/[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 166 by ph10, Wed May 9 14:48:28 2007 UTC revision 438 by ph10, Sun Sep 6 20:00:47 2009 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) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
# Line 58  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
# Line 82  are simple data values; negative values Line 97  are simple data values; negative values
97  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
98  is invalid. */  is invalid. */
99    
100  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  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,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
179  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    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 171  static const int posix_class_maps[] = { Line 257  static const int posix_class_maps[] = {
257  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
258  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
259  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
260  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
261    the number of relocations needed when a shared library is loaded dynamically,
262  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
263    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
265    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
266    "unrecognized character follows \\",  
267    "numbers out of order in {} quantifier",  static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number or name after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
319    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
325    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327    /* 50 */    /* 50 */
328    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
329    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
330    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
331    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
332    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
333    /* 55 */    /* 55 */
334    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
335    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
336    "\\g is not followed by an (optionally braced) non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
337    "(?+ or (?- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
338  };    "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 263  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 299  static const unsigned char digitab[] = Line 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 374  static const unsigned char ebcdic_charta Line 478  static const unsigned char ebcdic_charta
478  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
479    
480  static BOOL  static BOOL
481    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
482      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
483    
484    
485    
486    /*************************************************
487    *            Find an error text                  *
488    *************************************************/
489    
490    /* The error texts are now all in one long string, to save on relocations. As
491    some of the text is of unknown length, we can't use a table of offsets.
492    Instead, just count through the strings. This is not a performance issue
493    because it happens only when there has been a compilation error.
494    
495    Argument:   the error number
496    Returns:    pointer to the error string
497    */
498    
499    static const char *
500    find_error_text(int n)
501    {
502    const char *s = error_texts;
503    for (; n > 0; n--) while (*s++ != 0) {};
504    return s;
505    }
506    
507    
508  /*************************************************  /*************************************************
509  *            Handle escapes                      *  *            Handle escapes                      *
# Line 400  Arguments: Line 526  Arguments:
526    
527  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
528                   negative => a special escape sequence                   negative => a special escape sequence
529                   on error, errorptr is set                   on error, errorcodeptr is set
530  */  */
531    
532  static int  static int
# Line 418  ptr--;                            /* Set Line 544  ptr--;                            /* Set
544    
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* 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
548  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.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 443  else Line 569  else
569      /* 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
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
581      is an absolute backreference. If negative, it is a relative backreference.  
582      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
583        backreference. If negative, it is a relative backreference. This is a Perl
584        5.10 feature.
585    
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596      case 'g':      case CHAR_g:
597      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598        {        {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606          {
607          const uschar *p;
608          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611            {
612            c = -ESC_k;
613            break;
614            }
615        braced = TRUE;        braced = TRUE;
616        ptr++;        ptr++;
617        }        }
618      else braced = FALSE;      else braced = FALSE;
619    
620      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
621        {        {
622        negated = TRUE;        negated = TRUE;
623        ptr++;        ptr++;
# Line 472  else Line 626  else
626    
627      c = 0;      c = 0;
628      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
629        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
630    
631        if (c < 0)   /* Integer overflow */
632          {
633          *errorcodeptr = ERR61;
634          break;
635          }
636    
637      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638        {        {
639        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
640        return 0;        break;
641          }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647        }        }
648    
649      if (negated)      if (negated)
# Line 485  else Line 651  else
651        if (c > bracount)        if (c > bracount)
652          {          {
653          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
654          return 0;          break;
655          }          }
656        c = bracount - (c - 1);        c = bracount - (c - 1);
657        }        }
# Line 505  else Line 671  else
671      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
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      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:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683          if (c < 0)    /* Integer overflow */
684            {
685            *errorcodeptr = ERR61;
686            break;
687            }
688        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
689          {          {
690          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 526  else Line 697  else
697      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.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 539  else Line 710  else
710      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
711      than 3 octal digits. */      than 3 octal digits. */
712    
713      case '0':      case CHAR_0:
714      c -= '0';      c -= CHAR_0;
715      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
717      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
# Line 550  else Line 721  else
721      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
722      treated as a data character. */      treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        int count = 0;        int count = 0;
# Line 560  else Line 731  else
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          register int cc = *pt++;          register int cc = *pt++;
734          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736    
737  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
741          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745    
746        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
# Line 588  else Line 759  else
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
# Line 604  else Line 775  else
775      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
776      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
782        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
783        return 0;        break;
784        }        }
785    
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
788      c ^= 0x40;      c ^= 0x40;
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      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
798      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
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 678  if (c == 0) goto ERROR_RETURN; Line 849  if (c == 0) goto ERROR_RETURN;
849  /* \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
850  negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
858      }      }
859    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 714  top = _pcre_utt_size; Line 885  top = _pcre_utt_size;
885  while (bot < top)  while (bot < top)
886    {    {
887    i = (bot + top) >> 1;    i = (bot + top) >> 1;
888    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
889    if (c == 0)    if (c == 0)
890      {      {
891      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 757  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 798  int max = -1; Line 969  int max = -1;
969  /* 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
970  an integer overflow. */  an integer overflow. */
971    
972  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
974    {    {
975    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 808  if (min < 0 || min > 65535) Line 979  if (min < 0 || min > 65535)
979  /* 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.
980  Also, max must not be less than min. */  Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
989        {        {
990        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 838  return p; Line 1009  return p;
1009    
1010    
1011  /*************************************************  /*************************************************
1012  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1013  *************************************************/  *************************************************/
1014    
1015  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018  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
1019  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
1020  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1021  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1022  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025  Arguments:  Arguments:
1026    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1027    count        current count of capturing parens so far encountered    cd           compile background data
1028    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1029    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1030    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1034  */  */
1035    
1036  static int  static int
1037  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,
1038    BOOL xmode)    BOOL xmode, int *count)
1039  {  {
1040  const uschar *thisname;  uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051          ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          term++;
1104          }
1105        }
1106      }
1107    
1108    /* Past any initial parenthesis handling, scan for parentheses or vertical
1109    bars. */
1110    
1111    for (; *ptr != 0; ptr++)
1112      {
1113    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1114    
1115    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1116      {      {
1117      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1118      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1119        {        {
1120        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1121        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1122        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1123        }        }
1124      continue;      continue;
1125      }      }
1126    
1127    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1128      are handled for real. If the first character is '^', skip it. Also, if the
1129      first few characters (either before or after ^) are \Q\E or \E we skip them
1130      too. This makes for compatibility with Perl. Note the use of STR macros to
1131      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1132    
1133    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134      {      {
1135      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1136        for (;;)
1137        {        {
1138        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1139          {          {
1140          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1141          if (*ptr == 'Q') for (;;)            ptr+= 2;
1142            else if (strncmp((const char *)ptr+2,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 4;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1149            {
1150            negate_class = TRUE;
1151            ptr++;
1152            }
1153          else break;
1154          }
1155    
1156        /* If the next character is ']', it is a data character that must be
1157        skipped, except in JavaScript compatibility mode. */
1158    
1159        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1160            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1161          ptr++;
1162    
1163        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1164          {
1165          if (*ptr == 0) return -1;
1166          if (*ptr == CHAR_BACKSLASH)
1167            {
1168            if (*(++ptr) == 0) goto FAIL_EXIT;
1169            if (*ptr == CHAR_Q) for (;;)
1170            {            {
1171            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1172            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1173            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1174            }            }
1175          continue;          continue;
1176          }          }
# Line 905  for (; *ptr != 0; ptr++) Line 1180  for (; *ptr != 0; ptr++)
1180    
1181    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1182    
1183    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1184      {      {
1185      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1186      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1187      continue;      continue;
1188      }      }
1189    
1190    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1191    
1192    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1193      {      {
1194      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1195      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1196      continue;      if (*ptr == 0) goto FAIL_EXIT;
1197        }
1198    
1199      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200        {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204      }      }
1205    
1206    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1207    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1208        if (*count > hwm_count) hwm_count = *count;
1209        *count = start_count;
1210        }
1211      }
1212    
1213    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1214    *ptrptr = ptr;
1215    return -1;
1216    }
1217    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1218    
   count++;  
1219    
1220    if (name == NULL && count == lorn) return count;  
1221    term = *ptr++;  /*************************************************
1222    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1223    thisname = ptr;  *************************************************/
1224    while (*ptr != term) ptr++;  
1225    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1226        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1227      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1228    returns when it reaches a given numbered subpattern. This is used for forward
1229    references to subpatterns. We used to be able to start this scan from the
1230    current compiling point, using the current count value from cd->bracount, and
1231    do it all in a single loop, but the addition of the possibility of duplicate
1232    subpattern numbers means that we have to scan from the very start, in order to
1233    take account of such duplicates, and to use a recursive function to keep track
1234    of the different types of group.
1235    
1236    Arguments:
1237      cd           compile background data
1238      name         name to seek, or NULL if seeking a numbered subpattern
1239      lorn         name length, or subpattern number if name is NULL
1240      xmode        TRUE if we are in /x mode
1241    
1242    Returns:       the number of the found subpattern, or -1 if not found
1243    */
1244    
1245    static int
1246    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1247    {
1248    uschar *ptr = (uschar *)cd->start_pattern;
1249    int count = 0;
1250    int rc;
1251    
1252    /* If the pattern does not start with an opening parenthesis, the first call
1253    to find_parens_sub() will scan right to the end (if necessary). However, if it
1254    does start with a parenthesis, find_parens_sub() will return when it hits the
1255    matching closing parens. That is why we have to have a loop. */
1256    
1257    for (;;)
1258      {
1259      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1260      if (rc > 0 || *ptr++ == 0) break;
1261    }    }
1262    
1263  return -1;  return rc;
1264  }  }
1265    
1266    
1267    
1268    
1269  /*************************************************  /*************************************************
1270  *      Find first significant op code            *  *      Find first significant op code            *
1271  *************************************************/  *************************************************/
# Line 1043  for (;;) Line 1361  for (;;)
1361    {    {
1362    int d;    int d;
1363    register int op = *cc;    register int op = *cc;
   
1364    switch (op)    switch (op)
1365      {      {
1366      case OP_CBRA:      case OP_CBRA:
# Line 1109  for (;;) Line 1426  for (;;)
1426      branchlength++;      branchlength++;
1427      cc += 2;      cc += 2;
1428  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1429      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1430        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1431  #endif  #endif
1432      break;      break;
1433    
# Line 1123  for (;;) Line 1438  for (;;)
1438      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1439      cc += 4;      cc += 4;
1440  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1441      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1442        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1443  #endif  #endif
1444      break;      break;
1445    
1446      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1447      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1448        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1449      cc += 4;      cc += 4;
1450      break;      break;
1451    
# Line 1149  for (;;) Line 1463  for (;;)
1463      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1464      case OP_WORDCHAR:      case OP_WORDCHAR:
1465      case OP_ANY:      case OP_ANY:
1466        case OP_ALLANY:
1467      branchlength++;      branchlength++;
1468      cc++;      cc++;
1469      break;      break;
# Line 1240  for (;;) Line 1555  for (;;)
1555      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1556      }      }
1557    
1558    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1559    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1560    arrange to skip the extra bytes. */    two bytes of parameters. */
1561    
1562    else    else
1563      {      {
1564        switch(c)
1565          {
1566          case OP_TYPESTAR:
1567          case OP_TYPEMINSTAR:
1568          case OP_TYPEPLUS:
1569          case OP_TYPEMINPLUS:
1570          case OP_TYPEQUERY:
1571          case OP_TYPEMINQUERY:
1572          case OP_TYPEPOSSTAR:
1573          case OP_TYPEPOSPLUS:
1574          case OP_TYPEPOSQUERY:
1575          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1576          break;
1577    
1578          case OP_TYPEUPTO:
1579          case OP_TYPEMINUPTO:
1580          case OP_TYPEEXACT:
1581          case OP_TYPEPOSUPTO:
1582          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1583          break;
1584          }
1585    
1586        /* Add in the fixed length from the table */
1587    
1588      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1589    
1590      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1591      a multi-byte character. The length in the table is a minimum, so we have to
1592      arrange to skip the extra bytes. */
1593    
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      if (utf8) switch(c)      if (utf8) switch(c)
1596        {        {
# Line 1268  for (;;) Line 1612  for (;;)
1612        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1613        break;        break;
1614        }        }
1615    #else
1616        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1617  #endif  #endif
1618      }      }
1619    }    }
# Line 1304  for (;;) Line 1650  for (;;)
1650    
1651    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1652    
1653    /* 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
1654    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
1655    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1656    
1657    else    else
1658      {      {
1659        switch(c)
1660          {
1661          case OP_TYPESTAR:
1662          case OP_TYPEMINSTAR:
1663          case OP_TYPEPLUS:
1664          case OP_TYPEMINPLUS:
1665          case OP_TYPEQUERY:
1666          case OP_TYPEMINQUERY:
1667          case OP_TYPEPOSSTAR:
1668          case OP_TYPEPOSPLUS:
1669          case OP_TYPEPOSQUERY:
1670          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1671          break;
1672    
1673          case OP_TYPEPOSUPTO:
1674          case OP_TYPEUPTO:
1675          case OP_TYPEMINUPTO:
1676          case OP_TYPEEXACT:
1677          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1678          break;
1679          }
1680    
1681        /* Add in the fixed length from the table */
1682    
1683      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1684    
1685        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1686        by a multi-byte character. The length in the table is a minimum, so we have
1687        to arrange to skip the extra bytes. */
1688    
1689  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1690      if (utf8) switch(c)      if (utf8) switch(c)
1691        {        {
# Line 1333  for (;;) Line 1707  for (;;)
1707        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1708        break;        break;
1709        }        }
1710    #else
1711        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1712  #endif  #endif
1713      }      }
1714    }    }
# Line 1348  for (;;) Line 1724  for (;;)
1724  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()
1725  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
1726  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1727  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
1728  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
1729    bracket whose current branch will already have been scanned.
1730    
1731  Arguments:  Arguments:
1732    code        points to start of search    code        points to start of search
# Line 1371  for (code = first_significant_code(code Line 1748  for (code = first_significant_code(code
1748    
1749    c = *code;    c = *code;
1750    
1751    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1752      first_significant_code() with a TRUE final argument. */
1753    
1754      if (c == OP_ASSERT)
1755        {
1756        do code += GET(code, 1); while (*code == OP_ALT);
1757        c = *code;
1758        continue;
1759        }
1760    
1761      /* Groups with zero repeats can of course be empty; skip them. */
1762    
1763      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1764        {
1765        code += _pcre_OP_lengths[c];
1766        do code += GET(code, 1); while (*code == OP_ALT);
1767        c = *code;
1768        continue;
1769        }
1770    
1771      /* For other groups, scan the branches. */
1772    
1773      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1774      {      {
1775      BOOL empty_branch;      BOOL empty_branch;
1776      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1777    
1778      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1779        empty branch, so just skip over the conditional, because it could be empty.
1780        Otherwise, scan the individual branches of the group. */
1781    
1782      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1783        code += GET(code, 1);        code += GET(code, 1);
1784        else
1785          {
1786          empty_branch = FALSE;
1787          do
1788            {
1789            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1790              empty_branch = TRUE;
1791            code += GET(code, 1);
1792            }
1793          while (*code == OP_ALT);
1794          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1795        }        }
     while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1796    
1797      /* Move past the KET and fudge things so that the increment in the "for"      c = *code;
     above has no effect. */  
   
     c = OP_END;  
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1798      continue;      continue;
1799      }      }
1800    
# Line 1400  for (code = first_significant_code(code Line 1802  for (code = first_significant_code(code
1802    
1803    switch (c)    switch (c)
1804      {      {
1805      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1806        cannot be represented just by a bit map. This includes negated single
1807        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1808        actual length is stored in the compiled code, so we must update "code"
1809        here. */
1810    
1811  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1812      case OP_XCLASS:      case OP_XCLASS:
1813      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1814      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1815  #endif  #endif
1816    
# Line 1448  for (code = first_significant_code(code Line 1854  for (code = first_significant_code(code
1854      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1855      case OP_WORDCHAR:      case OP_WORDCHAR:
1856      case OP_ANY:      case OP_ANY:
1857        case OP_ALLANY:
1858      case OP_ANYBYTE:      case OP_ANYBYTE:
1859      case OP_CHAR:      case OP_CHAR:
1860      case OP_CHARNC:      case OP_CHARNC:
# Line 1466  for (code = first_significant_code(code Line 1873  for (code = first_significant_code(code
1873      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1874      return FALSE;      return FALSE;
1875    
1876        /* These are going to continue, as they may be empty, but we have to
1877        fudge the length for the \p and \P cases. */
1878    
1879        case OP_TYPESTAR:
1880        case OP_TYPEMINSTAR:
1881        case OP_TYPEPOSSTAR:
1882        case OP_TYPEQUERY:
1883        case OP_TYPEMINQUERY:
1884        case OP_TYPEPOSQUERY:
1885        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1886        break;
1887    
1888        /* Same for these */
1889    
1890        case OP_TYPEUPTO:
1891        case OP_TYPEMINUPTO:
1892        case OP_TYPEPOSUPTO:
1893        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1894        break;
1895    
1896      /* End of branch */      /* End of branch */
1897    
1898      case OP_KET:      case OP_KET:
# Line 1484  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      case OP_QUERY:      case OP_QUERY:
1912      case OP_MINQUERY:      case OP_MINQUERY:
1913      case OP_POSQUERY:      case OP_POSQUERY:
1914        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1915        break;
1916    
1917      case OP_UPTO:      case OP_UPTO:
1918      case OP_MINUPTO:      case OP_MINUPTO:
1919      case OP_POSUPTO:      case OP_POSUPTO:
1920      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1921      break;      break;
1922  #endif  #endif
1923      }      }
# Line 1535  return TRUE; Line 1965  return TRUE;
1965  *************************************************/  *************************************************/
1966    
1967  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1968  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
1969  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1970  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1971    
1972    Originally, this function only recognized a sequence of letters between the
1973    terminators, but it seems that Perl recognizes any sequence of characters,
1974    though of course unknown POSIX names are subsequently rejected. Perl gives an
1975    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1976    didn't consider this to be a POSIX class. Likewise for [:1234:].
1977    
1978    The problem in trying to be exactly like Perl is in the handling of escapes. We
1979    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1980    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1981    below handles the special case of \], but does not try to do any other escape
1982    processing. This makes it different from Perl for cases such as [:l\ower:]
1983    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1984    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1985    I think.
1986    
1987  Argument:  Arguments:
1988    ptr      pointer to the initial [    ptr      pointer to the initial [
1989    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1990    
1991  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1992  */  */
1993    
1994  static BOOL  static BOOL
1995  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1996  {  {
1997  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1998  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1999  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2000    {    {
2001    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2002    return TRUE;      {
2003        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2004        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2005          {
2006          *endptr = ptr;
2007          return TRUE;
2008          }
2009        }
2010    }    }
2011  return FALSE;  return FALSE;
2012  }  }
# Line 1582  Returns:     a value representing the na Line 2031  Returns:     a value representing the na
2031  static int  static int
2032  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2033  {  {
2034    const char *pn = posix_names;
2035  register int yield = 0;  register int yield = 0;
2036  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2037    {    {
2038    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2039      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2040      pn += posix_name_lengths[yield] + 1;
2041    yield++;    yield++;
2042    }    }
2043  return -1;  return -1;
# Line 1601  return -1; Line 2052  return -1;
2052  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2053  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2054  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
2055  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
2056  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
2057  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
2058  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
2059  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2060    OP_END.
2061    
2062  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2063  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 1628  adjust_recurse(uschar *group, int adjust Line 2080  adjust_recurse(uschar *group, int adjust
2080    uschar *save_hwm)    uschar *save_hwm)
2081  {  {
2082  uschar *ptr = group;  uschar *ptr = group;
2083    
2084  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2085    {    {
2086    int offset;    int offset;
# Line 1739  get_othercase_range(unsigned int *cptr, Line 2192  get_othercase_range(unsigned int *cptr,
2192  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2193    
2194  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2195    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2196    
2197  if (c > d) return FALSE;  if (c > d) return FALSE;
2198    
# Line 1748  next = othercase + 1; Line 2201  next = othercase + 1;
2201    
2202  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2203    {    {
2204    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2205    next++;    next++;
2206    }    }
2207    
# Line 1794  if ((options & PCRE_EXTENDED) != 0) Line 2247  if ((options & PCRE_EXTENDED) != 0)
2247    for (;;)    for (;;)
2248      {      {
2249      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2250      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2251        {        {
2252        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2253          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1806  if ((options & PCRE_EXTENDED) != 0) Line 2259  if ((options & PCRE_EXTENDED) != 0)
2259  /* 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
2260  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2261    
2262  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2263    {    {
2264    int temperrorcode = 0;    int temperrorcode = 0;
2265    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1831  if ((options & PCRE_EXTENDED) != 0) Line 2284  if ((options & PCRE_EXTENDED) != 0)
2284    for (;;)    for (;;)
2285      {      {
2286      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2287      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2288        {        {
2289        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2290          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1842  if ((options & PCRE_EXTENDED) != 0) Line 2295  if ((options & PCRE_EXTENDED) != 0)
2295    
2296  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2297    
2298  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2299    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2300        return FALSE;
2301    
2302  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2303  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1858  if (next >= 0) switch(op_code) Line 2312  if (next >= 0) switch(op_code)
2312    case OP_CHAR:    case OP_CHAR:
2313  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2314    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2315    #else
2316      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2317  #endif  #endif
2318    return item != next;    return item != next;
2319    
# Line 1876  if (next >= 0) switch(op_code) Line 2332  if (next >= 0) switch(op_code)
2332      unsigned int othercase;      unsigned int othercase;
2333      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2334  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2335      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2336  #else  #else
2337      othercase = NOTACHAR;      othercase = NOTACHAR;
2338  #endif  #endif
# Line 1889  if (next >= 0) switch(op_code) Line 2345  if (next >= 0) switch(op_code)
2345    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2346    
2347    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2348    if (item == next) return TRUE;    if (item == next) return TRUE;
2349    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2350  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1898  if (next >= 0) switch(op_code) Line 2353  if (next >= 0) switch(op_code)
2353      unsigned int othercase;      unsigned int othercase;
2354      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2355  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2356      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2357  #else  #else
2358      othercase = NOTACHAR;      othercase = NOTACHAR;
2359  #endif  #endif
# Line 1926  if (next >= 0) switch(op_code) Line 2381  if (next >= 0) switch(op_code)
2381    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2382    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2383    
2384      case OP_HSPACE:
2385      case OP_NOT_HSPACE:
2386      switch(next)
2387        {
2388        case 0x09:
2389        case 0x20:
2390        case 0xa0:
2391        case 0x1680:
2392        case 0x180e:
2393        case 0x2000:
2394        case 0x2001:
2395        case 0x2002:
2396        case 0x2003:
2397        case 0x2004:
2398        case 0x2005:
2399        case 0x2006:
2400        case 0x2007:
2401        case 0x2008:
2402        case 0x2009:
2403        case 0x200A:
2404        case 0x202f:
2405        case 0x205f:
2406        case 0x3000:
2407        return op_code != OP_HSPACE;
2408        default:
2409        return op_code == OP_HSPACE;
2410        }
2411    
2412      case OP_VSPACE:
2413      case OP_NOT_VSPACE:
2414      switch(next)
2415        {
2416        case 0x0a:
2417        case 0x0b:
2418        case 0x0c:
2419        case 0x0d:
2420        case 0x85:
2421        case 0x2028:
2422        case 0x2029:
2423        return op_code != OP_VSPACE;
2424        default:
2425        return op_code == OP_VSPACE;
2426        }
2427    
2428    default:    default:
2429    return FALSE;    return FALSE;
2430    }    }
# Line 1960  switch(op_code) Line 2459  switch(op_code)
2459      case ESC_W:      case ESC_W:
2460      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2461    
2462        case ESC_h:
2463        case ESC_H:
2464        switch(item)
2465          {
2466          case 0x09:
2467          case 0x20:
2468          case 0xa0:
2469          case 0x1680:
2470          case 0x180e:
2471          case 0x2000:
2472          case 0x2001:
2473          case 0x2002:
2474          case 0x2003:
2475          case 0x2004:
2476          case 0x2005:
2477          case 0x2006:
2478          case 0x2007:
2479          case 0x2008:
2480          case 0x2009:
2481          case 0x200A:
2482          case 0x202f:
2483          case 0x205f:
2484          case 0x3000:
2485          return -next != ESC_h;
2486          default:
2487          return -next == ESC_h;
2488          }
2489    
2490        case ESC_v:
2491        case ESC_V:
2492        switch(item)
2493          {
2494          case 0x0a:
2495          case 0x0b:
2496          case 0x0c:
2497          case 0x0d:
2498          case 0x85:
2499          case 0x2028:
2500          case 0x2029:
2501          return -next != ESC_v;
2502          default:
2503          return -next == ESC_v;
2504          }
2505    
2506      default:      default:
2507      return FALSE;      return FALSE;
2508      }      }
2509    
2510    case OP_DIGIT:    case OP_DIGIT:
2511    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2512             next == -ESC_h || next == -ESC_v;
2513    
2514    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2515    return next == -ESC_d;    return next == -ESC_d;
# Line 1974  switch(op_code) Line 2518  switch(op_code)
2518    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2519    
2520    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2521    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2522    
2523      case OP_HSPACE:
2524      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2525    
2526      case OP_NOT_HSPACE:
2527      return next == -ESC_h;
2528    
2529      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2530      case OP_VSPACE:
2531      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2532    
2533      case OP_NOT_VSPACE:
2534      return next == -ESC_v;
2535    
2536    case OP_WORDCHAR:    case OP_WORDCHAR:
2537    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2538    
2539    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2540    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2050  uschar classbits[32]; Line 2607  uschar classbits[32];
2607  BOOL class_utf8;  BOOL class_utf8;
2608  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2609  uschar *class_utf8data;  uschar *class_utf8data;
2610    uschar *class_utf8data_base;
2611  uschar utf8_char[6];  uschar utf8_char[6];
2612  #else  #else
2613  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2089  req_caseopt = ((options & PCRE_CASELESS) Line 2647  req_caseopt = ((options & PCRE_CASELESS)
2647  for (;; ptr++)  for (;; ptr++)
2648    {    {
2649    BOOL negate_class;    BOOL negate_class;
2650      BOOL should_flip_negation;
2651    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2652    BOOL is_quantifier;    BOOL is_quantifier;
2653    BOOL is_recurse;    BOOL is_recurse;
2654      BOOL reset_bracount;
2655    int class_charcount;    int class_charcount;
2656    int class_lastchar;    int class_lastchar;
2657    int newoptions;    int newoptions;
2658    int recno;    int recno;
2659      int refsign;
2660    int skipbytes;    int skipbytes;
2661    int subreqbyte;    int subreqbyte;
2662    int subfirstbyte;    int subfirstbyte;
# Line 2128  for (;; ptr++) Line 2689  for (;; ptr++)
2689      */      */
2690    
2691      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2692    
2693        /* Paranoid check for integer overflow */
2694    
2695        if (OFLOW_MAX - *lengthptr < code - last_code)
2696          {
2697          *errorcodeptr = ERR20;
2698          goto FAILED;
2699          }
2700    
2701      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2702      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2703    
# Line 2165  for (;; ptr++) Line 2735  for (;; ptr++)
2735    
2736    if (inescq && c != 0)    if (inescq && c != 0)
2737      {      {
2738      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2739        {        {
2740        inescq = FALSE;        inescq = FALSE;
2741        ptr++;        ptr++;
# Line 2191  for (;; ptr++) Line 2761  for (;; ptr++)
2761    /* 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
2762    a quantifier. */    a quantifier. */
2763    
2764    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2765      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2766        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2767    
2768    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2769         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2207  for (;; ptr++) Line 2778  for (;; ptr++)
2778    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2779      {      {
2780      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2781      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2782        {        {
2783        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2784          {          {
# Line 2232  for (;; ptr++) Line 2803  for (;; ptr++)
2803      {      {
2804      /* ===================================================================*/      /* ===================================================================*/
2805      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2806      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2807      case ')':      case CHAR_RIGHT_PARENTHESIS:
2808      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2809      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2810      *codeptr = code;      *codeptr = code;
2811      *ptrptr = ptr;      *ptrptr = ptr;
2812      if (lengthptr != NULL)      if (lengthptr != NULL)
2813        {        {
2814          if (OFLOW_MAX - *lengthptr < code - last_code)
2815            {
2816            *errorcodeptr = ERR20;
2817            goto FAILED;
2818            }
2819        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2820        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2821        }        }
# Line 2250  for (;; ptr++) Line 2826  for (;; ptr++)
2826      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2827      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2828    
2829      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2830      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2831        {        {
2832        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2259  for (;; ptr++) Line 2835  for (;; ptr++)
2835      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2836      break;      break;
2837    
2838      case '$':      case CHAR_DOLLAR_SIGN:
2839      previous = NULL;      previous = NULL;
2840      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2841      break;      break;
# Line 2267  for (;; ptr++) Line 2843  for (;; ptr++)
2843      /* 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
2844      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2845    
2846      case '.':      case CHAR_DOT:
2847      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2848      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2849      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2850      previous = code;      previous = code;
2851      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2852      break;      break;
2853    
2854    
# Line 2287  for (;; ptr++) Line 2863  for (;; ptr++)
2863      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,
2864      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2865      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.
     */  
2866    
2867      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2868        default (Perl) mode, it is treated as a data character. */
2869    
2870        case CHAR_RIGHT_SQUARE_BRACKET:
2871        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2872          {
2873          *errorcodeptr = ERR64;
2874          goto FAILED;
2875          }
2876        goto NORMAL_CHAR;
2877    
2878        case CHAR_LEFT_SQUARE_BRACKET:
2879      previous = code;      previous = code;
2880    
2881      /* 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
2882      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. */
2883    
2884      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2885          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2886            check_posix_syntax(ptr, &tempptr))
2887        {        {
2888        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2889        goto FAILED;        goto FAILED;
2890        }        }
2891    
2892      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2893        if the first few characters (either before or after ^) are \Q\E or \E we
2894        skip them too. This makes for compatibility with Perl. */
2895    
2896      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2897        for (;;)
2898        {        {
       negate_class = TRUE;  
2899        c = *(++ptr);        c = *(++ptr);
2900          if (c == CHAR_BACKSLASH)
2901            {
2902            if (ptr[1] == CHAR_E)
2903              ptr++;
2904            else if (strncmp((const char *)ptr+1,
2905                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2906              ptr += 3;
2907            else
2908              break;
2909            }
2910          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2911            negate_class = TRUE;
2912          else break;
2913        }        }
2914      else  
2915        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2916        an initial ']' is taken as a data character -- the code below handles
2917        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2918        [^] must match any character, so generate OP_ALLANY. */
2919    
2920        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2921            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2922        {        {
2923        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2924          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2925          zerofirstbyte = firstbyte;
2926          break;
2927        }        }
2928    
2929        /* If a class contains a negative special such as \S, we need to flip the
2930        negation flag at the end, so that support for characters > 255 works
2931        correctly (they are all included in the class). */
2932    
2933        should_flip_negation = FALSE;
2934    
2935      /* 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
2936      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
2937      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2331  for (;; ptr++) Line 2949  for (;; ptr++)
2949  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2950      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2951      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2952        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2953  #endif  #endif
2954    
2955      /* 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 2346  for (;; ptr++) Line 2965  for (;; ptr++)
2965          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2966          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2967          }          }
2968    
2969          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2970          data and reset the pointer. This is so that very large classes that
2971          contain a zillion UTF-8 characters no longer overwrite the work space
2972          (which is on the stack). */
2973    
2974          if (lengthptr != NULL)
2975            {
2976            *lengthptr += class_utf8data - class_utf8data_base;
2977            class_utf8data = class_utf8data_base;
2978            }
2979    
2980  #endif  #endif
2981    
2982        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2983    
2984        if (inescq)        if (inescq)
2985          {          {
2986          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2987            {            {
2988            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
2989            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2367  for (;; ptr++) Line 2998  for (;; ptr++)
2998        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2999        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3000    
3001        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3002            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3003            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3004          {          {
3005          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3006          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3007          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3008          uschar pbits[32];          uschar pbits[32];
3009    
3010          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3011            {            {
3012            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3013            goto FAILED;            goto FAILED;
3014            }            }
3015    
3016          ptr += 2;          ptr += 2;
3017          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3018            {            {
3019            local_negate = TRUE;            local_negate = TRUE;
3020              should_flip_negation = TRUE;  /* Note negative special */
3021            ptr++;            ptr++;
3022            }            }
3023    
# Line 2452  for (;; ptr++) Line 3084  for (;; ptr++)
3084        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
3085        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3086        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3087        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
3088        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3089    
3090        if (c == '\\')        if (c == CHAR_BACKSLASH)
3091          {          {
3092          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3093          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3094    
3095          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 */
3096          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3097          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3098          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3099            {            {
3100            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3101              {              {
3102              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3103              }              }
3104            else inescq = TRUE;            else inescq = TRUE;
3105            continue;            continue;
3106            }            }
3107            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3108    
3109          if (c < 0)          if (c < 0)
3110            {            {
# Line 2487  for (;; ptr++) Line 3120  for (;; ptr++)
3120              continue;              continue;
3121    
3122              case ESC_D:              case ESC_D:
3123                should_flip_negation = TRUE;
3124              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3125              continue;              continue;
3126    
# Line 2495  for (;; ptr++) Line 3129  for (;; ptr++)
3129              continue;              continue;
3130    
3131              case ESC_W:              case ESC_W:
3132                should_flip_negation = TRUE;
3133              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3134              continue;              continue;
3135    
# Line 2504  for (;; ptr++) Line 3139  for (;; ptr++)
3139              continue;              continue;
3140    
3141              case ESC_S:              case ESC_S:
3142                should_flip_negation = TRUE;
3143              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3144              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
             continue;  
   
             case ESC_E: /* Perl ignores an orphan \E */  
3145              continue;              continue;
3146    
3147              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
# Line 2520  for (;; ptr++) Line 3153  for (;; ptr++)
3153            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3154                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3155    
3156              /* We need to deal with \H, \h, \V, and \v in both phases because
3157              they use extra memory. */
3158    
3159              if (-c == ESC_h)
3160                {
3161                SETBIT(classbits, 0x09); /* VT */
3162                SETBIT(classbits, 0x20); /* SPACE */
3163                SETBIT(classbits, 0xa0); /* NSBP */
3164    #ifdef SUPPORT_UTF8
3165                if (utf8)
3166                  {
3167                  class_utf8 = TRUE;
3168                  *class_utf8data++ = XCL_SINGLE;
3169                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3170                  *class_utf8data++ = XCL_SINGLE;
3171                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3172                  *class_utf8data++ = XCL_RANGE;
3173                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3174                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3175                  *class_utf8data++ = XCL_SINGLE;
3176                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3177                  *class_utf8data++ = XCL_SINGLE;
3178                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3179                  *class_utf8data++ = XCL_SINGLE;
3180                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3181                  }
3182    #endif
3183                continue;
3184                }
3185    
3186              if (-c == ESC_H)
3187                {
3188                for (c = 0; c < 32; c++)
3189                  {
3190                  int x = 0xff;
3191                  switch (c)
3192                    {
3193                    case 0x09/8: x ^= 1 << (0x09%8); break;
3194                    case 0x20/8: x ^= 1 << (0x20%8); break;
3195                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3196                    default: break;
3197                    }
3198                  classbits[c] |= x;
3199                  }
3200    
3201    #ifdef SUPPORT_UTF8
3202                if (utf8)
3203                  {
3204                  class_utf8 = TRUE;
3205                  *class_utf8data++ = XCL_RANGE;
3206                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3207                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3208                  *class_utf8data++ = XCL_RANGE;
3209                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3210                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3211                  *class_utf8data++ = XCL_RANGE;
3212                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3213                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3214                  *class_utf8data++ = XCL_RANGE;
3215                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3216                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3217                  *class_utf8data++ = XCL_RANGE;
3218                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3219                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3220                  *class_utf8data++ = XCL_RANGE;
3221                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3222                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3223                  *class_utf8data++ = XCL_RANGE;
3224                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3225                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3226                  }
3227    #endif
3228                continue;
3229                }
3230    
3231              if (-c == ESC_v)
3232                {
3233                SETBIT(classbits, 0x0a); /* LF */
3234                SETBIT(classbits, 0x0b); /* VT */
3235                SETBIT(classbits, 0x0c); /* FF */
3236                SETBIT(classbits, 0x0d); /* CR */
3237                SETBIT(classbits, 0x85); /* NEL */
3238    #ifdef SUPPORT_UTF8
3239                if (utf8)
3240                  {
3241                  class_utf8 = TRUE;
3242                  *class_utf8data++ = XCL_RANGE;
3243                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3244                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3245                  }
3246    #endif
3247                continue;
3248                }
3249    
3250              if (-c == ESC_V)
3251                {
3252                for (c = 0; c < 32; c++)
3253                  {
3254                  int x = 0xff;
3255                  switch (c)
3256                    {
3257                    case 0x0a/8: x ^= 1 << (0x0a%8);
3258                                 x ^= 1 << (0x0b%8);
3259                                 x ^= 1 << (0x0c%8);
3260                                 x ^= 1 << (0x0d%8);
3261                                 break;
3262                    case 0x85/8: x ^= 1 << (0x85%8); break;
3263                    default: break;
3264                    }
3265                  classbits[c] |= x;
3266                  }
3267    
3268    #ifdef SUPPORT_UTF8
3269                if (utf8)
3270                  {
3271                  class_utf8 = TRUE;
3272                  *class_utf8data++ = XCL_RANGE;
3273                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3274                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3275                  *class_utf8data++ = XCL_RANGE;
3276                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3277                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3278                  }
3279    #endif
3280                continue;
3281                }
3282    
3283            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3284    
3285  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2563  for (;; ptr++) Line 3323  for (;; ptr++)
3323        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3324    
3325        CHECK_RANGE:        CHECK_RANGE:
3326        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3327          {          {
3328          inescq = FALSE;          inescq = FALSE;
3329          ptr += 2;          ptr += 2;
# Line 2571  for (;; ptr++) Line 3331  for (;; ptr++)
3331    
3332        oldptr = ptr;        oldptr = ptr;
3333    
3334        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3335    
3336          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3337    
3338          /* Check for range */
3339    
3340          if (!inescq && ptr[1] == CHAR_MINUS)
3341          {          {
3342          int d;          int d;
3343          ptr += 2;          ptr += 2;
3344          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3345    
3346          /* 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
3347          mode. */          mode. */
3348    
3349          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3350            {            {
3351            ptr += 2;            ptr += 2;
3352            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3353                { ptr += 2; continue; }
3354            inescq = TRUE;            inescq = TRUE;
3355            break;            break;
3356            }            }
3357    
3358          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3359            {            {
3360            ptr = oldptr;            ptr = oldptr;
3361            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2607  for (;; ptr++) Line 3374  for (;; ptr++)
3374          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
3375          in such circumstances. */          in such circumstances. */
3376    
3377          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3378            {            {
3379            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3380            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3381    
3382            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3383            special means the '-' was literal */            special means the '-' was literal */
3384    
3385            if (d < 0)            if (d < 0)
3386              {              {
3387              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3388              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3389              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3390                {                {
3391                ptr = oldptr;                ptr = oldptr;
3392                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2638  for (;; ptr++) Line 3405  for (;; ptr++)
3405    
3406          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3407    
3408            /* Remember \r or \n */
3409    
3410            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3411    
3412          /* 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
3413          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3414          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2660  for (;; ptr++) Line 3431  for (;; ptr++)
3431              unsigned int origd = d;              unsigned int origd = d;
3432              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3433                {                {
3434                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3435                      ocd <= (unsigned int)d)
3436                    continue;                          /* Skip embedded ranges */
3437    
3438                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3439                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3440                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3441                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3442                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3443                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3444                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3445                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3446                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3447                  d = ocd;                  d = ocd;
3448                  continue;                  continue;
# Line 2752  for (;; ptr++) Line 3527  for (;; ptr++)
3527          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3528            {            {
3529            unsigned int othercase;            unsigned int othercase;
3530            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3531              {              {
3532              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3533              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2779  for (;; ptr++) Line 3554  for (;; ptr++)
3554    
3555      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3556    
3557      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3558    
3559      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3560        {        {
# Line 2787  for (;; ptr++) Line 3562  for (;; ptr++)
3562        goto FAILED;        goto FAILED;
3563        }        }
3564    
3565    
3566    /* This code has been disabled because it would mean that \s counts as
3567    an explicit \r or \n reference, and that's not really what is wanted. Now
3568    we set the flag only if there is a literal "\r" or "\n" in the class. */
3569    
3570    #if 0
3571        /* Remember whether \r or \n are in this class */
3572    
3573        if (negate_class)
3574          {
3575          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3576          }
3577        else
3578          {
3579          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3580          }
3581    #endif
3582    
3583    
3584      /* 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
3585      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
3586      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
3587      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3588      single-bytes only. This is an historical hangover. Maybe one day we can  
3589      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3590        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3591        operate on single-bytes only. This is an historical hangover. Maybe one day
3592        we can tidy these opcodes to handle multi-byte characters.
3593    
3594      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
3595      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2802  for (;; ptr++) Line 3599  for (;; ptr++)
3599      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3600    
3601  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3602      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3603            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3604  #else  #else
3605      if (class_charcount == 1)      if (class_charcount == 1)
3606  #endif  #endif
# Line 2848  for (;; ptr++) Line 3643  for (;; ptr++)
3643      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3644    
3645      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3646      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3647      we can omit the bitmap in the actual compiled code. */      such as \S in the class, because in that case all characters > 255 are in
3648        the class, so any that were explicitly given as well can be ignored. If
3649        (when there are explicit characters > 255 that must be listed) there are no
3650        characters < 256, we can omit the bitmap in the actual compiled code. */
3651    
3652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3653      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3654        {        {
3655        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3656        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2878  for (;; ptr++) Line 3676  for (;; ptr++)
3676        }        }
3677  #endif  #endif
3678    
3679      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3680      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3681      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3682      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3683    
3684        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3685      if (negate_class)      if (negate_class)
3686        {        {
       *code++ = OP_NCLASS;  
3687        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3688          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3689        }        }
3690      else      else
3691        {        {
       *code++ = OP_CLASS;  
3692        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3693        }        }
3694      code += 32;      code += 32;
# Line 2902  for (;; ptr++) Line 3699  for (;; ptr++)
3699      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3700      has been tested above. */      has been tested above. */
3701    
3702      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3703      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3704      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3705      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3706      goto REPEAT;      goto REPEAT;
3707    
3708      case '*':      case CHAR_ASTERISK:
3709      repeat_min = 0;      repeat_min = 0;
3710      repeat_max = -1;      repeat_max = -1;
3711      goto REPEAT;      goto REPEAT;
3712    
3713      case '+':      case CHAR_PLUS:
3714      repeat_min = 1;      repeat_min = 1;
3715      repeat_max = -1;      repeat_max = -1;
3716      goto REPEAT;      goto REPEAT;
3717    
3718      case '?':      case CHAR_QUESTION_MARK:
3719      repeat_min = 0;      repeat_min = 0;
3720      repeat_max = 1;      repeat_max = 1;
3721    
# Line 2953  for (;; ptr++) Line 3750  for (;; ptr++)
3750      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
3751      repeat type to the non-default. */      repeat type to the non-default. */
3752    
3753      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3754        {        {
3755        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3756        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3757        ptr++;        ptr++;
3758        }        }
3759      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3760        {        {
3761        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3762        ptr++;        ptr++;
# Line 3074  for (;; ptr++) Line 3871  for (;; ptr++)
3871    
3872        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3873    
3874          /*--------------------------------------------------------------------*/
3875          /* This code is obsolete from release 8.00; the restriction was finally
3876          removed: */
3877    
3878        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3879        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3880    
3881        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3882          /*--------------------------------------------------------------------*/
3883    
3884        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3885    
# Line 3224  for (;; ptr++) Line 4026  for (;; ptr++)
4026          goto END_REPEAT;          goto END_REPEAT;
4027          }          }
4028    
4029          /*--------------------------------------------------------------------*/
4030          /* This code is obsolete from release 8.00; the restriction was finally
4031          removed: */
4032    
4033        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4034        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4035    
4036        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4037          /*--------------------------------------------------------------------*/
4038    
4039        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4040          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3263  for (;; ptr++) Line 4070  for (;; ptr++)
4070          goto FAILED;          goto FAILED;
4071          }          }
4072    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4073        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4074        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4075        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3293  for (;; ptr++) Line 4092  for (;; ptr++)
4092    
4093        if (repeat_min == 0)        if (repeat_min == 0)
4094          {          {
4095          /* 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
4096          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4097    
4098          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4099          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4100          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4101          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4102          the start of the whole regex. Temporarily terminate the pattern while          **   }
4103          doing this. */  
4104            However, that fails when a group is referenced as a subroutine from
4105            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4106            so that it is skipped on execution. As we don't have a list of which
4107            groups are referenced, we cannot do this selectively.
4108    
4109            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4110            and do no more at this point. However, we do need to adjust any
4111            OP_RECURSE calls inside the group that refer to the group itself or any
4112            internal or forward referenced group, because the offset is from the
4113            start of the whole regex. Temporarily terminate the pattern while doing
4114            this. */
4115    
4116          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4117            {            {
4118            *code = OP_END;            *code = OP_END;
4119            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4120            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4121            code++;            code++;
4122              if (repeat_max == 0)
4123                {
4124                *previous++ = OP_SKIPZERO;
4125                goto END_REPEAT;
4126                }
4127            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4128            }            }
4129    
# Line 3359  for (;; ptr++) Line 4168  for (;; ptr++)
4168          if (repeat_min > 1)          if (repeat_min > 1)
4169            {            {
4170            /* 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
4171            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4172              potential integer overflow. */
4173    
4174            if (lengthptr != NULL)            if (lengthptr != NULL)
4175              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4176                int delta = (repeat_min - 1)*length_prevgroup;
4177                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4178                                                                (double)INT_MAX ||
4179                    OFLOW_MAX - *lengthptr < delta)
4180                  {
4181                  *errorcodeptr = ERR20;
4182                  goto FAILED;
4183                  }
4184                *lengthptr += delta;
4185                }
4186    
4187            /* This is compiling for real */            /* This is compiling for real */
4188    
# Line 3400  for (;; ptr++) Line 4220  for (;; ptr++)
4220          /* 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
4221          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4222          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4223          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4224            paranoid checks to avoid integer overflow. */
4225    
4226          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4227            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4228              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4229                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4230              if ((double)repeat_max *
4231                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4232                      > (double)INT_MAX ||
4233                  OFLOW_MAX - *lengthptr < delta)
4234                {
4235                *errorcodeptr = ERR20;
4236                goto FAILED;
4237                }
4238              *lengthptr += delta;
4239              }
4240    
4241          /* This is compiling for real */          /* This is compiling for real */
4242    
# Line 3486  for (;; ptr++) Line 4318  for (;; ptr++)
4318          }          }
4319        }        }
4320    
4321        /* If previous is OP_FAIL, it was generated by an empty class [] in
4322        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4323        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4324        error above. We can just ignore the repeat in JS case. */
4325    
4326        else if (*previous == OP_FAIL) goto END_REPEAT;
4327    
4328      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4329    
4330      else      else
# Line 3510  for (;; ptr++) Line 4349  for (;; ptr++)
4349      if (possessive_quantifier)      if (possessive_quantifier)
4350        {        {
4351        int len;        int len;
4352        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4353            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4354            tempcode += _pcre_OP_lengths[*tempcode] +
4355              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4356    
4357          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4358            {
4359          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4360    #ifdef SUPPORT_UTF8
4361            if (utf8 && tempcode[-1] >= 0xc0)
4362              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4363    #endif
4364            }
4365    
4366        len = code - tempcode;        len = code - tempcode;
4367        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4368          {          {
# Line 3556  for (;; ptr++) Line 4406  for (;; ptr++)
4406      /* ===================================================================*/      /* ===================================================================*/
4407      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4408      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4409      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4410    
4411      case '(':      case CHAR_LEFT_PARENTHESIS:
4412      newoptions = options;      newoptions = options;
4413      skipbytes = 0;      skipbytes = 0;
4414      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4415      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4416        reset_bracount = FALSE;
4417    
4418        /* First deal with various "verbs" that can be introduced by '*'. */
4419    
4420        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4421          {
4422          int i, namelen;
4423          const char *vn = verbnames;
4424          const uschar *name = ++ptr;
4425          previous = NULL;
4426          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4427          if (*ptr == CHAR_COLON)
4428            {
4429            *errorcodeptr = ERR59;   /* Not supported */
4430            goto FAILED;
4431            }
4432          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4433            {
4434            *errorcodeptr = ERR60;
4435            goto FAILED;
4436            }
4437          namelen = ptr - name;
4438          for (i = 0; i < verbcount; i++)
4439            {
4440            if (namelen == verbs[i].len &&
4441                strncmp((char *)name, vn, namelen) == 0)
4442              {
4443              *code = verbs[i].op;
4444              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4445              break;
4446              }
4447            vn += verbs[i].len + 1;
4448            }
4449          if (i < verbcount) continue;
4450          *errorcodeptr = ERR60;
4451          goto FAILED;
4452          }
4453    
4454      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4455        appearance of any of them means that this is not a capturing group. */
4456    
4457        else if (*ptr == CHAR_QUESTION_MARK)
4458        {        {
4459        int i, set, unset, namelen;        int i, set, unset, namelen;
4460        int *optset;        int *optset;
# Line 3575  for (;; ptr++) Line 4463  for (;; ptr++)
4463    
4464        switch (*(++ptr))        switch (*(++ptr))
4465          {          {
4466          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4467          ptr++;          ptr++;
4468          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4469          if (*ptr == 0)          if (*ptr == 0)
4470            {            {
4471            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3587  for (;; ptr++) Line 4475  for (;; ptr++)
4475    
4476    
4477          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4478          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4479            reset_bracount = TRUE;
4480            /* Fall through */
4481    
4482            /* ------------------------------------------------------------ */
4483            case CHAR_COLON:          /* Non-capturing bracket */
4484          bravalue = OP_BRA;          bravalue = OP_BRA;
4485          ptr++;          ptr++;
4486          break;          break;
4487    
4488    
4489          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4490          case '(':          case CHAR_LEFT_PARENTHESIS:
4491          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4492    
4493          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 3614  for (;; ptr++) Line 4507  for (;; ptr++)
4507          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4508          including assertions, are processed. */          including assertions, are processed. */
4509    
4510          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4511                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4512            break;            break;
4513    
4514          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 3622  for (;; ptr++) Line 4516  for (;; ptr++)
4516    
4517          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4518          skipbytes = 3;          skipbytes = 3;
4519            refsign = -1;
4520    
4521          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4522    
4523          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4524            {            {
4525            terminator = -1;            terminator = -1;
4526            ptr += 2;            ptr += 2;
# Line 3635  for (;; ptr++) Line 4530  for (;; ptr++)
4530          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4531          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4532    
4533          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4534            {            {
4535            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4536            ptr++;            ptr++;
4537            }            }
4538          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4539            {            {
4540            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4541            ptr++;            ptr++;
4542            }            }
4543          else terminator = 0;          else
4544              {
4545              terminator = 0;
4546              if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4547              }
4548    
4549          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
4550    
# Line 3664  for (;; ptr++) Line 4563  for (;; ptr++)
4563            {            {
4564            if (recno >= 0)            if (recno >= 0)
4565              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4566                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4567            ptr++;            ptr++;
4568            }            }
4569          namelen = ptr - name;          namelen = ptr - name;
4570    
4571          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4572                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4573            {            {
4574            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4575            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3681  for (;; ptr++) Line 4581  for (;; ptr++)
4581          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4582    
4583          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
4584          reference. */          reference. If the string started with "+" or "-" we require the rest to
4585            be digits, in which case recno will be set. */
4586    
4587            if (refsign > 0)
4588              {
4589              if (recno <= 0)
4590                {
4591                *errorcodeptr = ERR58;
4592                goto FAILED;
4593                }
4594              recno = (refsign == CHAR_MINUS)?
4595                cd->bracount - recno + 1 : recno +cd->bracount;
4596              if (recno <= 0 || recno > cd->final_bracount)
4597                {
4598                *errorcodeptr = ERR15;
4599                goto FAILED;
4600                }
4601              PUT2(code, 2+LINK_SIZE, recno);
4602              break;
4603              }
4604    
4605            /* Otherwise (did not start with "+" or "-"), start by looking for the
4606            name. */
4607    
4608          slot = cd->name_table;          slot = cd->name_table;
4609          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3700  for (;; ptr++) Line 4622  for (;; ptr++)
4622    
4623          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4624    
4625          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4626                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4627            {            {
4628            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 3721  for (;; ptr++) Line 4643  for (;; ptr++)
4643          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4644          specific group number. */          specific group number. */
4645    
4646          else if (*name == 'R')          else if (*name == CHAR_R)
4647            {            {
4648            recno = 0;            recno = 0;
4649            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3731  for (;; ptr++) Line 4653  for (;; ptr++)
4653                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4654                goto FAILED;                goto FAILED;
4655                }                }
4656              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4657              }              }
4658            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4659            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3741  for (;; ptr++) Line 4663  for (;; ptr++)
4663          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4664          false. */          false. */
4665    
4666          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4667            {            {
4668            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4669            skipbytes = 1;            skipbytes = 1;
4670            }            }
4671    
4672          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4673            in the second pass here, so final_bracount is set. */
4674    
4675          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4676            {            {
4677            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4678            }            }
# Line 3765  for (;; ptr++) Line 4688  for (;; ptr++)
4688    
4689    
4690          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4691          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4692          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4693          ptr++;          ptr++;
4694          break;          break;
4695    
4696    
4697          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4698          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4699          ptr++;          ptr++;
4700            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4701              {
4702              *code++ = OP_FAIL;
4703              previous = NULL;
4704              continue;
4705              }
4706            bravalue = OP_ASSERT_NOT;
4707          break;          break;
4708    
4709    
4710          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4711          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4712          switch (ptr[1])          switch (ptr[1])
4713            {            {
4714            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4715            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4716            ptr += 2;            ptr += 2;
4717            break;            break;
4718    
4719            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4720            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4721            ptr += 2;            ptr += 2;
4722            break;            break;
# Line 3802  for (;; ptr++) Line 4731  for (;; ptr++)
4731    
4732    
4733          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4734          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4735          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4736          ptr++;          ptr++;
4737          break;          break;
4738    
4739    
4740          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4741          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4742          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4743          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4744          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4745            {            {
4746            int n = 0;            int n = 0;
4747            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4748              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4749            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4750              {              {
4751              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4752              goto FAILED;              goto FAILED;
# Line 3837  for (;; ptr++) Line 4766  for (;; ptr++)
4766    
4767    
4768          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4769          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4770          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4771                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4772            {            {
4773            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4774            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4775            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4776            }            }
4777          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4778            {            {
4779            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4780            goto FAILED;            goto FAILED;
# Line 3854  for (;; ptr++) Line 4784  for (;; ptr++)
4784    
4785          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4786          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4787          case '\'':          case CHAR_APOSTROPHE:
4788            {            {
4789            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4790                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4791            name = ++ptr;            name = ++ptr;
4792    
4793            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3930  for (;; ptr++) Line 4861  for (;; ptr++)
4861    
4862    
4863          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4864          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4865          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4866          is_recurse = TRUE;          is_recurse = TRUE;
4867          /* Fall through */          /* Fall through */
4868    
4869          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4870          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4871          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4872            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4873            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4874    
4875          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4876          name = ++ptr;          name = ++ptr;
# Line 3949  for (;; ptr++) Line 4882  for (;; ptr++)
4882    
4883          if (lengthptr != NULL)          if (lengthptr != NULL)
4884            {            {
4885              if (namelen == 0)
4886                {
4887                *errorcodeptr = ERR62;
4888                goto FAILED;
4889                }
4890            if (*ptr != terminator)            if (*ptr != terminator)
4891              {              {
4892              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3962  for (;; ptr++) Line 4900  for (;; ptr++)
4900            recno = 0;            recno = 0;
4901            }            }
4902    
4903          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4904            first, and then check that we have reached the end of the name in the
4905            table. That way, if the name that is longer than any in the table,
4906            the comparison will fail without reading beyond the table entry. */
4907    
4908          else          else
4909            {            {
4910            slot = cd->name_table;            slot = cd->name_table;
4911            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4912              {              {
4913              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4914                    slot[2+namelen] == 0)
4915                  break;
4916              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4917              }              }
4918    
# Line 3978  for (;; ptr++) Line 4921  for (;; ptr++)
4921              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4922              }              }
4923            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4924                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
4925                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4926              {              {
4927              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 3994  for (;; ptr++) Line 4937  for (;; ptr++)
4937    
4938    
4939          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4940          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4941          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4942          /* Fall through */          /* Fall through */
4943    
4944    
4945          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4946          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4947          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
4948          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
4949            {            {
4950            const uschar *called;            const uschar *called;
4951            int sign = *ptr;            terminator = CHAR_RIGHT_PARENTHESIS;
4952    
4953              /* Come here from the \g<...> and \g'...' code (Oniguruma
4954              compatibility). However, the syntax has been checked to ensure that
4955              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4956              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4957              ever be taken. */
4958    
4959            if (sign == '+') ptr++;            HANDLE_NUMERICAL_RECURSION:
4960            else if (sign == '-')  
4961              if ((refsign = *ptr) == CHAR_PLUS)
4962                {
4963                ptr++;
4964                if ((digitab[*ptr] & ctype_digit) == 0)
4965                  {
4966                  *errorcodeptr = ERR63;
4967                  goto FAILED;
4968                  }
4969                }
4970              else if (refsign == CHAR_MINUS)
4971              {              {
4972              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
4973                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
4974              ptr++;              ptr++;
4975              }              }
4976    
4977            recno = 0;            recno = 0;
4978            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4979              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
4980    
4981            if (*ptr != ')')            if (*ptr != terminator)
4982              {              {
4983              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4984              goto FAILED;              goto FAILED;
4985              }              }
4986    
4987            if (sign == '-')            if (refsign == CHAR_MINUS)
4988              {              {
4989              if (recno == 0)              if (recno == 0)
4990                {                {
4991                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
4992                goto FAILED;                goto FAILED;
4993                }                }
4994              recno = cd->bracount - recno + 1;              recno = cd->bracount - recno + 1;
4995              if (recno <= 0)              if (recno <= 0)
4996                {                {
4997                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4998                goto FAILED;                goto FAILED;
4999                }                }
5000              }              }
5001            else if (sign == '+')            else if (refsign == CHAR_PLUS)
5002              {              {
5003              if (recno == 0)              if (recno == 0)
5004                {                {
5005                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
5006                goto FAILED;                goto FAILED;
5007                }                }
5008              recno += cd->bracount;              recno += cd->bracount;
5009              }              }
5010    
5011            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
5012    
# Line 4072  for (;; ptr++) Line 5031  for (;; ptr++)
5031    
5032              if (called == NULL)              if (called == NULL)
5033                {                {
5034                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5035                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5036                  {                  {
5037                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5038                  goto FAILED;                  goto FAILED;
# Line 4121  for (;; ptr++) Line 5080  for (;; ptr++)
5080    
5081          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5082          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
5083          OTHER_CHAR_AFTER_QUERY:          OTHER_CHAR_AFTER_QUERY:
5084          set = unset = 0;          set = unset = 0;
5085          optset = &set;          optset = &set;
5086    
5087          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5088            {            {
5089            switch (*ptr++)            switch (*ptr++)
5090              {              {
5091              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5092    
5093              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5094              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5095              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5096              break;              break;
5097    
5098              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5099              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5100              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5101              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5102              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5103              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5104    
5105              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5106                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4172  for (;; ptr++) Line 5131  for (;; ptr++)
5131          both phases.          both phases.
5132    
5133          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5134          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5135          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5136    
5137          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5138            {            {
5139            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5140                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5141              {              {
5142              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5143              }              }
5144           else           else
5145              {              {
# Line 4192  for (;; ptr++) Line 5148  for (;; ptr++)
5148                *code++ = OP_OPT;                *code++ = OP_OPT;
5149                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5150                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5151              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5152              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5153              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5154              }              }
5155    
5156              /* Change options at this level, and pass them back for use
5157              in subsequent branches. When not at the start of the pattern, this
5158              information is also necessary so that a resetting item can be
5159              compiled at the end of a group (if we are in a group). */
5160    
5161              *optionsptr = options = newoptions;
5162            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5163            continue;              /* It is complete */            continue;              /* It is complete */
5164            }            }
# Line 4256  for (;; ptr++) Line 5212  for (;; ptr++)
5212           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5213           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5214            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5215             reset_bracount,               /* True if (?| group */
5216           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5217           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5218           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4272  for (;; ptr++) Line 5229  for (;; ptr++)
5229      is on the bracket. */      is on the bracket. */
5230    
5231      /* If this is a conditional bracket, check that there are no more than      /* If this is a conditional bracket, check that there are no more than
5232      two branches in the group, or just one if it's a DEFINE group. */      two branches in the group, or just one if it's a DEFINE group. We do this
5233        in the real compile phase, not in the pre-pass, where the whole group may
5234        not be available. */
5235    
5236      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5237        {        {
5238        uschar *tc = code;        uschar *tc = code;
5239        int condcount = 0;        int condcount = 0;
# Line 4315  for (;; ptr++) Line 5274  for (;; ptr++)
5274    
5275      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5276    
5277      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5278        {        {
5279        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5280        goto FAILED;        goto FAILED;
5281        }        }
5282    
5283      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
5284      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
5285      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
5286      a quantifier. */      duplicated by a quantifier.*/
5287    
5288      if (lengthptr != NULL)      if (lengthptr != NULL)
5289        {        {
5290          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5291            {
5292            *errorcodeptr = ERR20;
5293            goto FAILED;
5294            }
5295        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5296        code++;        *code++ = OP_BRA;
5297        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5298        *code++ = OP_KET;        *code++ = OP_KET;
5299        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5300          break;    /* No need to waste time with special character handling */
5301        }        }
5302    
5303      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5304    
5305      else code = tempcode;      code = tempcode;
5306    
5307      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5308      relevant. */      relevant. */
# Line 4407  for (;; ptr++) Line 5372  for (;; ptr++)
5372      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5373      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5374    
5375      case '\\':      case CHAR_BACKSLASH:
5376      tempptr = ptr;      tempptr = ptr;
5377      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5378      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4416  for (;; ptr++) Line 5381  for (;; ptr++)
5381        {        {
5382        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5383          {          {
5384          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5385            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5386                else inescq = TRUE;
5387          continue;          continue;
5388          }          }
5389    
# Line 4434  for (;; ptr++) Line 5400  for (;; ptr++)
5400        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5401        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5402    
5403        /* \k<name> or \k'name' is a back reference by name (Perl syntax) */        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5404          is a subroutine call by number (Oniguruma syntax). In fact, the value
5405          -ESC_g is returned only for these cases. So we don't need to check for <
5406          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5407          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5408          that is a synonym for a named back reference). */
5409    
5410          if (-c == ESC_g)
5411            {
5412            const uschar *p;
5413            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5414            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5415              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5416    
5417            /* These two statements stop the compiler for warning about possibly
5418            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5419            fact, because we actually check for a number below, the paths that
5420            would actually be in error are never taken. */
5421    
5422            skipbytes = 0;
5423            reset_bracount = FALSE;
5424    
5425            /* Test for a name */
5426    
5427            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5428              {
5429              BOOL isnumber = TRUE;
5430              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5431                {
5432                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5433                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5434                }
5435              if (*p != terminator)
5436                {
5437                *errorcodeptr = ERR57;
5438                break;
5439                }
5440              if (isnumber)
5441                {
5442                ptr++;
5443                goto HANDLE_NUMERICAL_RECURSION;
5444                }
5445              is_recurse = TRUE;
5446              goto NAMED_REF_OR_RECURSE;
5447              }
5448    
5449            /* Test a signed number in angle brackets or quotes. */
5450    
5451            p = ptr + 2;
5452            while ((digitab[*p] & ctype_digit) != 0) p++;
5453            if (*p != terminator)
5454              {
5455              *errorcodeptr = ERR57;
5456              break;
5457              }
5458            ptr++;
5459            goto HANDLE_NUMERICAL_RECURSION;
5460            }
5461    
5462          /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5463          We also support \k{name} (.NET syntax) */
5464    
5465        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5466              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5467          {          {
5468          is_recurse = FALSE;          is_recurse = FALSE;
5469          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5470              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5471              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5472          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5473          }          }
5474    
# Line 4540  for (;; ptr++) Line 5569  for (;; ptr++)
5569      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5570      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5571    
5572        /* Remember if \r or \n were seen */
5573    
5574        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5575          cd->external_flags |= PCRE_HASCRORLF;
5576    
5577      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5578      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5579      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 4605  This function is used during the pre-com Line 5639  This function is used during the pre-com
5639  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
5640  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5641    
5642  Argument:  Arguments:
5643    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5644    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5645    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5646    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5647    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5648    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5649      reset_bracount TRUE to reset the count for each branch
5650    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5651    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5652    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
# Line 4625  Returns:         TRUE on success Line 5660  Returns:         TRUE on success
5660    
5661  static BOOL  static BOOL
5662  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5663    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5664    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5665      int *lengthptr)
5666  {  {
5667  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5668  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 4636  uschar *reverse_count = NULL; Line 5672  uschar *reverse_count = NULL;
5672  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5673  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5674  int length;  int length;
5675    int orig_bracount;
5676    int max_bracount;
5677  branch_chain bc;  branch_chain bc;
5678    
5679  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4664  code += 1 + LINK_SIZE + skipbytes; Line 5702  code += 1 + LINK_SIZE + skipbytes;
5702    
5703  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5704    
5705    orig_bracount = max_bracount = cd->bracount;
5706  for (;;)  for (;;)
5707    {    {
5708      /* For a (?| group, reset the capturing bracket count so that each branch
5709      uses the same numbers. */
5710    
5711      if (reset_bracount) cd->bracount = orig_bracount;
5712    
5713    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5714    
5715    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4695  for (;;) Line 5739  for (;;)
5739      return FALSE;      return FALSE;
5740      }      }
5741    
5742      /* Keep the highest bracket count in case (?| was used and some branch
5743      has fewer than the rest. */
5744    
5745      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5746    
5747    /* In the real compile phase, there is some post-processing to be done. */    /* In the real compile phase, there is some post-processing to be done. */
5748    
5749    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4758  for (;;) Line 5807  for (;;)
5807        }        }
5808      }      }
5809    
5810    /* Reached end of expression, either ')' or end of pattern. Go back through    /* Reached end of expression, either ')' or end of pattern. In the real
5811    the alternative branches and reverse the chain of offsets, with the field in    compile phase, go back through the alternative branches and reverse the chain
5812    the BRA item now becoming an offset to the first alternative. If there are    of offsets, with the field in the BRA item now becoming an offset to the
5813    no alternatives, it points to the end of the group. The length in the    first alternative. If there are no alternatives, it points to the end of the
5814    terminating ket is always the length of the whole bracketed item. If any of    group. The length in the terminating ket is always the length of the whole
5815    the ims options were changed inside the group, compile a resetting op-code    bracketed item. If any of the ims options were changed inside the group,
5816    following, except at the very end of the pattern. Return leaving the pointer    compile a resetting op-code following, except at the very end of the pattern.
5817    at the terminating char. */    Return leaving the pointer at the terminating char. */
5818    
5819    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5820      {      {
5821      int branch_length = code - last_branch;      if (lengthptr == NULL)
5822      do        {
5823        {        int branch_length = code - last_branch;
5824        int prev_length = GET(last_branch, 1);        do
5825        PUT(last_branch, 1, branch_length);          {
5826        branch_length = prev_length;          int prev_length = GET(last_branch, 1);
5827        last_branch -= branch_length;          PUT(last_branch, 1, branch_length);
5828            branch_length = prev_length;
5829            last_branch -= branch_length;
5830            }
5831          while (branch_length > 0);
5832        }        }
     while (branch_length > 0);  
5833    
5834      /* Fill in the ket */      /* Fill in the ket */
5835    
# Line 4787  for (;;) Line 5839  for (;;)
5839    
5840      /* Resetting option if needed */      /* Resetting option if needed */
5841    
5842      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5843        {        {
5844        *code++ = OP_OPT;        *code++ = OP_OPT;
5845        *code++ = oldims;        *code++ = oldims;
5846        length += 2;        length += 2;
5847        }        }
5848    
5849        /* Retain the highest bracket number, in case resetting was used. */
5850    
5851        cd->bracount = max_bracount;
5852    
5853      /* Set values to pass back */      /* Set values to pass back */
5854    
5855      *codeptr = code;      *codeptr = code;
5856      *ptrptr = ptr;      *ptrptr = ptr;
5857      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5858      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5859      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5860          {
5861          if (OFLOW_MAX - *lengthptr < length)
5862            {
5863            *errorcodeptr = ERR20;
5864            return FALSE;
5865            }
5866          *lengthptr += length;
5867          }
5868      return TRUE;      return TRUE;
5869      }      }
5870    
5871    /* Another branch follows; insert an "or" node. Its length field points back    /* Another branch follows. In the pre-compile phase, we can move the code
5872      pointer back to where it was for the start of the first branch. (That is,
5873      pretend that each branch is the only one.)
5874    
5875      In the real compile phase, insert an ALT node. Its length field points back
5876    to the previous branch while the bracket remains open. At the end the chain    to the previous branch while the bracket remains open. At the end the chain
5877    is reversed. It's done like this so that the start of the bracket has a    is reversed. It's done like this so that the start of the bracket has a
5878    zero offset until it is closed, making it possible to detect recursion. */    zero offset until it is closed, making it possible to detect recursion. */
5879    
5880    *code = OP_ALT;    if (lengthptr != NULL)
5881    PUT(code, 1, code - last_branch);      {
5882    bc.current = last_branch = code;      code = *codeptr + 1 + LINK_SIZE + skipbytes;
5883    code += 1 + LINK_SIZE;      length += 1 + LINK_SIZE;
5884        }
5885      else
5886        {
5887        *code = OP_ALT;
5888        PUT(code, 1, code - last_branch);
5889        bc.current = last_branch = code;
5890        code += 1 + LINK_SIZE;
5891        }
5892    
5893    ptr++;    ptr++;
   length += 1 + LINK_SIZE;  
5894    }    }
5895  /* Control never reaches here */  /* Control never reaches here */
5896  }  }
# Line 4892  do { Line 5968  do {
5968       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5969       }       }
5970    
5971     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
5972     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5973    
5974     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5975               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5976       {       {
5977       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5978           return FALSE;
5979       }       }
5980    
5981     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 4945  do { Line 6021  do {
6021       NULL, 0, FALSE);       NULL, 0, FALSE);
6022     register int op = *scode;     register int op = *scode;
6023    
6024       /* If we are at the start of a conditional assertion group, *both* the
6025       conditional assertion *and* what follows the condition must satisfy the test
6026       for start of line. Other kinds of condition fail. Note that there may be an
6027       auto-callout at the start of a condition. */
6028    
6029       if (op == OP_COND)
6030         {
6031         scode += 1 + LINK_SIZE;
6032         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6033         switch (*scode)
6034           {
6035           case OP_CREF:
6036           case OP_RREF:
6037           case OP_DEF:
6038           return FALSE;
6039    
6040           default:     /* Assertion */
6041           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6042           do scode += GET(scode, 1); while (*scode == OP_ALT);
6043           scode += 1 + LINK_SIZE;
6044           break;
6045           }
6046         scode = first_significant_code(scode, NULL, 0, FALSE);
6047         op = *scode;
6048         }
6049    
6050     /* Non-capturing brackets */     /* Non-capturing brackets */
6051    
6052     if (op == OP_BRA)     if (op == OP_BRA)
# Line 4963  do { Line 6065  do {
6065    
6066     /* Other brackets */     /* Other brackets */
6067    
6068     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6069       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6070         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6071         }
6072    
6073     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
6074     may be referenced. */     may be referenced. */
# Line 5081  Returns:        pointer to compiled data Line 6185  Returns:        pointer to compiled data
6185                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6186  */  */
6187    
6188  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6189  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6190    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6191  {  {
# Line 5089  return pcre_compile2(pattern, options, N Line 6193  return pcre_compile2(pattern, options, N
6193  }  }
6194    
6195    
6196  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6197  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6198    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6199  {  {
# Line 5097  real_pcre *re; Line 6201  real_pcre *re;
6201  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
6202  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6203  int errorcode = 0;  int errorcode = 0;
6204    int skipatstart = 0;
6205  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
6206  BOOL utf8;  BOOL utf8;
6207  #endif  #endif
# Line 5115  to fill in forward references to subpatt Line 6220  to fill in forward references to subpatt
6220    
6221  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6222    
   
6223  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6224    
6225  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5143  if (erroroffset == NULL) Line 6247  if (erroroffset == NULL)
6247    
6248  *erroroffset = 0;  *erroroffset = 0;
6249    
6250    /* Set up pointers to the individual character tables */
6251    
6252    if (tables == NULL) tables = _pcre_default_tables;
6253    cd->lcc = tables + lcc_offset;
6254    cd->fcc = tables + fcc_offset;
6255    cd->cbits = tables + cbits_offset;
6256    cd->ctypes = tables + ctypes_offset;
6257    
6258    /* Check that all undefined public option bits are zero */
6259    
6260    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6261      {
6262      errorcode = ERR17;
6263      goto PCRE_EARLY_ERROR_RETURN;
6264      }
6265    
6266    /* Check for global one-time settings at the start of the pattern, and remember
6267    the offset for later. */
6268    
6269    while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6270           ptr[skipatstart+1] == CHAR_ASTERISK)
6271      {
6272      int newnl = 0;
6273      int newbsr = 0;
6274    
6275      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6276        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6277    
6278      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6279        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6280      else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6281        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6282      else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6283        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6284      else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6285        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6286      else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6287        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6288    
6289      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6290        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6291      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6292        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6293    
6294      if (newnl != 0)
6295        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6296      else if (newbsr != 0)
6297        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6298      else break;
6299      }
6300    
6301  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6302    
6303  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 5161  if ((options & PCRE_UTF8) != 0) Line 6316  if ((options & PCRE_UTF8) != 0)
6316    }    }
6317  #endif  #endif
6318    
6319  if ((options & ~PUBLIC_OPTIONS) != 0)  /* Check validity of \R options. */
6320    
6321    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6322    {    {
6323    errorcode = ERR17;    case 0:
6324    goto PCRE_EARLY_ERROR_RETURN;    case PCRE_BSR_ANYCRLF:
6325      case PCRE_BSR_UNICODE:
6326      break;
6327      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6328    }    }
6329    
 /* Set up pointers to the individual character tables */  
   
 if (tables == NULL) tables = _pcre_default_tables;  
 cd->lcc = tables + lcc_offset;  
 cd->fcc = tables + fcc_offset;  
 cd->cbits = tables + cbits_offset;  
 cd->ctypes = tables + ctypes_offset;  
   
6330  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6331  current code allows for fixed one- or two-byte sequences, plus "any" and  current code allows for fixed one- or two-byte sequences, plus "any" and
6332  "anycrlf". */  "anycrlf". */
6333    
6334  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6335    {    {
6336    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6337    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6338    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6339    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6340         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6341    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6342    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6343    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5234  to compile parts of the pattern into; th Line 6386  to compile parts of the pattern into; th
6386  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6387  is a test for its doing so. */  is a test for its doing so. */
6388    
6389  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6390  cd->names_found = 0;  cd->names_found = 0;
6391  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6392  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5244  cd->hwm = cworkspace; Line 6396  cd->hwm = cworkspace;
6396  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6397  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6398  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6399  cd->external_options = options;  cd->external_options = options;
6400    cd->external_flags = 0;
6401    
6402  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6403  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 5253  been put into the cd block so that they Line 6405  been put into the cd block so that they
6405  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6406  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6407    
6408    ptr += skipatstart;
6409  code = cworkspace;  code = cworkspace;
6410  *code = OP_BRA;  *code = OP_BRA;
6411  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
6412    &code, &ptr, &errorcode, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,
6413      &length);
6414  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
6415    
6416  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 5282  if (re == NULL) Line 6436  if (re == NULL)
6436    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6437    }    }
6438    
6439  /* Put in the magic number, and save the sizes, initial options, and character  /* Put in the magic number, and save the sizes, initial options, internal
6440  table pointer. NULL is used for the default character tables. The nullpad field  flags, and character table pointer. NULL is used for the default character
6441  is at the end; it's there to help in the case when a regex compiled on a system  tables. The nullpad field is at the end; it's there to help in the case when a
6442  with 4-byte pointers is run on another with 8-byte pointers. */  regex compiled on a system with 4-byte pointers is run on another with 8-byte
6443    pointers. */
6444    
6445  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6446  re->size = size;  re->size = size;
6447  re->options = cd->external_options;  re->options = cd->external_options;
6448    re->flags = cd->external_flags;
6449  re->dummy1 = 0;  re->dummy1 = 0;
6450  re->first_byte = 0;  re->first_byte = 0;
6451  re->req_byte = 0;  re->req_byte = 0;
# Line 5307  field. Reset the bracket count and the n Line 6463  field. Reset the bracket count and the n
6463  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6464  */  */
6465    
6466    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6467  cd->bracount = 0;  cd->bracount = 0;
6468  cd->names_found = 0;  cd->names_found = 0;
6469  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
#