
1300      }      }
1301    
1302    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1303      are handled for real. If the first character is '^', skip it. Also, if the
1304      first few characters (either before or after ^) are \Q\E or \E we skip them
1305      too. This makes for compatibility with Perl. Note the use of STR macros to
1306      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1307    
1308    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1309      {      {
1310      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1311        for (;;)
1312          {
1313          if (ptr[1] == CHAR_BACKSLASH)
1314            {
1315            if (ptr[2] == CHAR_E)
1316              ptr+= 2;
1317            else if (strncmp((const char *)ptr+2,
1318                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1319              ptr += 4;
1320            else
1321              break;
1322            }
1323          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1324            {
1325            negate_class = TRUE;
1326            ptr++;
1327            }
1328          else break;
1329          }
1330    
1331        /* If the next character is ']', it is a data character that must be
1332        skipped, except in JavaScript compatibility mode. */
1333    
1334        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1335            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1336          ptr++;
1337    
1338        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1339        {        {
1340        if (*ptr == '\\')        if (*ptr == 0) return -1;
1341          if (*ptr == CHAR_BACKSLASH)
1342          {          {
1343          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1344          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1345            {            {
1346            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1347            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1348            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1349            }            }
1350          continue;          continue;
1351          }          }
# Line 967  for (; *ptr != 0; ptr++) Line 1355  for (; *ptr != 0; ptr++)
1355    
1356    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1357    
1358    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1359      {      {
1360      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1361      if (*ptr == 0) return -1;      while (*ptr != 0)
1362          {
1363          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1364          ptr++;
1365    #ifdef SUPPORT_UTF8
1366          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1367    #endif
1368          }
1369        if (*ptr == 0) goto FAIL_EXIT;
1370      continue;      continue;
1371      }      }
1372    
1373    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1374    
1375    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1376      {      {
1377      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1378      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1379      continue;      if (*ptr == 0) goto FAIL_EXIT;
1380      }      }
1381    
1382    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1383    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1384        if (dup_parens && *count < hwm_count) *count = hwm_count;
1385        goto FAIL_EXIT;
1386        }
1387    
1388      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1389        {
1390        if (*count > hwm_count) hwm_count = *count;
1391        *count = start_count;
1392        }
1393      }
1394    
1395    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1396    *ptrptr = ptr;
1397    return -1;
1398    }
1399    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1400    
   count++;  
1401    
1402    if (name == NULL && count == lorn) return count;  
1403    term = *ptr++;  /*************************************************
1404    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1405    thisname = ptr;  *************************************************/
1406    while (*ptr != term) ptr++;  
1407    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1408        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1409      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1410    returns when it reaches a given numbered subpattern. This is used for forward
1411    references to subpatterns. We used to be able to start this scan from the
1412    current compiling point, using the current count value from cd->bracount, and
1413    do it all in a single loop, but the addition of the possibility of duplicate
1414    subpattern numbers means that we have to scan from the very start, in order to
1415    take account of such duplicates, and to use a recursive function to keep track
1416    of the different types of group.
1417    
1418    Arguments:
1419      cd           compile background data
1420      name         name to seek, or NULL if seeking a numbered subpattern
1421      lorn         name length, or subpattern number if name is NULL
1422      xmode        TRUE if we are in /x mode
1423      utf8         TRUE if we are in UTF-8 mode
1424    
1425    Returns:       the number of the found subpattern, or -1 if not found
1426    */
1427    
1428    static int
1429    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1430      BOOL utf8)
1431    {
1432    uschar *ptr = (uschar *)cd->start_pattern;
1433    int count = 0;
1434    int rc;
1435    
1436    /* If the pattern does not start with an opening parenthesis, the first call
1437    to find_parens_sub() will scan right to the end (if necessary). However, if it
1438    does start with a parenthesis, find_parens_sub() will return when it hits the
1439    matching closing parens. That is why we have to have a loop. */
1440    
1441    for (;;)
1442      {
1443      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1444      if (rc > 0 || *ptr++ == 0) break;
1445    }    }
1446    
1447  return -1;  return rc;
1448  }  }
1449    
1450    
1451    
1452    
1453  /*************************************************  /*************************************************
1454  *      Find first significant op code            *  *      Find first significant op code            *
1455  *************************************************/  *************************************************/
1456    
1457  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1458  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1459  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1460  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1461  assertions, and also the \b assertion; for others it does not.  does not.
1462    
1463  Arguments:  Arguments:
1464    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1465    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1466    
1467  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1468  */  */
1469    
1470  static const uschar*  static const uschar*
1471  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1472  {  {
1473  for (;;)  for (;;)
1474    {    {
1475    switch ((int)*code)    switch ((int)*code)
1476      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1477      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1478      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1479      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1059  for (;;) Line 1489  for (;;)
1489    
1490      case OP_CALLOUT:      case OP_CALLOUT:
1491      case OP_CREF:      case OP_CREF:
1492        case OP_NCREF:
1493      case OP_RREF:      case OP_RREF:
1494        case OP_NRREF:
1495      case OP_DEF:      case OP_DEF:
1496      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1497      break;      break;
# Line 1075  for (;;) Line 1507  for (;;)
1507    
1508    
1509  /*************************************************  /*************************************************
1510  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1511  *************************************************/  *************************************************/
1512    
1513  /* 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,
1514  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.
1515  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
1516    temporarily terminated with OP_END when this function is called.
1517    
1518    This function is called when a backward assertion is encountered, so that if it
1519    fails, the error message can point to the correct place in the pattern.
1520    However, we cannot do this when the assertion contains subroutine calls,
1521    because they can be forward references. We solve this by remembering this case
1522    and doing the check at the end; a flag specifies which mode we are running in.
1523    
1524  Arguments:  Arguments:
1525    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1526    options  the compiling options    utf8     TRUE in UTF-8 mode
1527      atend    TRUE if called when the pattern is complete
1528  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1529               or -2 if \C was encountered  
1530    Returns:   the fixed length,
1531                 or -1 if there is no fixed length,
1532                 or -2 if \C was encountered (in UTF-8 mode only)
1533                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1534                 or -4 if an unknown opcode was encountered (internal error)
1535  */  */
1536    
1537  static int  static int
1538  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1539  {  {
1540  int length = -1;  int length = -1;
1541    
# Line 1104  branch, check the length against that of Line 1548  branch, check the length against that of
1548  for (;;)  for (;;)
1549    {    {
1550    int d;    int d;
1551      uschar *ce, *cs;
1552    register int op = *cc;    register int op = *cc;
   
1553    switch (op)    switch (op)
1554      {      {
1555        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1556        OP_BRA (normal non-capturing bracket) because the other variants of these
1557        opcodes are all concerned with unlimited repeated groups, which of course
1558        are not of fixed length. */
1559    
1560      case OP_CBRA:      case OP_CBRA:
1561      case OP_BRA:      case OP_BRA:
1562      case OP_ONCE:      case OP_ONCE:
1563        case OP_ONCE_NC:
1564      case OP_COND:      case OP_COND:
1565      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1566      if (d < 0) return d;      if (d < 0) return d;
1567      branchlength += d;      branchlength += d;
1568      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1569      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1570      break;      break;
1571    
1572      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1573      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1574      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1575        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1576        because they all imply an unlimited repeat. */
1577    
1578      case OP_ALT:      case OP_ALT:
1579      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1580      case OP_END:      case OP_END:
1581        case OP_ACCEPT:
1582        case OP_ASSERT_ACCEPT:
1583      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1584        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1585      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1135  for (;;) Line 1587  for (;;)
1587      branchlength = 0;      branchlength = 0;
1588      break;      break;
1589    
1590        /* A true recursion implies not fixed length, but a subroutine call may
1591        be OK. If the subroutine is a forward reference, we can't deal with
1592        it until the end of the pattern, so return -3. */
1593    
1594        case OP_RECURSE:
1595        if (!atend) return -3;
1596        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1597        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1598        if (cc > cs && cc < ce) return -1;                /* Recursion */
1599        d = find_fixedlength(cs + 2, utf8, atend, cd);
1600        if (d < 0) return d;
1601        branchlength += d;
1602        cc += 1 + LINK_SIZE;
1603        break;
1604    
1605      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1606    
1607      case OP_ASSERT:      case OP_ASSERT:
# Line 1146  for (;;) Line 1613  for (;;)
1613    
1614      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1615    
1616      case OP_REVERSE:      case OP_MARK:
1617        case OP_PRUNE_ARG:
1618        case OP_SKIP_ARG:
1619        case OP_THEN_ARG:
1620        cc += cc[1] + _pcre_OP_lengths[*cc];
1621        break;
1622    
1623        case OP_CALLOUT:
1624        case OP_CIRC:
1625        case OP_CIRCM:
1626        case OP_CLOSE:
1627        case OP_COMMIT:
1628      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1629      case OP_DEF:      case OP_DEF:
1630      case OP_OPT:      case OP_DOLL:
1631      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1632      case OP_EOD:      case OP_EOD:
1633      case OP_EODN:      case OP_EODN:
1634      case OP_CIRC:      case OP_FAIL:
1635      case OP_DOLL:      case OP_NCREF:
1636        case OP_NRREF:
1637      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1638        case OP_PRUNE:
1639        case OP_REVERSE:
1640        case OP_RREF:
1641        case OP_SET_SOM:
1642        case OP_SKIP:
1643        case OP_SOD:
1644        case OP_SOM:
1645        case OP_THEN:
1646      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1647      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1648      break;      break;
# Line 1166  for (;;) Line 1650  for (;;)
1650      /* Handle literal characters */      /* Handle literal characters */
1651    
1652      case OP_CHAR:      case OP_CHAR:
1653      case OP_CHARNC:      case OP_CHARI:
1654      case OP_NOT:      case OP_NOT:
1655        case OP_NOTI:
1656      branchlength++;      branchlength++;
1657      cc += 2;      cc += 2;
1658  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1659      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1660  #endif  #endif
1661      break;      break;
1662    
# Line 1182  for (;;) Line 1664  for (;;)
1664      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1665    
1666      case OP_EXACT:      case OP_EXACT:
1667        case OP_EXACTI:
1668        case OP_NOTEXACT:
1669        case OP_NOTEXACTI:
1670      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1671      cc += 4;      cc += 4;
1672  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1673      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1674  #endif  #endif
1675      break;      break;
1676    
1677      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1678      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1679        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1680      cc += 4;      cc += 4;
1681      break;      break;
1682    
# Line 1204  for (;;) Line 1687  for (;;)
1687      cc += 2;      cc += 2;
1688      /* Fall through */      /* Fall through */
1689    
1690        case OP_HSPACE:
1691        case OP_VSPACE:
1692        case OP_NOT_HSPACE:
1693        case OP_NOT_VSPACE:
1694      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1695      case OP_DIGIT:      case OP_DIGIT:
1696      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1211  for (;;) Line 1698  for (;;)
1698      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1699      case OP_WORDCHAR:      case OP_WORDCHAR:
1700      case OP_ANY:      case OP_ANY:
1701        case OP_ALLANY:
1702      branchlength++;      branchlength++;
1703      cc++;      cc++;
1704      break;      break;
1705    
1706      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1707        otherwise \C is coded as OP_ALLANY. */
1708    
1709      case OP_ANYBYTE:      case OP_ANYBYTE:
1710      return -2;      return -2;
# Line 1234  for (;;) Line 1723  for (;;)
1723    
1724      switch (*cc)      switch (*cc)
1725        {        {
1726          case OP_CRPLUS:
1727          case OP_CRMINPLUS:
1728        case OP_CRSTAR:        case OP_CRSTAR:
1729        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1730        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1254  for (;;) Line 1745  for (;;)
1745    
1746      /* Anything else is variable length */      /* Anything else is variable length */
1747    
1748      default:      case OP_ANYNL:
1749        case OP_BRAMINZERO:
1750        case OP_BRAPOS:
1751        case OP_BRAPOSZERO:
1752        case OP_BRAZERO:
1753        case OP_CBRAPOS:
1754        case OP_EXTUNI:
1755        case OP_KETRMAX:
1756        case OP_KETRMIN:
1757        case OP_KETRPOS:
1758        case OP_MINPLUS:
1759        case OP_MINPLUSI:
1760        case OP_MINQUERY:
1761        case OP_MINQUERYI:
1762        case OP_MINSTAR:
1763        case OP_MINSTARI:
1764        case OP_MINUPTO:
1765        case OP_MINUPTOI:
1766        case OP_NOTMINPLUS:
1767        case OP_NOTMINPLUSI:
1768        case OP_NOTMINQUERY:
1769        case OP_NOTMINQUERYI:
1770        case OP_NOTMINSTAR:
1771        case OP_NOTMINSTARI:
1772        case OP_NOTMINUPTO:
1773        case OP_NOTMINUPTOI:
1774        case OP_NOTPLUS:
1775        case OP_NOTPLUSI:
1776        case OP_NOTPOSPLUS:
1777        case OP_NOTPOSPLUSI:
1778        case OP_NOTPOSQUERY:
1779        case OP_NOTPOSQUERYI:
1780        case OP_NOTPOSSTAR:
1781        case OP_NOTPOSSTARI:
1782        case OP_NOTPOSUPTO:
1783        case OP_NOTPOSUPTOI:
1784        case OP_NOTQUERY:
1785        case OP_NOTQUERYI:
1786        case OP_NOTSTAR:
1787        case OP_NOTSTARI:
1788        case OP_NOTUPTO:
1789        case OP_NOTUPTOI:
1790        case OP_PLUS:
1791        case OP_PLUSI:
1792        case OP_POSPLUS:
1793        case OP_POSPLUSI:
1794        case OP_POSQUERY:
1795        case OP_POSQUERYI:
1796        case OP_POSSTAR:
1797        case OP_POSSTARI:
1798        case OP_POSUPTO:
1799        case OP_POSUPTOI:
1800        case OP_QUERY:
1801        case OP_QUERYI:
1802        case OP_REF:
1803        case OP_REFI:
1804        case OP_SBRA:
1805        case OP_SBRAPOS:
1806        case OP_SCBRA:
1807        case OP_SCBRAPOS:
1808        case OP_SCOND:
1809        case OP_SKIPZERO:
1810        case OP_STAR:
1811        case OP_STARI:
1812        case OP_TYPEMINPLUS:
1813        case OP_TYPEMINQUERY:
1814        case OP_TYPEMINSTAR:
1815        case OP_TYPEMINUPTO:
1816        case OP_TYPEPLUS:
1817        case OP_TYPEPOSPLUS:
1818        case OP_TYPEPOSQUERY:
1819        case OP_TYPEPOSSTAR:
1820        case OP_TYPEPOSUPTO:
1821        case OP_TYPEQUERY:
1822        case OP_TYPESTAR:
1823        case OP_TYPEUPTO:
1824        case OP_UPTO:
1825        case OP_UPTOI:
1826      return -1;      return -1;
1827    
1828        /* Catch unrecognized opcodes so that when new ones are added they
1829        are not forgotten, as has happened in the past. */
1830    
1831        default:
1832        return -4;
1833      }      }
1834    }    }
1835  /* Control never gets here */  /* Control never gets here */
# Line 1265  for (;;) Line 1839  for (;;)
1839    
1840    
1841  /*************************************************  /*************************************************
1842  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1843  *************************************************/  *************************************************/
1844    
1845  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1846  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1847    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1848    so that it can be called from pcre_study() when finding the minimum matching
1849    length.
1850    
1851  Arguments:  Arguments:
1852    code        points to start of expression    code        points to start of expression
1853    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1854    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1855    
1856  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
1857  */  */
1858    
1859  static const uschar *  const uschar *
1860  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1861  {  {
1862  for (;;)  for (;;)
1863    {    {
1864    register int c = *code;    register int c = *code;
1865    
1866    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1867    
1868    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1293  for (;;) Line 1871  for (;;)
1871    
1872    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1873    
1874      /* Handle recursion */
1875    
1876      else if (c == OP_REVERSE)
1877        {
1878        if (number < 0) return (uschar *)code;
1879        code += _pcre_OP_lengths[c];
1880        }
1881    
1882    /* Handle capturing bracket */    /* Handle capturing bracket */
1883    
1884    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1885               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1886      {      {
1887      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1888      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1889      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1890      }      }
1891    
1892      /* Otherwise, we can get the item's length from the table, except that for
1893      repeated character types, we have to test for \p and \P, which have an extra
1894      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1895      must add in its length. */
1896    
1897      else
1898        {
1899        switch(c)
1900          {
1901          case OP_TYPESTAR:
1902          case OP_TYPEMINSTAR:
1903          case OP_TYPEPLUS:
1904          case OP_TYPEMINPLUS:
1905          case OP_TYPEQUERY:
1906          case OP_TYPEMINQUERY:
1907          case OP_TYPEPOSSTAR:
1908          case OP_TYPEPOSPLUS:
1909          case OP_TYPEPOSQUERY:
1910          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1911          break;
1912    
1913          case OP_TYPEUPTO:
1914          case OP_TYPEMINUPTO:
1915          case OP_TYPEEXACT:
1916          case OP_TYPEPOSUPTO:
1917          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1918          break;
1919    
1920          case OP_MARK:
1921          case OP_PRUNE_ARG:
1922          case OP_SKIP_ARG:
1923          code += code[1];
1924          break;
1925    
1926          case OP_THEN_ARG:
1927          code += code[1];
1928          break;
1929          }
1930    
1931        /* Add in the fixed length from the table */
1932    
1933        code += _pcre_OP_lengths[c];
1934    
1935    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1936    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
1937    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
1938    
   else  
     {  
     code += _pcre_OP_lengths[c];  
1939  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1940      if (utf8) switch(c)      if (utf8) switch(c)
1941        {        {
1942        case OP_CHAR:        case OP_CHAR:
1943        case OP_CHARNC:        case OP_CHARI:
1944        case OP_EXACT:        case OP_EXACT:
1945          case OP_EXACTI:
1946        case OP_UPTO:        case OP_UPTO:
1947          case OP_UPTOI:
1948        case OP_MINUPTO:        case OP_MINUPTO:
1949          case OP_MINUPTOI:
1950        case OP_POSUPTO:        case OP_POSUPTO:
1951          case OP_POSUPTOI:
1952        case OP_STAR:        case OP_STAR:
1953          case OP_STARI:
1954        case OP_MINSTAR:        case OP_MINSTAR:
1955          case OP_MINSTARI:
1956        case OP_POSSTAR:        case OP_POSSTAR:
1957          case OP_POSSTARI:
1958        case OP_PLUS:        case OP_PLUS:
1959          case OP_PLUSI:
1960        case OP_MINPLUS:        case OP_MINPLUS:
1961          case OP_MINPLUSI:
1962        case OP_POSPLUS:        case OP_POSPLUS:
1963          case OP_POSPLUSI:
1964        case OP_QUERY:        case OP_QUERY:
1965          case OP_QUERYI:
1966        case OP_MINQUERY:        case OP_MINQUERY:
1967          case OP_MINQUERYI:
1968        case OP_POSQUERY:        case OP_POSQUERY:
1969          case OP_POSQUERYI:
1970        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1971        break;        break;
1972        }        }
1973    #else
1974        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1975  #endif  #endif
1976      }      }
1977    }    }
# Line 1366  for (;;) Line 2008  for (;;)
2008    
2009    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2010    
2011    /* 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
2012    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
2013    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
2014    bytes. */    must add in its length. */
2015    
2016    else    else
2017      {      {
2018        switch(c)
2019          {
2020          case OP_TYPESTAR:
2021          case OP_TYPEMINSTAR:
2022          case OP_TYPEPLUS:
2023          case OP_TYPEMINPLUS:
2024          case OP_TYPEQUERY:
2025          case OP_TYPEMINQUERY:
2026          case OP_TYPEPOSSTAR:
2027          case OP_TYPEPOSPLUS:
2028          case OP_TYPEPOSQUERY:
2029          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2030          break;
2031    
2032          case OP_TYPEPOSUPTO:
2033          case OP_TYPEUPTO:
2034          case OP_TYPEMINUPTO:
2035          case OP_TYPEEXACT:
2036          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2037          break;
2038    
2039          case OP_MARK:
2040          case OP_PRUNE_ARG:
2041          case OP_SKIP_ARG:
2042          code += code[1];
2043          break;
2044    
2045          case OP_THEN_ARG:
2046          code += code[1];
2047          break;
2048          }
2049    
2050        /* Add in the fixed length from the table */
2051    
2052      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2053    
2054        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2055        by a multi-byte character. The length in the table is a minimum, so we have
2056        to arrange to skip the extra bytes. */
2057    
2058  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2059      if (utf8) switch(c)      if (utf8) switch(c)
2060        {        {
2061        case OP_CHAR:        case OP_CHAR:
2062        case OP_CHARNC:        case OP_CHARI:
2063        case OP_EXACT:        case OP_EXACT:
2064          case OP_EXACTI:
2065        case OP_UPTO:        case OP_UPTO:
2066          case OP_UPTOI:
2067        case OP_MINUPTO:        case OP_MINUPTO:
2068          case OP_MINUPTOI:
2069        case OP_POSUPTO:        case OP_POSUPTO:
2070          case OP_POSUPTOI:
2071        case OP_STAR:        case OP_STAR:
2072          case OP_STARI:
2073        case OP_MINSTAR:        case OP_MINSTAR:
2074          case OP_MINSTARI:
2075        case OP_POSSTAR:        case OP_POSSTAR:
2076          case OP_POSSTARI:
2077        case OP_PLUS:        case OP_PLUS:
2078          case OP_PLUSI:
2079        case OP_MINPLUS:        case OP_MINPLUS:
2080          case OP_MINPLUSI:
2081        case OP_POSPLUS:        case OP_POSPLUS:
2082          case OP_POSPLUSI:
2083        case OP_QUERY:        case OP_QUERY:
2084          case OP_QUERYI:
2085        case OP_MINQUERY:        case OP_MINQUERY:
2086          case OP_MINQUERYI:
2087        case OP_POSQUERY:        case OP_POSQUERY:
2088          case OP_POSQUERYI:
2089        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2090        break;        break;
2091        }        }
2092    #else
2093        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2094  #endif  #endif
2095      }      }
2096    }    }
# Line 1410  for (;;) Line 2106  for (;;)
2106  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()
2107  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
2108  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2109  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
2110  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
2111    bracket whose current branch will already have been scanned.
2112    
2113  Arguments:  Arguments:
2114    code        points to start of search    code        points to start of search
2115    endcode     points to where to stop    endcode     points to where to stop
2116    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2117      cd          contains pointers to tables etc.
2118    
2119  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2120  */  */
2121    
2122  static BOOL  static BOOL
2123  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2124      compile_data *cd)
2125  {  {
2126  register int c;  register int c;
2127  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2128       code < endcode;       code < endcode;
2129       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2130    {    {
2131    const uschar *ccode;    const uschar *ccode;
2132    
2133    c = *code;    c = *code;
2134    
2135    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2136      first_significant_code() with a TRUE final argument. */
2137    
2138    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2139      {      {
     code += _pcre_OP_lengths[c];  
2140      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2141      c = *code;      c = *code;
2142      continue;      continue;
2143      }      }
2144    
2145    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2146      implies a backward reference subroutine call, we can scan it. If it's a
2147      forward reference subroutine call, we can't. To detect forward reference
2148      we have to scan up the list that is kept in the workspace. This function is
2149      called only when doing the real compile, not during the pre-compile that
2150      measures the size of the compiled pattern. */
2151    
2152    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2153      {      {
2154        const uschar *scode;
2155      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2156    
2157      /* Scan a closed bracket */      /* Test for forward reference */
2158    
2159        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2160          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2161    
2162        /* Not a forward reference, test for completed backward reference */
2163    
2164      empty_branch = FALSE;      empty_branch = FALSE;
2165        scode = cd->start_code + GET(code, 1);
2166        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2167    
2168        /* Completed backwards reference */
2169    
2170      do      do
2171        {        {
2172        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2173            {
2174          empty_branch = TRUE;          empty_branch = TRUE;
2175            break;
2176            }
2177          scode += GET(scode, 1);
2178          }
2179        while (*scode == OP_ALT);
2180    
2181        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2182        continue;
2183        }
2184    
2185      /* Groups with zero repeats can of course be empty; skip them. */
2186    
2187      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2188          c == OP_BRAPOSZERO)
2189        {
2190        code += _pcre_OP_lengths[c];
2191        do code += GET(code, 1); while (*code == OP_ALT);
2192        c = *code;
2193        continue;
2194        }
2195    
2196      /* A nested group that is already marked as "could be empty" can just be
2197      skipped. */
2198    
2199      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2200          c == OP_SCBRA || c == OP_SCBRAPOS)
2201        {
2202        do code += GET(code, 1); while (*code == OP_ALT);
2203        c = *code;
2204        continue;
2205        }
2206    
2207      /* For other groups, scan the branches. */
2208    
2209      if (c == OP_BRA  || c == OP_BRAPOS ||
2210          c == OP_CBRA || c == OP_CBRAPOS ||
2211          c == OP_ONCE || c == OP_ONCE_NC ||
2212          c == OP_COND)
2213        {
2214        BOOL empty_branch;
2215        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2216    
2217        /* If a conditional group has only one branch, there is a second, implied,
2218        empty branch, so just skip over the conditional, because it could be empty.
2219        Otherwise, scan the individual branches of the group. */
2220    
2221        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2222        code += GET(code, 1);        code += GET(code, 1);
2223        else
2224          {
2225          empty_branch = FALSE;
2226          do
2227            {
2228            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2229              empty_branch = TRUE;
2230            code += GET(code, 1);
2231            }
2232          while (*code == OP_ALT);
2233          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2234        }        }
2235      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2236      c = *code;      c = *code;
2237      continue;      continue;
2238      }      }
# Line 1469  for (code = first_significant_code(code Line 2241  for (code = first_significant_code(code
2241    
2242    switch (c)    switch (c)
2243      {      {
2244      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2245        cannot be represented just by a bit map. This includes negated single
2246        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2247        actual length is stored in the compiled code, so we must update "code"
2248        here. */
2249    
2250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2251      case OP_XCLASS:      case OP_XCLASS:
2252      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2253      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2254  #endif  #endif
2255    
# Line 1517  for (code = first_significant_code(code Line 2293  for (code = first_significant_code(code
2293      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2294      case OP_WORDCHAR:      case OP_WORDCHAR:
2295      case OP_ANY:      case OP_ANY:
2296        case OP_ALLANY:
2297      case OP_ANYBYTE:      case OP_ANYBYTE:
2298      case OP_CHAR:      case OP_CHAR:
2299      case OP_CHARNC:      case OP_CHARI:
2300      case OP_NOT:      case OP_NOT:
2301        case OP_NOTI:
2302      case OP_PLUS:      case OP_PLUS:
2303      case OP_MINPLUS:      case OP_MINPLUS:
2304      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1535  for (code = first_significant_code(code Line 2313  for (code = first_significant_code(code
2313      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2314      return FALSE;      return FALSE;
2315    
2316        /* These are going to continue, as they may be empty, but we have to
2317        fudge the length for the \p and \P cases. */
2318    
2319        case OP_TYPESTAR:
2320        case OP_TYPEMINSTAR:
2321        case OP_TYPEPOSSTAR:
2322        case OP_TYPEQUERY:
2323        case OP_TYPEMINQUERY:
2324        case OP_TYPEPOSQUERY:
2325        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2326        break;
2327    
2328        /* Same for these */
2329    
2330        case OP_TYPEUPTO:
2331        case OP_TYPEMINUPTO:
2332        case OP_TYPEPOSUPTO:
2333        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2334        break;
2335    
2336      /* End of branch */      /* End of branch */
2337    
2338      case OP_KET:      case OP_KET:
2339      case OP_KETRMAX:      case OP_KETRMAX:
2340      case OP_KETRMIN:      case OP_KETRMIN:
2341        case OP_KETRPOS:
2342      case OP_ALT:      case OP_ALT:
2343      return TRUE;      return TRUE;
2344    
# Line 1548  for (code = first_significant_code(code Line 2347  for (code = first_significant_code(code
2347    
2348  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2349      case OP_STAR:      case OP_STAR:
2350        case OP_STARI:
2351      case OP_MINSTAR:      case OP_MINSTAR:
2352        case OP_MINSTARI:
2353      case OP_POSSTAR:      case OP_POSSTAR:
2354        case OP_POSSTARI:
2355      case OP_QUERY:      case OP_QUERY:
2356        case OP_QUERYI:
2357      case OP_MINQUERY:      case OP_MINQUERY:
2358        case OP_MINQUERYI:
2359      case OP_POSQUERY:      case OP_POSQUERY:
2360        case OP_POSQUERYI:
2361        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2362        break;
2363    
2364      case OP_UPTO:      case OP_UPTO:
2365        case OP_UPTOI:
2366      case OP_MINUPTO:      case OP_MINUPTO:
2367        case OP_MINUPTOI:
2368      case OP_POSUPTO:      case OP_POSUPTO:
2369      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2370        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2371      break;      break;
2372  #endif  #endif
2373    
2374        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2375        string. */
2376    
2377        case OP_MARK:
2378        case OP_PRUNE_ARG:
2379        case OP_SKIP_ARG:
2380        code += code[1];
2381        break;
2382    
2383        case OP_THEN_ARG:
2384        code += code[1];
2385        break;
2386    
2387        /* None of the remaining opcodes are required to match a character. */
2388    
2389        default:
2390        break;
2391      }      }
2392    }    }
2393    
# Line 1575  return TRUE; Line 2404  return TRUE;
2404  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2405  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2406  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2407    This function is called only during the real compile, not during the
2408    pre-compile.
2409    
2410  Arguments:  Arguments:
2411    code        points to start of the recursion    code        points to start of the recursion
2412    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2413    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2414    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2415      cd          pointers to tables etc
2416    
2417  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2418  */  */
2419    
2420  static BOOL  static BOOL
2421  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2422    BOOL utf8)    BOOL utf8, compile_data *cd)
2423  {  {
2424  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2425    {    {
2426    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2427        return FALSE;
2428    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2429    }    }
2430  return TRUE;  return TRUE;
# Line 1604  return TRUE; Line 2437  return TRUE;
2437  *************************************************/  *************************************************/
2438    
2439  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2440  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
2441  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2442  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2443    
2444    Originally, this function only recognized a sequence of letters between the
2445    terminators, but it seems that Perl recognizes any sequence of characters,
2446    though of course unknown POSIX names are subsequently rejected. Perl gives an
2447    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2448    didn't consider this to be a POSIX class. Likewise for [:1234:].
2449    
2450    The problem in trying to be exactly like Perl is in the handling of escapes. We
2451    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2452    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2453    below handles the special case of \], but does not try to do any other escape
2454    processing. This makes it different from Perl for cases such as [:l\ower:]
2455    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2456    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2457    I think.
2458    
2459    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2460    It seems that the appearance of a nested POSIX class supersedes an apparent
2461    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2462    a digit.
2463    
2464    In Perl, unescaped square brackets may also appear as part of class names. For
2465    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2466    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2467    seem right at all. PCRE does not allow closing square brackets in POSIX class
2468    names.
2469    
2470  Argument:  Arguments:
2471    ptr      pointer to the initial [    ptr      pointer to the initial [
2472    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2473    
2474  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2475  */  */
2476    
2477  static BOOL  static BOOL
2478  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2479  {  {
2480  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2481  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2482  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2483    {    {
2484    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2485    return TRUE;      ptr++;
2486      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2487      else
2488        {
2489        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2490          {
2491          *endptr = ptr;
2492          return TRUE;
2493          }
2494        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2495             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2496              ptr[1] == CHAR_EQUALS_SIGN) &&
2497            check_posix_syntax(ptr, endptr))
2498          return FALSE;
2499        }
2500    }    }
2501  return FALSE;  return FALSE;
2502  }  }
# Line 1651  Returns:     a value representing the na Line 2521  Returns:     a value representing the na
2521  static int  static int
2522  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2523  {  {
2524    const char *pn = posix_names;
2525  register int yield = 0;  register int yield = 0;
2526  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2527    {    {
2528    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2529      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2530      pn += posix_name_lengths[yield] + 1;
2531    yield++;    yield++;
2532    }    }
2533  return -1;  return -1;
# Line 1670  return -1; Line 2542  return -1;
2542  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2543  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2544  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
2545  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
2546  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
2547  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
2548  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
2549  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2550    OP_END.
2551    
2552  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2553  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 1697  adjust_recurse(uschar *group, int adjust Line 2570  adjust_recurse(uschar *group, int adjust
2570    uschar *save_hwm)    uschar *save_hwm)
2571  {  {
2572  uschar *ptr = group;  uschar *ptr = group;
2573    
2574  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2575    {    {
2576    int offset;    int offset;
# Line 1750  auto_callout(uschar *code, const uschar Line 2624  auto_callout(uschar *code, const uschar
2624  {  {
2625  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2626  *code++ = 255;  *code++ = 255;
2627  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2628  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2629  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2630  }  }
2631    
# Line 1776  Returns:             nothing Line 2650  Returns:             nothing
2650  static void  static void
2651  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2652  {  {
2653  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2654  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2655  }  }
2656    
# Line 1808  get_othercase_range(unsigned int *cptr, Line 2682  get_othercase_range(unsigned int *cptr,
2682  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2683    
2684  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2685    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2686    
2687  if (c > d) return FALSE;  if (c > d) return FALSE;
2688    
# Line 1817  next = othercase + 1; Line 2691  next = othercase + 1;
2691    
2692  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2693    {    {
2694    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2695    next++;    next++;
2696    }    }
2697    
# Line 1826  for (++c; c <= d; c++) Line 2700  for (++c; c <= d; c++)
2700    
2701  return TRUE;  return TRUE;
2702  }  }
2703    
2704    
2705    
2706    /*************************************************
2707    *        Check a character and a property        *
2708    *************************************************/
2709    
2710    /* This function is called by check_auto_possessive() when a property item
2711    is adjacent to a fixed character.
2712    
2713    Arguments:
2714      c            the character
2715      ptype        the property type
2716      pdata        the data for the type
2717      negated      TRUE if it's a negated property (\P or \p{^)
2718    
2719    Returns:       TRUE if auto-possessifying is OK
2720    */
2721    
2722    static BOOL
2723    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2724    {
2725    const ucd_record *prop = GET_UCD(c);
2726    switch(ptype)
2727      {
2728      case PT_LAMP:
2729      return (prop->chartype == ucp_Lu ||
2730              prop->chartype == ucp_Ll ||
2731              prop->chartype == ucp_Lt) == negated;
2732    
2733      case PT_GC:
2734      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2735    
2736      case PT_PC:
2737      return (pdata == prop->chartype) == negated;
2738    
2739      case PT_SC:
2740      return (pdata == prop->script) == negated;
2741    
2742      /* These are specials */
2743    
2744      case PT_ALNUM:
2745      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2746              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2747    
2748      case PT_SPACE:    /* Perl space */
2749      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2750              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2751              == negated;
2752    
2753      case PT_PXSPACE:  /* POSIX space */
2754      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2755              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2756              c == CHAR_FF || c == CHAR_CR)
2757              == negated;
2758    
2759      case PT_WORD:
2760      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2761              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2762              c == CHAR_UNDERSCORE) == negated;
2763      }
2764    return FALSE;
2765    }
2766  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2767    
2768    
# Line 1839  whether the next thing could possibly ma Line 2776  whether the next thing could possibly ma
2776  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2777    
2778  Arguments:  Arguments:
2779    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2780    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2781    ptr           next character in pattern    ptr           next character in pattern
2782    options       options bits    options       options bits
2783    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1851  Returns:        TRUE if possessifying is Line 2786  Returns:        TRUE if possessifying is
2786  */  */
2787    
2788  static BOOL  static BOOL
2789  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2790    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2791  {  {
2792  int next;  int c, next;
2793    int op_code = *previous++;
2794    
2795  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2796    
# Line 1863  if ((options & PCRE_EXTENDED) != 0) Line 2799  if ((options & PCRE_EXTENDED) != 0)
2799    for (;;)    for (;;)
2800      {      {
2801      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2802      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2803        {        {
2804        while (*(++ptr) != 0)        ptr++;
2805          while (*ptr != 0)
2806            {
2807          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2808            ptr++;
2809    #ifdef SUPPORT_UTF8
2810            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2811    #endif
2812            }
2813        }        }
2814      else break;      else break;
2815      }      }
# Line 1875  if ((options & PCRE_EXTENDED) != 0) Line 2818  if ((options & PCRE_EXTENDED) != 0)
2818  /* 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
2819  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2820    
2821  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2822    {    {
2823    int temperrorcode = 0;    int temperrorcode = 0;
2824    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1900  if ((options & PCRE_EXTENDED) != 0) Line 2843  if ((options & PCRE_EXTENDED) != 0)
2843    for (;;)    for (;;)
2844      {      {
2845      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2846      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2847        {        {
2848        while (*(++ptr) != 0)        ptr++;
2849          while (*ptr != 0)
2850            {
2851          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2852            ptr++;
2853    #ifdef SUPPORT_UTF8
2854            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2855    #endif
2856            }
2857        }        }
2858      else break;      else break;
2859      }      }
# Line 1911  if ((options & PCRE_EXTENDED) != 0) Line 2861  if ((options & PCRE_EXTENDED) != 0)
2861    
2862  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2863    
2864  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2865    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2866        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2867    
2868  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2869    the next item is a character. */
2870    
2871  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2872    {    {
2873    case OP_CHAR:    case OP_CHAR:
2874  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2875    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2876    #else
2877      c = *previous;
2878  #endif  #endif
2879    return item != next;    return c != next;
2880    
2881    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2882    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2883    high-valued characters. */    high-valued characters. */
2884    
2885    case OP_CHARNC:    case OP_CHARI:
2886  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2887    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2888    #else
2889      c = *previous;
2890  #endif  #endif
2891    if (item == next) return FALSE;    if (c == next) return FALSE;
2892  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2893    if (utf8)    if (utf8)
2894      {      {
2895      unsigned int othercase;      unsigned int othercase;
2896      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2897  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2898      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2899  #else  #else
2900      othercase = NOTACHAR;      othercase = NOTACHAR;
2901  #endif  #endif
2902      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2903      }      }
2904    else    else
2905  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2906    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2907    
2908    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2909      opcodes are not used for multi-byte characters, because they are coded using
2910      an XCLASS instead. */
2911    
2912    case OP_NOT:    case OP_NOT:
2913    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2914    if (item == next) return TRUE;  
2915    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2916      if ((c = *previous) == next) return TRUE;
2917  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2918    if (utf8)    if (utf8)
2919      {      {
2920      unsigned int othercase;      unsigned int othercase;
2921      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2922  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2923      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2924  #else  #else
2925      othercase = NOTACHAR;      othercase = NOTACHAR;
2926  #endif  #endif
2927      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2928      }      }
2929    else    else
2930  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2931    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2932    
2933      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2934      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2935    
2936    case OP_DIGIT:    case OP_DIGIT:
2937    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2018  if (next >= 0) switch(op_code) Line 2974  if (next >= 0) switch(op_code)
2974      case 0x202f:      case 0x202f:
2975      case 0x205f:      case 0x205f:
2976      case 0x3000:      case 0x3000:
2977      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2978      default:      default:
2979      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2980      }      }
2981    
2982      case OP_ANYNL:
2983    case OP_VSPACE:    case OP_VSPACE:
2984    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2985    switch(next)    switch(next)
# Line 2034  if (next >= 0) switch(op_code) Line 2991  if (next >= 0) switch(op_code)
2991      case 0x85:      case 0x85:
2992      case 0x2028:      case 0x2028:
2993      case 0x2029:      case 0x2029:
2994      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2995      default:      default:
2996      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2997      }      }
2998    
2999    #ifdef SUPPORT_UCP
3000      case OP_PROP:
3001      return check_char_prop(next, previous[0], previous[1], FALSE);
3002    
3003      case OP_NOTPROP:
3004      return check_char_prop(next, previous[0], previous[1], TRUE);
3005    #endif
3006    
3007    default:    default:
3008    return FALSE;    return FALSE;
3009    }    }
3010    
3011    
3012  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3013    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3014    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3015    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3016    replaced by OP_PROP codes when PCRE_UCP is set. */
3017    
3018  switch(op_code)  switch(op_code)
3019    {    {
3020    case OP_CHAR:    case OP_CHAR:
3021    case OP_CHARNC:    case OP_CHARI:
3022  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3023    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3024    #else
3025      c = *previous;
3026  #endif  #endif
3027    switch(-next)    switch(-next)
3028      {      {
3029      case ESC_d:      case ESC_d:
3030      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3031    
3032      case ESC_D:      case ESC_D:
3033      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3034    
3035      case ESC_s:      case ESC_s:
3036      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3037    
3038      case ESC_S:      case ESC_S:
3039      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3040    
3041      case ESC_w:      case ESC_w:
3042      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3043    
3044      case ESC_W:      case ESC_W:
3045      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3046    
3047      case ESC_h:      case ESC_h:
3048      case ESC_H:      case ESC_H:
3049      switch(item)      switch(c)
3050        {        {
3051        case 0x09:        case 0x09:
3052        case 0x20:        case 0x20:
# Line 2103  switch(op_code) Line 3074  switch(op_code)
3074    
3075      case ESC_v:      case ESC_v:
3076      case ESC_V:      case ESC_V:
3077      switch(item)      switch(c)
3078        {        {
3079        case 0x0a:        case 0x0a:
3080        case 0x0b:        case 0x0b:
# Line 2117  switch(op_code) Line 3088  switch(op_code)
3088        return -next == ESC_v;        return -next == ESC_v;
3089        }        }
3090    
3091        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3092        their substitutions and process them. The result will always be either
3093        -ESC_p or -ESC_P. Then fall through to process those values. */
3094    
3095    #ifdef SUPPORT_UCP
3096        case ESC_du:
3097        case ESC_DU:
3098        case ESC_wu:
3099        case ESC_WU:
3100        case ESC_su:
3101        case ESC_SU:
3102          {
3103          int temperrorcode = 0;
3104          ptr = substitutes[-next - ESC_DU];
3105          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3106          if (temperrorcode != 0) return FALSE;
3107          ptr++;    /* For compatibility */
3108          }
3109        /* Fall through */
3110    
3111        case ESC_p:
3112        case ESC_P:
3113          {
3114          int ptype, pdata, errorcodeptr;
3115          BOOL negated;
3116    
3117          ptr--;      /* Make ptr point at the p or P */
3118          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3119          if (ptype < 0) return FALSE;
3120          ptr++;      /* Point past the final curly ket */
3121    
3122          /* If the property item is optional, we have to give up. (When generated
3123          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3124          to the original \d etc. At this point, ptr will point to a zero byte. */
3125    
3126          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3127            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3128              return FALSE;
3129    
3130          /* Do the property check. */
3131    
3132          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3133          }
3134    #endif
3135    
3136      default:      default:
3137      return FALSE;      return FALSE;
3138      }      }
3139    
3140      /* In principle, support for Unicode properties should be integrated here as
3141      well. It means re-organizing the above code so as to get hold of the property
3142      values before switching on the op-code. However, I wonder how many patterns
3143      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3144      these op-codes are never generated.) */
3145    
3146    case OP_DIGIT:    case OP_DIGIT:
3147    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3148           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3149    
3150    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3151    return next == -ESC_d;    return next == -ESC_d;
3152    
3153    case OP_WHITESPACE:    case OP_WHITESPACE:
3154    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3155    
3156    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3157    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3158    
3159    case OP_HSPACE:    case OP_HSPACE:
3160    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3161             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3162    
3163    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3164    return next == -ESC_h;    return next == -ESC_h;
3165    
3166    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3167      case OP_ANYNL:
3168    case OP_VSPACE:    case OP_VSPACE:
3169    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3170    
3171    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3172    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3173    
3174    case OP_WORDCHAR:    case OP_WORDCHAR:
3175    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3176             next == -ESC_v || next == -ESC_R;
3177    
3178    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3179    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2180  Arguments: Line 3205  Arguments:
3205    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3206    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3207    bcptr          points to current branch chain    bcptr          points to current branch chain
3208      cond_depth     conditional nesting depth
3209    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3210    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3211                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2191  Returns:         TRUE on success Line 3217  Returns:         TRUE on success
3217  static BOOL  static BOOL
3218  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3219    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3220    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3221  {  {
3222  int repeat_type, op_type;  int repeat_type, op_type;
3223  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2200  int greedy_default, greedy_non_default; Line 3226  int greedy_default, greedy_non_default;
3226  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3227  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3228  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3229  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3230  int after_manual_callout = 0;  int after_manual_callout = 0;
3231  int length_prevgroup = 0;  int length_prevgroup = 0;
3232  register int c;  register int c;
# Line 2212  BOOL inescq = FALSE; Line 3238  BOOL inescq = FALSE;
3238  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3239  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3240  const uschar *tempptr;  const uschar *tempptr;
3241    const uschar *nestptr = NULL;
3242  uschar *previous = NULL;  uschar *previous = NULL;
3243  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3244  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3245  uschar classbits[32];  uschar classbits[32];
3246    
3247    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3248    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3249    dynamically as we process the pattern. */
3250    
3251  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3252  BOOL class_utf8;  BOOL class_utf8;
3253  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3254  uschar *class_utf8data;  uschar *class_utf8data;
3255    uschar *class_utf8data_base;
3256  uschar utf8_char[6];  uschar utf8_char[6];
3257  #else  #else
3258  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3259  #endif  #endif
3260    
3261  #ifdef DEBUG  #ifdef PCRE_DEBUG
3262  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3263  #endif  #endif
3264    
# Line 2260  req_caseopt = ((options & PCRE_CASELESS) Line 3291  req_caseopt = ((options & PCRE_CASELESS)
3291  for (;; ptr++)  for (;; ptr++)
3292    {    {
3293    BOOL negate_class;    BOOL negate_class;
3294      BOOL should_flip_negation;
3295    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3296    BOOL is_quantifier;    BOOL is_quantifier;
3297    BOOL is_recurse;    BOOL is_recurse;
# Line 2274  for (;; ptr++) Line 3306  for (;; ptr++)
3306    int subfirstbyte;    int subfirstbyte;
3307    int terminator;    int terminator;
3308    int mclength;    int mclength;
3309      int tempbracount;
3310    uschar mcbuffer[8];    uschar mcbuffer[8];
3311    
3312    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3313    
3314    c = *ptr;    c = *ptr;
3315    
3316      /* If we are at the end of a nested substitution, revert to the outer level
3317      string. Nesting only happens one level deep. */
3318    
3319      if (c == 0 && nestptr != NULL)
3320        {
3321        ptr = nestptr;
3322        nestptr = NULL;
3323        c = *ptr;
3324        }
3325    
3326    /* 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
3327    previous cycle of this loop. */    previous cycle of this loop. */
3328    
3329    if (lengthptr != NULL)    if (lengthptr != NULL)
3330      {      {
3331  #ifdef DEBUG  #ifdef PCRE_DEBUG
3332      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3333  #endif  #endif
3334      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3335        {        {
3336        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3337        goto FAILED;        goto FAILED;
# Line 2310  for (;; ptr++) Line 3353  for (;; ptr++)
3353        goto FAILED;        goto FAILED;
3354        }        }
3355    
3356      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3357      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3358          c));
3359    
3360      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3361      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2337  for (;; ptr++) Line 3381  for (;; ptr++)
3381    /* 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
3382    reference list. */    reference list. */
3383    
3384    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3385      {      {
3386      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3387      goto FAILED;      goto FAILED;
# Line 2347  for (;; ptr++) Line 3391  for (;; ptr++)
3391    
3392    if (inescq && c != 0)    if (inescq && c != 0)
3393      {      {
3394      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3395        {        {
3396        inescq = FALSE;        inescq = FALSE;
3397        ptr++;        ptr++;
# Line 2373  for (;; ptr++) Line 3417  for (;; ptr++)
3417    /* 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
3418    a quantifier. */    a quantifier. */
3419    
3420    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3421      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3422        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3423    
3424    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3425         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2384  for (;; ptr++) Line 3429  for (;; ptr++)
3429      previous_callout = NULL;      previous_callout = NULL;
3430      }      }
3431    
3432    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3433    
3434    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3435      {      {
3436      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3437      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3438        {        {
3439        while (*(++ptr) != 0)        ptr++;
3440          while (*ptr != 0)
3441          {          {
3442          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3443            ptr++;
3444    #ifdef SUPPORT_UTF8
3445            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3446    #endif
3447          }          }
3448        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3449    
# Line 2414  for (;; ptr++) Line 3464  for (;; ptr++)
3464      {      {
3465      /* ===================================================================*/      /* ===================================================================*/
3466      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3467      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3468      case ')':      case CHAR_RIGHT_PARENTHESIS:
3469      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3470      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3471      *codeptr = code;      *codeptr = code;
# Line 2427  for (;; ptr++) Line 3477  for (;; ptr++)
3477          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3478          goto FAILED;          goto FAILED;
3479          }          }
3480        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3481        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3482        }        }
3483      return TRUE;      return TRUE;
# Line 2437  for (;; ptr++) Line 3487  for (;; ptr++)
3487      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3488      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3489    
3490      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3491        previous = NULL;
3492      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3493        {        {
3494        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3495          *code++ = OP_CIRCM;
3496        }        }
3497      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3498      break;      break;
3499    
3500      case '$':      case CHAR_DOLLAR_SIGN:
3501      previous = NULL;      previous = NULL;
3502      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3503      break;      break;
3504    
3505      /* 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
3506      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3507    
3508      case '.':      case CHAR_DOT:
3509      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3510      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3511      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3512      previous = code;      previous = code;
3513      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3514      break;      break;
3515    
3516    
# Line 2474  for (;; ptr++) Line 3525  for (;; ptr++)
3525      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,
3526      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3527      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.
     */  
3528    
3529      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3530        default (Perl) mode, it is treated as a data character. */
3531    
3532        case CHAR_RIGHT_SQUARE_BRACKET:
3533        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3534          {
3535          *errorcodeptr = ERR64;
3536          goto FAILED;
3537          }
3538        goto NORMAL_CHAR;
3539    
3540        case CHAR_LEFT_SQUARE_BRACKET:
3541      previous = code;      previous = code;
3542    
3543      /* 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
3544      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. */
3545    
3546      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3547          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3548            check_posix_syntax(ptr, &tempptr))
3549        {        {
3550        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3551        goto FAILED;        goto FAILED;
3552        }        }
3553    
# Line 2497  for (;; ptr++) Line 3559  for (;; ptr++)
3559      for (;;)      for (;;)
3560        {        {
3561        c = *(++ptr);        c = *(++ptr);
3562        if (c == '\\')        if (c == CHAR_BACKSLASH)
3563          {          {
3564          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3565            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3566              else break;          else if (strncmp((const char *)ptr+1,
3567                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3568              ptr += 3;
3569            else
3570              break;
3571          }          }
3572        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3573          negate_class = TRUE;          negate_class = TRUE;
3574        else break;        else break;
3575        }        }
3576    
3577        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3578        an initial ']' is taken as a data character -- the code below handles
3579        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3580        [^] must match any character, so generate OP_ALLANY. */
3581    
3582        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3583            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3584          {
3585          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3586          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3587          zerofirstbyte = firstbyte;
3588          break;
3589          }
3590    
3591        /* If a class contains a negative special such as \S, we need to flip the
3592        negation flag at the end, so that support for characters > 255 works
3593        correctly (they are all included in the class). */
3594    
3595        should_flip_negation = FALSE;
3596    
3597      /* 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
3598      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
3599      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2525  for (;; ptr++) Line 3611  for (;; ptr++)
3611  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3612      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3613      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3614        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3615  #endif  #endif
3616    
3617      /* 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 2540  for (;; ptr++) Line 3627  for (;; ptr++)
3627          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3628          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3629          }          }
3630    
3631          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3632          data and reset the pointer. This is so that very large classes that
3633          contain a zillion UTF-8 characters no longer overwrite the work space
3634          (which is on the stack). */
3635    
3636          if (lengthptr != NULL)
3637            {
3638            *lengthptr += class_utf8data - class_utf8data_base;
3639            class_utf8data = class_utf8data_base;
3640            }
3641    
3642  #endif  #endif
3643    
3644        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3645    
3646        if (inescq)        if (inescq)
3647          {          {
3648          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3649            {            {
3650            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3651            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2561  for (;; ptr++) Line 3660  for (;; ptr++)
3660        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3661        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3662    
3663        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3664            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3665            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3666          {          {
3667          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3668          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3669          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3670          uschar pbits[32];          uschar pbits[32];
3671    
3672          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3673            {            {
3674            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3675            goto FAILED;            goto FAILED;
3676            }            }
3677    
3678          ptr += 2;          ptr += 2;
3679          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3680            {            {
3681            local_negate = TRUE;            local_negate = TRUE;
3682              should_flip_negation = TRUE;  /* Note negative special */
3683            ptr++;            ptr++;
3684            }            }
3685    
3686          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3687          if (posix_class < 0)          if (posix_class < 0)
3688            {            {
3689            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2597  for (;; ptr++) Line 3697  for (;; ptr++)
3697          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3698            posix_class = 0;            posix_class = 0;
3699    
3700          /* 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
3701          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3702          subtract bits that may be in the main map already. At the end we or the  
3703          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3704            if ((options & PCRE_UCP) != 0)
3705              {
3706              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3707              if (posix_substitutes[pc] != NULL)
3708                {
3709                nestptr = tempptr + 1;
3710                ptr = posix_substitutes[pc] - 1;
3711                continue;
3712                }
3713              }
3714    #endif
3715            /* In the non-UCP case, we build the bit map for the POSIX class in a
3716            chunk of local store because we may be adding and subtracting from it,
3717            and we don't want to subtract bits that may be in the main map already.
3718            At the end we or the result into the bit map that is being built. */
3719    
3720          posix_class *= 3;          posix_class *= 3;
3721    
# Line 2644  for (;; ptr++) Line 3759  for (;; ptr++)
3759    
3760        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3761        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
3762        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
3763        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
3764        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
3765        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3766          PCRE_EXTRA is set. */
3767    
3768        if (c == '\\')        if (c == CHAR_BACKSLASH)
3769          {          {
3770          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3771          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3772    
3773          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 */
3774          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_N)            /* \N is not supported in a class */
3775          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */            {
3776              *errorcodeptr = ERR71;
3777              goto FAILED;
3778              }
3779          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3780            {            {
3781            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3782              {              {
3783              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3784              }              }
3785            else inescq = TRUE;            else inescq = TRUE;
3786            continue;            continue;
3787            }            }
3788            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3789    
3790          if (c < 0)          if (c < 0)
3791            {            {
3792            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3793            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3794    
3795            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3796              {              {
3797    #ifdef SUPPORT_UCP
3798                case ESC_du:     /* These are the values given for \d etc */
3799                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3800                case ESC_wu:     /* escape sequence with an appropriate \p */
3801                case ESC_WU:     /* or \P to test Unicode properties instead */
3802                case ESC_su:     /* of the default ASCII testing. */
3803                case ESC_SU:
3804                nestptr = ptr;
3805                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3806                class_charcount -= 2;                /* Undo! */
3807                continue;
3808    #endif
3809              case ESC_d:              case ESC_d:
3810              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3811              continue;              continue;
3812    
3813              case ESC_D:              case ESC_D:
3814                should_flip_negation = TRUE;
3815              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3816              continue;              continue;
3817    
# Line 2689  for (;; ptr++) Line 3820  for (;; ptr++)
3820              continue;              continue;
3821    
3822              case ESC_W:              case ESC_W:
3823                should_flip_negation = TRUE;
3824              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3825              continue;              continue;
3826    
3827                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3828                if it was previously set by something earlier in the character
3829                class. */
3830    
3831              case ESC_s:              case ESC_s:
3832              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3833              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3834                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3835              continue;              continue;
3836    
3837              case ESC_S:              case ESC_S:
3838                should_flip_negation = TRUE;
3839              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3840              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3841              continue;              continue;
3842    
3843              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)  
             {  
3844              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3845              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3846              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2742  for (;; ptr++) Line 3864  for (;; ptr++)
3864                }                }
3865  #endif  #endif
3866              continue;              continue;
             }  
3867    
3868            if (-c == ESC_H)              case ESC_H:
             {  
3869              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3870                {                {
3871                int x = 0xff;                int x = 0xff;
# Line 2787  for (;; ptr++) Line 3907  for (;; ptr++)
3907                }                }
3908  #endif  #endif
3909              continue;              continue;
             }  
3910    
3911            if (-c == ESC_v)              case ESC_v:
             {  
3912              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3913              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3914              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2806  for (;; ptr++) Line 3924  for (;; ptr++)
3924                }                }
3925  #endif  #endif
3926              continue;              continue;
             }  
3927    
3928            if (-c == ESC_V)              case ESC_V:
             {  
3929              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3930                {                {
3931                int x = 0xff;                int x = 0xff;
# Line 2839  for (;; ptr++) Line 3955  for (;; ptr++)
3955                }                }
3956  #endif  #endif
3957              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3958    
3959  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3960            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3961              {              case ESC_P:
3962              BOOL negated;                {
3963              int pdata;                BOOL negated;
3964              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3965              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3966              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3967              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3968                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3969              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3970              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3971              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3972              continue;                class_charcount -= 2;   /* Not a < 256 character */
3973              }                continue;
3974                  }
3975  #endif  #endif
3976            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3977            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3978            treated as literals. */              treated as literals. */
3979    
3980            if ((options & PCRE_EXTRA) != 0)              default:
3981              {              if ((options & PCRE_EXTRA) != 0)
3982              *errorcodeptr = ERR7;                {
3983              goto FAILED;                *errorcodeptr = ERR7;
3984                  goto FAILED;
3985                  }
3986                class_charcount -= 2;  /* Undo the default count from above */
3987                c = *ptr;              /* Get the final character and fall through */
3988                break;
3989              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3990            }            }
3991    
3992          /* 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 2884  for (;; ptr++) Line 4000  for (;; ptr++)
4000        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
4001    
4002        CHECK_RANGE:        CHECK_RANGE:
4003        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4004          {          {
4005          inescq = FALSE;          inescq = FALSE;
4006          ptr += 2;          ptr += 2;
# Line 2892  for (;; ptr++) Line 4008  for (;; ptr++)
4008    
4009        oldptr = ptr;        oldptr = ptr;
4010    
4011        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
4012    
4013          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4014    
4015          /* Check for range */
4016    
4017          if (!inescq && ptr[1] == CHAR_MINUS)
4018          {          {
4019          int d;          int d;
4020          ptr += 2;          ptr += 2;
4021          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4022    
4023          /* 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
4024          mode. */          mode. */
4025    
4026          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4027            {            {
4028            ptr += 2;            ptr += 2;
4029            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4030                { ptr += 2; continue; }
4031            inescq = TRUE;            inescq = TRUE;
4032            break;            break;
4033            }            }
4034    
4035          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4036            {            {
4037            ptr = oldptr;            ptr = oldptr;
4038            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2928  for (;; ptr++) Line 4051  for (;; ptr++)
4051          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
4052          in such circumstances. */          in such circumstances. */
4053    
4054          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
4055            {            {
4056            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4057            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4058    
4059            /* \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 */  
4060    
4061            if (d < 0)            if (d < 0)
4062              {              {
4063              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  
4064                {                {
4065                ptr = oldptr;                ptr = oldptr;
4066                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2959  for (;; ptr++) Line 4079  for (;; ptr++)
4079    
4080          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
4081    
4082            /* Remember \r or \n */
4083    
4084            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4085    
4086          /* 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
4087          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
4088          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3077  for (;; ptr++) Line 4201  for (;; ptr++)
4201          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4202            {            {
4203            unsigned int othercase;            unsigned int othercase;
4204            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4205              {              {
4206              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4207              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3102  for (;; ptr++) Line 4226  for (;; ptr++)
4226          }          }
4227        }        }
4228    
4229      /* 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.
4230        If we are at the end of an internal nested string, revert to the outer
4231        string. */
4232    
4233        while (((c = *(++ptr)) != 0 ||
4234               (nestptr != NULL &&
4235                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4236               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4237    
4238      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4239    
4240      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4241        {        {
4242        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4243        goto FAILED;        goto FAILED;
4244        }        }
4245    
4246      /* 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
4247      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
4248      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
4249      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4250      single-bytes only. This is an historical hangover. Maybe one day we can  
4251      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4252        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4253        operate on single-bytes characters only. This is an historical hangover.
4254        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4255    
4256      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
4257      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4258      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4259      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4260      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4261      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4262    
4263  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4264      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4265            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4266  #else  #else
4267      if (class_charcount == 1)      if (class_charcount == 1)
4268  #endif  #endif
4269        {        {
4270        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4271    
4272        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4273    
4274        if (negate_class)        if (negate_class)
4275          {          {
4276          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4277          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4278          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4279          *code++ = class_lastchar;          *code++ = class_lastchar;
4280          break;          break;
4281          }          }
# Line 3173  for (;; ptr++) Line 4305  for (;; ptr++)
4305      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4306    
4307      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4308      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4309      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
4310        characters > 255 are in the class, so any that were explicitly given as
4311        well can be ignored. If (when there are explicit characters > 255 that must
4312        be listed) there are no characters < 256, we can omit the bitmap in the
4313        actual compiled code. */
4314    
4315  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4316      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4317        {        {
4318        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4319        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3203  for (;; ptr++) Line 4339  for (;; ptr++)
4339        }        }
4340  #endif  #endif
4341    
4342      /* 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
4343      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
4344      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
4345      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4346        negating it if necessary. */
4347    
4348        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4349      if (negate_class)      if (negate_class)
4350        {        {
       *code++ = OP_NCLASS;  
4351        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4352          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4353        }        }
4354      else      else
4355        {        {
       *code++ = OP_CLASS;  
4356        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4357        }        }
4358      code += 32;      code += 32;
# Line 3227  for (;; ptr++) Line 4363  for (;; ptr++)
4363      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4364      has been tested above. */      has been tested above. */
4365    
4366      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4367      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4368      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4369      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4370      goto REPEAT;      goto REPEAT;
4371    
4372      case '*':      case CHAR_ASTERISK:
4373      repeat_min = 0;      repeat_min = 0;
4374      repeat_max = -1;      repeat_max = -1;
4375      goto REPEAT;      goto REPEAT;
4376    
4377      case '+':      case CHAR_PLUS:
4378      repeat_min = 1;      repeat_min = 1;
4379      repeat_max = -1;      repeat_max = -1;
4380      goto REPEAT;      goto REPEAT;
4381    
4382      case '?':      case CHAR_QUESTION_MARK:
4383      repeat_min = 0;      repeat_min = 0;
4384      repeat_max = 1;      repeat_max = 1;
4385    
# Line 3267  for (;; ptr++) Line 4403  for (;; ptr++)
4403      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4404      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4405    
4406      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4407      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4408    
4409      tempcode = previous;      tempcode = previous;
4410    
# Line 3278  for (;; ptr++) Line 4414  for (;; ptr++)
4414      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
4415      repeat type to the non-default. */      repeat type to the non-default. */
4416    
4417      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4418        {        {
4419        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4420        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4421        ptr++;        ptr++;
4422        }        }
4423      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4424        {        {
4425        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4426        ptr++;        ptr++;
4427        }        }
4428      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4429    
4430        /* If previous was a recursion call, wrap it in atomic brackets so that
4431        previous becomes the atomic group. All recursions were so wrapped in the
4432        past, but it no longer happens for non-repeated recursions. In fact, the
4433        repeated ones could be re-implemented independently so as not to need this,
4434        but for the moment we rely on the code for repeating groups. */
4435    
4436        if (*previous == OP_RECURSE)
4437          {
4438          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4439          *previous = OP_ONCE;
4440          PUT(previous, 1, 2 + 2*LINK_SIZE);
4441          previous[2 + 2*LINK_SIZE] = OP_KET;
4442          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4443          code += 2 + 2 * LINK_SIZE;
4444          length_prevgroup = 3 + 3*LINK_SIZE;
4445    
4446          /* When actually compiling, we need to check whether this was a forward
4447          reference, and if so, adjust the offset. */
4448    
4449          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4450            {
4451            int offset = GET(cd->hwm, -LINK_SIZE);
4452            if (offset == previous + 1 - cd->start_code)
4453              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4454            }
4455          }
4456    
4457        /* Now handle repetition for the different types of item. */
4458    
4459      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4460      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4461      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4462      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4463      instead.  */      instead.  */
4464    
4465      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4466        {        {
4467          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4468    
4469        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4470        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4471        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3331  for (;; ptr++) Line 4498  for (;; ptr++)
4498    
4499        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4500            repeat_max < 0 &&            repeat_max < 0 &&
4501            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4502          {          {
4503          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4504          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3344  for (;; ptr++) Line 4510  for (;; ptr++)
4510      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4511      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4512      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4513      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4514      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4515    
4516      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4517        {        {
4518        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4519        c = previous[1];        c = previous[1];
4520        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4521            repeat_max < 0 &&            repeat_max < 0 &&
4522            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4523          {          {
4524          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4525          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3377  for (;; ptr++) Line 4543  for (;; ptr++)
4543    
4544        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4545            repeat_max < 0 &&            repeat_max < 0 &&
4546            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4547          {          {
4548          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4549          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3399  for (;; ptr++) Line 4565  for (;; ptr++)
4565    
4566        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4567    
4568          /*--------------------------------------------------------------------*/
4569          /* This code is obsolete from release 8.00; the restriction was finally
4570          removed: */
4571    
4572        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4573        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4574    
4575        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4576          /*--------------------------------------------------------------------*/
4577    
4578        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4579    
# Line 3541  for (;; ptr++) Line 4712  for (;; ptr++)
4712  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4713               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4714  #endif  #endif
4715               *previous == OP_REF)               *previous == OP_REF ||
4716                 *previous == OP_REFI)
4717        {        {
4718        if (repeat_max == 0)        if (repeat_max == 0)
4719          {          {
# Line 3549  for (;; ptr++) Line 4721  for (;; ptr++)
4721          goto END_REPEAT;          goto END_REPEAT;
4722          }          }
4723    
4724          /*--------------------------------------------------------------------*/
4725          /* This code is obsolete from release 8.00; the restriction was finally
4726          removed: */
4727    
4728        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4729        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4730    
4731        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4732          /*--------------------------------------------------------------------*/
4733    
4734        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4735          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3570  for (;; ptr++) Line 4747  for (;; ptr++)
4747        }        }
4748    
4749      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4750      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4751        opcodes such as BRA and CBRA, as this is the place where they get converted
4752        into the more special varieties such as BRAPOS and SBRA. A test for >=
4753        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4754        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4755        repetition of assertions, but now it does, for Perl compatibility. */
4756    
4757      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4758        {        {
4759        register int i;        register int i;
4760        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4761        uschar *bralink = NULL;        uschar *bralink = NULL;
4762          uschar *brazeroptr = NULL;
4763    
4764        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4765          we just ignore the repeat. */
4766    
4767        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4768          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4769    
4770        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4771        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4772        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4773        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4774        pointer. */  
4775          if (*previous < OP_ONCE)    /* Assertion */
4776        if (repeat_max == -1)          {
4777          {          if (repeat_min > 0) goto END_REPEAT;
4778          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4779          }          }
4780    
4781        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3610  for (;; ptr++) Line 4787  for (;; ptr++)
4787    
4788        if (repeat_min == 0)        if (repeat_min == 0)
4789          {          {
4790          /* 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
4791          altogether. */          output altogether, like this:
4792    
4793          if (repeat_max == 0)          ** if (repeat_max == 0)
4794            {          **   {
4795            code = previous;          **   code = previous;
4796            goto END_REPEAT;          **   goto END_REPEAT;
4797            }          **   }
4798    
4799            However, that fails when a group or a subgroup within it is referenced
4800            as a subroutine from elsewhere in the pattern, so now we stick in
4801            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4802            don't have a list of which groups are referenced, we cannot do this
4803            selectively.
4804    
4805            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4806            and do no more at this point. However, we do need to adjust any
4807            OP_RECURSE calls inside the group that refer to the group itself or any
4808            internal or forward referenced group, because the offset is from the
4809            start of the whole regex. Temporarily terminate the pattern while doing
4810            this. */
4811    
4812          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4813            {            {
4814            *code = OP_END;            *code = OP_END;
4815            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4816            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4817            code++;            code++;
4818              if (repeat_max == 0)
4819                {
4820                *previous++ = OP_SKIPZERO;
4821                goto END_REPEAT;
4822                }
4823              brazeroptr = previous;    /* Save for possessive optimizing */
4824            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4825            }            }
4826    
# Line 3656  for (;; ptr++) Line 4845  for (;; ptr++)
4845            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4846            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4847    
4848            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4849            bralink = previous;            bralink = previous;
4850            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4851            }            }
# Line 3677  for (;; ptr++) Line 4866  for (;; ptr++)
4866            {            {
4867            /* 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
4868            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4869            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4870              integer type when available, otherwise double. */
4871    
4872            if (lengthptr != NULL)            if (lengthptr != NULL)
4873              {              {
4874              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4875              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4876                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4877