/[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 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 97 by ph10, Mon Mar 5 12:36:47 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) 1997-2005 University of Cambridge             Copyright (c) 1997-2006 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    #define NLBLOCK cd             /* Block containing newline information */
46    #define PSSTART start_pattern  /* Field containing processed string start */
47    #define PSEND   end_pattern    /* Field containing processed string end */
48    
49    
50  #include "pcre_internal.h"  #include "pcre_internal.h"
51    
52    
53    /* When DEBUG is defined, we need the pcre_printint() function, which is also
54    used by pcretest. DEBUG is not defined when building a production library. */
55    
56    #ifdef DEBUG
57    #include "pcre_printint.src"
58    #endif
59    
60    
61  /*************************************************  /*************************************************
62  *      Code parameters and static tables         *  *      Code parameters and static tables         *
63  *************************************************/  *************************************************/
64    
65  /* 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
66  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
67  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
68  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
69  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
70    so this number is very generous.
71    
72    The same workspace is used during the second, actual compile phase for
73    remembering forward references to groups so that they can be filled in at the
74    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
75    is 4 there is plenty of room. */
76    
77  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
78    
79    
80  /* 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 63  are simple data values; negative values Line 82  are simple data values; negative values
82  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
83  is invalid. */  is invalid. */
84    
85  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
86  static const short int escapes[] = {  static const short int escapes[] = {
87       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
88       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
89     '@', -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 */
90       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */
91  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */
92  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
93     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
94       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
95  -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,      0, -ESC_w,   /* p - w */
96       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
97  };  };
98    
99  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
100  static const short int escapes[] = {  static const short int escapes[] = {
101  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
102  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 88  static const short int escapes[] = { Line 107  static const short int escapes[] = {
107  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
108  /*  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,
109  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,
110  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
111  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
112  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,
113  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
# Line 97  static const short int escapes[] = { Line 116  static const short int escapes[] = {
116  /*  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,
117  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,
118  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,
119  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
120  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,
121  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
122  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
# Line 107  static const short int escapes[] = { Line 126  static const short int escapes[] = {
126    
127    
128  /* 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
129  terminated by a zero length entry. The first three must be alpha, upper, lower,  terminated by a zero length entry. The first three must be alpha, lower, upper,
130  as this is assumed for handling case independence. */  as this is assumed for handling case independence. */
131    
132  static const char *const posix_names[] = {  static const char *const posix_names[] = {
# Line 118  static const char *const posix_names[] = Line 137  static const char *const posix_names[] =
137  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
138    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
139    
140  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
141  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
142  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
143    characters are removed, and for [:alpha:] and [:alnum:] the underscore
144    character is removed. The triples in the table consist of the base map offset,
145    second map offset or -1 if no second map, and a non-negative value for map
146    addition or a negative value for map subtraction (if there are two maps). The
147    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
148    remove vertical space characters, 2 => remove underscore. */
149    
150  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
151    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
152    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
153    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
154    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
155    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
156    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
157    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
158    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
159    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
160    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
161    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
162    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
163    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
164    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
165  };  };
166    
167    
168    #define STRING(a)  # a
169    #define XSTRING(s) STRING(s)
170    
171  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
172  are passed to the outside world. */  are passed to the outside world. Do not ever re-use any error number, because
173    they are documented. Always add a new error instead. Messages marked DEAD below
174    are no longer used. */
175    
176  static const char *error_texts[] = {  static const char *error_texts[] = {
177    "no error",    "no error",
# Line 156  static const char *error_texts[] = { Line 186  static const char *error_texts[] = {
186    "range out of order in character class",    "range out of order in character class",
187    "nothing to repeat",    "nothing to repeat",
188    /* 10 */    /* 10 */
189    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string",  /** DEAD **/
190    "internal error: unexpected repeat",    "internal error: unexpected repeat",
191    "unrecognized character after (?",    "unrecognized character after (?",
192    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class",
# Line 166  static const char *error_texts[] = { Line 196  static const char *error_texts[] = {
196    "erroffset passed as NULL",    "erroffset passed as NULL",
197    "unknown option bit(s) set",    "unknown option bit(s) set",
198    "missing ) after comment",    "missing ) after comment",
199    "parentheses nested too deeply",    "parentheses nested too deeply",  /** DEAD **/
200    /* 20 */    /* 20 */
201    "regular expression too large",    "regular expression too large",
202    "failed to get memory",    "failed to get memory",
# Line 175  static const char *error_texts[] = { Line 205  static const char *error_texts[] = {
205    "unrecognized character after (?<",    "unrecognized character after (?<",
206    /* 25 */    /* 25 */
207    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length",
208    "malformed number after (?(",    "malformed number or name after (?(",
209    "conditional group contains more than two branches",    "conditional group contains more than two branches",
210    "assertion expected after (?(",    "assertion expected after (?(",
211    "(?R or (?digits must be followed by )",    "(?R or (?digits must be followed by )",
# Line 183  static const char *error_texts[] = { Line 213  static const char *error_texts[] = {
213    "unknown POSIX class name",    "unknown POSIX class name",
214    "POSIX collating elements are not supported",    "POSIX collating elements are not supported",
215    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support",
216    "spare error",    "spare error",  /** DEAD **/
217    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large",
218    /* 35 */    /* 35 */
219    "invalid condition (?(0)",    "invalid condition (?(0)",
# Line 194  static const char *error_texts[] = { Line 224  static const char *error_texts[] = {
224    /* 40 */    /* 40 */
225    "recursive call could loop indefinitely",    "recursive call could loop indefinitely",
226    "unrecognized character after (?P",    "unrecognized character after (?P",
227    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)",
228    "two named groups have the same name",    "two named subpatterns have the same name",
229    "invalid UTF-8 string",    "invalid UTF-8 string",
230    /* 45 */    /* 45 */
231    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled",
232    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence",
233    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p",
234      "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",
235      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",
236      /* 50 */
237      "repeated subpattern is too long",
238      "octal value is greater than \\377 (not in UTF-8 mode)",
239      "internal error: overran compiling workspace",
240      "internal error: previously-checked referenced subpattern not found",
241      "DEFINE group contains more than one branch",
242      /* 55 */
243      "repeating a DEFINE group is not allowed",
244      "inconsistent NEWLINE options",
245      "\\g is not followed by an (optionally braced) non-zero number"
246  };  };
247    
248    
# Line 220  For convenience, we use the same bit def Line 262  For convenience, we use the same bit def
262    
263  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
264    
265  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
266  static const unsigned char digitab[] =  static const unsigned char digitab[] =
267    {    {
268    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 298  static const unsigned char digitab[] =
298    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
299    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
300    
301  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
302  static const unsigned char digitab[] =  static const unsigned char digitab[] =
303    {    {
304    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 312  static const unsigned char digitab[] =
312    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
313    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
314    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
315    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
316    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
317    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
318    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 346  static const unsigned char ebcdic_charta
346    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
347    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
348    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
349    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
350    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
351    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
352    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 373  static const unsigned char ebcdic_charta
373  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
374    
375  static BOOL  static BOOL
376    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,
377      int *, int *, branch_chain *, compile_data *);      int *, branch_chain *, compile_data *, int *);
378    
379    
380    
# Line 342  static BOOL Line 384  static BOOL
384    
385  /* 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
386  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
387  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
388  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
389  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,
390    ptr is pointing at the \. On exit, it is on the final character of the escape
391    sequence.
392    
393  Arguments:  Arguments:
394    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 362  static int Line 406  static int
406  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
407    int options, BOOL isclass)    int options, BOOL isclass)
408  {  {
409  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
410    const uschar *ptr = *ptrptr + 1;
411  int c, i;  int c, i;
412    
413    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
414    ptr--;                            /* Set pointer back to the last byte */
415    
416  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
417    
 c = *(++ptr);  
418  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
419    
420  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in
421  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.
422  Otherwise further processing may be required. */  Otherwise further processing may be required. */
423    
424  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
425  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
426  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
427    
428  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
429  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
430  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
431  #endif  #endif
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 435  else if ((i = escapes[c - 0x48]) != 0)
435  else  else
436    {    {
437    const uschar *oldptr;    const uschar *oldptr;
438      BOOL braced, negated;
439    
440    switch (c)    switch (c)
441      {      {
442      /* 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 401  else Line 450  else
450      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
451      break;      break;
452    
453        /* \g must be followed by a number, either plain or braced. If positive, it
454        is an absolute backreference. If negative, it is a relative backreference.
455        This is a Perl 5.10 feature. */
456    
457        case 'g':
458        if (ptr[1] == '{')
459          {
460          braced = TRUE;
461          ptr++;
462          }
463        else braced = FALSE;
464    
465        if (ptr[1] == '-')
466          {
467          negated = TRUE;
468          ptr++;
469          }
470        else negated = FALSE;
471    
472        c = 0;
473        while ((digitab[ptr[1]] & ctype_digit) != 0)
474          c = c * 10 + *(++ptr) - '0';
475    
476        if (c == 0 || (braced && *(++ptr) != '}'))
477          {
478          *errorcodeptr = ERR57;
479          return 0;
480          }
481    
482        if (negated)
483          {
484          if (c > bracount)
485            {
486            *errorcodeptr = ERR15;
487            return 0;
488            }
489          c = bracount - (c - 1);
490          }
491    
492        c = -(ESC_REF + c);
493        break;
494    
495      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
496      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
497      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 442  else Line 533  else
533        }        }
534    
535      /* \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
536      larger first octal digit. */      larger first octal digit. The original code used just to take the least
537        significant 8 bits of octal numbers (I think this is what early Perls used
538        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
539        than 3 octal digits. */
540    
541      case '0':      case '0':
542      c -= '0';      c -= '0';
543      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
544          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
545      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
546      break;      break;
547    
548      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
549      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
550        treated as a data character. */
551    
552      case 'x':      case 'x':
553  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
554        {        {
555        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
556        register int count = 0;        int count = 0;
557    
558        c = 0;        c = 0;
559        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
560          {          {
561          int cc = *pt++;          register int cc = *pt++;
562            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
563          count++;          count++;
564  #if !EBCDIC    /* ASCII coding */  
565    #ifndef EBCDIC  /* ASCII coding */
566          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
567          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
568  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
569          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
570          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
571  #endif  #endif
572          }          }
573    
574        if (*pt == '}')        if (*pt == '}')
575          {          {
576          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
577          ptr = pt;          ptr = pt;
578          break;          break;
579          }          }
580    
581        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
582        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
583        }        }
 #endif  
584    
585      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
586    
587      c = 0;      c = 0;
588      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
589        {        {
590        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
591        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
592  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
593        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
594        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
595  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
596        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
597        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
598  #endif  #endif
599        }        }
600      break;      break;
601    
602      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
603        This coding is ASCII-specific, but then the whole concept of \cx is
604        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
605    
606      case 'c':      case 'c':
607      c = *(++ptr);      c = *(++ptr);
# Line 511  else Line 611  else
611        return 0;        return 0;
612        }        }
613    
614      /* 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 */  
615      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
616      c ^= 0x40;      c ^= 0x40;
617  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
618      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
619      c ^= 0xC0;      c ^= 0xC0;
620  #endif  #endif
# Line 560  escape sequence. Line 656  escape sequence.
656  Argument:  Argument:
657    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
658    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
659      dptr           points to an int that is set to the detailed property value
660    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
661    
662  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
663  */  */
664    
665  static int  static int
666  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
667  {  {
668  int c, i, bot, top;  int c, i, bot, top;
669  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
670  char name[4];  char name[32];
671    
672  c = *(++ptr);  c = *(++ptr);
673  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
674    
675  *negptr = FALSE;  *negptr = FALSE;
676    
677  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
678  preceded by ^ for negation. */  negation. */
679    
680  if (c == '{')  if (c == '{')
681    {    {
# Line 587  if (c == '{') Line 684  if (c == '{')
684      *negptr = TRUE;      *negptr = TRUE;
685      ptr++;      ptr++;
686      }      }
687    for (i = 0; i <= 2; i++)    for (i = 0; i < sizeof(name) - 1; i++)
688      {      {
689      c = *(++ptr);      c = *(++ptr);
690      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
691      if (c == '}') break;      if (c == '}') break;
692      name[i] = c;      name[i] = c;
693      }      }
694    if (c !='}')   /* Try to distinguish error cases */    if (c !='}') goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
695    name[i] = 0;    name[i] = 0;
696    }    }
697    
# Line 619  top = _pcre_utt_size; Line 712  top = _pcre_utt_size;
712    
713  while (bot < top)  while (bot < top)
714    {    {
715    i = (bot + top)/2;    i = (bot + top) >> 1;
716    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt[i].name);
717    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
718        {
719        *dptr = _pcre_utt[i].value;
720        return _pcre_utt[i].type;
721        }
722    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
723    }    }
724    
 UNKNOWN_RETURN:  
725  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
726  *ptrptr = ptr;  *ptrptr = ptr;
727  return -1;  return -1;
# Line 698  read_repeat_counts(const uschar *p, int Line 794  read_repeat_counts(const uschar *p, int
794  int min = 0;  int min = 0;
795  int max = -1;  int max = -1;
796    
797    /* Read the minimum value and do a paranoid check: a negative value indicates
798    an integer overflow. */
799    
800  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
801    if (min < 0 || min > 65535)
802      {
803      *errorcodeptr = ERR5;
804      return p;
805      }
806    
807    /* Read the maximum value if there is one, and again do a paranoid on its size.
808    Also, max must not be less than min. */
809    
810  if (*p == '}') max = min; else  if (*p == '}') max = min; else
811    {    {
# Line 706  if (*p == '}') max = min; else Line 813  if (*p == '}') max = min; else
813      {      {
814      max = 0;      max = 0;
815      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
816        if (max < 0 || max > 65535)
817          {
818          *errorcodeptr = ERR5;
819          return p;
820          }
821      if (max < min)      if (max < min)
822        {        {
823        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 826  if (*p == '}') max = min; else
826      }      }
827    }    }
828    
829  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
830  pointer to the terminating '}'. */  '}'. */
831    
832  if (min > 65535 || max > 65535)  *minp = min;
833    *errorcodeptr = ERR5;  *maxp = max;
834  else  return p;
835    }
836    
837    
838    
839    /*************************************************
840    *       Find forward referenced subpattern       *
841    *************************************************/
842    
843    /* This function scans along a pattern's text looking for capturing
844    subpatterns, and counting them. If it finds a named pattern that matches the
845    name it is given, it returns its number. Alternatively, if the name is NULL, it
846    returns when it reaches a given numbered subpattern. This is used for forward
847    references to subpatterns. We know that if (?P< is encountered, the name will
848    be terminated by '>' because that is checked in the first pass.
849    
850    Arguments:
851      ptr          current position in the pattern
852      count        current count of capturing parens so far encountered
853      name         name to seek, or NULL if seeking a numbered subpattern
854      lorn         name length, or subpattern number if name is NULL
855      xmode        TRUE if we are in /x mode
856    
857    Returns:       the number of the named subpattern, or -1 if not found
858    */
859    
860    static int
861    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
862      BOOL xmode)
863    {
864    const uschar *thisname;
865    
866    for (; *ptr != 0; ptr++)
867    {    {
868    *minp = min;    int term;
869    *maxp = max;  
870      /* Skip over backslashed characters and also entire \Q...\E */
871    
872      if (*ptr == '\\')
873        {
874        if (*(++ptr) == 0) return -1;
875        if (*ptr == 'Q') for (;;)
876          {
877          while (*(++ptr) != 0 && *ptr != '\\');
878          if (*ptr == 0) return -1;
879          if (*(++ptr) == 'E') break;
880          }
881        continue;
882        }
883    
884      /* Skip over character classes */
885    
886      if (*ptr == '[')
887        {
888        while (*(++ptr) != ']')
889          {
890          if (*ptr == '\\')
891            {
892            if (*(++ptr) == 0) return -1;
893            if (*ptr == 'Q') for (;;)
894              {
895              while (*(++ptr) != 0 && *ptr != '\\');
896              if (*ptr == 0) return -1;
897              if (*(++ptr) == 'E') break;
898              }
899            continue;
900            }
901          }
902        continue;
903        }
904    
905      /* Skip comments in /x mode */
906    
907      if (xmode && *ptr == '#')
908        {
909        while (*(++ptr) != 0 && *ptr != '\n');
910        if (*ptr == 0) return -1;
911        continue;
912        }
913    
914      /* An opening parens must now be a real metacharacter */
915    
916      if (*ptr != '(') continue;
917      if (ptr[1] != '?')
918        {
919        count++;
920        if (name == NULL && count == lorn) return count;
921        continue;
922        }
923    
924      ptr += 2;
925      if (*ptr == 'P') ptr++;                      /* Allow optional P */
926    
927      /* We have to disambiguate (?<! and (?<= from (?<name> */
928    
929      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
930           *ptr != '\'')
931        continue;
932    
933      count++;
934    
935      if (name == NULL && count == lorn) return count;
936      term = *ptr++;
937      if (term == '<') term = '>';
938      thisname = ptr;
939      while (*ptr != term) ptr++;
940      if (name != NULL && lorn == ptr - thisname &&
941          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
942        return count;
943    }    }
944  return p;  
945    return -1;
946  }  }
947    
948    
# Line 778  for (;;) Line 996  for (;;)
996    
997      case OP_CALLOUT:      case OP_CALLOUT:
998      case OP_CREF:      case OP_CREF:
999      case OP_BRANUMBER:      case OP_RREF:
1000        case OP_DEF:
1001      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1002      break;      break;
1003    
# Line 823  for (;;) Line 1042  for (;;)
1042    {    {
1043    int d;    int d;
1044    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
1045    
1046    switch (op)    switch (op)
1047      {      {
1048        case OP_CBRA:
1049      case OP_BRA:      case OP_BRA:
1050      case OP_ONCE:      case OP_ONCE:
1051      case OP_COND:      case OP_COND:
1052      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1053      if (d < 0) return d;      if (d < 0) return d;
1054      branchlength += d;      branchlength += d;
1055      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1084  for (;;)
1084      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1085    
1086      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1087      case OP_CREF:      case OP_CREF:
1088        case OP_RREF:
1089        case OP_DEF:
1090      case OP_OPT:      case OP_OPT:
1091      case OP_CALLOUT:      case OP_CALLOUT:
1092      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1104  for (;;)
1104    
1105      case OP_CHAR:      case OP_CHAR:
1106      case OP_CHARNC:      case OP_CHARNC:
1107        case OP_NOT:
1108      branchlength++;      branchlength++;
1109      cc += 2;      cc += 2;
1110  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 917  for (;;) Line 1138  for (;;)
1138    
1139      case OP_PROP:      case OP_PROP:
1140      case OP_NOTPROP:      case OP_NOTPROP:
1141      cc++;      cc += 2;
1142      /* Fall through */      /* Fall through */
1143    
1144      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 998  Returns: pointer to the opcode for Line 1219  Returns: pointer to the opcode for
1219  static const uschar *  static const uschar *
1220  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1221  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1222  for (;;)  for (;;)
1223    {    {
1224    register int c = *code;    register int c = *code;
1225    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1226    else if (c > OP_BRA)  
1227      /* XCLASS is used for classes that cannot be represented just by a bit
1228      map. This includes negated single high-valued characters. The length in
1229      the table is zero; the actual length is stored in the compiled code. */
1230    
1231      if (c == OP_XCLASS) code += GET(code, 1);
1232    
1233      /* Handle capturing bracket */
1234    
1235      else if (c == OP_CBRA)
1236      {      {
1237      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1238      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1239      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1240      }      }
1241    
1242      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1243      a multi-byte character. The length in the table is a minimum, so we have to
1244      arrange to skip the extra bytes. */
1245    
1246    else    else
1247      {      {
1248      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
   
 #ifdef SUPPORT_UTF8  
   
     /* In UTF-8 mode, opcodes that are followed by a character may be followed  
     by a multi-byte character. The length in the table is a minimum, so we have  
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
   
1249      if (utf8) switch(c)      if (utf8) switch(c)
1250        {        {
1251        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1253  for (;;)
1253        case OP_EXACT:        case OP_EXACT:
1254        case OP_UPTO:        case OP_UPTO:
1255        case OP_MINUPTO:        case OP_MINUPTO:
1256          case OP_POSUPTO:
1257        case OP_STAR:        case OP_STAR:
1258        case OP_MINSTAR:        case OP_MINSTAR:
1259          case OP_POSSTAR:
1260        case OP_PLUS:        case OP_PLUS:
1261        case OP_MINPLUS:        case OP_MINPLUS:
1262          case OP_POSPLUS:
1263        case OP_QUERY:        case OP_QUERY:
1264        case OP_MINQUERY:        case OP_MINQUERY:
1265        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1266        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;  
1267        break;        break;
1268        }        }
 #endif  
1269      }      }
1270    }    }
1271  }  }
# Line 1072  Returns: pointer to the opcode for Line 1289  Returns: pointer to the opcode for
1289  static const uschar *  static const uschar *
1290  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1291  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1292  for (;;)  for (;;)
1293    {    {
1294    register int c = *code;    register int c = *code;
1295    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1296    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1297    else if (c > OP_BRA)  
1298      {    /* XCLASS is used for classes that cannot be represented just by a bit
1299      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1300      }    the table is zero; the actual length is stored in the compiled code. */
1301    
1302      if (c == OP_XCLASS) code += GET(code, 1);
1303    
1304      /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes
1305      that are followed by a character may be followed by a multi-byte character.
1306      The length in the table is a minimum, so we have to arrange to skip the extra
1307      bytes. */
1308    
1309    else    else
1310      {      {
1311      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
   
 #ifdef SUPPORT_UTF8  
   
     /* In UTF-8 mode, opcodes that are followed by a character may be followed  
     by a multi-byte character. The length in the table is a minimum, so we have  
     to scan along to skip the extra bytes. All opcodes are less than 128, so we  
     can use relatively efficient code. */  
   
1312      if (utf8) switch(c)      if (utf8) switch(c)
1313        {        {
1314        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1316  for (;;)
1316        case OP_EXACT:        case OP_EXACT:
1317        case OP_UPTO:        case OP_UPTO:
1318        case OP_MINUPTO:        case OP_MINUPTO:
1319          case OP_POSUPTO:
1320        case OP_STAR:        case OP_STAR:
1321        case OP_MINSTAR:        case OP_MINSTAR:
1322          case OP_POSSTAR:
1323        case OP_PLUS:        case OP_PLUS:
1324        case OP_MINPLUS:        case OP_MINPLUS:
1325          case OP_POSPLUS:
1326        case OP_QUERY:        case OP_QUERY:
1327        case OP_MINQUERY:        case OP_MINQUERY:
1328        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1329        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;  
1330        break;        break;
1331        }        }
 #endif  
1332      }      }
1333    }    }
1334  }  }
# Line 1132  for (;;) Line 1340  for (;;)
1340  *************************************************/  *************************************************/
1341    
1342  /* 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
1343  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()
1344  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
1345  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
1346  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1347    struck an inner bracket whose current branch will already have been scanned.
1348    
1349  Arguments:  Arguments:
1350    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1358  static BOOL
1358  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1359  {  {
1360  register int c;  register int c;
1361  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);
1362       code < endcode;       code < endcode;
1363       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1364    {    {
# Line 1157  for (code = first_significant_code(code Line 1366  for (code = first_significant_code(code
1366    
1367    c = *code;    c = *code;
1368    
1369    if (c >= OP_BRA)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)
1370      {      {
1371      BOOL empty_branch;      BOOL empty_branch;
1372      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1173  for (code = first_significant_code(code Line 1382  for (code = first_significant_code(code
1382        }        }
1383      while (*code == OP_ALT);      while (*code == OP_ALT);
1384      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
1385      code += 1 + LINK_SIZE;  
1386      c = *code;      /* Move past the KET and fudge things so that the increment in the "for"
1387        above has no effect. */
1388    
1389        c = OP_END;
1390        code += 1 + LINK_SIZE - _pcre_OP_lengths[c];
1391        continue;
1392      }      }
1393    
1394    else switch (c)    /* Handle the other opcodes */
1395    
1396      switch (c)
1397      {      {
1398      /* Check for quantifiers after a class */      /* Check for quantifiers after a class */
1399    
# Line 1233  for (code = first_significant_code(code Line 1449  for (code = first_significant_code(code
1449      case OP_NOT:      case OP_NOT:
1450      case OP_PLUS:      case OP_PLUS:
1451      case OP_MINPLUS:      case OP_MINPLUS:
1452        case OP_POSPLUS:
1453      case OP_EXACT:      case OP_EXACT:
1454      case OP_NOTPLUS:      case OP_NOTPLUS:
1455      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1456        case OP_NOTPOSPLUS:
1457      case OP_NOTEXACT:      case OP_NOTEXACT:
1458      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1459      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1460        case OP_TYPEPOSPLUS:
1461      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1462      return FALSE;      return FALSE;
1463    
# Line 1250  for (code = first_significant_code(code Line 1469  for (code = first_significant_code(code
1469      case OP_ALT:      case OP_ALT:
1470      return TRUE;      return TRUE;
1471    
1472      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1473      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1474    
1475  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1476      case OP_STAR:      case OP_STAR:
1477      case OP_MINSTAR:      case OP_MINSTAR:
1478        case OP_POSSTAR:
1479      case OP_QUERY:      case OP_QUERY:
1480      case OP_MINQUERY:      case OP_MINQUERY:
1481        case OP_POSQUERY:
1482      case OP_UPTO:      case OP_UPTO:
1483      case OP_MINUPTO:      case OP_MINUPTO:
1484        case OP_POSUPTO:
1485      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1486      break;      break;
1487  #endif  #endif
# Line 1377  earlier groups that are outside the curr Line 1599  earlier groups that are outside the curr
1599  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
1600  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
1601  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
1602  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,
1603  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1604    
1605    This function has been extended with the possibility of forward references for
1606    recursions and subroutine calls. It must also check the list of such references
1607    for the group we are dealing with. If it finds that one of the recursions in
1608    the current group is on this list, it adjusts the offset in the list, not the
1609    value in the reference (which is a group number).
1610    
1611  Arguments:  Arguments:
1612    group      points to the start of the group    group      points to the start of the group
1613    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1614    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1615    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1616      save_hwm   the hwm forward reference pointer at the start of the group
1617    
1618  Returns:     nothing  Returns:     nothing
1619  */  */
1620    
1621  static void  static void
1622  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1623      uschar *save_hwm)
1624  {  {
1625  uschar *ptr = group;  uschar *ptr = group;
1626  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1627    {    {
1628    int offset = GET(ptr, 1);    int offset;
1629    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1630    
1631      /* See if this recursion is on the forward reference list. If so, adjust the
1632      reference. */
1633    
1634      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1635        {
1636        offset = GET(hc, 0);
1637        if (cd->start_code + offset == ptr + 1)
1638          {
1639          PUT(hc, 0, offset + adjust);
1640          break;
1641          }
1642        }
1643    
1644      /* Otherwise, adjust the recursion offset if it's after the start of this
1645      group. */
1646    
1647      if (hc >= cd->hwm)
1648        {
1649        offset = GET(ptr, 1);
1650        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1651        }
1652    
1653    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1654    }    }
1655  }  }
# Line 1475  Yield: TRUE when range returned; Line 1728  Yield: TRUE when range returned;
1728  */  */
1729    
1730  static BOOL  static BOOL
1731  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1732      unsigned int *odptr)
1733  {  {
1734  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1735    
1736  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1737    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1738    
1739  if (c > d) return FALSE;  if (c > d) return FALSE;
1740    
# Line 1492  next = othercase + 1; Line 1743  next = othercase + 1;
1743    
1744  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1745    {    {
1746    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1747    next++;    next++;
1748    }    }
1749    
# Line 1506  return TRUE; Line 1755  return TRUE;
1755  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1756    
1757    
1758    
1759  /*************************************************  /*************************************************
1760  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1761  *************************************************/  *************************************************/
1762    
1763  /* 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
1764  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
1765  bits.  sense to automatically possessify the repeated item.
1766    
1767  Arguments:  Arguments:
1768    optionsptr     pointer to the option bits    op_code       the repeated op code
1769    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1770    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1771    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1772    errorcodeptr   points to error code variable    ptr           next character in pattern
1773    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1774    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.  
1775    
1776  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1777  */  */
1778    
1779  static BOOL  static BOOL
1780  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1781    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1782  {  {
1783  int repeat_type, op_type;  int next;
1784  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
1785  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
1786  int greedy_default, greedy_non_default;  
1787  int firstbyte, reqbyte;  if ((options & PCRE_EXTENDED) != 0)
1788  int zeroreqbyte, zerofirstbyte;    {
1789  int req_caseopt, reqvary, tempreqvary;    for (;;)
1790  int condcount = 0;      {
1791  int options = *optionsptr;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1792  int after_manual_callout = 0;      if (*ptr == '#')
1793  register int c;        {
1794  register uschar *code = *codeptr;        while (*(++ptr) != 0)
1795  uschar *tempcode;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1796  BOOL inescq = FALSE;        }
1797  BOOL groupsetfirstbyte = FALSE;      else break;
1798  const uschar *ptr = *ptrptr;      }
1799  const uschar *tempptr;    }
 uschar *previous = NULL;  
 uschar *previous_callout = NULL;  
 uschar classbits[32];  
1800    
1801    /* If the next item is one that we can handle, get its value. A non-negative
1802    value is a character, a negative value is an escape value. */
1803    
1804    if (*ptr == '\\')
1805      {
1806      int temperrorcode = 0;
1807      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
1808      if (temperrorcode != 0) return FALSE;
1809      ptr++;    /* Point after the escape sequence */
1810      }
1811    
1812    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
1813      {
1814  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1815  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;  
1816  #endif  #endif
1817      next = *ptr++;
1818      }
1819    
1820  /* Set up the default and non-default settings for greediness */  else return FALSE;
1821    
1822  greedy_default = ((options & PCRE_UNGREEDY) != 0);  /* Skip whitespace and comments in extended mode */
 greedy_non_default = greedy_default ^ 1;  
1823    
1824  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  if ((options & PCRE_EXTENDED) != 0)
1825  matching encountered yet". It gets changed to REQ_NONE if we hit something that    {
1826  matches a non-fixed char first char; reqbyte just remains unset if we never    for (;;)
1827  find one.      {
1828        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1829        if (*ptr == '#')
1830          {
1831          while (*(++ptr) != 0)
1832            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1833          }
1834        else break;
1835        }
1836      }
1837    
1838  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. */  
1839    
1840  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
1841      return FALSE;
1842    
1843  /* 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
1844  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
1845  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
1846  case status of the value. This is used only for ASCII characters. */  utf8_char. */
1847    
 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  
1848    
1849  /* Switch on next character until the end of the branch */  /* Handle cases when the next item is a character. */
1850    
1851  for (;; ptr++)  if (next >= 0) switch(op_code)
1852    {    {
1853    BOOL negate_class;    case OP_CHAR:
1854    BOOL possessive_quantifier;  #ifdef SUPPORT_UTF8
1855    BOOL is_quantifier;    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
1856    int class_charcount;  #endif
1857    int class_lastchar;    return item != next;
1858    int newoptions;  
1859    int recno;    /* For CHARNC (caseless character) we must check the other case. If we have
1860    int skipbytes;    Unicode property support, we can use it to test the other case of
1861    int subreqbyte;    high-valued characters. */
1862    int subfirstbyte;  
1863    int mclength;    case OP_CHARNC:
1864    #ifdef SUPPORT_UTF8
1865      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
1866    #endif
1867      if (item == next) return FALSE;
1868    #ifdef SUPPORT_UTF8
1869      if (utf8)
1870        {
1871        unsigned int othercase;
1872        if (next < 128) othercase = cd->fcc[next]; else
1873    #ifdef SUPPORT_UCP
1874        othercase = _pcre_ucp_othercase((unsigned int)next);
1875    #else
1876        othercase = NOTACHAR;
1877    #endif
1878        return (unsigned int)item != othercase;
1879        }
1880      else
1881    #endif  /* SUPPORT_UTF8 */
1882      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
1883    
1884      /* For OP_NOT, "item" must be a single-byte character. */
1885    
1886      case OP_NOT:
1887      if (next < 0) return FALSE;  /* Not a character */
1888      if (item == next) return TRUE;
1889      if ((options & PCRE_CASELESS) == 0) return FALSE;
1890    #ifdef SUPPORT_UTF8
1891      if (utf8)
1892        {
1893        unsigned int othercase;
1894        if (next < 128) othercase = cd->fcc[next]; else
1895    #ifdef SUPPORT_UCP
1896        othercase = _pcre_ucp_othercase(next);
1897    #else
1898        othercase = NOTACHAR;
1899    #endif
1900        return (unsigned int)item == othercase;
1901        }
1902      else
1903    #endif  /* SUPPORT_UTF8 */
1904      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
1905    
1906      case OP_DIGIT:
1907      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
1908    
1909      case OP_NOT_DIGIT:
1910      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
1911    
1912      case OP_WHITESPACE:
1913      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
1914    
1915      case OP_NOT_WHITESPACE:
1916      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
1917    
1918      case OP_WORDCHAR:
1919      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
1920    
1921      case OP_NOT_WORDCHAR:
1922      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
1923    
1924      default:
1925      return FALSE;
1926      }
1927    
1928    
1929    /* Handle the case when the next item is \d, \s, etc. */
1930    
1931    switch(op_code)
1932      {
1933      case OP_CHAR:
1934      case OP_CHARNC:
1935    #ifdef SUPPORT_UTF8
1936      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
1937    #endif
1938      switch(-next)
1939        {
1940        case ESC_d:
1941        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
1942    
1943        case ESC_D:
1944        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
1945    
1946        case ESC_s:
1947        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
1948    
1949        case ESC_S:
1950        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
1951    
1952        case ESC_w:
1953        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
1954    
1955        case ESC_W:
1956        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
1957    
1958        default:
1959        return FALSE;
1960        }
1961    
1962      case OP_DIGIT:
1963      return next == -ESC_D || next == -ESC_s || next == -ESC_W;
1964    
1965      case OP_NOT_DIGIT:
1966      return next == -ESC_d;
1967    
1968      case OP_WHITESPACE:
1969      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
1970    
1971      case OP_NOT_WHITESPACE:
1972      return next == -ESC_s;
1973    
1974      case OP_WORDCHAR:
1975      return next == -ESC_W || next == -ESC_s;
1976    
1977      case OP_NOT_WORDCHAR:
1978      return next == -ESC_w || next == -ESC_d;
1979    
1980      default:
1981      return FALSE;
1982      }
1983    
1984    /* Control does not reach here */
1985    }
1986    
1987    
1988    
1989    /*************************************************
1990    *           Compile one branch                   *
1991    *************************************************/
1992    
1993    /* Scan the pattern, compiling it into the a vector. If the options are
1994    changed during the branch, the pointer is used to change the external options
1995    bits. This function is used during the pre-compile phase when we are trying
1996    to find out the amount of memory needed, as well as during the real compile
1997    phase. The value of lengthptr distinguishes the two phases.
1998    
1999    Arguments:
2000      optionsptr     pointer to the option bits
2001      codeptr        points to the pointer to the current code point
2002      ptrptr         points to the current pattern pointer
2003      errorcodeptr   points to error code variable
2004      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2005      reqbyteptr     set to the last literal character required, else < 0
2006      bcptr          points to current branch chain
2007      cd             contains pointers to tables etc.
2008      lengthptr      NULL during the real compile phase
2009                     points to length accumulator during pre-compile phase
2010    
2011    Returns:         TRUE on success
2012                     FALSE, with *errorcodeptr set non-zero on error
2013    */
2014    
2015    static BOOL
2016    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2017      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2018      compile_data *cd, int *lengthptr)
2019    {
2020    int repeat_type, op_type;
2021    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2022    int bravalue = 0;
2023    int greedy_default, greedy_non_default;
2024    int firstbyte, reqbyte;
2025    int zeroreqbyte, zerofirstbyte;
2026    int req_caseopt, reqvary, tempreqvary;
2027    int options = *optionsptr;
2028    int after_manual_callout = 0;
2029    int length_prevgroup = 0;
2030    register int c;
2031    register uschar *code = *codeptr;
2032    uschar *last_code = code;
2033    uschar *orig_code = code;
2034    uschar *tempcode;
2035    BOOL inescq = FALSE;
2036    BOOL groupsetfirstbyte = FALSE;
2037    const uschar *ptr = *ptrptr;
2038    const uschar *tempptr;
2039    uschar *previous = NULL;
2040    uschar *previous_callout = NULL;
2041    uschar *save_hwm = NULL;
2042    uschar classbits[32];
2043    
2044    #ifdef SUPPORT_UTF8
2045    BOOL class_utf8;
2046    BOOL utf8 = (options & PCRE_UTF8) != 0;
2047    uschar *class_utf8data;
2048    uschar utf8_char[6];
2049    #else
2050    BOOL utf8 = FALSE;
2051    uschar *utf8_char = NULL;
2052    #endif
2053    
2054    #ifdef DEBUG
2055    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2056    #endif
2057    
2058    /* Set up the default and non-default settings for greediness */
2059    
2060    greedy_default = ((options & PCRE_UNGREEDY) != 0);
2061    greedy_non_default = greedy_default ^ 1;
2062    
2063    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2064    matching encountered yet". It gets changed to REQ_NONE if we hit something that
2065    matches a non-fixed char first char; reqbyte just remains unset if we never
2066    find one.
2067    
2068    When we hit a repeat whose minimum is zero, we may have to adjust these values
2069    to take the zero repeat into account. This is implemented by setting them to
2070    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2071    item types that can be repeated set these backoff variables appropriately. */
2072    
2073    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
2074    
2075    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
2076    according to the current setting of the caseless flag. REQ_CASELESS is a bit
2077    value > 255. It is added into the firstbyte or reqbyte variables to record the
2078    case status of the value. This is used only for ASCII characters. */
2079    
2080    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
2081    
2082    /* Switch on next character until the end of the branch */
2083    
2084    for (;; ptr++)
2085      {
2086      BOOL negate_class;
2087      BOOL possessive_quantifier;
2088      BOOL is_quantifier;
2089      BOOL is_recurse;
2090      int class_charcount;
2091      int class_lastchar;
2092      int newoptions;
2093      int recno;
2094      int skipbytes;
2095      int subreqbyte;
2096      int subfirstbyte;
2097      int terminator;
2098      int mclength;
2099    uschar mcbuffer[8];    uschar mcbuffer[8];
2100    
2101    /* Next byte in the pattern */    /* Get next byte in the pattern */
2102    
2103    c = *ptr;    c = *ptr;
2104    
2105      /* If we are in the pre-compile phase, accumulate the length used for the
2106      previous cycle of this loop. */
2107    
2108      if (lengthptr != NULL)
2109        {
2110    #ifdef DEBUG
2111        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2112    #endif
2113        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2114          {
2115          *errorcodeptr = ERR52;
2116          goto FAILED;
2117          }
2118    
2119        /* There is at least one situation where code goes backwards: this is the
2120        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2121        the class is simply eliminated. However, it is created first, so we have to
2122        allow memory for it. Therefore, don't ever reduce the length at this point.
2123        */
2124    
2125        if (code < last_code) code = last_code;
2126        *lengthptr += code - last_code;
2127        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2128    
2129        /* If "previous" is set and it is not at the start of the work space, move
2130        it back to there, in order to avoid filling up the work space. Otherwise,
2131        if "previous" is NULL, reset the current code pointer to the start. */
2132    
2133        if (previous != NULL)
2134          {
2135          if (previous > orig_code)
2136            {
2137            memmove(orig_code, previous, code - previous);
2138            code -= previous - orig_code;
2139            previous = orig_code;
2140            }
2141          }
2142        else code = orig_code;
2143    
2144        /* Remember where this code item starts so we can pick up the length
2145        next time round. */
2146    
2147        last_code = code;
2148        }
2149    
2150      /* In the real compile phase, just check the workspace used by the forward
2151      reference list. */
2152    
2153      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2154        {
2155        *errorcodeptr = ERR52;
2156        goto FAILED;
2157        }
2158    
2159    /* 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 */
2160    
2161    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1623  for (;; ptr++) Line 2170  for (;; ptr++)
2170        {        {
2171        if (previous_callout != NULL)        if (previous_callout != NULL)
2172          {          {
2173          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2174              complete_callout(previous_callout, ptr, cd);
2175          previous_callout = NULL;          previous_callout = NULL;
2176          }          }
2177        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2192  for (;; ptr++)
2192    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2193         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2194      {      {
2195      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2196          complete_callout(previous_callout, ptr, cd);
2197      previous_callout = NULL;      previous_callout = NULL;
2198      }      }
2199    
# Line 1655  for (;; ptr++) Line 2204  for (;; ptr++)
2204      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2205      if (c == '#')      if (c == '#')
2206        {        {
2207        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2208        on the Macintosh. */          {
2209        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2210        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2211          if (*ptr != 0) continue;
2212    
2213          /* Else fall through to handle end of string */
2214          c = 0;
2215        }        }
2216      }      }
2217    
# Line 1672  for (;; ptr++) Line 2225  for (;; ptr++)
2225    
2226    switch(c)    switch(c)
2227      {      {
2228      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2229        case 0:                        /* The branch terminates at string end */
2230      case 0:      case '|':                      /* or | or ) */
     case '|':  
2231      case ')':      case ')':
2232      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2233      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2234      *codeptr = code;      *codeptr = code;
2235      *ptrptr = ptr;      *ptrptr = ptr;
2236        if (lengthptr != NULL)
2237          {
2238          *lengthptr += code - last_code;   /* To include callout length */
2239          DPRINTF((">> end branch\n"));
2240          }
2241      return TRUE;      return TRUE;
2242    
2243    
2244        /* ===================================================================*/
2245      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2246      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2247    
# Line 1711  for (;; ptr++) Line 2270  for (;; ptr++)
2270      *code++ = OP_ANY;      *code++ = OP_ANY;
2271      break;      break;
2272    
2273      /* Character classes. If the included characters are all < 255 in value, we  
2274      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2275      case where there is only one such character. For negated classes, we build      /* Character classes. If the included characters are all < 256, we build a
2276      the map as usual, then invert it at the end. However, we use a different      32-byte bitmap of the permitted characters, except in the special case
2277      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2278        map as usual, then invert it at the end. However, we use a different opcode
2279        so that data characters > 255 can be handled correctly.
2280    
2281      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2282      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
# Line 1749  for (;; ptr++) Line 2310  for (;; ptr++)
2310        }        }
2311    
2312      /* 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
2313      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
2314      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2315    
2316      class_charcount = 0;      class_charcount = 0;
2317      class_lastchar = -1;      class_lastchar = -1;
2318    
2319        /* Initialize the 32-char bit map to all zeros. We build the map in a
2320        temporary bit of memory, in case the class contains only 1 character (less
2321        than 256), because in that case the compiled code doesn't use the bit map.
2322        */
2323    
2324        memset(classbits, 0, 32 * sizeof(uschar));
2325    
2326  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2327      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2328      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2329  #endif  #endif
2330    
     /* 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));  
   
2331      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2332      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
2333      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. */  
2334    
2335      do      if (c != 0) do
2336        {        {
2337          const uschar *oldptr;
2338    
2339  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2340        if (utf8 && c > 127)        if (utf8 && c > 127)
2341          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1786  for (;; ptr++) Line 2347  for (;; ptr++)
2347    
2348        if (inescq)        if (inescq)
2349          {          {
2350          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2351            {            {
2352            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2353            ptr++;            ptr++;                            /* Skip the 'E' */
2354            continue;            continue;                         /* Carry on with next */
2355            }            }
2356          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2357          }          }
2358    
2359        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1806  for (;; ptr++) Line 2367  for (;; ptr++)
2367            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2368          {          {
2369          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2370          int posix_class, i;          int posix_class, taboffset, tabopt;
2371          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2372            uschar pbits[32];
2373    
2374          if (ptr[1] != ':')          if (ptr[1] != ':')
2375            {            {
# Line 1836  for (;; ptr++) Line 2398  for (;; ptr++)
2398          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2399            posix_class = 0;            posix_class = 0;
2400    
2401          /* Or into the map we are building up to 3 of the static class          /* We build the bit map for the POSIX class in a chunk of local store
2402          tables, or their negations. The [:blank:] class sets up the same          because we may be adding and subtracting from it, and we don't want to
2403          chars as the [:space:] class (all white space). We remove the vertical          subtract bits that may be in the main map already. At the end we or the
2404          white space chars afterwards. */          result into the bit map that is being built. */
2405    
2406          posix_class *= 3;          posix_class *= 3;
2407          for (i = 0; i < 3; i++)  
2408            /* Copy in the first table (always present) */
2409    
2410            memcpy(pbits, cbits + posix_class_maps[posix_class],
2411              32 * sizeof(uschar));
2412    
2413            /* If there is a second table, add or remove it as required. */
2414    
2415            taboffset = posix_class_maps[posix_class + 1];
2416            tabopt = posix_class_maps[posix_class + 2];
2417    
2418            if (taboffset >= 0)
2419            {            {
2420            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2421            int taboffset = posix_class_maps[posix_class + i];              for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
           if (taboffset < 0) break;  
           if (local_negate)  
             {  
             if (i == 0)  
               for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];  
             else  
               for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];  
             if (blankclass) classbits[1] |= 0x3c;  
             }  
2422            else            else
2423              {              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
2424            }            }
2425    
2426            /* Not see if we need to remove any special characters. An option
2427            value of 1 removes vertical space and 2 removes underscore. */
2428    
2429            if (tabopt < 0) tabopt = -tabopt;
2430            if (tabopt == 1) pbits[1] &= ~0x3c;
2431              else if (tabopt == 2) pbits[11] &= 0x7f;
2432    
2433            /* Add the POSIX table or its complement into the main table that is
2434            being built and we are done. */
2435    
2436            if (local_negate)
2437              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2438            else
2439              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2440    
2441          ptr = tempptr + 1;          ptr = tempptr + 1;
2442          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2443          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2444          }          }
2445    
2446        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2447        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
2448        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.
2449        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2450        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  
2451        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2452    
2453        if (c == '\\')        if (c == '\\')
2454          {          {
2455          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2456            if (*errorcodeptr != 0) goto FAILED;
2457    
2458          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2459          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 */
2460            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2461          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2462            {            {
2463            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1895  for (;; ptr++) Line 2472  for (;; ptr++)
2472            {            {
2473            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2474            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2475            switch (-c)  
2476              /* Save time by not doing this in the pre-compile phase. */
2477    
2478              if (lengthptr == NULL) switch (-c)
2479              {              {
2480              case ESC_d:              case ESC_d:
2481              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 1923  for (;; ptr++) Line 2503  for (;; ptr++)
2503              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2504              continue;              continue;
2505    
2506                case ESC_E: /* Perl ignores an orphan \E */
2507                continue;
2508    
2509                default:    /* Not recognized; fall through */
2510                break;      /* Need "default" setting to stop compiler warning. */
2511                }
2512    
2513              /* In the pre-compile phase, just do the recognition. */
2514    
2515              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2516                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2517    
2518              /* We need to deal with \P and \p in both phases. */
2519    
2520  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2521              case ESC_p:            if (-c == ESC_p || -c == ESC_P)
2522              case ESC_P:              {
2523                {              BOOL negated;
2524                BOOL negated;              int pdata;
2525                int property = get_ucp(&ptr, &negated, errorcodeptr);              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2526                if (property < 0) goto FAILED;              if (ptype < 0) goto FAILED;
2527                class_utf8 = TRUE;              class_utf8 = TRUE;
2528                *class_utf8data++ = ((-c == ESC_p) != negated)?              *class_utf8data++ = ((-c == ESC_p) != negated)?
2529                  XCL_PROP : XCL_NOTPROP;                XCL_PROP : XCL_NOTPROP;
2530                *class_utf8data++ = property;              *class_utf8data++ = ptype;
2531                class_charcount -= 2;   /* Not a < 256 character */              *class_utf8data++ = pdata;
2532                }              class_charcount -= 2;   /* Not a < 256 character */
2533              continue;              continue;
2534                }
2535  #endif  #endif
2536              /* Unrecognized escapes are faulted if PCRE is running in its
2537              strict mode. By default, for compatibility with Perl, they are
2538              treated as literals. */
2539    
2540              /* Unrecognized escapes are faulted if PCRE is running in its            if ((options & PCRE_EXTRA) != 0)
2541              strict mode. By default, for compatibility with Perl, they are              {
2542              treated as literals. */              *errorcodeptr = ERR7;
2543                goto FAILED;
             default:  
             if ((options & PCRE_EXTRA) != 0)  
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
2544              }              }
2545    
2546              class_charcount -= 2;  /* Undo the default count from above */
2547              c = *ptr;              /* Get the final character and fall through */
2548            }            }
2549    
2550          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
2551          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
2552    
2553          }   /* End of backslash handling */          }   /* End of backslash handling */
2554    
2555        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
2556        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
2557        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
2558          entirely. The code for handling \Q and \E is messy. */
2559    
2560          CHECK_RANGE:
2561          while (ptr[1] == '\\' && ptr[2] == 'E')
2562            {
2563            inescq = FALSE;
2564            ptr += 2;
2565            }
2566    
2567          oldptr = ptr;
2568    
2569        if (ptr[1] == '-' && ptr[2] != ']')        if (!inescq && ptr[1] == '-')
2570          {          {
2571          int d;          int d;
2572          ptr += 2;          ptr += 2;
2573            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
2574    
2575            /* If we hit \Q (not followed by \E) at this point, go into escaped
2576            mode. */
2577    
2578            while (*ptr == '\\' && ptr[1] == 'Q')
2579              {
2580              ptr += 2;
2581              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
2582              inescq = TRUE;
2583              break;
2584              }
2585    
2586            if (*ptr == 0 || (!inescq && *ptr == ']'))
2587              {
2588              ptr = oldptr;
2589              goto LONE_SINGLE_CHARACTER;
2590              }
2591    
2592  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2593          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 2602  for (;; ptr++)
2602          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
2603          in such circumstances. */          in such circumstances. */
2604    
2605          if (d == '\\')          if (!inescq && d == '\\')
2606            {            {
2607            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2608            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
2609    
2610            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backslash; \X is literal X; \R is literal R; any other
2611            was literal */            special means the '-' was literal */
2612    
2613            if (d < 0)            if (d < 0)
2614              {              {
2615              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
2616              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
2617                else if (d == -ESC_R) d = 'R'; else
2618                {                {
2619                ptr = oldptr - 2;                ptr = oldptr;
2620                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
2621                }                }
2622              }              }
2623            }            }
2624    
2625          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
2626          the pre-pass. Optimize one-character ranges */          one-character ranges */
2627    
2628            if (d < c)
2629              {
2630              *errorcodeptr = ERR8;
2631              goto FAILED;
2632              }
2633    
2634          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
2635    
# Line 2022  for (;; ptr++) Line 2650  for (;; ptr++)
2650  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2651            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
2652              {              {
2653              int occ, ocd;              unsigned int occ, ocd;
2654              int cc = c;              unsigned int cc = c;
2655              int origd = d;              unsigned int origd = d;
2656              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
2657                {                {
2658                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */
# Line 2082  for (;; ptr++) Line 2710  for (;; ptr++)
2710          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
2711          for partial ranges without UCP support. */          for partial ranges without UCP support. */
2712    
2713          for (; c <= d; c++)          class_charcount += d - c + 1;
2714            class_lastchar = d;
2715    
2716            /* We can save a bit of time by skipping this in the pre-compile. */
2717    
2718            if (lengthptr == NULL) for (; c <= d; c++)
2719            {            {
2720            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
2721            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 2723  for (;; ptr++)
2723              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
2724              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
2725              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
2726            }            }
2727    
2728          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 2746  for (;; ptr++)
2746  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2747          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
2748            {            {
2749            int chartype;            unsigned int othercase;
2750            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
2751              {              {
2752              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
2753              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 2772  for (;; ptr++)
2772          }          }
2773        }        }
2774    
2775      /* 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. */  
2776    
2777      while ((c = *(++ptr)) != ']' || inescq);      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
2778    
2779      /* If class_charcount is 1, we saw precisely one character whose value is      if (c == 0)                          /* Missing terminating ']' */
2780          {
2781          *errorcodeptr = ERR6;
2782          goto FAILED;
2783          }
2784    
2785        /* If class_charcount is 1, we saw precisely one character whose value is
2786      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we
2787      can optimize the negative case only if there were no characters >= 128      can optimize the negative case only if there were no characters >= 128
2788      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      because OP_NOT and the related opcodes like OP_NOTSTAR operate on
# Line 2210  for (;; ptr++) Line 2844  for (;; ptr++)
2844    
2845      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
2846      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode. If there are no characters < 256,
2847      we can omit the bitmap. */      we can omit the bitmap in the actual compiled code. */
2848    
2849  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2850      if (class_utf8)      if (class_utf8)
# Line 2220  for (;; ptr++) Line 2854  for (;; ptr++)
2854        code += LINK_SIZE;        code += LINK_SIZE;
2855        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
2856    
2857        /* 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;
2858        the extra data */        otherwise just move the code pointer to the end of the extra data. */
2859    
2860        if (class_charcount > 0)        if (class_charcount > 0)
2861          {          {
2862          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
2863            memmove(code + 32, code, class_utf8data - code);
2864          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
2865          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;  
2866          }          }
2867          else code = class_utf8data;
2868    
2869        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
2870    
# Line 2254  for (;; ptr++) Line 2881  for (;; ptr++)
2881      if (negate_class)      if (negate_class)
2882        {        {
2883        *code++ = OP_NCLASS;        *code++ = OP_NCLASS;
2884        for (c = 0; c < 32; c++) code[c] = ~classbits[c];        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
2885            for (c = 0; c < 32; c++) code[c] = ~classbits[c];
2886        }        }
2887      else      else
2888        {        {
# Line 2264  for (;; ptr++) Line 2892  for (;; ptr++)
2892      code += 32;      code += 32;
2893      break;      break;
2894    
2895    
2896        /* ===================================================================*/
2897      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
2898      has been tested above. */      has been tested above. */
2899    
# Line 2331  for (;; ptr++) Line 2961  for (;; ptr++)
2961        }        }
2962      else repeat_type = greedy_default;      else repeat_type = greedy_default;
2963    
     /* 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;  
       }  
   
2964      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
2965      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
2966      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 2378  for (;; ptr++) Line 2994  for (;; ptr++)
2994          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
2995          }          }
2996    
2997          /* If the repetition is unlimited, it pays to see if the next thing on
2998          the line is something that cannot possibly match this character. If so,
2999          automatically possessifying this item gains some performance in the case
3000          where the match fails. */
3001    
3002          if (!possessive_quantifier &&
3003              repeat_max < 0 &&
3004              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3005                options, cd))
3006            {
3007            repeat_type = 0;    /* Force greedy */
3008            possessive_quantifier = TRUE;
3009            }
3010    
3011        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3012        }        }
3013    
3014      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3015      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-
3016      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3017      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3018        currently used only for single-byte chars. */
3019    
3020      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3021        {        {
3022        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3023        c = previous[1];        c = previous[1];
3024          if (!possessive_quantifier &&
3025              repeat_max < 0 &&
3026              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3027            {
3028            repeat_type = 0;    /* Force greedy */
3029            possessive_quantifier = TRUE;
3030            }
3031        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3032        }        }
3033    
# Line 2403  for (;; ptr++) Line 3041  for (;; ptr++)
3041      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3042        {        {
3043        uschar *oldcode;        uschar *oldcode;
3044        int prop_type;        int prop_type, prop_value;
3045        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3046        c = *previous;        c = *previous;
3047    
3048          if (!possessive_quantifier &&
3049              repeat_max < 0 &&
3050              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3051            {
3052            repeat_type = 0;    /* Force greedy */
3053            possessive_quantifier = TRUE;
3054            }
3055    
3056        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3057        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3058          previous[1] : -1;          {
3059            prop_type = previous[1];
3060            prop_value = previous[2];
3061            }
3062          else prop_type = prop_value = -1;
3063    
3064        oldcode = code;        oldcode = code;
3065        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2443  for (;; ptr++) Line 3093  for (;; ptr++)
3093          }          }
3094    
3095        /* 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
3096        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3097        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
3098        one less than the maximum. */        one less than the maximum. */
3099    
# Line 2470  for (;; ptr++) Line 3120  for (;; ptr++)
3120    
3121          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3122          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
3123          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3124          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3125          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3126    
# Line 2486  for (;; ptr++) Line 3136  for (;; ptr++)
3136  #endif  #endif
3137              {              {
3138              *code++ = c;              *code++ = c;
3139              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3140                  {
3141                  *code++ = prop_type;
3142                  *code++ = prop_value;
3143                  }
3144              }              }
3145            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3146            }            }
3147    
3148          /* 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
3149          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3150            UPTO is just for 1 instance, we can use QUERY instead. */
3151    
3152          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3153            {            {
# Line 2505  for (;; ptr++) Line 3160  for (;; ptr++)
3160            else            else
3161  #endif  #endif
3162            *code++ = c;            *code++ = c;
3163            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3164                {
3165                *code++ = prop_type;
3166                *code++ = prop_value;
3167                }
3168            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3169            *code++ = OP_UPTO + repeat_type;  
3170            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3171                {
3172                *code++ = OP_QUERY + repeat_type;
3173                }
3174              else
3175                {
3176                *code++ = OP_UPTO + repeat_type;
3177                PUT2INC(code, 0, repeat_max);
3178                }
3179            }            }
3180          }          }
3181    
# Line 2524  for (;; ptr++) Line 3191  for (;; ptr++)
3191  #endif  #endif
3192        *code++ = c;        *code++ = c;
3193    
3194        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3195        defines the required property. */        define the required property. */
3196    
3197  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3198        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3199            {
3200            *code++ = prop_type;
3201            *code++ = prop_value;
3202            }
3203  #endif  #endif
3204        }        }
3205    
# Line 2571  for (;; ptr++) Line 3242  for (;; ptr++)
3242      /* 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
3243      cases. */      cases. */
3244    
3245      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3246               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3247        {        {
3248        register int i;        register int i;
3249        int ketoffset = 0;        int ketoffset = 0;
3250        int len = code - previous;        int len = code - previous;
3251        uschar *bralink = NULL;        uschar *bralink = NULL;
3252    
3253          /* Repeating a DEFINE group is pointless */
3254    
3255          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3256            {
3257            *errorcodeptr = ERR55;
3258            goto FAILED;
3259            }
3260    
3261          /* This is a paranoid check to stop integer overflow later on */
3262    
3263          if (len > MAX_DUPLENGTH)
3264            {
3265            *errorcodeptr = ERR50;
3266            goto FAILED;
3267            }
3268    
3269        /* 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
3270        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
3271        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 2613  for (;; ptr++) Line 3300  for (;; ptr++)
3300          /* 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
3301          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
3302          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
3303          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3304          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3305            doing this. */
3306    
3307          if (repeat_max <= 1)          if (repeat_max <= 1)
3308            {            {
3309            *code = OP_END;            *code = OP_END;
3310            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3311            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3312            code++;            code++;
3313            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2637  for (;; ptr++) Line 3325  for (;; ptr++)
3325            {            {
3326            int offset;            int offset;
3327            *code = OP_END;            *code = OP_END;
3328            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3329            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3330            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3331            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3345  for (;; ptr++)
3345        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3346        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3347        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
3348        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3349          forward reference subroutine calls in the group, there will be entries on
3350          the workspace list; replicate these with an appropriate increment. */
3351    
3352        else        else
3353          {          {
3354          if (repeat_min > 1)          if (repeat_min > 1)
3355            {            {
3356            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3357            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. */
3358    
3359              if (lengthptr != NULL)
3360                *lengthptr += (repeat_min - 1)*length_prevgroup;
3361    
3362              /* This is compiling for real */
3363    
3364              else
3365              {              {
3366              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3367              code += len;              for (i = 1; i < repeat_min; i++)
3368                  {
3369                  uschar *hc;
3370                  uschar *this_hwm = cd->hwm;
3371                  memcpy(code, previous, len);
3372                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3373                    {
3374                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3375                    cd->hwm += LINK_SIZE;
3376                    }
3377                  save_hwm = this_hwm;
3378                  code += len;
3379                  }
3380              }              }
3381            }            }
3382    
3383          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3384          }          }
3385    
# Line 2677  for (;; ptr++) Line 3387  for (;; ptr++)
3387        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3388        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,
3389        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
3390        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3391          replicate entries on the forward reference list. */
3392    
3393        if (repeat_max >= 0)        if (repeat_max >= 0)
3394          {          {
3395          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3396            just adjust the length as if we had. For each repetition we must add 1
3397            to the length for BRAZERO and for all but the last repetition we must
3398            add 2 + 2*LINKSIZE to allow for the nesting that occurs. */
3399    
3400            if (lengthptr != NULL && repeat_max > 0)
3401              *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3402                2 - 2*LINK_SIZE;  /* Last one doesn't nest */
3403    
3404            /* This is compiling for real */
3405    
3406            else for (i = repeat_max - 1; i >= 0; i--)
3407            {            {
3408              uschar *hc;
3409              uschar *this_hwm = cd->hwm;
3410    
3411            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3412    
3413            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2698  for (;; ptr++) Line 3423  for (;; ptr++)
3423              }              }
3424    
3425            memcpy(code, previous, len);            memcpy(code, previous, len);
3426              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3427                {
3428                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3429                cd->hwm += LINK_SIZE;
3430                }
3431              save_hwm = this_hwm;
3432            code += len;            code += len;
3433            }            }
3434    
# Line 2720  for (;; ptr++) Line 3451  for (;; ptr++)
3451        /* 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
3452        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3453        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
3454        correct offset was computed above. */        correct offset was computed above.
3455    
3456          Then, when we are doing the actual compile phase, check to see whether
3457          this group is a non-atomic one that could match an empty string. If so,
3458          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3459          that runtime checking can be done. [This check is also applied to
3460          atomic groups at runtime, but in a different way.] */
3461    
3462        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
3463            {
3464            uschar *ketcode = code - ketoffset;
3465            uschar *bracode = ketcode - GET(ketcode, 1);
3466            *ketcode = OP_KETRMAX + repeat_type;
3467            if (lengthptr == NULL && *bracode != OP_ONCE)
3468              {
3469              uschar *scode = bracode;
3470              do
3471                {
3472                if (could_be_empty_branch(scode, ketcode, utf8))
3473                  {
3474                  *bracode += OP_SBRA - OP_BRA;
3475                  break;
3476                  }
3477                scode += GET(scode, 1);
3478                }
3479              while (*scode == OP_ALT);
3480              }
3481            }
3482        }        }
3483    
3484      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 3489  for (;; ptr++)
3489        goto FAILED;        goto FAILED;
3490        }        }
3491    
3492      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
3493      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
3494      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
3495      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.
3496      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
3497        but the special opcodes can optimize it a bit. The repeated item starts at
3498        tempcode, not at previous, which might be the first part of a string whose
3499        (former) last char we repeated.
3500    
3501        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
3502        an 'upto' may follow. We skip over an 'exact' item, and then test the
3503        length of what remains before proceeding. */
3504    
3505      if (possessive_quantifier)      if (possessive_quantifier)
3506        {        {
3507        int len = code - tempcode;        int len;
3508        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
3509        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
3510        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
3511        tempcode[0] = OP_ONCE;        len = code - tempcode;
3512        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
3513        PUTINC(code, 0, len);          {
3514        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
3515            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
3516            case OP_QUERY: *tempcode = OP_POSQUERY; break;
3517            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
3518    
3519            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
3520            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
3521            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
3522            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
3523    
3524            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
3525            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
3526            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
3527            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
3528    
3529            default:
3530            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
3531            code += 1 + LINK_SIZE;
3532            len += 1 + LINK_SIZE;
3533            tempcode[0] = OP_ONCE;
3534            *code++ = OP_KET;
3535            PUTINC(code, 0, len);
3536            PUT(tempcode, 1, len);
3537            break;
3538            }
3539        }        }
3540    
3541      /* 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 2761  for (;; ptr++) Line 3548  for (;; ptr++)
3548      break;      break;
3549    
3550    
3551      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
3552      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
3553      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
3554      of any of them means that this is not a referencing group. They were      parenthesis forms. First deal with the specials; all are introduced by ?,
3555      checked for validity in the first pass over the string, so we don't have to      and the appearance of any of them means that this is not a capturing
3556      check for syntax errors here.  */      group. */
3557    
3558      case '(':      case '(':
3559      newoptions = options;      newoptions = options;
3560      skipbytes = 0;      skipbytes = 0;
3561        bravalue = OP_CBRA;
3562        save_hwm = cd->hwm;
3563    
3564      if (*(++ptr) == '?')      if (*(++ptr) == '?')
3565        {        {
3566        int set, unset;        int i, set, unset, namelen;
3567        int *optset;        int *optset;
3568          const uschar *name;
3569          uschar *slot;
3570    
3571        switch (*(++ptr))        switch (*(++ptr))
3572          {          {
3573          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
3574          ptr++;          ptr++;
3575          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
3576            if (*ptr == 0)
3577              {
3578              *errorcodeptr = ERR18;
3579              goto FAILED;
3580              }
3581          continue;          continue;
3582    
3583          case ':':                 /* Non-extracting bracket */  
3584            /* ------------------------------------------------------------ */
3585            case ':':                 /* Non-capturing bracket */
3586          bravalue = OP_BRA;          bravalue = OP_BRA;
3587          ptr++;          ptr++;
3588          break;          break;
3589    
3590    
3591            /* ------------------------------------------------------------ */
3592          case '(':          case '(':
3593          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
3594    
3595          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
3596            group), a name (referring to a named group), or 'R', referring to
3597            recursion. R<digits> and R&name are also permitted for recursion tests.
3598    
3599            There are several syntaxes for testing a named group: (?(name)) is used
3600            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
3601    
3602            There are two unfortunate ambiguities, caused by history. (a) 'R' can
3603            be the recursive thing or the name 'R' (and similarly for 'R' followed
3604            by digits), and (b) a number could be a name that consists of digits.
3605            In both cases, we look for a name first; if not found, we try the other
3606            cases. */
3607    
3608            /* For conditions that are assertions, check the syntax, and then exit
3609            the switch. This will take control down to where bracketed groups,
3610            including assertions, are processed. */
3611    
3612            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
3613              break;
3614    
3615            /* Most other conditions use OP_CREF (a couple change to OP_RREF
3616            below), and all need to skip 3 bytes at the start of the group. */
3617    
3618            code[1+LINK_SIZE] = OP_CREF;
3619            skipbytes = 3;
3620    
3621            /* Check for a test for recursion in a named group. */
3622    
3623            if (ptr[1] == 'R' && ptr[2] == '&')
3624              {
3625              terminator = -1;
3626              ptr += 2;
3627              code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
3628              }
3629    
3630            /* Check for a test for a named group's having been set, using the Perl
3631            syntax (?(<name>) or (?('name') */
3632    
3633            else if (ptr[1] == '<')
3634              {
3635              terminator = '>';
3636              ptr++;
3637              }
3638            else if (ptr[1] == '\'')
3639              {
3640              terminator = '\'';
3641              ptr++;
3642              }
3643            else terminator = 0;
3644    
3645            /* We now expect to read a name; any thing else is an error */
3646    
3647            if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
3648              {
3649              ptr += 1;  /* To get the right offset */
3650              *errorcodeptr = ERR28;
3651              goto FAILED;
3652              }
3653    
3654            /* Read the name, but also get it as a number if it's all digits */
3655    
3656            recno = 0;
3657            name = ++ptr;
3658            while ((cd->ctypes[*ptr] & ctype_word) != 0)
3659              {
3660              if (recno >= 0)
3661                recno = ((digitab[*ptr] & ctype_digit) != 0)?
3662                  recno * 10 + *ptr - '0' : -1;
3663              ptr++;
3664              }
3665            namelen = ptr - name;
3666    
3667            if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
3668              {
3669              ptr--;      /* Error offset */
3670              *errorcodeptr = ERR26;
3671              goto FAILED;
3672              }
3673    
3674            /* Do no further checking in the pre-compile phase. */
3675    
3676            if (lengthptr != NULL) break;
3677    
3678            /* In the real compile we do the work of looking for the actual
3679            reference. */
3680    
3681            slot = cd->name_table;
3682            for (i = 0; i < cd->names_found; i++)
3683              {
3684              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
3685              slot += cd->name_entry_size;
3686              }
3687    
3688            /* Found a previous named subpattern */
3689    
3690            if (i < cd->names_found)
3691              {
3692              recno = GET2(slot, 0);
3693              PUT2(code, 2+LINK_SIZE, recno);
3694              }
3695    
3696            /* Search the pattern for a forward reference */
3697    
3698            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
3699                            (options & PCRE_EXTENDED) != 0)) > 0)
3700              {
3701              PUT2(code, 2+LINK_SIZE, i);
3702              }
3703    
3704            /* If terminator == 0 it means that the name followed directly after
3705            the opening parenthesis [e.g. (?(abc)...] and in this case there are
3706            some further alternatives to try. For the cases where terminator != 0
3707            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
3708            now checked all the possibilities, so give an error. */
3709    
3710          if (ptr[1] == 'R')          else if (terminator != 0)
3711            {            {
3712            code[1+LINK_SIZE] = OP_CREF;            *errorcodeptr = ERR15;
3713            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            goto FAILED;
           skipbytes = 3;  
           ptr += 3;  
3714            }            }
3715    
3716          /* Condition to test for a numbered subpattern match. We know that          /* Check for (?(R) for recursion. Allow digits after R to specify a
3717          if a digit follows ( then there will just be digits until ) because          specific group number. */
         the syntax was checked in the first pass. */  
3718    
3719          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (*name == 'R')
3720            {            {
3721            int condref;                 /* Don't amalgamate; some compilers */            recno = 0;
3722            condref = *(++ptr) - '0';    /* grumble at autoincrement in declaration */            for (i = 1; i < namelen; i++)
           while (*(++ptr) != ')') condref = condref*10 + *ptr - '0';  
           if (condref == 0)  
3723              {              {
3724              *errorcodeptr = ERR35;              if ((digitab[name[i]] & ctype_digit) == 0)
3725              goto FAILED;                {
3726                  *errorcodeptr = ERR15;
3727                  goto FAILED;
3728                  }
3729                recno = recno * 10 + name[i] - '0';
3730              }              }
3731            ptr++;            if (recno == 0) recno = RREF_ANY;
3732            code[1+LINK_SIZE] = OP_CREF;            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
3733            PUT2(code, 2+LINK_SIZE, condref);            PUT2(code, 2+LINK_SIZE, recno);
3734            skipbytes = 3;            }
3735    
3736            /* Similarly, check for the (?(DEFINE) "condition", which is always
3737            false. */
3738    
3739            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
3740              {
3741              code[1+LINK_SIZE] = OP_DEF;
3742              skipbytes = 1;
3743              }
3744    
3745            /* Check for the "name" actually being a subpattern number. */
3746    
3747            else if (recno > 0)
3748              {
3749              PUT2(code, 2+LINK_SIZE, recno);
3750              }
3751    
3752            /* Either an unidentified subpattern, or a reference to (?(0) */
3753    
3754            else
3755              {
3756              *errorcodeptr = (recno == 0)? ERR35: ERR15;
3757              goto FAILED;
3758            }            }
         /* For conditions that are assertions, we just fall through, having  
         set bravalue above. */  
3759          break;          break;
3760    
3761    
3762            /* ------------------------------------------------------------ */
3763          case '=':                 /* Positive lookahead */          case '=':                 /* Positive lookahead */
3764          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
3765          ptr++;          ptr++;
3766          break;          break;
3767    
3768    
3769            /* ------------------------------------------------------------ */
3770          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
3771          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
3772          ptr++;          ptr++;
3773          break;          break;
3774    
3775          case '<':                 /* Lookbehinds */  
3776          switch (*(++ptr))          /* ------------------------------------------------------------ */
3777            case '<':                 /* Lookbehind or named define */
3778            switch (ptr[1])
3779            {            {
3780            case '=':               /* Positive lookbehind */            case '=':               /* Positive lookbehind */
3781            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
3782            ptr++;            ptr += 2;
3783            break;            break;
3784    
3785            case '!':               /* Negative lookbehind */            case '!':               /* Negative lookbehind */
3786            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
3787            ptr++;            ptr += 2;
3788            break;            break;
3789    
3790              default:                /* Could be name define, else bad */
3791              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
3792              ptr++;                  /* Correct offset for error */
3793              *errorcodeptr = ERR24;
3794              goto FAILED;
3795            }            }
3796          break;          break;
3797    
3798    
3799            /* ------------------------------------------------------------ */
3800          case '>':                 /* One-time brackets */          case '>':                 /* One-time brackets */
3801          bravalue = OP_ONCE;          bravalue = OP_ONCE;
3802          ptr++;          ptr++;
3803          break;          break;
3804    
3805    
3806            /* ------------------------------------------------------------ */
3807          case 'C':                 /* Callout - may be followed by digits; */          case 'C':                 /* Callout - may be followed by digits; */
3808          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
3809          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
3810          *code++ = OP_CALLOUT;     /* Already checked that the terminating */          *code++ = OP_CALLOUT;
3811            {                       /* closing parenthesis is present. */            {
3812            int n = 0;            int n = 0;
3813            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
3814              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - '0';
3815              if (*ptr != ')')
3816                {
3817                *errorcodeptr = ERR39;
3818                goto FAILED;
3819                }
3820            if (n > 255)            if (n > 255)
3821              {              {
3822              *errorcodeptr = ERR38;              *errorcodeptr = ERR38;
# Line 2876  for (;; ptr++) Line 3830  for (;; ptr++)
3830          previous = NULL;          previous = NULL;
3831          continue;          continue;
3832    
3833          case 'P':                 /* Named subpattern handling */  
3834          if (*(++ptr) == '<')      /* Definition */          /* ------------------------------------------------------------ */
3835            case 'P':                 /* Python-style named subpattern handling */
3836            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
3837              {
3838              is_recurse = *ptr == '>';
3839              terminator = ')';
3840              goto NAMED_REF_OR_RECURSE;
3841              }
3842            else if (*ptr != '<')    /* Test for Python-style definition */
3843            {            {
3844            int i, namelen;            *errorcodeptr = ERR41;
3845            uschar *slot = cd->name_table;            goto FAILED;
3846            const uschar *name;     /* Don't amalgamate; some compilers */            }
3847            name = ++ptr;           /* grumble at autoincrement in declaration */          /* Fall through to handle (?P< as (?< is handled */
3848    
           while (*ptr++ != '>');  
           namelen = ptr - name - 1;  
3849    
3850            for (i = 0; i < cd->names_found; i++)          /* ------------------------------------------------------------ */
3851            DEFINE_NAME:    /* Come here from (?< handling */
3852            case '\'':
3853              {
3854              terminator = (*ptr == '<')? '>' : '\'';
3855              name = ++ptr;
3856    
3857              while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
3858              namelen = ptr - name;
3859    
3860              /* In the pre-compile phase, just do a syntax check. */
3861    
3862              if (lengthptr != NULL)
3863              {              {
3864              int crc = memcmp(name, slot+2, namelen);              if (*ptr != terminator)
3865              if (crc == 0)                {
3866                  *errorcodeptr = ERR42;
3867                  goto FAILED;
3868                  }
3869                if (cd->names_found >= MAX_NAME_COUNT)
3870                  {
3871                  *errorcodeptr = ERR49;
3872                  goto FAILED;
3873                  }
3874                if (namelen + 3 > cd->name_entry_size)
3875                {                {
3876                if (slot[2+namelen] == 0)                cd->name_entry_size = namelen + 3;
3877                  if (namelen > MAX_NAME_SIZE)
3878                  {                  {
3879                  *errorcodeptr = ERR43;                  *errorcodeptr = ERR48;
3880                  goto FAILED;                  goto FAILED;
3881                  }                  }
               crc = -1;             /* Current name is substring */  
3882                }                }
3883              if (crc < 0)              }
3884    
3885              /* In the real compile, create the entry in the table */
3886    
3887              else
3888                {
3889                slot = cd->name_table;
3890                for (i = 0; i < cd->names_found; i++)
3891                {                {
3892                memmove(slot + cd->name_entry_size, slot,                int crc = memcmp(name, slot+2, namelen);
3893                  (cd->names_found - i) * cd->name_entry_size);                if (crc == 0)
3894                break;                  {
3895                    if (slot[2+namelen] == 0)
3896                      {
3897                      if ((options & PCRE_DUPNAMES) == 0)
3898                        {
3899                        *errorcodeptr = ERR43;
3900                        goto FAILED;
3901                        }
3902                      }
3903                    else crc = -1;      /* Current name is substring */
3904                    }
3905                  if (crc < 0)
3906                    {
3907                    memmove(slot + cd->name_entry_size, slot,
3908                      (cd->names_found - i) * cd->name_entry_size);
3909                    break;
3910                    }
3911                  slot += cd->name_entry_size;
3912                }                }
             slot += cd->name_entry_size;  
             }  
3913    
3914            PUT2(slot, 0, *brackets + 1);              PUT2(slot, 0, cd->bracount + 1);
3915            memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
3916            slot[2+namelen] = 0;              slot[2+namelen] = 0;
3917            cd->names_found++;              }
           goto NUMBERED_GROUP;  
3918            }            }
3919    
3920          if (*ptr == '=' || *ptr == '>')  /* Reference or recursion */          /* In both cases, count the number of names we've encountered. */
3921    
3922            ptr++;                    /* Move past > or ' */
3923            cd->names_found++;
3924            goto NUMBERED_GROUP;
3925    
3926    
3927            /* ------------------------------------------------------------ */
3928            case '&':                 /* Perl recursion/subroutine syntax */
3929            terminator = ')';
3930            is_recurse = TRUE;
3931            /* Fall through */
3932    
3933            /* We come here from the Python syntax above that handles both
3934            references (?P=name) and recursion (?P>name), as well as falling
3935            through from the Perl recursion syntax (?&name). */
3936    
3937            NAMED_REF_OR_RECURSE:
3938            name = ++ptr;
3939            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
3940            namelen = ptr - name;
3941    
3942            /* In the pre-compile phase, do a syntax check and set a dummy
3943            reference number. */
3944    
3945            if (lengthptr != NULL)
3946            {            {
3947            int i, namelen;            if (*ptr != terminator)
3948            int type = *ptr++;              {
3949            const uschar *name = ptr;              *errorcodeptr = ERR42;
3950            uschar *slot = cd->name_table;              goto FAILED;
3951                }
3952              if (namelen > MAX_NAME_SIZE)
3953                {
3954                *errorcodeptr = ERR48;
3955                goto FAILED;
3956                }
3957              recno = 0;
3958              }
3959    
3960            while (*ptr != ')') ptr++;          /* In the real compile, seek the name in the table */
           namelen = ptr - name;  
3961    
3962            else
3963              {
3964              slot = cd->name_table;
3965            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
3966              {              {
3967              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
3968              slot += cd->name_entry_size;              slot += cd->name_entry_size;
3969              }              }
3970            if (i >= cd->names_found)  
3971              if (i < cd->names_found)         /* Back reference */
3972                {
3973                recno = GET2(slot, 0);
3974                }
3975              else if ((recno =                /* Forward back reference */
3976                        find_parens(ptr, cd->bracount, name, namelen,
3977                          (options & PCRE_EXTENDED) != 0)) <= 0)
3978              {              {
3979              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
3980              goto FAILED;              goto FAILED;
3981              }              }
3982              }
3983    
3984            recno = GET2(slot, 0);          /* In both phases, we can now go to the code than handles numerical
3985            recursion or backreferences. */
           if (type == '>') goto HANDLE_RECURSION;  /* A few lines below */  
   
           /* Back reference */  
3986    
3987            previous = code;          if (is_recurse) goto HANDLE_RECURSION;
3988            *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;  
           }  
3989    
         /* Should never happen */  
         break;  
3990    
3991          case 'R':                 /* Pattern recursion */          /* ------------------------------------------------------------ */
3992            case 'R':                 /* Recursion */
3993          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
3994          /* Fall through */          /* Fall through */
3995    
         /* Recursion or "subroutine" call */  
3996    
3997          case '0': case '1': case '2': case '3': case '4':          /* ------------------------------------------------------------ */
3998          case '5': case '6': case '7': case '8': case '9':          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */
3999            case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4000            {            {
4001            const uschar *called;            const uschar *called;
4002            recno = 0;            recno = 0;
4003            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4004              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4005              if (*ptr != ')')
4006                {
4007                *errorcodeptr = ERR29;
4008                goto FAILED;
4009                }
4010    
4011            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
4012    
4013            HANDLE_RECURSION:            HANDLE_RECURSION:
4014    
4015            previous = code;            previous = code;
4016              called = cd->start_code;
4017    
4018            /* Find the bracket that is being referenced. Temporarily end the            /* When we are actually compiling, find the bracket that is being
4019            regex in case it doesn't exist. */            referenced. Temporarily end the regex in case it doesn't exist before
4020              this point. If we end up with a forward reference, first check that
4021            *code = OP_END;            the bracket does occur later so we can give the error (and position)
4022            called = (recno == 0)?            now. Then remember this forward reference in the workspace so it can
4023              cd->start_code : find_bracket(cd->start_code, utf8, recno);            be filled in at the end. */
4024    
4025            if (called == NULL)            if (lengthptr == NULL)
4026              {              {
4027              *errorcodeptr = ERR15;              *code = OP_END;
4028              goto FAILED;              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);
             }  
4029    
4030            /* 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. */  
4031    
4032            if (GET(called, 1) == 0 && could_be_empty(called, code, bcptr, utf8))              if (called == NULL)
4033              {                {
4034              *errorcodeptr = ERR40;                if (find_parens(ptr, cd->bracount, NULL, recno,
4035              goto FAILED;                     (options & PCRE_EXTENDED) != 0) < 0)
4036                    {
4037                    *errorcodeptr = ERR15;
4038                    goto FAILED;
4039                    }
4040                  called = cd->start_code + recno;
4041                  PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
4042                  }
4043    
4044                /* If not a forward reference, and the subpattern is still open,
4045                this is a recursive call. We check to see if this is a left
4046                recursion that could loop for ever, and diagnose that case. */
4047    
4048                else if (GET(called, 1) == 0 &&
4049                         could_be_empty(called, code, bcptr, utf8))
4050                  {
4051                  *errorcodeptr = ERR40;
4052                  goto FAILED;
4053                  }
4054              }              }
4055    
4056            /* Insert the recursion/subroutine item */            /* Insert the recursion/subroutine item, automatically wrapped inside
4057              "once" brackets. Set up a "previous group" length so that a
4058              subsequent quantifier will work. */
4059    
4060              *code = OP_ONCE;
4061              PUT(code, 1, 2 + 2*LINK_SIZE);
4062              code += 1 + LINK_SIZE;
4063    
4064            *code = OP_RECURSE;            *code = OP_RECURSE;
4065            PUT(code, 1, called - cd->start_code);            PUT(code, 1, called - cd->start_code);
4066            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
           }  
         continue;  
4067    
4068          /* Character after (? not specially recognized */            *code = OP_KET;
4069              PUT(code, 1, 2 + 2*LINK_SIZE);
4070              code += 1 + LINK_SIZE;
4071    
4072          default:                  /* Option setting */            length_prevgroup = 3 + 3*LINK_SIZE;
4073              }
4074    
4075            /* Can't determine a first byte now */
4076    
4077            if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4078            continue;
4079    
4080    
4081            /* ------------------------------------------------------------ */
4082            default:              /* Other characters: check option setting */
4083          set = unset = 0;          set = unset = 0;
4084          optset = &set;          optset = &set;
4085    
# Line 3016  for (;; ptr++) Line 4089  for (;; ptr++)
4089              {              {
4090              case '-': optset = &unset; break;              case '-': optset = &unset; break;
4091    
4092                case 'J':    /* Record that it changed in the external options */
4093                *optset |= PCRE_DUPNAMES;
4094                cd->external_options |= PCRE_JCHANGED;
4095                break;
4096    
4097              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
4098              case 'm': *optset |= PCRE_MULTILINE; break;              case 'm': *optset |= PCRE_MULTILINE; break;
4099              case 's': *optset |= PCRE_DOTALL; break;              case 's': *optset |= PCRE_DOTALL; break;
4100              case 'x': *optset |= PCRE_EXTENDED; break;              case 'x': *optset |= PCRE_EXTENDED; break;
4101              case 'U': *optset |= PCRE_UNGREEDY; break;              case 'U': *optset |= PCRE_UNGREEDY; break;
4102              case 'X': *optset |= PCRE_EXTRA; break;              case 'X': *optset |= PCRE_EXTRA; break;
4103    
4104                default:  *errorcodeptr = ERR12;
4105                          ptr--;    /* Correct the offset */
4106                          goto FAILED;
4107              }              }
4108            }            }
4109    
# Line 3030  for (;; ptr++) Line 4112  for (;; ptr++)
4112          newoptions = (options | set) & (~unset);          newoptions = (options | set) & (~unset);
4113    
4114          /* 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
4115          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
4116          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
4117          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
4118          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
4119          a group), a resetting item can be compiled.          caseless checking of required bytes.
4120    
4121          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
4122          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
4123          change to compile. */          compiled. In the pre-compile phase, however, the code pointer can have
4124            that value after the start, because it gets reset as code is discarded
4125            during the pre-compile. However, this can happen only at top level - if
4126            we are within parentheses, the starting BRA will still be present. At
4127            any parenthesis level, the length value can be used to test if anything
4128            has been compiled at that level. Thus, a test for both these conditions
4129            is necessary to ensure we correctly detect the start of the pattern in
4130            both phases.
4131    
4132            If we are not at the pattern start, compile code to change the ims
4133            options if this setting actually changes any of them. We also pass the
4134            new setting back so that it can be put at the start of any following
4135            branches, and when this group ends (if we are in a group), a resetting
4136            item can be compiled. */
4137    
4138          if (*ptr == ')')          if (*ptr == ')')
4139            {            {
4140            if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))            if (code == cd->start_code + 1 + LINK_SIZE &&
4141                   (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
4142              {              {
4143              *code++ = OP_OPT;              cd->external_options = newoptions;
4144              *code++ = newoptions & PCRE_IMS;              options = newoptions;
4145              }              }
4146             else
4147                {
4148                if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
4149                  {
4150                  *code++ = OP_OPT;
4151                  *code++ = newoptions & PCRE_IMS;
4152                  }
4153    
4154            /* Change options at this level, and pass them back for use              /* Change options at this level, and pass them back for use
4155            in subsequent branches. Reset the greedy defaults and the case              in subsequent branches. Reset the greedy defaults and the case
4156            value for firstbyte and reqbyte. */              value for firstbyte and reqbyte. */
4157    
4158            *optionsptr = options = newoptions;              *optionsptr = options = newoptions;
4159            greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
4160            greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
4161            req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
4162                }
4163    
4164            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
4165            continue;              /* It is complete */            continue;              /* It is complete */
# Line 3068  for (;; ptr++) Line 4172  for (;; ptr++)
4172    
4173          bravalue = OP_BRA;          bravalue = OP_BRA;
4174          ptr++;          ptr++;
4175          }          }     /* End of switch for character following (? */
4176        }        }       /* End of (? handling */
4177    
4178      /* If PCRE_NO_AUTO_CAPTURE is set, all unadorned brackets become      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,
4179      non-capturing and behave like (?:...) brackets */      all unadorned brackets become non-capturing and behave like (?:...)
4180        brackets. */
4181    
4182      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
4183        {        {
4184        bravalue = OP_BRA;        bravalue = OP_BRA;
4185        }        }
4186    
4187      /* 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. */  
4188    
4189      else      else
4190        {        {
4191        NUMBERED_GROUP:        NUMBERED_GROUP:
4192        if (++(*brackets) > EXTRACT_BASIC_MAX)        cd->bracount += 1;
4193          {        PUT2(code, 1+LINK_SIZE, cd->bracount);
4194          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;  
4195        }        }
4196    
4197      /* Process nested bracketed re. Assertions may not be repeated, but other      /* Process nested bracketed regex. Assertions may not be repeated, but
4198      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
4199      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
4200      new setting for the ims options if they have changed. */      compilers complain otherwise. Pass in a new setting for the ims options if
4201        they have changed. */
4202    
4203      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
4204      *code = bravalue;      *code = bravalue;
4205      tempcode = code;      tempcode = code;
4206      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;     /* Save value before bracket */
4207        length_prevgroup = 0;              /* Initialize for pre-compile phase */
4208    
4209      if (!compile_regex(      if (!compile_regex(
4210           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
4211           options & PCRE_IMS,           /* The previous ims option state */           options & PCRE_IMS,           /* The previous ims option state */
          brackets,                     /* Extracting bracket count */  
4212           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
4213           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
4214           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
4215           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
4216            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
4217           skipbytes,                    /* Skip over OP_COND/OP_BRANUMBER */           skipbytes,                    /* Skip over bracket number */
4218           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
4219           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
4220           bcptr,                        /* Current branch chain */           bcptr,                        /* Current branch chain */
4221           cd))                          /* Tables block */           cd,                           /* Tables block */
4222             (lengthptr == NULL)? NULL :   /* Actual compile phase */
4223               &length_prevgroup           /* Pre-compile phase */
4224             ))
4225        goto FAILED;        goto FAILED;
4226    
4227      /* 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 3128  for (;; ptr++) Line 4230  for (;; ptr++)
4230      is on the bracket. */      is on the bracket. */
4231    
4232      /* 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
4233      two branches in the group. */      two branches in the group, or just one if it's a DEFINE group. */
4234    
4235      else if (bravalue == OP_COND)      if (bravalue == OP_COND)
4236        {        {
4237        uschar *tc = code;        uschar *tc = code;
4238        condcount = 0;        int condcount = 0;
4239    
4240        do {        do {
4241           condcount++;           condcount++;
# Line 3141  for (;; ptr++) Line 4243  for (;; ptr++)
4243           }           }
4244        while (*tc != OP_KET);        while (*tc != OP_KET);
4245    
4246        if (condcount > 2)        /* A DEFINE group is never obeyed inline (the "condition" is always
4247          false). It must have only one branch. */
4248    
4249          if (code[LINK_SIZE+1] == OP_DEF)
4250          {          {
4251          *errorcodeptr = ERR27;          if (condcount > 1)
4252          goto FAILED;            {
4253              *errorcodeptr = ERR54;
4254              goto FAILED;
4255              }
4256            bravalue = OP_DEF;   /* Just a flag to suppress char handling below */
4257            }
4258    
4259          /* A "normal" conditional group. If there is just one branch, we must not
4260          make use of its firstbyte or reqbyte, because this is equivalent to an
4261          empty second branch. */
4262    
4263          else
4264            {
4265            if (condcount > 2)
4266              {
4267              *errorcodeptr = ERR27;
4268              goto FAILED;
4269              }
4270            if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;
4271          }          }
4272          }
4273    
4274        /* If there is just one branch, we must not make use of its firstbyte or      /* Error if hit end of pattern */
4275        reqbyte, because this is equivalent to an empty second branch. */  
4276        if (*ptr != ')')
4277          {
4278          *errorcodeptr = ERR14;
4279          goto FAILED;
4280          }
4281    
4282        if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;      /* In the pre-compile phase, update the length by the length of the nested
4283        group, less the brackets at either end. Then reduce the compiled code to
4284        just the brackets so that it doesn't use much memory if it is duplicated by
4285        a quantifier. */
4286    
4287        if (lengthptr != NULL)
4288          {
4289          *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
4290          code++;
4291          PUTINC(code, 0, 1 + LINK_SIZE);
4292          *code++ = OP_KET;
4293          PUTINC(code, 0, 1 + LINK_SIZE);
4294        }        }
4295    
4296      /* Handle updating of the required and first characters. Update for normal      /* Otherwise update the main code pointer to the end of the group. */
4297      brackets of all kinds, and conditions with two branches (see code above).  
4298      If the bracket is followed by a quantifier with zero repeat, we have to      else code = tempcode;
4299      back off. Hence the definition of zeroreqbyte and zerofirstbyte outside the  
4300      main loop so that they can be accessed for the back off. */      /* For a DEFINE group, required and first character settings are not
4301        relevant. */
4302    
4303        if (bravalue == OP_DEF) break;
4304    
4305        /* Handle updating of the required and first characters for other types of
4306        group. Update for normal brackets of all kinds, and conditions with two
4307        branches (see code above). If the bracket is followed by a quantifier with
4308        zero repeat, we have to back off. Hence the definition of zeroreqbyte and
4309        zerofirstbyte outside the main loop so that they can be accessed for the
4310        back off. */
4311    
4312      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4313      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
4314      groupsetfirstbyte = FALSE;      groupsetfirstbyte = FALSE;
4315    
4316      if (bravalue >= OP_BRA || bravalue == OP_ONCE || bravalue == OP_COND)      if (bravalue >= OP_ONCE)
4317        {        {
4318        /* 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
4319        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 3204  for (;; ptr++) Line 4354  for (;; ptr++)
4354      firstbyte, looking for an asserted first char. */      firstbyte, looking for an asserted first char. */
4355    
4356      else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;      else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;
4357        break;     /* End of processing '(' */
4358    
     /* Now update the main code pointer to the end of the group. */  
   
     code = tempcode;  
   
     /* Error if hit end of pattern */  
   
     if (*ptr != ')')  
       {  
       *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);  
4359    
4360      /* Handle metacharacters introduced by \. For ones like \d, the ESC_ values      /* ===================================================================*/
4361        /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
4362      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
4363      back references, the values are ESC_REF plus the reference number. Only      back references, the values are ESC_REF plus the reference number. Only
4364      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
4365      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
4366      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
4367    
4368        case '\\':
4369        tempptr = ptr;
4370        c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
4371        if (*errorcodeptr != 0) goto FAILED;
4372    
4373      if (c < 0)      if (c < 0)
4374        {        {
4375        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 3242  for (;; ptr++) Line 4379  for (;; ptr++)
4379          continue;          continue;
4380          }          }
4381    
4382          if (-c == ESC_E) continue;  /* Perl ignores an orphan \E */
4383    
4384        /* For metasequences that actually match a character, we disable the        /* For metasequences that actually match a character, we disable the
4385        setting of a first character if it hasn't already been set. */        setting of a first character if it hasn't already been set. */
4386    
# Line 3253  for (;; ptr++) Line 4392  for (;; ptr++)
4392        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
4393        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4394    
4395        /* Back references are handled specially */        /* \k<name> or \k'name' is a back reference by name (Perl syntax) */
4396    
4397          if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))
4398            {
4399            is_recurse = FALSE;
4400            terminator = (*(++ptr) == '<')? '>' : '\'';
4401            goto NAMED_REF_OR_RECURSE;
4402            }
4403    
4404          /* Back references are handled specially; must disable firstbyte if
4405          not set to cope with cases like (?=(\w+))\1: which would otherwise set
4406          ':' later. */
4407    
4408        if (-c >= ESC_REF)        if (-c >= ESC_REF)
4409          {          {
4410          int number = -c - ESC_REF;          recno = -c - ESC_REF;
4411    
4412            HANDLE_REFERENCE:    /* Come here from named backref handling */
4413            if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4414          previous = code;          previous = code;
4415          *code++ = OP_REF;          *code++ = OP_REF;
4416          PUT2INC(code, 0, number);          PUT2INC(code, 0, recno);
4417            cd->backref_map |= (recno < 32)? (1 << recno) : 1;
4418            if (recno > cd->top_backref) cd->top_backref = recno;
4419          }          }
4420    
4421        /* So are Unicode property matches, if supported. We know that get_ucp        /* So are Unicode property matches, if supported. */
       won't fail because it was tested in the pre-pass. */  
4422    
4423  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4424        else if (-c == ESC_P || -c == ESC_p)        else if (-c == ESC_P || -c == ESC_p)
4425          {          {
4426          BOOL negated;          BOOL negated;
4427          int value = get_ucp(&ptr, &negated, errorcodeptr);          int pdata;
4428            int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4429            if (ptype < 0) goto FAILED;
4430          previous = code;          previous = code;
4431          *code++ = ((-c == ESC_p) != negated)? OP_PROP : OP_NOTPROP;          *code++ = ((-c == ESC_p) != negated)? OP_PROP : OP_NOTPROP;
4432          *code++ = value;          *code++ = ptype;
4433            *code++ = pdata;
4434            }
4435    #else
4436    
4437          /* If Unicode properties are not supported, \X, \P, and \p are not
4438          allowed. */
4439    
4440          else if (-c == ESC_X || -c == ESC_P || -c == ESC_p)
4441            {
4442            *errorcodeptr = ERR45;
4443            goto FAILED;
4444          }          }
4445  #endif  #endif
4446    
4447        /* For the rest, we can obtain the OP value by negating the escape        /* For the rest (including \X when Unicode properties are supported), we
4448        value */        can obtain the OP value by negating the escape value. */
4449    
4450        else        else
4451          {          {
# Line 3302  for (;; ptr++) Line 4469  for (;; ptr++)
4469       mcbuffer[0] = c;       mcbuffer[0] = c;
4470       mclength = 1;       mclength = 1;
4471       }       }
   
4472      goto ONE_CHAR;      goto ONE_CHAR;
4473    
4474    
4475        /* ===================================================================*/
4476      /* Handle a literal character. It is guaranteed not to be whitespace or #      /* Handle a literal character. It is guaranteed not to be whitespace or #
4477      when the extended flag is set. If we are in UTF-8 mode, it may be a      when the extended flag is set. If we are in UTF-8 mode, it may be a
4478      multi-byte literal character. */      multi-byte literal character. */
# Line 3315  for (;; ptr++) Line 4483  for (;; ptr++)
4483      mcbuffer[0] = c;      mcbuffer[0] = c;
4484    
4485  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4486      if (utf8 && (c & 0xc0) == 0xc0)      if (utf8 && c >= 0xc0)
4487        {        {
4488        while ((ptr[1] & 0xc0) == 0x80)        while ((ptr[1] & 0xc0) == 0x80)
4489          mcbuffer[mclength++] = *(++ptr);          mcbuffer[mclength++] = *(++ptr);
# Line 3366  for (;; ptr++) Line 4534  for (;; ptr++)
4534      }      }
4535    }                   /* end of big loop */    }                   /* end of big loop */
4536    
4537