/[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 87 by nigel, Sat Feb 24 21:41:21 2007 UTC revision 226 by ph10, Tue Aug 21 11:46:08 2007 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) 1997-2006 University of Cambridge             Copyright (c) 1997-2007 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 */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
# Line 53  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  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  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  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  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,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  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,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143    /* Table of special "verbs" like (*PRUNE) */
144    
145    typedef struct verbitem {
146      const char *name;
147      int   len;
148      int   op;
149    } verbitem;
150    
151    static verbitem verbs[] = {
152      { "ACCEPT", 6, OP_ACCEPT },
153      { "COMMIT", 6, OP_COMMIT },
154      { "F",      1, OP_FAIL },
155      { "FAIL",   4, OP_FAIL },
156      { "PRUNE",  5, OP_PRUNE },
157      { "SKIP",   4, OP_SKIP  },
158      { "THEN",   4, OP_THEN  }
159    };
160    
161    static int verbcount = sizeof(verbs)/sizeof(verbitem);
162    
163    
164  /* Tables of names of POSIX character classes and their lengths. The list is  /* Tables of names of POSIX character classes and their lengths. The list is
165  terminated by a zero length entry. The first three must be alpha, lower, upper,  terminated by a zero length entry. The first three must be alpha, lower, upper,
166  as this is assumed for handling case independence. */  as this is assumed for handling case independence. */
# Line 155  static const int posix_class_maps[] = { Line 201  static const int posix_class_maps[] = {
201  };  };
202    
203    
204    #define STRING(a)  # a
205    #define XSTRING(s) STRING(s)
206    
207  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
208  are passed to the outside world. */  are passed to the outside world. Do not ever re-use any error number, because
209    they are documented. Always add a new error instead. Messages marked DEAD below
210    are no longer used. */
211    
212  static const char *error_texts[] = {  static const char *error_texts[] = {
213    "no error",    "no error",
# Line 171  static const char *error_texts[] = { Line 222  static const char *error_texts[] = {
222    "range out of order in character class",    "range out of order in character class",
223    "nothing to repeat",    "nothing to repeat",
224    /* 10 */    /* 10 */
225    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string",  /** DEAD **/
226    "internal error: unexpected repeat",    "internal error: unexpected repeat",
227    "unrecognized character after (?",    "unrecognized character after (?",
228    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class",
# Line 181  static const char *error_texts[] = { Line 232  static const char *error_texts[] = {
232    "erroffset passed as NULL",    "erroffset passed as NULL",
233    "unknown option bit(s) set",    "unknown option bit(s) set",
234    "missing ) after comment",    "missing ) after comment",
235    "parentheses nested too deeply",    "parentheses nested too deeply",  /** DEAD **/
236    /* 20 */    /* 20 */
237    "regular expression too large",    "regular expression is too large",
238    "failed to get memory",    "failed to get memory",
239    "unmatched parentheses",    "unmatched parentheses",
240    "internal error: code overflow",    "internal error: code overflow",
241    "unrecognized character after (?<",    "unrecognized character after (?<",
242    /* 25 */    /* 25 */
243    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length",
244    "malformed number after (?(",    "malformed number or name after (?(",
245    "conditional group contains more than two branches",    "conditional group contains more than two branches",
246    "assertion expected after (?(",    "assertion expected after (?(",
247    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )",
248    /* 30 */    /* 30 */
249    "unknown POSIX class name",    "unknown POSIX class name",
250    "POSIX collating elements are not supported",    "POSIX collating elements are not supported",
251    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support",
252    "spare error",    "spare error",  /** DEAD **/
253    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large",
254    /* 35 */    /* 35 */
255    "invalid condition (?(0)",    "invalid condition (?(0)",
# Line 209  static const char *error_texts[] = { Line 260  static const char *error_texts[] = {
260    /* 40 */    /* 40 */
261    "recursive call could loop indefinitely",    "recursive call could loop indefinitely",
262    "unrecognized character after (?P",    "unrecognized character after (?P",
263    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)",
264    "two named groups have the same name",    "two named subpatterns have the same name",
265    "invalid UTF-8 string",    "invalid UTF-8 string",
266    /* 45 */    /* 45 */
267    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled",
268    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence",
269    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p",
270      "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",
271      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",
272      /* 50 */
273      "repeated subpattern is too long",    /** DEAD **/
274      "octal value is greater than \\377 (not in UTF-8 mode)",
275      "internal error: overran compiling workspace",
276      "internal error: previously-checked referenced subpattern not found",
277      "DEFINE group contains more than one branch",
278      /* 55 */
279      "repeating a DEFINE group is not allowed",
280      "inconsistent NEWLINE options",
281      "\\g is not followed by a braced name or an optionally braced non-zero number",
282      "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",
283      "(*VERB) with an argument is not supported",
284      /* 60 */
285      "(*VERB) not recognized",
286      "number is too big"
287  };  };
288    
289    
# Line 235  For convenience, we use the same bit def Line 303  For convenience, we use the same bit def
303    
304  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
305    
306  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
307  static const unsigned char digitab[] =  static const unsigned char digitab[] =
308    {    {
309    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 271  static const unsigned char digitab[] = Line 339  static const unsigned char digitab[] =
339    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
340    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
341    
342  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
343  static const unsigned char digitab[] =  static const unsigned char digitab[] =
344    {    {
345    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 285  static const unsigned char digitab[] = Line 353  static const unsigned char digitab[] =
353    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
354    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
355    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
356    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
357    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
358    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
359    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 319  static const unsigned char ebcdic_charta Line 387  static const unsigned char ebcdic_charta
387    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
388    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
389    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
390    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
391    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
392    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
393    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 346  static const unsigned char ebcdic_charta Line 414  static const unsigned char ebcdic_charta
414  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
415    
416  static BOOL  static BOOL
417    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
418      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
419    
420    
421    
# Line 357  static BOOL Line 425  static BOOL
425    
426  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
427  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
428  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
429  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
430  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
431    ptr is pointing at the \. On exit, it is on the final character of the escape
432    sequence.
433    
434  Arguments:  Arguments:
435    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 370  Arguments: Line 440  Arguments:
440    
441  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
442                   negative => a special escape sequence                   negative => a special escape sequence
443                   on error, errorptr is set                   on error, errorcodeptr is set
444  */  */
445    
446  static int  static int
# Line 392  if (c == 0) *errorcodeptr = ERR1; Line 462  if (c == 0) *errorcodeptr = ERR1;
462  a table. A non-zero result is something that can be returned immediately.  a table. A non-zero result is something that can be returned immediately.
463  Otherwise further processing may be required. */  Otherwise further processing may be required. */
464    
465  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
466  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
467  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
468    
469  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
470  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
471  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
472  #endif  #endif
# Line 406  else if ((i = escapes[c - 0x48]) != 0) Line 476  else if ((i = escapes[c - 0x48]) != 0)
476  else  else
477    {    {
478    const uschar *oldptr;    const uschar *oldptr;
479      BOOL braced, negated;
480    
481    switch (c)    switch (c)
482      {      {
483      /* 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
# Line 419  else Line 491  else
491      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
492      break;      break;
493    
494        /* \g must be followed by a number, either plain or braced. If positive, it
495        is an absolute backreference. If negative, it is a relative backreference.
496        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
497        reference to a named group. This is part of Perl's movement towards a
498        unified syntax for back references. As this is synonymous with \k{name}, we
499        fudge it up by pretending it really was \k. */
500    
501        case 'g':
502        if (ptr[1] == '{')
503          {
504          const uschar *p;
505          for (p = ptr+2; *p != 0 && *p != '}'; p++)
506            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
507          if (*p != 0 && *p != '}')
508            {
509            c = -ESC_k;
510            break;
511            }
512          braced = TRUE;
513          ptr++;
514          }
515        else braced = FALSE;
516    
517        if (ptr[1] == '-')
518          {
519          negated = TRUE;
520          ptr++;
521          }
522        else negated = FALSE;
523    
524        c = 0;
525        while ((digitab[ptr[1]] & ctype_digit) != 0)
526          c = c * 10 + *(++ptr) - '0';
527    
528        if (c < 0)
529          {
530          *errorcodeptr = ERR61;
531          break;
532          }
533    
534        if (c == 0 || (braced && *(++ptr) != '}'))
535          {
536          *errorcodeptr = ERR57;
537          break;
538          }
539    
540        if (negated)
541          {
542          if (c > bracount)
543            {
544            *errorcodeptr = ERR15;
545            break;
546            }
547          c = bracount - (c - 1);
548          }
549    
550        c = -(ESC_REF + c);
551        break;
552    
553      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
554      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
555      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 440  else Line 571  else
571        c -= '0';        c -= '0';
572        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
573          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
574          if (c < 0)
575            {
576            *errorcodeptr = ERR61;
577            break;
578            }
579        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
580          {          {
581          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 460  else Line 596  else
596        }        }
597    
598      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
599      larger first octal digit. */      larger first octal digit. The original code used just to take the least
600        significant 8 bits of octal numbers (I think this is what early Perls used
601        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
602        than 3 octal digits. */
603    
604      case '0':      case '0':
605      c -= '0';      c -= '0';
606      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
607          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
608      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
609      break;      break;
610    
611      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
# Line 486  else Line 625  else
625          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == '0') continue;     /* Leading zeroes */
626          count++;          count++;
627    
628  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
629          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
630          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
631  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
632          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
633          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
634  #endif  #endif
# Line 513  else Line 652  else
652        {        {
653        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
654        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
655  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
656        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
657        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
658  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
659        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
660        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
661  #endif  #endif
662        }        }
663      break;      break;
664    
665      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
666        This coding is ASCII-specific, but then the whole concept of \cx is
667        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
668    
669      case 'c':      case 'c':
670      c = *(++ptr);      c = *(++ptr);
671      if (c == 0)      if (c == 0)
672        {        {
673        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
674        return 0;        break;
675        }        }
676    
677      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII coding */
     is ASCII-specific, but then the whole concept of \cx is ASCII-specific.  
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
678      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
679      c ^= 0x40;      c ^= 0x40;
680  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
681      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
682      c ^= 0xC0;      c ^= 0xC0;
683  #endif  #endif
# Line 610  if (c == '{') Line 747  if (c == '{')
747      *negptr = TRUE;      *negptr = TRUE;
748      ptr++;      ptr++;
749      }      }
750    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
751      {      {
752      c = *(++ptr);      c = *(++ptr);
753      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 763  return p; Line 900  return p;
900    
901    
902  /*************************************************  /*************************************************
903    *       Find forward referenced subpattern       *
904    *************************************************/
905    
906    /* This function scans along a pattern's text looking for capturing
907    subpatterns, and counting them. If it finds a named pattern that matches the
908    name it is given, it returns its number. Alternatively, if the name is NULL, it
909    returns when it reaches a given numbered subpattern. This is used for forward
910    references to subpatterns. We know that if (?P< is encountered, the name will
911    be terminated by '>' because that is checked in the first pass.
912    
913    Arguments:
914      ptr          current position in the pattern
915      count        current count of capturing parens so far encountered
916      name         name to seek, or NULL if seeking a numbered subpattern
917      lorn         name length, or subpattern number if name is NULL
918      xmode        TRUE if we are in /x mode
919    
920    Returns:       the number of the named subpattern, or -1 if not found
921    */
922    
923    static int
924    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
925      BOOL xmode)
926    {
927    const uschar *thisname;
928    
929    for (; *ptr != 0; ptr++)
930      {
931      int term;
932    
933      /* Skip over backslashed characters and also entire \Q...\E */
934    
935      if (*ptr == '\\')
936        {
937        if (*(++ptr) == 0) return -1;
938        if (*ptr == 'Q') for (;;)
939          {
940          while (*(++ptr) != 0 && *ptr != '\\');
941          if (*ptr == 0) return -1;
942          if (*(++ptr) == 'E') break;
943          }
944        continue;
945        }
946    
947      /* Skip over character classes */
948    
949      if (*ptr == '[')
950        {
951        while (*(++ptr) != ']')
952          {
953          if (*ptr == 0) return -1;
954          if (*ptr == '\\')
955            {
956            if (*(++ptr) == 0) return -1;
957            if (*ptr == 'Q') for (;;)
958              {
959              while (*(++ptr) != 0 && *ptr != '\\');
960              if (*ptr == 0) return -1;
961              if (*(++ptr) == 'E') break;
962              }
963            continue;
964            }
965          }
966        continue;
967        }
968    
969      /* Skip comments in /x mode */
970    
971      if (xmode && *ptr == '#')
972        {
973        while (*(++ptr) != 0 && *ptr != '\n');
974        if (*ptr == 0) return -1;
975        continue;
976        }
977    
978      /* An opening parens must now be a real metacharacter */
979    
980      if (*ptr != '(') continue;
981      if (ptr[1] != '?' && ptr[1] != '*')
982        {
983        count++;
984        if (name == NULL && count == lorn) return count;
985        continue;
986        }
987    
988      ptr += 2;
989      if (*ptr == 'P') ptr++;                      /* Allow optional P */
990    
991      /* We have to disambiguate (?<! and (?<= from (?<name> */
992    
993      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
994           *ptr != '\'')
995        continue;
996    
997      count++;
998    
999      if (name == NULL && count == lorn) return count;
1000      term = *ptr++;
1001      if (term == '<') term = '>';
1002      thisname = ptr;
1003      while (*ptr != term) ptr++;
1004      if (name != NULL && lorn == ptr - thisname &&
1005          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1006        return count;
1007      }
1008    
1009    return -1;
1010    }
1011    
1012    
1013    
1014    /*************************************************
1015  *      Find first significant op code            *  *      Find first significant op code            *
1016  *************************************************/  *************************************************/
1017    
# Line 811  for (;;) Line 1060  for (;;)
1060    
1061      case OP_CALLOUT:      case OP_CALLOUT:
1062      case OP_CREF:      case OP_CREF:
1063      case OP_BRANUMBER:      case OP_RREF:
1064        case OP_DEF:
1065      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1066      break;      break;
1067    
# Line 856  for (;;) Line 1106  for (;;)
1106    {    {
1107    int d;    int d;
1108    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1109    switch (op)    switch (op)
1110      {      {
1111        case OP_CBRA:
1112      case OP_BRA:      case OP_BRA:
1113      case OP_ONCE:      case OP_ONCE:
1114      case OP_COND:      case OP_COND:
1115      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1116      if (d < 0) return d;      if (d < 0) return d;
1117      branchlength += d;      branchlength += d;
1118      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 898  for (;;) Line 1147  for (;;)
1147      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1148    
1149      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1150      case OP_CREF:      case OP_CREF:
1151        case OP_RREF:
1152        case OP_DEF:
1153      case OP_OPT:      case OP_OPT:
1154      case OP_CALLOUT:      case OP_CALLOUT:
1155      case OP_SOD:      case OP_SOD:
# Line 917  for (;;) Line 1167  for (;;)
1167    
1168      case OP_CHAR:      case OP_CHAR:
1169      case OP_CHARNC:      case OP_CHARNC:
1170        case OP_NOT:
1171      branchlength++;      branchlength++;
1172      cc += 2;      cc += 2;
1173  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 943  for (;;) Line 1194  for (;;)
1194    
1195      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1196      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1197        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1198      cc += 4;      cc += 4;
1199      break;      break;
1200    
# Line 1031  Returns:      pointer to the opcode for Line 1283  Returns:      pointer to the opcode for
1283  static const uschar *  static const uschar *
1284  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1285  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1286  for (;;)  for (;;)
1287    {    {
1288    register int c = *code;    register int c = *code;
1289    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1290    else if (c > OP_BRA)  
1291      /* XCLASS is used for classes that cannot be represented just by a bit
1292      map. This includes negated single high-valued characters. The length in
1293      the table is zero; the actual length is stored in the compiled code. */
1294    
1295      if (c == OP_XCLASS) code += GET(code, 1);
1296    
1297      /* Handle capturing bracket */
1298    
1299      else if (c == OP_CBRA)
1300      {      {
1301      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1302      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1303      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1304      }      }
1305    
1306      /* Otherwise, we can get the item's length from the table, except that for
1307      repeated character types, we have to test for \p and \P, which have an extra
1308      two bytes of parameters. */
1309    
1310    else    else
1311      {      {
1312      code += _pcre_OP_lengths[c];      switch(c)
1313          {
1314          case OP_TYPESTAR:
1315          case OP_TYPEMINSTAR:
1316          case OP_TYPEPLUS:
1317          case OP_TYPEMINPLUS:
1318          case OP_TYPEQUERY:
1319          case OP_TYPEMINQUERY:
1320          case OP_TYPEPOSSTAR:
1321          case OP_TYPEPOSPLUS:
1322          case OP_TYPEPOSQUERY:
1323          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1324          break;
1325    
1326  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1327          case OP_TYPEMINUPTO:
1328          case OP_TYPEEXACT:
1329          case OP_TYPEPOSUPTO:
1330          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1331          break;
1332          }
1333    
1334      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1335      by a multi-byte character. The length in the table is a minimum, so we have  
1336      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1337      can use relatively efficient code. */  
1338      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1339      a multi-byte character. The length in the table is a minimum, so we have to
1340      arrange to skip the extra bytes. */
1341    
1342    #ifdef SUPPORT_UTF8
1343      if (utf8) switch(c)      if (utf8) switch(c)
1344        {        {
1345        case OP_CHAR:        case OP_CHAR:
# Line 1064  for (;;) Line 1347  for (;;)
1347        case OP_EXACT:        case OP_EXACT:
1348        case OP_UPTO:        case OP_UPTO:
1349        case OP_MINUPTO:        case OP_MINUPTO:
1350          case OP_POSUPTO:
1351        case OP_STAR:        case OP_STAR:
1352        case OP_MINSTAR:        case OP_MINSTAR:
1353          case OP_POSSTAR:
1354        case OP_PLUS:        case OP_PLUS:
1355        case OP_MINPLUS:        case OP_MINPLUS:
1356          case OP_POSPLUS:
1357        case OP_QUERY:        case OP_QUERY:
1358        case OP_MINQUERY:        case OP_MINQUERY:
1359        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1360        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1361        break;        break;
1362        }        }
1363  #endif  #endif
# Line 1105  Returns:      pointer to the opcode for Line 1384  Returns:      pointer to the opcode for
1384  static const uschar *  static const uschar *
1385  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1386  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1387  for (;;)  for (;;)
1388    {    {
1389    register int c = *code;    register int c = *code;
1390    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1391    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1392    else if (c > OP_BRA)  
1393      {    /* XCLASS is used for classes that cannot be represented just by a bit
1394      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1395      }    the table is zero; the actual length is stored in the compiled code. */
1396    
1397      if (c == OP_XCLASS) code += GET(code, 1);
1398    
1399      /* Otherwise, we can get the item's length from the table, except that for
1400      repeated character types, we have to test for \p and \P, which have an extra
1401      two bytes of parameters. */
1402    
1403    else    else
1404      {      {
1405      code += _pcre_OP_lengths[c];      switch(c)
1406          {
1407          case OP_TYPESTAR:
1408          case OP_TYPEMINSTAR:
1409          case OP_TYPEPLUS:
1410          case OP_TYPEMINPLUS:
1411          case OP_TYPEQUERY:
1412          case OP_TYPEMINQUERY:
1413          case OP_TYPEPOSSTAR:
1414          case OP_TYPEPOSPLUS:
1415          case OP_TYPEPOSQUERY:
1416          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1417          break;
1418    
1419  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1420          case OP_TYPEUPTO:
1421          case OP_TYPEMINUPTO:
1422          case OP_TYPEEXACT:
1423          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1424          break;
1425          }
1426    
1427        /* Add in the fixed length from the table */
1428    
1429        code += _pcre_OP_lengths[c];
1430    
1431      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1432      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1433      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1434    
1435    #ifdef SUPPORT_UTF8
1436      if (utf8) switch(c)      if (utf8) switch(c)
1437        {        {
1438        case OP_CHAR:        case OP_CHAR:
# Line 1136  for (;;) Line 1440  for (;;)
1440        case OP_EXACT:        case OP_EXACT:
1441        case OP_UPTO:        case OP_UPTO:
1442        case OP_MINUPTO:        case OP_MINUPTO:
1443          case OP_POSUPTO:
1444        case OP_STAR:        case OP_STAR:
1445        case OP_MINSTAR:        case OP_MINSTAR:
1446          case OP_POSSTAR:
1447        case OP_PLUS:        case OP_PLUS:
1448        case OP_MINPLUS:        case OP_MINPLUS:
1449          case OP_POSPLUS:
1450        case OP_QUERY:        case OP_QUERY:
1451        case OP_MINQUERY:        case OP_MINQUERY:
1452        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1453        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1454        break;        break;
1455        }        }
1456  #endif  #endif
# Line 1165  for (;;) Line 1465  for (;;)
1465  *************************************************/  *************************************************/
1466    
1467  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
1468  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1469  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
1470  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
1471  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1472    struck an inner bracket whose current branch will already have been scanned.
1473    
1474  Arguments:  Arguments:
1475    code        points to start of search    code        points to start of search
# Line 1182  static BOOL Line 1483  static BOOL
1483  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1484  {  {
1485  register int c;  register int c;
1486  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1487       code < endcode;       code < endcode;
1488       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1489    {    {
# Line 1190  for (code = first_significant_code(code Line 1491  for (code = first_significant_code(code
1491    
1492    c = *code;    c = *code;
1493    
1494    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1495    
1496      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1497        {
1498        code += _pcre_OP_lengths[c];
1499        do code += GET(code, 1); while (*code == OP_ALT);
1500        c = *code;
1501        continue;
1502        }
1503    
1504      /* For other groups, scan the branches. */
1505    
1506      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1507      {      {
1508      BOOL empty_branch;      BOOL empty_branch;
1509      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1206  for (code = first_significant_code(code Line 1519  for (code = first_significant_code(code
1519        }        }
1520      while (*code == OP_ALT);      while (*code == OP_ALT);
1521      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1522      c = *code;      c = *code;
1523        continue;
1524      }      }
1525    
1526    else switch (c)    /* Handle the other opcodes */
1527    
1528      switch (c)
1529      {      {
1530      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1531        cannot be represented just by a bit map. This includes negated single
1532        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1533        actual length is stored in the compiled code, so we must update "code"
1534        here. */
1535    
1536  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1537      case OP_XCLASS:      case OP_XCLASS:
1538      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1539      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1540  #endif  #endif
1541    
# Line 1266  for (code = first_significant_code(code Line 1585  for (code = first_significant_code(code
1585      case OP_NOT:      case OP_NOT:
1586      case OP_PLUS:      case OP_PLUS:
1587      case OP_MINPLUS:      case OP_MINPLUS:
1588        case OP_POSPLUS:
1589      case OP_EXACT:      case OP_EXACT:
1590      case OP_NOTPLUS:      case OP_NOTPLUS:
1591      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1592        case OP_NOTPOSPLUS:
1593      case OP_NOTEXACT:      case OP_NOTEXACT:
1594      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1595      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1596        case OP_TYPEPOSPLUS:
1597      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1598      return FALSE;      return FALSE;
1599    
1600        /* These are going to continue, as they may be empty, but we have to
1601        fudge the length for the \p and \P cases. */
1602    
1603        case OP_TYPESTAR:
1604        case OP_TYPEMINSTAR:
1605        case OP_TYPEPOSSTAR:
1606        case OP_TYPEQUERY:
1607        case OP_TYPEMINQUERY:
1608        case OP_TYPEPOSQUERY:
1609        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1610        break;
1611    
1612        /* Same for these */
1613    
1614        case OP_TYPEUPTO:
1615        case OP_TYPEMINUPTO:
1616        case OP_TYPEPOSUPTO:
1617        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1618        break;
1619    
1620      /* End of branch */      /* End of branch */
1621    
# Line 1283  for (code = first_significant_code(code Line 1625  for (code = first_significant_code(code
1625      case OP_ALT:      case OP_ALT:
1626      return TRUE;      return TRUE;
1627    
1628      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1629      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1630    
1631  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1632      case OP_STAR:      case OP_STAR:
1633      case OP_MINSTAR:      case OP_MINSTAR:
1634        case OP_POSSTAR:
1635      case OP_QUERY:      case OP_QUERY:
1636      case OP_MINQUERY:      case OP_MINQUERY:
1637        case OP_POSQUERY:
1638      case OP_UPTO:      case OP_UPTO:
1639      case OP_MINUPTO:      case OP_MINUPTO:
1640        case OP_POSUPTO:
1641      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1642      break;      break;
1643  #endif  #endif
# Line 1410  earlier groups that are outside the curr Line 1755  earlier groups that are outside the curr
1755  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before
1756  it, after it has been compiled. This means that any OP_RECURSE items within it  it, after it has been compiled. This means that any OP_RECURSE items within it
1757  that refer to the group itself or any contained groups have to have their  that refer to the group itself or any contained groups have to have their
1758  offsets adjusted. That is the job of this function. Before it is called, the  offsets adjusted. That one of the jobs of this function. Before it is called,
1759  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1760    
1761    This function has been extended with the possibility of forward references for
1762    recursions and subroutine calls. It must also check the list of such references
1763    for the group we are dealing with. If it finds that one of the recursions in
1764    the current group is on this list, it adjusts the offset in the list, not the
1765    value in the reference (which is a group number).
1766    
1767  Arguments:  Arguments:
1768    group      points to the start of the group    group      points to the start of the group
1769    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1770    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1771    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1772      save_hwm   the hwm forward reference pointer at the start of the group
1773    
1774  Returns:     nothing  Returns:     nothing
1775  */  */
1776    
1777  static void  static void
1778  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1779      uschar *save_hwm)
1780  {  {
1781  uschar *ptr = group;  uschar *ptr = group;
1782    
1783  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1784    {    {
1785    int offset = GET(ptr, 1);    int offset;
1786    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1787    
1788      /* See if this recursion is on the forward reference list. If so, adjust the
1789      reference. */
1790    
1791      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1792        {
1793        offset = GET(hc, 0);
1794        if (cd->start_code + offset == ptr + 1)
1795          {
1796          PUT(hc, 0, offset + adjust);
1797          break;
1798          }
1799        }
1800    
1801      /* Otherwise, adjust the recursion offset if it's after the start of this
1802      group. */
1803    
1804      if (hc >= cd->hwm)
1805        {
1806        offset = GET(ptr, 1);
1807        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1808        }
1809    
1810    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1811    }    }
1812  }  }
# Line 1508  Yield:        TRUE when range returned; Line 1885  Yield:        TRUE when range returned;
1885  */  */
1886    
1887  static BOOL  static BOOL
1888  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1889      unsigned int *odptr)
1890  {  {
1891  int c, othercase, next;  unsigned int c, othercase, next;
1892    
1893  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1894    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
1895    
1896  if (c > d) return FALSE;  if (c > d) return FALSE;
1897    
# Line 1534  return TRUE; Line 1912  return TRUE;
1912  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1913    
1914    
1915    
1916  /*************************************************  /*************************************************
1917  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1918  *************************************************/  *************************************************/
1919    
1920  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called for unlimited repeats of certain items, to see
1921  changed during the branch, the pointer is used to change the external options  whether the next thing could possibly match the repeated item. If not, it makes
1922  bits.  sense to automatically possessify the repeated item.
1923    
1924  Arguments:  Arguments:
1925    optionsptr     pointer to the option bits    op_code       the repeated op code
1926    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1927    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1928    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1929    errorcodeptr   points to error code variable    ptr           next character in pattern
1930    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1931    reqbyteptr     set to the last literal character required, else < 0    cd            contains pointers to tables etc.
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
1932    
1933  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1934  */  */
1935    
1936  static BOOL  static BOOL
1937  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1938    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1939  {  {
1940  int repeat_type, op_type;  int next;
1941  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
1942  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
1943  int greedy_default, greedy_non_default;  
1944  int firstbyte, reqbyte;  if ((options & PCRE_EXTENDED) != 0)
1945  int zeroreqbyte, zerofirstbyte;    {
1946  int req_caseopt, reqvary, tempreqvary;    for (;;)
1947  int condcount = 0;      {
1948  int options = *optionsptr;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1949  int after_manual_callout = 0;      if (*ptr == '#')
1950  register int c;        {
1951  register uschar *code = *codeptr;        while (*(++ptr) != 0)
1952  uschar *tempcode;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1953  BOOL inescq = FALSE;        }
1954  BOOL groupsetfirstbyte = FALSE;      else break;
1955  const uschar *ptr = *ptrptr;      }
1956  const uschar *tempptr;    }
1957  uschar *previous = NULL;  
1958  uschar *previous_callout = NULL;  /* If the next item is one that we can handle, get its value. A non-negative
1959  uschar classbits[32];  value is a character, a negative value is an escape value. */
1960    
1961    if (*ptr == '\\')
1962      {
1963      int temperrorcode = 0;
1964      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
1965      if (temperrorcode != 0) return FALSE;
1966      ptr++;    /* Point after the escape sequence */
1967      }
1968    
1969    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
1970      {
1971  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1972  BOOL class_utf8;    if (utf8) { GETCHARINC(next, ptr); } else
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
1973  #endif  #endif
1974      next = *ptr++;
1975      }
1976    
1977  /* Set up the default and non-default settings for greediness */  else return FALSE;
1978    
1979  greedy_default = ((options & PCRE_UNGREEDY) != 0);  /* Skip whitespace and comments in extended mode */
 greedy_non_default = greedy_default ^ 1;  
1980    
1981  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  if ((options & PCRE_EXTENDED) != 0)
1982  matching encountered yet". It gets changed to REQ_NONE if we hit something that    {
1983  matches a non-fixed char first char; reqbyte just remains unset if we never    for (;;)
1984  find one.      {
1985        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1986        if (*ptr == '#')
1987          {
1988          while (*(++ptr) != 0)
1989            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1990          }
1991        else break;
1992        }
1993      }
1994    
1995  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* If the next thing is itself optional, we have to give up. */
 to take the zero repeat into account. This is implemented by setting them to  
 zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  
 item types that can be repeated set these backoff variables appropriately. */  
1996    
1997  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
1998      return FALSE;
1999    
2000  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* Now compare the next item with the previous opcode. If the previous is a
2001  according to the current setting of the caseless flag. REQ_CASELESS is a bit  positive single character match, "item" either contains the character or, if
2002  value > 255. It is added into the firstbyte or reqbyte variables to record the  "item" is greater than 127 in utf8 mode, the character's bytes are in
2003  case status of the value. This is used only for ASCII characters. */  utf8_char. */
2004    
2005  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  
2006    /* Handle cases when the next item is a character. */
2007    
2008    if (next >= 0) switch(op_code)
2009      {
2010      case OP_CHAR:
2011    #ifdef SUPPORT_UTF8
2012      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2013    #endif
2014      return item != next;
2015    
2016      /* For CHARNC (caseless character) we must check the other case. If we have
2017      Unicode property support, we can use it to test the other case of
2018      high-valued characters. */
2019    
2020      case OP_CHARNC:
2021    #ifdef SUPPORT_UTF8
2022      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2023    #endif
2024      if (item == next) return FALSE;
2025    #ifdef SUPPORT_UTF8
2026      if (utf8)
2027        {
2028        unsigned int othercase;
2029        if (next < 128) othercase = cd->fcc[next]; else
2030    #ifdef SUPPORT_UCP
2031        othercase = _pcre_ucp_othercase((unsigned int)next);
2032    #else
2033        othercase = NOTACHAR;
2034    #endif
2035        return (unsigned int)item != othercase;
2036        }
2037      else
2038    #endif  /* SUPPORT_UTF8 */
2039      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2040    
2041      /* For OP_NOT, "item" must be a single-byte character. */
2042    
2043      case OP_NOT:
2044      if (next < 0) return FALSE;  /* Not a character */
2045      if (item == next) return TRUE;
2046      if ((options & PCRE_CASELESS) == 0) return FALSE;
2047    #ifdef SUPPORT_UTF8
2048      if (utf8)
2049        {
2050        unsigned int othercase;
2051        if (next < 128) othercase = cd->fcc[next]; else
2052    #ifdef SUPPORT_UCP
2053        othercase = _pcre_ucp_othercase(next);
2054    #else
2055        othercase = NOTACHAR;
2056    #endif
2057        return (unsigned int)item == othercase;
2058        }
2059      else
2060    #endif  /* SUPPORT_UTF8 */
2061      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2062    
2063      case OP_DIGIT:
2064      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2065    
2066      case OP_NOT_DIGIT:
2067      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2068    
2069      case OP_WHITESPACE:
2070      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2071    
2072      case OP_NOT_WHITESPACE:
2073      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2074    
2075      case OP_WORDCHAR:
2076      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2077    
2078      case OP_NOT_WORDCHAR:
2079      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2080    
2081      case OP_HSPACE:
2082      case OP_NOT_HSPACE:
2083      switch(next)
2084        {
2085        case 0x09:
2086        case 0x20:
2087        case 0xa0:
2088        case 0x1680:
2089        case 0x180e:
2090        case 0x2000:
2091        case 0x2001:
2092        case 0x2002:
2093        case 0x2003:
2094        case 0x2004:
2095        case 0x2005:
2096        case 0x2006:
2097        case 0x2007:
2098        case 0x2008:
2099        case 0x2009:
2100        case 0x200A:
2101        case 0x202f:
2102        case 0x205f:
2103        case 0x3000:
2104        return op_code != OP_HSPACE;
2105        default:
2106        return op_code == OP_HSPACE;
2107        }
2108    
2109      case OP_VSPACE:
2110      case OP_NOT_VSPACE:
2111      switch(next)
2112        {
2113        case 0x0a:
2114        case 0x0b:
2115        case 0x0c:
2116        case 0x0d:
2117        case 0x85:
2118        case 0x2028:
2119        case 0x2029:
2120        return op_code != OP_VSPACE;
2121        default:
2122        return op_code == OP_VSPACE;
2123        }
2124    
2125      default:
2126      return FALSE;
2127      }
2128    
2129    
2130    /* Handle the case when the next item is \d, \s, etc. */
2131    
2132    switch(op_code)
2133      {
2134      case OP_CHAR:
2135      case OP_CHARNC:
2136    #ifdef SUPPORT_UTF8
2137      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2138    #endif
2139      switch(-next)
2140        {
2141        case ESC_d:
2142        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2143    
2144        case ESC_D:
2145        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2146    
2147        case ESC_s:
2148        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2149    
2150        case ESC_S:
2151        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2152    
2153        case ESC_w:
2154        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2155    
2156        case ESC_W:
2157        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2158    
2159        case ESC_h:
2160        case ESC_H:
2161        switch(item)
2162          {
2163          case 0x09:
2164          case 0x20:
2165          case 0xa0:
2166          case 0x1680:
2167          case 0x180e:
2168          case 0x2000:
2169          case 0x2001:
2170          case 0x2002:
2171          case 0x2003:
2172          case 0x2004:
2173          case 0x2005:
2174          case 0x2006:
2175          case 0x2007:
2176          case 0x2008:
2177          case 0x2009:
2178          case 0x200A:
2179          case 0x202f:
2180          case 0x205f:
2181          case 0x3000:
2182          return -next != ESC_h;
2183          default:
2184          return -next == ESC_h;
2185          }
2186    
2187        case ESC_v:
2188        case ESC_V:
2189        switch(item)
2190          {
2191          case 0x0a:
2192          case 0x0b:
2193          case 0x0c:
2194          case 0x0d:
2195          case 0x85:
2196          case 0x2028:
2197          case 0x2029:
2198          return -next != ESC_v;
2199          default:
2200          return -next == ESC_v;
2201          }
2202    
2203        default:
2204        return FALSE;
2205        }
2206    
2207      case OP_DIGIT:
2208      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2209             next == -ESC_h || next == -ESC_v;
2210    
2211      case OP_NOT_DIGIT:
2212      return next == -ESC_d;
2213    
2214      case OP_WHITESPACE:
2215      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2216    
2217      case OP_NOT_WHITESPACE:
2218      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2219    
2220      case OP_HSPACE:
2221      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2222    
2223      case OP_NOT_HSPACE:
2224      return next == -ESC_h;
2225    
2226      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2227      case OP_VSPACE:
2228      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2229    
2230      case OP_NOT_VSPACE:
2231      return next == -ESC_v;
2232    
2233      case OP_WORDCHAR:
2234      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2235    
2236      case OP_NOT_WORDCHAR:
2237      return next == -ESC_w || next == -ESC_d;
2238    
2239      default:
2240      return FALSE;
2241      }
2242    
2243    /* Control does not reach here */
2244    }
2245    
2246    
2247    
2248    /*************************************************
2249    *           Compile one branch                   *
2250    *************************************************/
2251    
2252    /* Scan the pattern, compiling it into the a vector. If the options are
2253    changed during the branch, the pointer is used to change the external options
2254    bits. This function is used during the pre-compile phase when we are trying
2255    to find out the amount of memory needed, as well as during the real compile
2256    phase. The value of lengthptr distinguishes the two phases.
2257    
2258    Arguments:
2259      optionsptr     pointer to the option bits
2260      codeptr        points to the pointer to the current code point
2261      ptrptr         points to the current pattern pointer
2262      errorcodeptr   points to error code variable
2263      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2264      reqbyteptr     set to the last literal character required, else < 0
2265      bcptr          points to current branch chain
2266      cd             contains pointers to tables etc.
2267      lengthptr      NULL during the real compile phase
2268                     points to length accumulator during pre-compile phase
2269    
2270    Returns:         TRUE on success
2271                     FALSE, with *errorcodeptr set non-zero on error
2272    */
2273    
2274    static BOOL
2275    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2276      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2277      compile_data *cd, int *lengthptr)
2278    {
2279    int repeat_type, op_type;
2280    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2281    int bravalue = 0;
2282    int greedy_default, greedy_non_default;
2283    int firstbyte, reqbyte;
2284    int zeroreqbyte, zerofirstbyte;
2285    int req_caseopt, reqvary, tempreqvary;
2286    int options = *optionsptr;
2287    int after_manual_callout = 0;
2288    int length_prevgroup = 0;
2289    register int c;
2290    register uschar *code = *codeptr;
2291    uschar *last_code = code;
2292    uschar *orig_code = code;
2293    uschar *tempcode;
2294    BOOL inescq = FALSE;
2295    BOOL groupsetfirstbyte = FALSE;
2296    const uschar *ptr = *ptrptr;
2297    const uschar *tempptr;
2298    uschar *previous = NULL;
2299    uschar *previous_callout = NULL;
2300    uschar *save_hwm = NULL;
2301    uschar classbits[32];
2302    
2303    #ifdef SUPPORT_UTF8
2304    BOOL class_utf8;
2305    BOOL utf8 = (options & PCRE_UTF8) != 0;
2306    uschar *class_utf8data;
2307    uschar utf8_char[6];
2308    #else
2309    BOOL utf8 = FALSE;
2310    uschar *utf8_char = NULL;
2311    #endif
2312    
2313    #ifdef DEBUG
2314    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2315    #endif
2316    
2317    /* Set up the default and non-default settings for greediness */
2318    
2319    greedy_default = ((options & PCRE_UNGREEDY) != 0);
2320    greedy_non_default = greedy_default ^ 1;
2321    
2322    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2323    matching encountered yet". It gets changed to REQ_NONE if we hit something that
2324    matches a non-fixed char first char; reqbyte just remains unset if we never
2325    find one.
2326    
2327    When we hit a repeat whose minimum is zero, we may have to adjust these values
2328    to take the zero repeat into account. This is implemented by setting them to
2329    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2330    item types that can be repeated set these backoff variables appropriately. */
2331    
2332    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
2333    
2334    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
2335    according to the current setting of the caseless flag. REQ_CASELESS is a bit
2336    value > 255. It is added into the firstbyte or reqbyte variables to record the
2337    case status of the value. This is used only for ASCII characters. */
2338    
2339    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
2340    
2341  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
2342    
# Line 1623  for (;; ptr++) Line 2345  for (;; ptr++)
2345    BOOL negate_class;    BOOL negate_class;
2346    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2347    BOOL is_quantifier;    BOOL is_quantifier;
2348      BOOL is_recurse;
2349      BOOL reset_bracount;
2350    int class_charcount;    int class_charcount;
2351    int class_lastchar;    int class_lastchar;
2352    int newoptions;    int newoptions;
2353    int recno;    int recno;
2354      int refsign;
2355    int skipbytes;    int skipbytes;
2356    int subreqbyte;    int subreqbyte;
2357    int subfirstbyte;    int subfirstbyte;
2358      int terminator;
2359    int mclength;    int mclength;
2360    uschar mcbuffer[8];    uschar mcbuffer[8];
2361    
2362    /* Next byte in the pattern */    /* Get next byte in the pattern */
2363    
2364    c = *ptr;    c = *ptr;
2365    
2366      /* If we are in the pre-compile phase, accumulate the length used for the
2367      previous cycle of this loop. */
2368    
2369      if (lengthptr != NULL)
2370        {
2371    #ifdef DEBUG
2372        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2373    #endif
2374        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2375          {
2376          *errorcodeptr = ERR52;
2377          goto FAILED;
2378          }
2379    
2380        /* There is at least one situation where code goes backwards: this is the
2381        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2382        the class is simply eliminated. However, it is created first, so we have to
2383        allow memory for it. Therefore, don't ever reduce the length at this point.
2384        */
2385    
2386        if (code < last_code) code = last_code;
2387    
2388        /* Paranoid check for integer overflow */
2389    
2390        if (OFLOW_MAX - *lengthptr < code - last_code)
2391          {
2392          *errorcodeptr = ERR20;
2393          goto FAILED;
2394          }
2395    
2396        *lengthptr += code - last_code;
2397        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2398    
2399        /* If "previous" is set and it is not at the start of the work space, move
2400        it back to there, in order to avoid filling up the work space. Otherwise,
2401        if "previous" is NULL, reset the current code pointer to the start. */
2402    
2403        if (previous != NULL)
2404          {
2405          if (previous > orig_code)
2406            {
2407            memmove(orig_code, previous, code - previous);
2408            code -= previous - orig_code;
2409            previous = orig_code;
2410            }
2411          }
2412        else code = orig_code;
2413    
2414        /* Remember where this code item starts so we can pick up the length
2415        next time round. */
2416    
2417        last_code = code;
2418        }
2419    
2420      /* In the real compile phase, just check the workspace used by the forward
2421      reference list. */
2422    
2423      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2424        {
2425        *errorcodeptr = ERR52;
2426        goto FAILED;
2427        }
2428    
2429    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
2430    
2431    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1651  for (;; ptr++) Line 2440  for (;; ptr++)
2440        {        {
2441        if (previous_callout != NULL)        if (previous_callout != NULL)
2442          {          {
2443          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2444              complete_callout(previous_callout, ptr, cd);
2445          previous_callout = NULL;          previous_callout = NULL;
2446          }          }
2447        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1672  for (;; ptr++) Line 2462  for (;; ptr++)
2462    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2463         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2464      {      {
2465      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2466          complete_callout(previous_callout, ptr, cd);
2467      previous_callout = NULL;      previous_callout = NULL;
2468      }      }
2469    
# Line 1683  for (;; ptr++) Line 2474  for (;; ptr++)
2474      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2475      if (c == '#')      if (c == '#')
2476        {        {
2477        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2478        on the Macintosh. */          {
2479        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2480        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2481          if (*ptr != 0) continue;
2482    
2483          /* Else fall through to handle end of string */
2484          c = 0;
2485        }        }
2486      }      }
2487    
# Line 1700  for (;; ptr++) Line 2495  for (;; ptr++)
2495    
2496    switch(c)    switch(c)
2497      {      {
2498      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2499        case 0:                        /* The branch terminates at string end */
2500      case 0:      case '|':                      /* or | or ) */
     case '|':  
2501      case ')':      case ')':
2502      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2503      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2504      *codeptr = code;      *codeptr = code;
2505      *ptrptr = ptr;      *ptrptr = ptr;
2506        if (lengthptr != NULL)
2507          {
2508          if (OFLOW_MAX - *lengthptr < code - last_code)
2509            {
2510            *errorcodeptr = ERR20;
2511            goto FAILED;
2512            }
2513          *lengthptr += code - last_code;   /* To include callout length */
2514          DPRINTF((">> end branch\n"));
2515          }
2516      return TRUE;      return TRUE;
2517    
2518    
2519        /* ===================================================================*/
2520      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2521      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2522    
# Line 1739  for (;; ptr++) Line 2545  for (;; ptr++)
2545      *code++ = OP_ANY;      *code++ = OP_ANY;
2546      break;      break;
2547    
2548    
2549        /* ===================================================================*/
2550      /* Character classes. If the included characters are all < 256, we build a      /* Character classes. If the included characters are all < 256, we build a
2551      32-byte bitmap of the permitted characters, except in the special case      32-byte bitmap of the permitted characters, except in the special case
2552      where there is only one such character. For negated classes, we build the      where there is only one such character. For negated classes, we build the
# Line 1764  for (;; ptr++) Line 2572  for (;; ptr++)
2572        goto FAILED;        goto FAILED;
2573        }        }
2574    
2575      /* 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,
2576        if the first few characters (either before or after ^) are \Q\E or \E we
2577        skip them too. This makes for compatibility with Perl. */
2578    
2579      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2580        for (;;)
2581        {        {
       negate_class = TRUE;  
2582        c = *(++ptr);        c = *(++ptr);
2583        }        if (c == '\\')
2584      else          {
2585        {          if (ptr[1] == 'E') ptr++;
2586        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2587                else break;
2588            }
2589          else if (!negate_class && c == '^')
2590            negate_class = TRUE;
2591          else break;
2592        }        }
2593    
2594      /* 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
2595      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
2596      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2597    
2598      class_charcount = 0;      class_charcount = 0;
2599      class_lastchar = -1;      class_lastchar = -1;
2600    
2601        /* Initialize the 32-char bit map to all zeros. We build the map in a
2602        temporary bit of memory, in case the class contains only 1 character (less
2603        than 256), because in that case the compiled code doesn't use the bit map.
2604        */
2605    
2606        memset(classbits, 0, 32 * sizeof(uschar));
2607    
2608  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2609      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2610      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2611  #endif  #endif
2612    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
2613      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2614      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
2615      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
2616    
2617      do      if (c != 0) do
2618        {        {
2619          const uschar *oldptr;
2620    
2621  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2622        if (utf8 && c > 127)        if (utf8 && c > 127)
2623          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1814  for (;; ptr++) Line 2629  for (;; ptr++)
2629    
2630        if (inescq)        if (inescq)
2631          {          {
2632          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2633            {            {
2634            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2635            ptr++;            ptr++;                            /* Skip the 'E' */
2636            continue;            continue;                         /* Carry on with next */
2637            }            }
2638          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2639          }          }
2640    
2641        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1911  for (;; ptr++) Line 2726  for (;; ptr++)
2726          }          }
2727    
2728        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2729        of the specials, which just set a flag. Escaped items are checked for        of the specials, which just set a flag. The sequence \b is a special
2730        validity in the pre-compiling pass. The sequence \b is a special case.        case. Inside a class (and only there) it is treated as backspace.
2731        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2732        it marks a word boundary. Other escapes have preset maps ready to        to 'or' into the one we are building. We assume they have more than one
       or into the one we are building. We assume they have more than one  
2733        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2734    
2735        if (c == '\\')        if (c == '\\')
2736          {          {
2737          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2738            if (*errorcodeptr != 0) goto FAILED;
2739    
2740          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2741          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2742            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2743          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2744            {            {
2745            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1933  for (;; ptr++) Line 2749  for (;; ptr++)
2749            else inescq = TRUE;            else inescq = TRUE;
2750            continue;            continue;
2751            }            }
2752            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2753    
2754          if (c < 0)          if (c < 0)
2755            {            {
2756            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2757            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2758            switch (-c)  
2759              /* Save time by not doing this in the pre-compile phase. */
2760    
2761              if (lengthptr == NULL) switch (-c)
2762              {              {
2763              case ESC_d:              case ESC_d:
2764              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
# Line 1966  for (;; ptr++) Line 2786  for (;; ptr++)
2786              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2787              continue;              continue;
2788    
2789  #ifdef SUPPORT_UCP              case ESC_E: /* Perl ignores an orphan \E */
             case ESC_p:  
             case ESC_P:  
               {  
               BOOL negated;  
               int pdata;  
               int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);  
               if (ptype < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = ptype;  
               *class_utf8data++ = pdata;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
2790              continue;              continue;
 #endif  
   
             /* Unrecognized escapes are faulted if PCRE is running in its  
             strict mode. By default, for compatibility with Perl, they are  
             treated as literals. */  
2791    
2792              default:              default:    /* Not recognized; fall through */
2793              if ((options & PCRE_EXTRA) != 0)              break;      /* Need "default" setting to stop compiler warning. */
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
2794              }              }
           }  
   
         /* Fall through if we have a single character (c >= 0). This may be  
         > 256 in UTF-8 mode. */  
2795    
2796          }   /* End of backslash handling */            /* In the pre-compile phase, just do the recognition. */
2797    
2798        /* A single character may be followed by '-' to form a range. However,            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2799        Perl does not permit ']' to be the end of the range. A '-' character                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
       here is treated as a literal. */  
2800    
2801        if (ptr[1] == '-' && ptr[2] != ']')            /* We need to deal with \H, \h, \V, and \v in both phases because
2802          {            they use extra memory. */
         int d;  
         ptr += 2;  
2803    
2804              if (-c == ESC_h)
2805                {
2806                SETBIT(classbits, 0x09); /* VT */
2807                SETBIT(classbits, 0x20); /* SPACE */
2808                SETBIT(classbits, 0xa0); /* NSBP */
2809  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2810          if (utf8)              if (utf8)
2811            {                           /* Braces are required because the */                {
2812            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */                class_utf8 = TRUE;
2813            }                *class_utf8data++ = XCL_SINGLE;
2814          else                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2815                  *class_utf8data++ = XCL_SINGLE;
2816                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2817                  *class_utf8data++ = XCL_RANGE;
2818                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2819                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2820                  *class_utf8data++ = XCL_SINGLE;
2821                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2822                  *class_utf8data++ = XCL_SINGLE;
2823                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2824                  *class_utf8data++ = XCL_SINGLE;
2825                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2826                  }
2827  #endif  #endif
2828          d = *ptr;  /* Not UTF-8 mode */              continue;
2829                }
         /* The second part of a range can be a single-character escape, but  
         not any of the other escapes. Perl 5.6 treats a hyphen as a literal  
         in such circumstances. */  
   
         if (d == '\\')  
           {  
           const uschar *oldptr = ptr;  
           d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
   
           /* \b is backslash; \X is literal X; any other special means the '-'  
           was literal */  
2830    
2831            if (d < 0)            if (-c == ESC_H)
2832              {              {
2833              if (d == -ESC_b) d = '\b';              for (c = 0; c < 32; c++)
             else if (d == -ESC_X) d = 'X'; else  
2834                {                {
2835                ptr = oldptr - 2;                int x = 0xff;
2836                  switch (c)
2837                    {
2838                    case 0x09/8: x ^= 1 << (0x09%8); break;
2839                    case 0x20/8: x ^= 1 << (0x20%8); break;
2840                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2841                    default: break;
2842                    }
2843                  classbits[c] |= x;
2844                  }
2845    
2846    #ifdef SUPPORT_UTF8
2847                if (utf8)
2848                  {
2849                  class_utf8 = TRUE;
2850                  *class_utf8data++ = XCL_RANGE;
2851                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2852                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2853                  *class_utf8data++ = XCL_RANGE;
2854                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2855                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2856                  *class_utf8data++ = XCL_RANGE;
2857                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2858                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2859                  *class_utf8data++ = XCL_RANGE;
2860                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2861                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2862                  *class_utf8data++ = XCL_RANGE;
2863                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2864                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2865                  *class_utf8data++ = XCL_RANGE;
2866                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2867                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2868                  *class_utf8data++ = XCL_RANGE;
2869                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2870                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2871                  }
2872    #endif
2873                continue;
2874                }
2875    
2876              if (-c == ESC_v)
2877                {
2878                SETBIT(classbits, 0x0a); /* LF */
2879                SETBIT(classbits, 0x0b); /* VT */
2880                SETBIT(classbits, 0x0c); /* FF */
2881                SETBIT(classbits, 0x0d); /* CR */
2882                SETBIT(classbits, 0x85); /* NEL */
2883    #ifdef SUPPORT_UTF8
2884                if (utf8)
2885                  {
2886                  class_utf8 = TRUE;
2887                  *class_utf8data++ = XCL_RANGE;
2888                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2889                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2890                  }
2891    #endif
2892                continue;
2893                }
2894    
2895              if (-c == ESC_V)
2896                {
2897                for (c = 0; c < 32; c++)
2898                  {
2899                  int x = 0xff;
2900                  switch (c)
2901                    {
2902                    case 0x0a/8: x ^= 1 << (0x0a%8);
2903                                 x ^= 1 << (0x0b%8);
2904                                 x ^= 1 << (0x0c%8);
2905                                 x ^= 1 << (0x0d%8);
2906                                 break;
2907                    case 0x85/8: x ^= 1 << (0x85%8); break;
2908                    default: break;
2909                    }
2910                  classbits[c] |= x;
2911                  }
2912    
2913    #ifdef SUPPORT_UTF8
2914                if (utf8)
2915                  {
2916                  class_utf8 = TRUE;
2917                  *class_utf8data++ = XCL_RANGE;
2918                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2919                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2920                  *class_utf8data++ = XCL_RANGE;
2921                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2922                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2923                  }
2924    #endif
2925                continue;
2926                }
2927    
2928              /* We need to deal with \P and \p in both phases. */
2929    
2930    #ifdef SUPPORT_UCP
2931              if (-c == ESC_p || -c == ESC_P)
2932                {
2933                BOOL negated;
2934                int pdata;
2935                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2936                if (ptype < 0) goto FAILED;
2937                class_utf8 = TRUE;
2938                *class_utf8data++ = ((-c == ESC_p) != negated)?
2939                  XCL_PROP : XCL_NOTPROP;
2940                *class_utf8data++ = ptype;
2941                *class_utf8data++ = pdata;
2942                class_charcount -= 2;   /* Not a < 256 character */
2943                continue;
2944                }
2945    #endif
2946              /* Unrecognized escapes are faulted if PCRE is running in its
2947              strict mode. By default, for compatibility with Perl, they are
2948              treated as literals. */
2949    
2950              if ((options & PCRE_EXTRA) != 0)
2951                {
2952                *errorcodeptr = ERR7;
2953                goto FAILED;
2954                }
2955    
2956              class_charcount -= 2;  /* Undo the default count from above */
2957              c = *ptr;              /* Get the final character and fall through */
2958              }
2959    
2960            /* Fall through if we have a single character (c >= 0). This may be
2961            greater than 256 in UTF-8 mode. */
2962    
2963            }   /* End of backslash handling */
2964    
2965          /* A single character may be followed by '-' to form a range. However,
2966          Perl does not permit ']' to be the end of the range. A '-' character
2967          at the end is treated as a literal. Perl ignores orphaned \E sequences
2968          entirely. The code for handling \Q and \E is messy. */
2969    
2970          CHECK_RANGE:
2971          while (ptr[1] == '\\' && ptr[2] == 'E')
2972            {
2973            inescq = FALSE;
2974            ptr += 2;
2975            }
2976    
2977          oldptr = ptr;
2978    
2979          if (!inescq && ptr[1] == '-')
2980            {
2981            int d;
2982            ptr += 2;
2983            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
2984    
2985            /* If we hit \Q (not followed by \E) at this point, go into escaped
2986            mode. */
2987    
2988            while (*ptr == '\\' && ptr[1] == 'Q')
2989              {
2990              ptr += 2;
2991              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
2992              inescq = TRUE;
2993              break;
2994              }
2995    
2996            if (*ptr == 0 || (!inescq && *ptr == ']'))
2997              {
2998              ptr = oldptr;
2999              goto LONE_SINGLE_CHARACTER;
3000              }
3001    
3002    #ifdef SUPPORT_UTF8
3003            if (utf8)
3004              {                           /* Braces are required because the */
3005              GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
3006              }
3007            else
3008    #endif
3009            d = *ptr;  /* Not UTF-8 mode */
3010    
3011            /* The second part of a range can be a single-character escape, but
3012            not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3013            in such circumstances. */
3014    
3015            if (!inescq && d == '\\')
3016              {
3017              d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3018              if (*errorcodeptr != 0) goto FAILED;
3019    
3020              /* \b is backslash; \X is literal X; \R is literal R; any other
3021              special means the '-' was literal */
3022    
3023              if (d < 0)
3024                {
3025                if (d == -ESC_b) d = '\b';
3026                else if (d == -ESC_X) d = 'X';
3027                else if (d == -ESC_R) d = 'R'; else
3028                  {
3029                  ptr = oldptr;
3030                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3031                }                }
3032              }              }
3033            }            }
3034    
3035          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3036          the pre-pass. Optimize one-character ranges */          one-character ranges */
3037    
3038            if (d < c)
3039              {
3040              *errorcodeptr = ERR8;
3041              goto FAILED;
3042              }
3043    
3044          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3045    
# Line 2067  for (;; ptr++) Line 3060  for (;; ptr++)
3060  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3061            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3062              {              {
3063              int occ, ocd;              unsigned int occ, ocd;
3064              int cc = c;              unsigned int cc = c;
3065              int origd = d;              unsigned int origd = d;
3066              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3067                {                {
3068                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3069                      ocd <= (unsigned int)d)
3070                    continue;                          /* Skip embedded ranges */
3071    
3072                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3073                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3074                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3075                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3076                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3077                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3078                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3079                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3080                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3081                  d = ocd;                  d = ocd;
3082                  continue;                  continue;
# Line 2127  for (;; ptr++) Line 3124  for (;; ptr++)
3124          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
3125          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3126    
3127          for (; c <= d; c++)          class_charcount += d - c + 1;
3128            class_lastchar = d;
3129    
3130            /* We can save a bit of time by skipping this in the pre-compile. */
3131    
3132            if (lengthptr == NULL) for (; c <= d; c++)
3133            {            {
3134            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3135            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2135  for (;; ptr++) Line 3137  for (;; ptr++)
3137              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3138              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3139              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3140            }            }
3141    
3142          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2160  for (;; ptr++) Line 3160  for (;; ptr++)
3160  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3161          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3162            {            {
3163            int othercase;            unsigned int othercase;
3164            if ((othercase = _pcre_ucp_othercase(c)) >= 0)            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
3165              {              {
3166              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3167              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2186  for (;; ptr++) Line 3186  for (;; ptr++)
3186          }          }
3187        }        }
3188    
3189      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
     loop. This "while" is the end of the "do" above. */  
3190    
3191      while ((c = *(++ptr)) != ']' || inescq);      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3192    
3193        if (c == 0)                          /* Missing terminating ']' */
3194          {
3195          *errorcodeptr = ERR6;
3196          goto FAILED;
3197          }
3198    
3199        /* Remember whether \r or \n are in this class */
3200    
3201        if (negate_class)
3202          {
3203          if ((classbits[1] & 0x24) != 0x24) cd->external_options |= PCRE_HASCRORLF;
3204          }
3205        else
3206          {
3207          if ((classbits[1] & 0x24) != 0) cd->external_options |= PCRE_HASCRORLF;
3208          }
3209    
3210      /* 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
3211      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
3212      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
3213      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3214      single-bytes only. This is an historical hangover. Maybe one day we can  
3215      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3216        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3217        operate on single-bytes only. This is an historical hangover. Maybe one day
3218        we can tidy these opcodes to handle multi-byte characters.
3219    
3220      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
3221      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 2206  for (;; ptr++) Line 3225  for (;; ptr++)
3225      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3226    
3227  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3228      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3229            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3230  #else  #else
3231      if (class_charcount == 1)      if (class_charcount == 1)
3232  #endif  #endif
# Line 2253  for (;; ptr++) Line 3270  for (;; ptr++)
3270    
3271      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3272      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode. If there are no characters < 256,
3273      we can omit the bitmap. */      we can omit the bitmap in the actual compiled code. */
3274    
3275  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3276      if (class_utf8)      if (class_utf8)
# Line 2263  for (;; ptr++) Line 3280  for (;; ptr++)
3280        code += LINK_SIZE;        code += LINK_SIZE;
3281        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3282    
3283        /* If the map is required, install it, and move on to the end of        /* If the map is required, move up the extra data to make room for it;
3284        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3285    
3286        if (class_charcount > 0)        if (class_charcount > 0)
3287          {          {
3288          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3289            memmove(code + 32, code, class_utf8data - code);
3290          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3291          code = class_utf8data;          code = class_utf8data + 32;
         }  
   
       /* If the map is not required, slide down the extra data. */  
   
       else  
         {  
         int len = class_utf8data - (code + 33);  
         memmove(code + 1, code + 33, len);  
         code += len + 1;  
3292          }          }
3293          else code = class_utf8data;
3294    
3295        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3296    
# Line 2297  for (;; ptr++) Line 3307  for (;; ptr++)
3307      if (negate_class)      if (negate_class)
3308        {        {
3309        *code++ = OP_NCLASS;        *code++ = OP_NCLASS;
3310        for (c = 0; c < 32; c++) code[c] = ~classbits[c];        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3311            for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3312        }        }
3313      else      else
3314        {        {
# Line 2307  for (;; ptr++) Line 3318  for (;; ptr++)
3318      code += 32;      code += 32;
3319      break;      break;
3320    
3321    
3322        /* ===================================================================*/
3323      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3324      has been tested above. */      has been tested above. */
3325    
# Line 2374  for (;; ptr++) Line 3387  for (;; ptr++)
3387        }        }
3388      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3389    
     /* If previous was a recursion, we need to wrap it inside brackets so that  
     it can be replicated if necessary. */  
   
     if (*previous == OP_RECURSE)  
       {  
       memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);  
       code += 1 + LINK_SIZE;  
       *previous = OP_BRA;  
       PUT(previous, 1, code - previous);  
       *code = OP_KET;  
       PUT(code, 1, code - previous);  
       code += 1 + LINK_SIZE;  
       }  
   
3390      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3391      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
3392      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2421  for (;; ptr++) Line 3420  for (;; ptr++)
3420          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3421          }          }
3422    
3423          /* If the repetition is unlimited, it pays to see if the next thing on
3424          the line is something that cannot possibly match this character. If so,
3425          automatically possessifying this item gains some performance in the case
3426          where the match fails. */
3427    
3428          if (!possessive_quantifier &&
3429              repeat_max < 0 &&
3430              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3431                options, cd))
3432            {
3433            repeat_type = 0;    /* Force greedy */
3434            possessive_quantifier = TRUE;
3435            }
3436    
3437        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3438        }        }
3439    
3440      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3441      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
3442      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3443      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3444        currently used only for single-byte chars. */
3445    
3446      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3447        {        {
3448        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3449        c = previous[1];        c = previous[1];
3450          if (!possessive_quantifier &&
3451              repeat_max < 0 &&
3452              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3453            {
3454            repeat_type = 0;    /* Force greedy */
3455            possessive_quantifier = TRUE;
3456            }
3457        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3458        }        }
3459    
# Line 2450  for (;; ptr++) Line 3471  for (;; ptr++)
3471        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3472        c = *previous;        c = *previous;
3473    
3474          if (!possessive_quantifier &&
3475              repeat_max < 0 &&
3476              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3477            {
3478            repeat_type = 0;    /* Force greedy */
3479            possessive_quantifier = TRUE;
3480            }
3481    
3482        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3483        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3484          {          {
# Line 2490  for (;; ptr++) Line 3519  for (;; ptr++)
3519          }          }
3520    
3521        /* A repeat minimum of 1 is optimized into some special cases. If the        /* A repeat minimum of 1 is optimized into some special cases. If the
3522        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3523        left in place and, if the maximum is greater than 1, we use OP_UPTO with        left in place and, if the maximum is greater than 1, we use OP_UPTO with
3524        one less than the maximum. */        one less than the maximum. */
3525    
# Line 2543  for (;; ptr++) Line 3572  for (;; ptr++)
3572            }            }
3573    
3574          /* Else insert an UPTO if the max is greater than the min, again          /* Else insert an UPTO if the max is greater than the min, again
3575          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3576            UPTO is just for 1 instance, we can use QUERY instead. */
3577    
3578          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3579            {            {
# Line 2562  for (;; ptr++) Line 3592  for (;; ptr++)
3592              *code++ = prop_value;              *code++ = prop_value;
3593              }              }
3594            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3595            *code++ = OP_UPTO + repeat_type;  
3596            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3597                {
3598                *code++ = OP_QUERY + repeat_type;
3599                }
3600              else
3601                {
3602                *code++ = OP_UPTO + repeat_type;
3603                PUT2INC(code, 0, repeat_max);
3604                }
3605            }            }
3606          }          }
3607    
# Line 2630  for (;; ptr++) Line 3668  for (;; ptr++)
3668      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
3669      cases. */      cases. */
3670    
3671      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3672               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3673        {        {
3674        register int i;        register int i;
3675        int ketoffset = 0;        int ketoffset = 0;
3676        int len = code - previous;        int len = code - previous;
3677        uschar *bralink = NULL;        uschar *bralink = NULL;
3678    
3679          /* Repeating a DEFINE group is pointless */
3680    
3681          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3682            {
3683            *errorcodeptr = ERR55;
3684            goto FAILED;
3685            }
3686    
3687        /* 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
3688        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
3689        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 2672  for (;; ptr++) Line 3718  for (;; ptr++)
3718          /* If the maximum is 1 or unlimited, we just have to stick in the          /* If the maximum is 1 or unlimited, we just have to stick in the
3719          BRAZERO and do no more at this point. However, we do need to adjust          BRAZERO and do no more at this point. However, we do need to adjust
3720          any OP_RECURSE calls inside the group that refer to the group itself or          any OP_RECURSE calls inside the group that refer to the group itself or
3721          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3722          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3723            doing this. */
3724    
3725          if (repeat_max <= 1)          if (repeat_max <= 1)
3726            {            {
3727            *code = OP_END;            *code = OP_END;
3728            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3729            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3730            code++;            code++;
3731            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2696  for (;; ptr++) Line 3743  for (;; ptr++)
3743            {            {
3744            int offset;            int offset;
3745            *code = OP_END;            *code = OP_END;
3746            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3747            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3748            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3749            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2716  for (;; ptr++) Line 3763  for (;; ptr++)
3763        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3764        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3765        copies that we need. If we set a first char from the group, and didn't        copies that we need. If we set a first char from the group, and didn't
3766        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3767          forward reference subroutine calls in the group, there will be entries on
3768          the workspace list; replicate these with an appropriate increment. */
3769    
3770        else        else
3771          {          {
3772          if (repeat_min > 1)          if (repeat_min > 1)
3773            {            {
3774            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3775            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3776              potential integer overflow. */
3777    
3778              if (lengthptr != NULL)
3779                {
3780                int delta = (repeat_min - 1)*length_prevgroup;
3781                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3782                                                                (double)INT_MAX ||
3783                    OFLOW_MAX - *lengthptr < delta)
3784                  {
3785                  *errorcodeptr = ERR20;
3786                  goto FAILED;
3787                  }
3788                *lengthptr += delta;
3789                }
3790    
3791              /* This is compiling for real */
3792    
3793              else
3794              {              {
3795              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3796              code += len;              for (i = 1; i < repeat_min; i++)
3797                  {
3798                  uschar *hc;
3799                  uschar *this_hwm = cd->hwm;
3800                  memcpy(code, previous, len);
3801                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3802                    {
3803                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3804                    cd->hwm += LINK_SIZE;
3805                    }
3806                  save_hwm = this_hwm;
3807                  code += len;
3808                  }
3809              }              }
3810            }            }
3811    
3812          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3813          }          }
3814    
# Line 2736  for (;; ptr++) Line 3816  for (;; ptr++)
3816        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3817        remembering the bracket starts on a stack. In the case of a zero minimum,        remembering the bracket starts on a stack. In the case of a zero minimum,
3818        the first one was set up above. In all cases the repeat_max now specifies        the first one was set up above. In all cases the repeat_max now specifies
3819        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3820          replicate entries on the forward reference list. */
3821    
3822        if (repeat_max >= 0)        if (repeat_max >= 0)
3823          {          {
3824          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3825            just adjust the length as if we had. For each repetition we must add 1
3826            to the length for BRAZERO and for all but the last repetition we must
3827            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3828            paranoid checks to avoid integer overflow. */
3829    
3830            if (lengthptr != NULL && repeat_max > 0)
3831              {
3832              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3833                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3834              if ((double)repeat_max *
3835                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3836                      > (double)INT_MAX ||
3837                  OFLOW_MAX - *lengthptr < delta)
3838                {
3839                *errorcodeptr = ERR20;
3840                goto FAILED;
3841                }
3842              *lengthptr += delta;
3843              }
3844    
3845            /* This is compiling for real */
3846    
3847            else for (i = repeat_max - 1; i >= 0; i--)
3848            {            {
3849              uschar *hc;
3850              uschar *this_hwm = cd->hwm;
3851    
3852            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3853    
3854            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2757  for (;; ptr++) Line 3864  for (;; ptr++)
3864              }              }
3865    
3866            memcpy(code, previous, len);            memcpy(code, previous, len);
3867              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3868                {
3869                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3870                cd->hwm += LINK_SIZE;
3871                }
3872              save_hwm = this_hwm;
3873            code += len;            code += len;
3874            }            }
3875    
# Line 2779  for (;; ptr++) Line 3892  for (;; ptr++)
3892        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. We
3893        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3894        don't know if there's been an options resetting after the ket. The        don't know if there's been an options resetting after the ket. The
3895        correct offset was computed above. */        correct offset was computed above.
3896    
3897          Then, when we are doing the actual compile phase, check to see whether
3898          this group is a non-atomic one that could match an empty string. If so,
3899          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3900          that runtime checking can be done. [This check is also applied to
3901          atomic groups at runtime, but in a different way.] */
3902    
3903        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
3904            {
3905            uschar *ketcode = code - ketoffset;
3906            uschar *bracode = ketcode - GET(ketcode, 1);
3907            *ketcode = OP_KETRMAX + repeat_type;
3908            if (lengthptr == NULL && *bracode != OP_ONCE)
3909              {
3910              uschar *scode = bracode;
3911              do
3912                {
3913                if (could_be_empty_branch(scode, ketcode, utf8))
3914                  {
3915                  *bracode += OP_SBRA - OP_BRA;
3916                  break;
3917                  }
3918                scode += GET(scode, 1);
3919                }
3920              while (*scode == OP_ALT);
3921              }
3922            }
3923        }        }
3924    
3925      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2792  for (;; ptr++) Line 3930  for (;; ptr++)
3930        goto FAILED;        goto FAILED;
3931        }        }
3932    
3933      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
3934      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
3935      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
3936      which might be the first part of a string whose (former) last char we      anything else, we wrap the entire repeated item inside OP_ONCE brackets.
3937      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
3938        but the special opcodes can optimize it a bit. The repeated item starts at
3939        tempcode, not at previous, which might be the first part of a string whose
3940        (former) last char we repeated.
3941    
3942        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
3943        an 'upto' may follow. We skip over an 'exact' item, and then test the
3944        length of what remains before proceeding. */
3945    
3946      if (possessive_quantifier)      if (possessive_quantifier)
3947        {        {
3948        int len = code - tempcode;        int len;
3949        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
3950        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
3951        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
3952        tempcode[0] = OP_ONCE;        len = code - tempcode;
3953        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
3954        PUTINC(code, 0, len);          {
3955        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
3956            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
3957            case OP_QUERY: *tempcode = OP_POSQUERY; break;
3958            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
3959    
3960            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
3961            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
3962            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
3963            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
3964    
3965            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
3966            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
3967            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
3968            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
3969    
3970            default:
3971            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
3972            code += 1 + LINK_SIZE;
3973            len += 1 + LINK_SIZE;
3974            tempcode[0] = OP_ONCE;
3975            *code++ = OP_KET;
3976            PUTINC(code, 0, len);
3977            PUT(tempcode, 1, len);
3978            break;
3979            }
3980        }        }
3981    
3982      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 2820  for (;; ptr++) Line 3989  for (;; ptr++)
3989      break;      break;
3990    
3991    
3992      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
3993      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
3994      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
3995      of any of them means that this is not a referencing group. They were      parenthesis forms.  */
     checked for validity in the first pass over the string, so we don't have to  
     check for syntax errors here.  */  
3996    
3997      case '(':      case '(':
3998      newoptions = options;      newoptions = options;
3999      skipbytes = 0;      skipbytes = 0;
4000        bravalue = OP_CBRA;
4001        save_hwm = cd->hwm;
4002        reset_bracount = FALSE;
4003    
4004        /* First deal with various "verbs" that can be introduced by '*'. */
4005    
4006        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4007          {
4008          int i, namelen;
4009          const uschar *name = ++ptr;
4010          previous = NULL;
4011          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4012          if (*ptr == ':')
4013            {
4014            *errorcodeptr = ERR59;   /* Not supported */
4015            goto FAILED;
4016            }
4017          if (*ptr != ')')
4018            {
4019            *errorcodeptr = ERR60;
4020            goto FAILED;
4021            }
4022          namelen = ptr - name;
4023          for (i = 0; i < verbcount; i++)
4024            {
4025            if (namelen == verbs[i].len &&
4026                strncmp((char *)name, verbs[i].name, namelen) == 0)
4027              {
4028              *code = verbs[i].op;
4029              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4030              break;
4031              }
4032            }
4033          if (i < verbcount) continue;
4034          *errorcodeptr = ERR60;
4035          goto FAILED;
4036          }
4037    
4038        /* Deal with the extended parentheses; all are introduced by '?', and the
4039        appearance of any of them means that this is not a capturing group. */
4040    
4041      if (*(++ptr) == '?')      else if (*ptr == '?')
4042        {        {
4043        int set, unset;        int i, set, unset, namelen;
4044        int *optset;        int *optset;
4045          const uschar *name;
4046          uschar *slot;
4047    
4048        switch (*(++ptr))        switch (*(++ptr))
4049          {          {
4050          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4051          ptr++;          ptr++;
4052          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4053            if (*ptr == 0)
4054              {
4055              *errorcodeptr = ERR18;
4056              goto FAILED;
4057              }
4058          continue;          continue;
4059    
4060          case ':':                 /* Non-extracting bracket */  
4061            /* ------------------------------------------------------------ */
4062            case '|':                 /* Reset capture count for each branch */
4063            reset_bracount = TRUE;
4064            /* Fall through */
4065    
4066            /* ------------------------------------------------------------ */
4067            case ':':                 /* Non-capturing bracket */
4068          bravalue = OP_BRA;          bravalue = OP_BRA;
4069          ptr++;          ptr++;
4070          break;          break;
4071    
4072    
4073            /* ------------------------------------------------------------ */
4074          case '(':          case '(':
4075          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4076    
4077          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4078            group), a name (referring to a named group), or 'R', referring to
4079            recursion. R<digits> and R&name are also permitted for recursion tests.
4080    
4081            There are several syntaxes for testing a named group: (?(name)) is used
4082            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4083    
4084            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4085            be the recursive thing or the name 'R' (and similarly for 'R' followed
4086            by digits), and (b) a number could be a name that consists of digits.
4087            In both cases, we look for a name first; if not found, we try the other
4088            cases. */
4089    
4090            /* For conditions that are assertions, check the syntax, and then exit
4091            the switch. This will take control down to where bracketed groups,
4092            including assertions, are processed. */
4093    
4094            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4095              break;
4096    
4097            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4098            below), and all need to skip 3 bytes at the start of the group. */
4099    
4100            code[1+LINK_SIZE] = OP_CREF;
4101            skipbytes = 3;
4102            refsign = -1;
4103    
4104            /* Check for a test for recursion in a named group. */
4105    
4106          if (ptr[1] == 'R')          if (ptr[1] == 'R' && ptr[2] == '&')
4107            {            {
4108            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4109            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4110            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4111            }            }
4112    
4113          /* Condition to test for a numbered subpattern match. We know that          /* Check for a test for a named group's having been set, using the Perl
4114          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4115    
4116          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4117            {            {
4118            int condref;                 /* Don't amalgamate; some compilers */            terminator = '>';
           condref = *(++ptr) - '0';    /* grumble at autoincrement in declaration */  
           while (*(++ptr) != ')') condref = condref*10 + *ptr - '0';  
           if (condref == 0)  
             {  
             *errorcodeptr = ERR35;  
             goto FAILED;  
             }  
4119            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4120            }            }
4121          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
4122          set bravalue above. */            {
4123          break;            terminator = '\'';
4124              ptr++;
4125          case '=':                 /* Positive lookahead */            }
4126          bravalue = OP_ASSERT;          else
4127          ptr++;            {
4128          break;            terminator = 0;
4129              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4130              }
4131    
4132          case '!':                 /* Negative lookahead */          /* We now expect to read a name; any thing else is an error */
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
4133    
4134          case '<':                 /* Lookbehinds */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
         switch (*(++ptr))  
4135            {            {
4136            case '=':               /* Positive lookbehind */            ptr += 1;  /* To get the right offset */
4137            bravalue = OP_ASSERTBACK;            *errorcodeptr = ERR28;
4138            ptr++;            goto FAILED;
4139            break;            }
4140    
4141            case '!':               /* Negative lookbehind */          /* Read the name, but also get it as a number if it's all digits */
4142            bravalue = OP_ASSERTBACK_NOT;  
4143            recno = 0;
4144            name = ++ptr;
4145            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4146              {
4147              if (recno >= 0)
4148                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4149                  recno * 10 + *ptr - '0' : -1;
4150            ptr++;            ptr++;
           break;  
4151            }            }
4152          break;          namelen = ptr - name;
4153    
4154          case '>':                 /* One-time brackets */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4155          bravalue = OP_ONCE;            {
4156          ptr++;            ptr--;      /* Error offset */
4157          break;            *errorcodeptr = ERR26;
4158              goto FAILED;
4159              }
4160    
4161          case 'C':                 /* Callout - may be followed by digits; */          /* Do no further checking in the pre-compile phase. */
4162          previous_callout = code;  /* Save for later completion */  
4163            if (lengthptr != NULL) break;
4164    
4165            /* In the real compile we do the work of looking for the actual
4166            reference. If the string started with "+" or "-" we require the rest to
4167            be digits, in which case recno will be set. */
4168    
4169            if (refsign > 0)
4170              {
4171              if (recno <= 0)
4172                {
4173                *errorcodeptr = ERR58;
4174                goto FAILED;
4175                }
4176              if (refsign == '-')
4177                {
4178                recno = cd->bracount - recno + 1;
4179                if (recno <= 0)
4180                  {
4181                  *errorcodeptr = ERR15;
4182                  goto FAILED;
4183                  }
4184                }
4185              else recno += cd->bracount;
4186              PUT2(code, 2+LINK_SIZE, recno);
4187              break;
4188              }
4189    
4190            /* Otherwise (did not start with "+" or "-"), start by looking for the
4191            name. */
4192    
4193            slot = cd->name_table;
4194            for (i = 0; i < cd->names_found; i++)
4195              {
4196              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4197              slot += cd->name_entry_size;
4198              }
4199    
4200            /* Found a previous named subpattern */
4201    
4202            if (i < cd->names_found)
4203              {
4204              recno = GET2(slot, 0);
4205              PUT2(code, 2+LINK_SIZE, recno);
4206              }
4207    
4208            /* Search the pattern for a forward reference */
4209    
4210            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4211                            (options & PCRE_EXTENDED) != 0)) > 0)
4212              {
4213              PUT2(code, 2+LINK_SIZE, i);
4214              }
4215    
4216            /* If terminator == 0 it means that the name followed directly after
4217            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4218            some further alternatives to try. For the cases where terminator != 0
4219            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4220            now checked all the possibilities, so give an error. */
4221    
4222            else if (terminator != 0)
4223              {
4224              *errorcodeptr = ERR15;
4225              goto FAILED;
4226              }
4227    
4228            /* Check for (?(R) for recursion. Allow digits after R to specify a
4229            specific group number. */
4230    
4231            else if (*name == 'R')
4232              {
4233              recno = 0;
4234              for (i = 1; i < namelen; i++)
4235                {
4236                if ((digitab[name[i]] & ctype_digit) == 0)
4237                  {
4238                  *errorcodeptr = ERR15;
4239                  goto FAILED;
4240                  }
4241                recno = recno * 10 + name[i] - '0';
4242                }
4243              if (recno == 0) recno = RREF_ANY;
4244              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4245              PUT2(code, 2+LINK_SIZE, recno);
4246              }
4247    
4248            /* Similarly, check for the (?(DEFINE) "condition", which is always
4249            false. */
4250    
4251            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4252              {
4253              code[1+LINK_SIZE] = OP_DEF;
4254              skipbytes = 1;
4255              }
4256    
4257            /* Check for the "name" actually being a subpattern number. */
4258    
4259            else if (recno > 0)
4260              {
4261              PUT2(code, 2+LINK_SIZE, recno);
4262              }
4263    
4264            /* Either an unidentified subpattern, or a reference to (?(0) */
4265    
4266            else
4267              {
4268              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4269              goto FAILED;
4270              }
4271            break;
4272    
4273    
4274            /* ------------------------------------------------------------ */
4275            case '=':                 /* Positive lookahead */
4276            bravalue = OP_ASSERT;
4277            ptr++;
4278            break;
4279    
4280    
4281            /* ------------------------------------------------------------ */
4282            case '!':                 /* Negative lookahead */
4283            ptr++;
4284            if (*ptr == ')')          /* Optimize (?!) */
4285              {
4286              *code++ = OP_FAIL;
4287              previous = NULL;
4288              continue;
4289              }
4290            bravalue = OP_ASSERT_NOT;
4291            break;
4292    
4293    
4294            /* ------------------------------------------------------------ */
4295            case '<':                 /* Lookbehind or named define */
4296            switch (ptr[1])
4297              {
4298              case '=':               /* Positive lookbehind */
4299              bravalue = OP_ASSERTBACK;
4300              ptr += 2;
4301              break;
4302    
4303              case '!':               /* Negative lookbehind */
4304              bravalue = OP_ASSERTBACK_NOT;
4305              ptr += 2;
4306              break;
4307    
4308              default:                /* Could be name define, else bad */
4309              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4310              ptr++;                  /* Correct offset for error */
4311              *errorcodeptr = ERR24;
4312              goto FAILED;
4313              }
4314            break;
4315    
4316    
4317            /* ------------------------------------------------------------ */
4318            case '>':                 /* One-time brackets */
4319            bravalue = OP_ONCE;
4320            ptr++;
4321            break;
4322    
4323    
4324            /* ------------------------------------------------------------ */
4325            case 'C':                 /* Callout - may be followed by digits; */
4326            previous_callout = code;  /* Save for later completion */
4327          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4328          *code++ = OP_CALLOUT;     /* Already checked that the terminating */          *code++ = OP_CALLOUT;
4329            {                       /* closing parenthesis is present. */            {
4330            int n = 0;            int n = 0;
4331            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4332              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - '0';
4333              if (*ptr != ')')
4334                {
4335                *errorcodeptr = ERR39;
4336                goto FAILED;
4337                }
4338            if (n > 255)            if (n > 255)
4339              {              {
4340              *errorcodeptr = ERR38;              *errorcodeptr = ERR38;
# Line 2935  for (;; ptr++) Line 4348  for (;; ptr++)
4348          previous = NULL;          previous = NULL;
4349          continue;          continue;
4350    
4351          case 'P':                 /* Named subpattern handling */  
4352          if (*(++ptr) == '<')      /* Definition */          /* ------------------------------------------------------------ */
4353            case 'P':                 /* Python-style named subpattern handling */
4354            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
4355              {
4356              is_recurse = *ptr == '>';
4357              terminator = ')';
4358              goto NAMED_REF_OR_RECURSE;
4359              }
4360            else if (*ptr != '<')    /* Test for Python-style definition */
4361              {
4362              *errorcodeptr = ERR41;
4363              goto FAILED;
4364              }
4365            /* Fall through to handle (?P< as (?< is handled */
4366    
4367    
4368            /* ------------------------------------------------------------ */
4369            DEFINE_NAME:    /* Come here from (?< handling */
4370            case '\'':
4371            {            {
4372            int i, namelen;            terminator = (*ptr == '<')? '>' : '\'';
4373            uschar *slot = cd->name_table;            name = ++ptr;
4374            const uschar *name;     /* Don't amalgamate; some compilers */  
4375            name = ++ptr;           /* grumble at autoincrement in declaration */            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
4376              namelen = ptr - name;
4377    
4378            while (*ptr++ != '>');            /* In the pre-compile phase, just do a syntax check. */
           namelen = ptr - name - 1;  
4379    
4380            for (i = 0; i < cd->names_found; i++)            if (lengthptr != NULL)
4381              {              {
4382              int crc = memcmp(name, slot+2, namelen);              if (*ptr != terminator)
             if (crc == 0)  
4383                {                {
4384                if (slot[2+namelen] == 0)                *errorcodeptr = ERR42;
4385                  goto FAILED;
4386                  }
4387                if (cd->names_found >= MAX_NAME_COUNT)
4388                  {
4389                  *errorcodeptr = ERR49;
4390                  goto FAILED;
4391                  }
4392                if (namelen + 3 > cd->name_entry_size)
4393                  {
4394                  cd->name_entry_size = namelen + 3;
4395                  if (namelen > MAX_NAME_SIZE)
4396                  {                  {
4397                  *errorcodeptr = ERR43;                  *errorcodeptr = ERR48;
4398                  goto FAILED;                  goto FAILED;
4399                  }                  }
               crc = -1;             /* Current name is substring */  
4400                }                }
4401              if (crc < 0)              }
4402    
4403              /* In the real compile, create the entry in the table */
4404    
4405              else
4406                {
4407                slot = cd->name_table;
4408                for (i = 0; i < cd->names_found; i++)
4409                {                {
4410                memmove(slot + cd->name_entry_size, slot,                int crc = memcmp(name, slot+2, namelen);
4411                  (cd->names_found - i) * cd->name_entry_size);                if (crc == 0)
4412                break;                  {
4413                    if (slot[2+namelen] == 0)
4414                      {
4415                      if ((options & PCRE_DUPNAMES) == 0)
4416                        {
4417                        *errorcodeptr = ERR43;
4418                        goto FAILED;
4419                        }
4420                      }
4421                    else crc = -1;      /* Current name is substring */
4422                    }
4423                  if (crc < 0)
4424                    {
4425                    memmove(slot + cd->name_entry_size, slot,
4426                      (cd->names_found - i) * cd->name_entry_size);
4427                    break;
4428                    }
4429                  slot += cd->name_entry_size;
4430                }                }
             slot += cd->name_entry_size;  
             }  
4431    
4432            PUT2(slot, 0, *brackets + 1);              PUT2(slot, 0, cd->bracount + 1);
4433            memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4434            slot[2+namelen] = 0;              slot[2+namelen] = 0;
4435            cd->names_found++;              }
           goto NUMBERED_GROUP;  
4436            }            }
4437    
4438          if (*ptr == '=' || *ptr == '>')  /* Reference or recursion */          /* In both cases, count the number of names we've encountered. */
4439    
4440            ptr++;                    /* Move past > or ' */
4441            cd->names_found++;
4442            goto NUMBERED_GROUP;
4443    
4444    
4445            /* ------------------------------------------------------------ */
4446            case '&':                 /* Perl recursion/subroutine syntax */
4447            terminator = ')';
4448            is_recurse = TRUE;
4449            /* Fall through */
4450    
4451            /* We come here from the Python syntax above that handles both
4452            references (?P=name) and recursion (?P>name), as well as falling
4453            through from the Perl recursion syntax (?&name). */
4454    
4455            NAMED_REF_OR_RECURSE:
4456            name = ++ptr;
4457            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
4458            namelen = ptr - name;
4459    
4460            /* In the pre-compile phase, do a syntax check and set a dummy
4461            reference number. */
4462    
4463            if (lengthptr != NULL)
4464            {            {
4465            int i, namelen;            if (*ptr != terminator)
4466            int type = *ptr++;              {
4467            const uschar *name = ptr;              *errorcodeptr = ERR42;
4468            uschar *slot = cd->name_table;              goto FAILED;
4469                }
4470              if (namelen > MAX_NAME_SIZE)
4471                {
4472                *errorcodeptr = ERR48;
4473                goto FAILED;
4474                }
4475              recno = 0;
4476              }
4477    
4478            while (*ptr != ')') ptr++;          /* In the real compile, seek the name in the table */
           namelen = ptr - name;  
4479    
4480            else
4481              {
4482              slot = cd->name_table;
4483            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4484              {              {
4485              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4486              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4487              }              }
4488            if (i >= cd->names_found)  
4489              if (i < cd->names_found)         /* Back reference */
4490                {
4491                recno = GET2(slot, 0);
4492                }
4493              else if ((recno =                /* Forward back reference */
4494                        find_parens(ptr, cd->bracount, name, namelen,
4495                          (options & PCRE_EXTENDED) != 0)) <= 0)
4496              {              {
4497              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
4498              goto FAILED;              goto FAILED;
4499              }              }
4500              }
4501    
4502            recno = GET2(slot, 0);          /* In both phases, we can now go to the code than handles numerical
4503            recursion or backreferences. */
           if (type == '>') goto HANDLE_RECURSION;  /* A few lines below */  
   
           /* Back reference */  
4504    
4505            previous = code;          if (is_recurse) goto HANDLE_RECURSION;
4506            *code++ = OP_REF;            else goto HANDLE_REFERENCE;
           PUT2INC(code, 0, recno);  
           cd->backref_map |= (recno < 32)? (1 << recno) : 1;  
           if (recno > cd->top_backref) cd->top_backref = recno;  
           continue;  
           }  
4507    
         /* Should never happen */  
         break;  
4508    
4509          case 'R':                 /* Pattern recursion */          /* ------------------------------------------------------------ */
4510            case 'R':                 /* Recursion */
4511          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4512          /* Fall through */          /* Fall through */
4513    
         /* Recursion or "subroutine" call */  
4514    
4515          case '0': case '1': case '2': case '3': case '4':          /* ------------------------------------------------------------ */
4516          case '5': case '6': case '7': case '8': case '9':          case '-': case '+':
4517            case '0': case '1': case '2': case '3': case '4':   /* Recursion or */
4518            case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4519            {            {
4520            const uschar *called;            const uschar *called;
4521    
4522              if ((refsign = *ptr) == '+') ptr++;
4523              else if (refsign == '-')
4524                {
4525                if ((digitab[ptr[1]] & ctype_digit) == 0)
4526                  goto OTHER_CHAR_AFTER_QUERY;
4527                ptr++;
4528                }
4529    
4530            recno = 0;            recno = 0;
4531            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4532              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4533    
4534              if (*ptr != ')')
4535                {
4536                *errorcodeptr = ERR29;
4537                goto FAILED;
4538                }
4539    
4540              if (refsign == '-')
4541                {
4542                if (recno == 0)
4543                  {
4544                  *errorcodeptr = ERR58;
4545                  goto FAILED;
4546                  }
4547                recno = cd->bracount - recno + 1;
4548                if (recno <= 0)
4549                  {
4550                  *errorcodeptr = ERR15;
4551                  goto FAILED;
4552                  }
4553                }
4554              else if (refsign == '+')
4555                {
4556                if (recno == 0)
4557                  {
4558                  *errorcodeptr = ERR58;
4559                  goto FAILED;
4560                  }
4561                recno += cd->bracount;
4562                }
4563    
4564            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
4565    
4566            HANDLE_RECURSION:            HANDLE_RECURSION:
4567    
4568            previous = code;            previous = code;
4569              called = cd->start_code;
4570    
4571            /* Find the bracket that is being referenced. Temporarily end the            /* When we are actually compiling, find the bracket that is being
4572            regex in case it doesn't exist. */            referenced. Temporarily end the regex in case it doesn't exist before
4573              this point. If we end up with a forward reference, first check that
4574            *code = OP_END;            the bracket does occur later so we can give the error (and position)
4575            called = (recno == 0)?            now. Then remember this forward reference in the workspace so it can
4576              cd->start_code : find_bracket(cd->start_code, utf8, recno);            be filled in at the end. */
4577    
4578            if (called == NULL)            if (lengthptr == NULL)
4579              {              {
4580              *errorcodeptr = ERR15;              *code = OP_END;
4581              goto FAILED;              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);
             }  
4582    
4583            /* If the subpattern is still open, this is a recursive call. We              /* Forward reference */
           check to see if this is a left recursion that could loop for ever,  
           and diagnose that case. */  
4584    
4585            if (GET(called, 1) == 0 && could_be_empty(called, code, bcptr, utf8))              if (called == NULL)
4586              {                {
4587              *errorcodeptr = ERR40;                if (find_parens(ptr, cd->bracount, NULL, recno,
4588              goto FAILED;                     (options & PCRE_EXTENDED) != 0) < 0)
4589                    {
4590                    *errorcodeptr = ERR15;
4591                    goto FAILED;
4592                    }
4593                  called = cd->start_code + recno;
4594                  PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
4595                  }
4596    
4597                /* If not a forward reference, and the subpattern is still open,
4598                this is a recursive call. We check to see if this is a left
4599                recursion that could loop for ever, and diagnose that case. */
4600    
4601                else if (GET(called, 1) == 0 &&
4602                         could_be_empty(called, code, bcptr, utf8))
4603                  {
4604                  *errorcodeptr = ERR40;
4605                  goto FAILED;
4606                  }
4607              }              }
4608    
4609            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item, automatically wrapped inside
4610            "once" brackets. */            "once" brackets. Set up a "previous group" length so that a
4611              subsequent quantifier will work. */
4612    
4613            *code = OP_ONCE;            *code = OP_ONCE;
4614            PUT(code, 1, 2 + 2*LINK_SIZE);            PUT(code, 1, 2 + 2*LINK_SIZE);
# Line 3069  for (;; ptr++) Line 4621  for (;; ptr++)
4621            *code = OP_KET;            *code = OP_KET;
4622            PUT(code, 1, 2 + 2*LINK_SIZE);            PUT(code, 1, 2 + 2*LINK_SIZE);
4623            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
4624    
4625              length_prevgroup = 3 + 3*LINK_SIZE;
4626            }            }
4627    
4628            /* Can't determine a first byte now */
4629    
4630            if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4631          continue;          continue;
4632    
         /* Character after (? not specially recognized */  
4633    
4634          default:                  /* Option setting */          /* ------------------------------------------------------------ */
4635            default:              /* Other characters: check option setting */
4636            OTHER_CHAR_AFTER_QUERY:
4637          set = unset = 0;          set = unset = 0;
4638          optset = &set;          optset = &set;
4639    
# Line 3084  for (;; ptr++) Line 4643  for (;; ptr++)
4643              {              {
4644              case '-': optset = &unset; break;              case '-': optset = &unset; break;
4645    
4646                case 'J':    /* Record that it changed in the external options */
4647                *optset |= PCRE_DUPNAMES;
4648                cd->external_options |= PCRE_JCHANGED;
4649                break;
4650    
4651              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
4652              case 'm': *optset |= PCRE_MULTILINE; break;              case 'm': *optset |= PCRE_MULTILINE; break;
4653              case 's': *optset |= PCRE_DOTALL; break;              case 's': *optset |= PCRE_DOTALL; break;
4654              case 'x': *optset |= PCRE_EXTENDED; break;              case 'x': *optset |= PCRE_EXTENDED; break;
4655              case 'U': *optset |= PCRE_UNGREEDY; break;              case 'U': *optset |= PCRE_UNGREEDY; break;
4656              case 'X': *optset |= PCRE_EXTRA; break;              case 'X': *optset |= PCRE_EXTRA; break;
4657    
4658                default:  *errorcodeptr = ERR12;
4659                          ptr--;    /* Correct the offset */
4660                          goto FAILED;
4661              }              }
4662            }            }
4663    
# Line 3098  for (;; ptr++) Line 4666  for (;; ptr++)
4666          newoptions = (options | set) & (~unset);          newoptions = (options | set) & (~unset);
4667    
4668          /* If the options ended with ')' this is not the start of a nested          /* If the options ended with ')' this is not the start of a nested
4669          group with option changes, so the options change at this level. Compile          group with option changes, so the options change at this level. If this
4670          code to change the ims options if this setting actually changes any of          item is right at the start of the pattern, the options can be
4671          them. We also pass the new setting back so that it can be put at the          abstracted and made external in the pre-compile phase, and ignored in
4672          start of any following branches, and when this group ends (if we are in          the compile phase. This can be helpful when matching -- for instance in
4673          a group), a resetting item can be compiled.          caseless checking of required bytes.
4674    
4675          Note that if this item is right at the start of the pattern, the          If the code pointer is not (cd->start_code + 1 + LINK_SIZE), we are
4676          options will have been abstracted and made global, so there will be no          definitely *not* at the start of the pattern because something has been
4677          change to compile. */          compiled. In the pre-compile phase, however, the code pointer can have
4678            that value after the start, because it gets reset as code is discarded
4679            during the pre-compile. However, this can happen only at top level - if
4680            we are within parentheses, the starting BRA will still be present. At
4681            any parenthesis level, the length value can be used to test if anything
4682            has been compiled at that level. Thus, a test for both these conditions
4683            is necessary to ensure we correctly detect the start of the pattern in
4684            both phases.
4685    
4686            If we are not at the pattern start, compile code to change the ims
4687            options if this setting actually changes any of them. We also pass the
4688            new setting back so that it can be put at the start of any following
4689            branches, and when this group ends (if we are in a group), a resetting
4690            item can be compiled. */
4691    
4692          if (*ptr == ')')          if (*ptr == ')')
4693            {            {
4694            if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))            if (code == cd->start_code + 1 + LINK_SIZE &&
4695                   (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
4696              {              {
4697              *code++ = OP_OPT;              cd->external_options = newoptions;
4698              *code++ = newoptions & PCRE_IMS;              options = newoptions;
4699              }              }
4700             else
4701                {
4702                if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
4703                  {
4704                  *code++ = OP_OPT;
4705                  *code++ = newoptions & PCRE_IMS;
4706                  }
4707    
4708            /* Change options at this level, and pass them back for use              /* Change options at this level, and pass them back for use
4709            in subsequent branches. Reset the greedy defaults and the case              in subsequent branches. Reset the greedy defaults and the case
4710            value for firstbyte and reqbyte. */              value for firstbyte and reqbyte. */
4711    
4712            *optionsptr = options = newoptions;              *optionsptr = options = newoptions;
4713            greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
4714            greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
4715            req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
4716                }
4717    
4718            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
4719            continue;              /* It is complete */            continue;              /* It is complete */
# Line 3136  for (;; ptr++) Line 4726  for (;; ptr++)
4726    
4727          bravalue = OP_BRA;          bravalue = OP_BRA;
4728          ptr++;          ptr++;
4729          }          }     /* End of switch for character following (? */
4730        }        }       /* End of (? handling */
4731    
4732      /* If PCRE_NO_AUTO_CAPTURE is set, all unadorned brackets become      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,
4733      non-capturing and behave like (?:...) brackets */      all unadorned brackets become non-capturing and behave like (?:...)
4734        brackets. */
4735    
4736      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
4737        {        {
4738        bravalue = OP_BRA;        bravalue = OP_BRA;
4739        }        }
4740    
4741      /* Else we have a referencing group; adjust the opcode. If the bracket      /* Else we have a capturing group. */
     number is greater than EXTRACT_BASIC_MAX, we set the opcode one higher, and  
     arrange for the true number to follow later, in an OP_BRANUMBER item. */  
4742    
4743      else      else
4744        {        {
4745        NUMBERED_GROUP:        NUMBERED_GROUP:
4746        if (++(*brackets) > EXTRACT_BASIC_MAX)        cd->bracount += 1;
4747          {        PUT2(code, 1+LINK_SIZE, cd->bracount);
4748          bravalue = OP_BRA + EXTRACT_BASIC_MAX + 1;        skipbytes = 2;
         code[1+LINK_SIZE] = OP_BRANUMBER;  
         PUT2(code, 2+LINK_SIZE, *brackets);  
         skipbytes = 3;  
         }  
       else bravalue = OP_BRA + *brackets;  
4749        }        }
4750    
4751      /* Process nested bracketed re. Assertions may not be repeated, but other      /* Process nested bracketed regex. Assertions may not be repeated, but
4752      kinds can be. We copy code into a non-register variable in order to be able      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a
4753      to pass its address because some compilers complain otherwise. Pass in a      non-register variable in order to be able to pass its address because some
4754      new setting for the ims options if they have changed. */      compilers complain otherwise. Pass in a new setting for the ims options if
4755        they have changed. */
4756    
4757      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
4758      *code = bravalue;      *code = bravalue;
4759      tempcode = code;      tempcode = code;
4760      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;     /* Save value before bracket */
4761        length_prevgroup = 0;              /* Initialize for pre-compile phase */
4762    
4763      if (!compile_regex(      if (!compile_regex(
4764           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
4765           options & PCRE_IMS,           /* The previous ims option state */           options & PCRE_IMS,           /* The previous ims option state */
          brackets,                     /* Extracting bracket count */  
4766           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
4767           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
4768           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
4769           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
4770            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
4771           skipbytes,                    /* Skip over OP_COND/OP_BRANUMBER */           reset_bracount,               /* True if (?| group */
4772             skipbytes,                    /* Skip over bracket number */
4773           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
4774           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
4775           bcptr,                        /* Current branch chain */           bcptr,                        /* Current branch chain */
4776           cd))                          /* Tables block */           cd,                           /* Tables block */
4777             (lengthptr == NULL)? NULL :   /* Actual compile phase */
4778               &length_prevgroup           /* Pre-compile phase */
4779             ))
4780        goto FAILED;        goto FAILED;
4781    
4782      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
# Line 3196  for (;; ptr++) Line 4785  for (;; ptr++)
4785      is on the bracket. */      is on the bracket. */
4786    
4787      /* 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
4788      two branches in the group. */      two branches in the group, or just one if it's a DEFINE group. We do this
4789        in the real compile phase, not in the pre-pass, where the whole group may
4790        not be available. */
4791    
4792      else if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
4793        {        {
4794        uschar *tc = code;        uschar *tc = code;
4795        condcount = 0;        int condcount = 0;
4796    
4797        do {        do {
4798           condcount++;           condcount++;
# Line 3209  for (;; ptr++) Line 4800  for (;; ptr++)
4800           }           }
4801        while (*tc != OP_KET);        while (*tc != OP_KET);
4802    
4803        if (condcount > 2)        /* A DEFINE group is never obeyed inline (the "condition" is always
4804          false). It must have only one branch. */
4805    
4806          if (code[LINK_SIZE+1] == OP_DEF)
4807          {          {
4808          *errorcodeptr = ERR27;          if (condcount > 1)
4809          goto FAILED;            {
4810              *errorcodeptr = ERR54;
4811              goto FAILED;
4812              }
4813            bravalue = OP_DEF;   /* Just a flag to suppress char handling below */
4814          }          }
4815    
4816        /* If there is just one branch, we must not make use of its firstbyte or        /* A "normal" conditional group. If there is just one branch, we must not
4817        reqbyte, because this is equivalent to an empty second branch. */        make use of its firstbyte or reqbyte, because this is equivalent to an
4818          empty second branch. */
4819    
4820        if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;        else
4821            {
4822            if (condcount > 2)
4823              {
4824              *errorcodeptr = ERR27;
4825              goto FAILED;
4826              }
4827            if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;
4828            }
4829          }
4830    
4831        /* Error if hit end of pattern */
4832    
4833        if (*ptr != ')')
4834          {
4835          *errorcodeptr = ERR14;
4836          goto FAILED;
4837          }
4838    
4839        /* In the pre-compile phase, update the length by the length of the group,
4840        less the brackets at either end. Then reduce the compiled code to just a
4841        set of non-capturing brackets so that it doesn't use much memory if it is
4842        duplicated by a quantifier.*/
4843    
4844        if (lengthptr != NULL)
4845          {
4846          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
4847            {
4848            *errorcodeptr = ERR20;
4849            goto FAILED;
4850            }
4851          *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
4852          *code++ = OP_BRA;
4853          PUTINC(code, 0, 1 + LINK_SIZE);
4854          *code++ = OP_KET;
4855          PUTINC(code, 0, 1 + LINK_SIZE);
4856          break;    /* No need to waste time with special character handling */
4857        }        }
4858    
4859      /* Handle updating of the required and first characters. Update for normal      /* Otherwise update the main code pointer to the end of the group. */
4860      brackets of all kinds, and conditions with two branches (see code above).  
4861      If the bracket is followed by a quantifier with zero repeat, we have to      code = tempcode;
4862      back off. Hence the definition of zeroreqbyte and zerofirstbyte outside the  
4863      main loop so that they can be accessed for the back off. */      /* For a DEFINE group, required and first character settings are not
4864        relevant. */
4865    
4866        if (bravalue == OP_DEF) break;
4867    
4868        /* Handle updating of the required and first characters for other types of
4869        group. Update for normal brackets of all kinds, and conditions with two
4870        branches (see code above). If the bracket is followed by a quantifier with
4871        zero repeat, we have to back off. Hence the definition of zeroreqbyte and
4872        zerofirstbyte outside the main loop so that they can be accessed for the
4873        back off. */
4874    
4875      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4876      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
4877      groupsetfirstbyte = FALSE;      groupsetfirstbyte = FALSE;
4878    
4879      if (bravalue >= OP_BRA || bravalue == OP_ONCE || bravalue == OP_COND)      if (bravalue >= OP_ONCE)
4880        {        {
4881        /* If we have not yet set a firstbyte in this branch, take it from the        /* If we have not yet set a firstbyte in this branch, take it from the
4882        subpattern, remembering that it was set here so that a repeat of more        subpattern, remembering that it was set here so that a repeat of more
# Line 3272  for (;; ptr++) Line 4917  for (;; ptr++)
4917      firstbyte, looking for an asserted first char. */      firstbyte, looking for an asserted first char. */
4918    
4919      else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;      else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;
4920        break;     /* End of processing '(' */
4921    
     /* Now update the main code pointer to the end of the group. */  
   
     code = tempcode;  
   
     /* Error if hit end of pattern */  
4922    
4923      if (*ptr != ')')      /* ===================================================================*/
4924        {      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
       *errorcodeptr = ERR14;  
       goto FAILED;  
       }  
     break;  
   
     /* Check \ for being a real metacharacter; if not, fall through and handle  
     it as a data character at the start of a string. Escape items are checked  
     for validity in the pre-compiling pass. */  
   
     case '\\':  
     tempptr = ptr;  
     c = check_escape(&ptr, errorcodeptr, *brackets, options, FALSE);  
   
     /* Handle metacharacters introduced by \. For ones like \d, the ESC_ values  
4925      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values. For the
4926      back references, the values are ESC_REF plus the reference number. Only      back references, the values are ESC_REF plus the reference number. Only
4927      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
4928      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
4929      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
4930    
4931        case '\\':
4932        tempptr = ptr;
4933        c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
4934        if (*errorcodeptr != 0) goto FAILED;
4935    
4936      if (c < 0)      if (c < 0)
4937        {        {
4938        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 3310  for (;; ptr++) Line 4942  for (;; ptr++)
4942          continue;          continue;
4943          }          }
4944    
4945