/[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 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 224 by ph10, Mon Aug 20 12:58:29 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) 1997-2007 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include <config.h>
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When DEBUG is defined, we need the pcre_printint() function, which is also
57    used by pcretest. DEBUG is not defined when building a production library. */
58    
59    #ifdef DEBUG
60    #include "pcre_printint.src"
61    #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143    /* Table of special "verbs" like (*PRUNE) */
144    
145    typedef struct verbitem {
146      const char *name;
147      int   len;
148      int   op;
149    } verbitem;
150    
151    static verbitem verbs[] = {
152      { "ACCEPT", 6, OP_ACCEPT },
153      { "COMMIT", 6, OP_COMMIT },
154      { "F",      1, OP_FAIL },
155      { "FAIL",   4, OP_FAIL },
156      { "PRUNE",  5, OP_PRUNE },
157      { "SKIP",   4, OP_SKIP  },
158      { "THEN",   4, OP_THEN  }
159    };
160    
161    static int verbcount = sizeof(verbs)/sizeof(verbitem);
162    
163    
164  /* Tables of names of POSIX character classes and their lengths. The list is  /* Tables of names of POSIX character classes and their lengths. The list is
165  terminated by a zero length entry. The first three must be alpha, upper, lower,  terminated by a zero length entry. The first three must be alpha, lower, upper,
166  as this is assumed for handling case independence. */  as this is assumed for handling case independence. */
167    
168  static const char *const posix_names[] = {  static const char *const posix_names[] = {
# Line 118  static const char *const posix_names[] = Line 173  static const char *const posix_names[] =
173  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
174    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 };
175    
176  /* 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
177  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
178  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
179    characters are removed, and for [:alpha:] and [:alnum:] the underscore
180    character is removed. The triples in the table consist of the base map offset,
181    second map offset or -1 if no second map, and a non-negative value for map
182    addition or a negative value for map subtraction (if there are two maps). The
183    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
184    remove vertical space characters, 2 => remove underscore. */
185    
186  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
187    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
188    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
189    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
190    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
191    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
192    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
193    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
194    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
195    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
196    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
197    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
198    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
199    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
200    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
201  };  };
202    
203    
204    #define STRING(a)  # a
205    #define XSTRING(s) STRING(s)
206    
207  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
208  are passed to the outside world. */  are passed to the outside world. Do not ever re-use any error number, because
209    they are documented. Always add a new error instead. Messages marked DEAD below
210    are no longer used. */
211    
212  static const char *error_texts[] = {  static const char *error_texts[] = {
213    "no error",    "no error",
# Line 156  static const char *error_texts[] = { Line 222  static const char *error_texts[] = {
222    "range out of order in character class",    "range out of order in character class",
223    "nothing to repeat",    "nothing to repeat",
224    /* 10 */    /* 10 */
225    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string",  /** DEAD **/
226    "internal error: unexpected repeat",    "internal error: unexpected repeat",
227    "unrecognized character after (?",    "unrecognized character after (?",
228    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class",
# Line 166  static const char *error_texts[] = { Line 232  static const char *error_texts[] = {
232    "erroffset passed as NULL",    "erroffset passed as NULL",
233    "unknown option bit(s) set",    "unknown option bit(s) set",
234    "missing ) after comment",    "missing ) after comment",
235    "parentheses nested too deeply",    "parentheses nested too deeply",  /** DEAD **/
236    /* 20 */    /* 20 */
237    "regular expression too large",    "regular expression is too large",
238    "failed to get memory",    "failed to get memory",
239    "unmatched parentheses",    "unmatched parentheses",
240    "internal error: code overflow",    "internal error: code overflow",
241    "unrecognized character after (?<",    "unrecognized character after (?<",
242    /* 25 */    /* 25 */
243    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length",
244    "malformed number after (?(",    "malformed number or name after (?(",
245    "conditional group contains more than two branches",    "conditional group contains more than two branches",
246    "assertion expected after (?(",    "assertion expected after (?(",
247    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )",
248    /* 30 */    /* 30 */
249    "unknown POSIX class name",    "unknown POSIX class name",
250    "POSIX collating elements are not supported",    "POSIX collating elements are not supported",
251    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support",
252    "spare error",    "spare error",  /** DEAD **/
253    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large",
254    /* 35 */    /* 35 */
255    "invalid condition (?(0)",    "invalid condition (?(0)",
# Line 194  static const char *error_texts[] = { Line 260  static const char *error_texts[] = {
260    /* 40 */    /* 40 */
261    "recursive call could loop indefinitely",    "recursive call could loop indefinitely",
262    "unrecognized character after (?P",    "unrecognized character after (?P",
263    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)",
264    "two named groups have the same name",    "two named subpatterns have the same name",
265    "invalid UTF-8 string",    "invalid UTF-8 string",
266    /* 45 */    /* 45 */
267    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled",
268    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence",
269    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p",
270      "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",
271      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",
272      /* 50 */
273      "repeated subpattern is too long",    /** DEAD **/
274      "octal value is greater than \\377 (not in UTF-8 mode)",
275      "internal error: overran compiling workspace",
276      "internal error: previously-checked referenced subpattern not found",
277      "DEFINE group contains more than one branch",
278      /* 55 */
279      "repeating a DEFINE group is not allowed",
280      "inconsistent NEWLINE options",
281      "\\g is not followed by a braced name or an optionally braced non-zero number",
282      "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",
283      "(*VERB) with an argument is not supported",
284      /* 60 */
285      "(*VERB) not recognized",
286      "number is too big"
287  };  };
288    
289    
# Line 220  For convenience, we use the same bit def Line 303  For convenience, we use the same bit def
303    
304  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
305    
306  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
307  static const unsigned char digitab[] =  static const unsigned char digitab[] =
308    {    {
309    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 339  static const unsigned char digitab[] =
339    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
340    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
341    
342  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
343  static const unsigned char digitab[] =  static const unsigned char digitab[] =
344    {    {
345    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 353  static const unsigned char digitab[] =
353    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
354    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
355    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
356    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
357    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
358    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
359    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 387  static const unsigned char ebcdic_charta
387    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
388    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
389    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
390    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
391    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
392    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
393    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 414  static const unsigned char ebcdic_charta
414  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
415    
416  static BOOL  static BOOL
417    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
418      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
419    
420    
421    
# Line 342  static BOOL Line 425  static BOOL
425    
426  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
427  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
428  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
429  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
430  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
431    ptr is pointing at the \. On exit, it is on the final character of the escape
432    sequence.
433    
434  Arguments:  Arguments:
435    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 440  Arguments:
440    
441  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
442                   negative => a special escape sequence                   negative => a special escape sequence
443                   on error, errorptr is set                   on error, errorcodeptr is set
444  */  */
445    
446  static int  static int
447  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
448    int options, BOOL isclass)    int options, BOOL isclass)
449  {  {
450  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
451    const uschar *ptr = *ptrptr + 1;
452  int c, i;  int c, i;
453    
454    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
455    ptr--;                            /* Set pointer back to the last byte */
456    
457  /* 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. */
458    
 c = *(++ptr);  
459  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
460    
461  /* 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
462  a table. A non-zero result is something that can be returned immediately.  a table. A non-zero result is something that can be returned immediately.
463  Otherwise further processing may be required. */  Otherwise further processing may be required. */
464    
465  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
466  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
467  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
468    
469  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
470  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
471  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
472  #endif  #endif
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 476  else if ((i = escapes[c - 0x48]) != 0)
476  else  else
477    {    {
478    const uschar *oldptr;    const uschar *oldptr;
479      BOOL braced, negated;
480    
481    switch (c)    switch (c)
482      {      {
483      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
# Line 401  else Line 491  else
491      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
492      break;      break;
493    
494        /* \g must be followed by a number, either plain or braced. If positive, it
495        is an absolute backreference. If negative, it is a relative backreference.
496        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
497        reference to a named group. This is part of Perl's movement towards a
498        unified syntax for back references. As this is synonymous with \k{name}, we
499        fudge it up by pretending it really was \k. */
500    
501        case 'g':
502        if (ptr[1] == '{')
503          {
504          const uschar *p;
505          for (p = ptr+2; *p != 0 && *p != '}'; p++)
506            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
507          if (*p != 0 && *p != '}')
508            {
509            c = -ESC_k;
510            break;
511            }
512          braced = TRUE;
513          ptr++;
514          }
515        else braced = FALSE;
516    
517        if (ptr[1] == '-')
518          {
519          negated = TRUE;
520          ptr++;
521          }
522        else negated = FALSE;
523    
524        c = 0;
525        while ((digitab[ptr[1]] & ctype_digit) != 0)
526          c = c * 10 + *(++ptr) - '0';
527    
528        if (c < 0)
529          {
530          *errorcodeptr = ERR61;
531          break;
532          }
533    
534        if (c == 0 || (braced && *(++ptr) != '}'))
535          {
536          *errorcodeptr = ERR57;
537          break;
538          }
539    
540        if (negated)
541          {
542          if (c > bracount)
543            {
544            *errorcodeptr = ERR15;
545            break;
546            }
547          c = bracount - (c - 1);
548          }
549    
550        c = -(ESC_REF + c);
551        break;
552    
553      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
554      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
555      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 571  else
571        c -= '0';        c -= '0';
572        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
573          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
574          if (c < 0)
575            {
576            *errorcodeptr = ERR61;
577            break;
578            }
579        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
580          {          {
581          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 442  else Line 596  else
596        }        }
597    
598      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
599      larger first octal digit. */      larger first octal digit. The original code used just to take the least
600        significant 8 bits of octal numbers (I think this is what early Perls used
601        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
602        than 3 octal digits. */
603    
604      case '0':      case '0':
605      c -= '0';      c -= '0';
606      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
607          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
608      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
609      break;      break;
610    
611      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
612      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
613        treated as a data character. */
614    
615      case 'x':      case 'x':
616  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
617        {        {
618        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
619        register int count = 0;        int count = 0;
620    
621        c = 0;        c = 0;
622        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
623          {          {
624          int cc = *pt++;          register int cc = *pt++;
625            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
626          count++;          count++;
627  #if !EBCDIC    /* ASCII coding */  
628    #ifndef EBCDIC  /* ASCII coding */
629          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
630          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
631  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
632          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
633          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
634  #endif  #endif
635          }          }
636    
637        if (*pt == '}')        if (*pt == '}')
638          {          {
639          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
640          ptr = pt;          ptr = pt;
641          break;          break;
642          }          }
643    
644        /* 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
645        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
646        }        }
 #endif  
647    
648      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
649    
650      c = 0;      c = 0;
651      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
652        {        {
653        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
654        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
655  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
656        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
657        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
658  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
659        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
660        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
661  #endif  #endif
662        }        }
663      break;      break;
664    
665      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
666        This coding is ASCII-specific, but then the whole concept of \cx is
667        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
668    
669      case 'c':      case 'c':
670      c = *(++ptr);      c = *(++ptr);
671      if (c == 0)      if (c == 0)
672        {        {
673        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
674        return 0;        break;
675        }        }
676    
677      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII coding */
     is ASCII-specific, but then the whole concept of \cx is ASCII-specific.  
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
678      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
679      c ^= 0x40;      c ^= 0x40;
680  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
681      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
682      c ^= 0xC0;      c ^= 0xC0;
683  #endif  #endif
# Line 560  escape sequence. Line 719  escape sequence.
719  Argument:  Argument:
720    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
721    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
722      dptr           points to an int that is set to the detailed property value
723    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
724    
725  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
726  */  */
727    
728  static int  static int
729  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
730  {  {
731  int c, i, bot, top;  int c, i, bot, top;
732  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
733  char name[4];  char name[32];
734    
735  c = *(++ptr);  c = *(++ptr);
736  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
737    
738  *negptr = FALSE;  *negptr = FALSE;
739    
740  /* \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
741  preceded by ^ for negation. */  negation. */
742    
743  if (c == '{')  if (c == '{')
744    {    {
# Line 587  if (c == '{') Line 747  if (c == '{')
747      *negptr = TRUE;      *negptr = TRUE;
748      ptr++;      ptr++;
749      }      }
750    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
751      {      {
752      c = *(++ptr);      c = *(++ptr);
753      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
754      if (c == '}') break;      if (c == '}') break;
755      name[i] = c;      name[i] = c;
756      }      }
757    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;  
     }  
758    name[i] = 0;    name[i] = 0;
759    }    }
760    
# Line 619  top = _pcre_utt_size; Line 775  top = _pcre_utt_size;
775    
776  while (bot < top)  while (bot < top)
777    {    {
778    i = (bot + top)/2;    i = (bot + top) >> 1;
779    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt[i].name);
780    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
781        {
782        *dptr = _pcre_utt[i].value;
783        return _pcre_utt[i].type;
784        }
785    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
786    }    }
787    
 UNKNOWN_RETURN:  
788  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
789  *ptrptr = ptr;  *ptrptr = ptr;
790  return -1;  return -1;
# Line 698  read_repeat_counts(const uschar *p, int Line 857  read_repeat_counts(const uschar *p, int
857  int min = 0;  int min = 0;
858  int max = -1;  int max = -1;
859    
860    /* Read the minimum value and do a paranoid check: a negative value indicates
861    an integer overflow. */
862    
863  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
864    if (min < 0 || min > 65535)
865      {
866      *errorcodeptr = ERR5;
867      return p;
868      }
869    
870    /* Read the maximum value if there is one, and again do a paranoid on its size.
871    Also, max must not be less than min. */
872    
873  if (*p == '}') max = min; else  if (*p == '}') max = min; else
874    {    {
# Line 706  if (*p == '}') max = min; else Line 876  if (*p == '}') max = min; else
876      {      {
877      max = 0;      max = 0;
878      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
879        if (max < 0 || max > 65535)
880          {
881          *errorcodeptr = ERR5;
882          return p;
883          }
884      if (max < min)      if (max < min)
885        {        {
886        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 889  if (*p == '}') max = min; else
889      }      }
890    }    }
891    
892  /* 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
893  pointer to the terminating '}'. */  '}'. */
894    
895  if (min > 65535 || max > 65535)  *minp = min;
896    *errorcodeptr = ERR5;  *maxp = max;
897  else  return p;
898    }
899    
900    
901    
902    /*************************************************
903    *       Find forward referenced subpattern       *
904    *************************************************/
905    
906    /* This function scans along a pattern's text looking for capturing
907    subpatterns, and counting them. If it finds a named pattern that matches the
908    name it is given, it returns its number. Alternatively, if the name is NULL, it
909    returns when it reaches a given numbered subpattern. This is used for forward
910    references to subpatterns. We know that if (?P< is encountered, the name will
911    be terminated by '>' because that is checked in the first pass.
912    
913    Arguments:
914      ptr          current position in the pattern
915      count        current count of capturing parens so far encountered
916      name         name to seek, or NULL if seeking a numbered subpattern
917      lorn         name length, or subpattern number if name is NULL
918      xmode        TRUE if we are in /x mode
919    
920    Returns:       the number of the named subpattern, or -1 if not found
921    */
922    
923    static int
924    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
925      BOOL xmode)
926    {
927    const uschar *thisname;
928    
929    for (; *ptr != 0; ptr++)
930    {    {
931    *minp = min;    int term;
932    *maxp = max;  
933      /* Skip over backslashed characters and also entire \Q...\E */
934    
935      if (*ptr == '\\')
936        {
937        if (*(++ptr) == 0) return -1;
938        if (*ptr == 'Q') for (;;)
939          {
940          while (*(++ptr) != 0 && *ptr != '\\');
941          if (*ptr == 0) return -1;
942          if (*(++ptr) == 'E') break;
943          }
944        continue;
945        }
946    
947      /* Skip over character classes */
948    
949      if (*ptr == '[')
950        {
951        while (*(++ptr) != ']')
952          {
953          if (*ptr == 0) return -1;
954          if (*ptr == '\\')
955            {
956            if (*(++ptr) == 0) return -1;
957            if (*ptr == 'Q') for (;;)
958              {
959              while (*(++ptr) != 0 && *ptr != '\\');
960              if (*ptr == 0) return -1;
961              if (*(++ptr) == 'E') break;
962              }
963            continue;
964            }
965          }
966        continue;
967        }
968    
969      /* Skip comments in /x mode */
970    
971      if (xmode && *ptr == '#')
972        {
973        while (*(++ptr) != 0 && *ptr != '\n');
974        if (*ptr == 0) return -1;
975        continue;
976        }
977    
978      /* An opening parens must now be a real metacharacter */
979    
980      if (*ptr != '(') continue;
981      if (ptr[1] != '?' && ptr[1] != '*')
982        {
983        count++;
984        if (name == NULL && count == lorn) return count;
985        continue;
986        }
987    
988      ptr += 2;
989      if (*ptr == 'P') ptr++;                      /* Allow optional P */
990    
991      /* We have to disambiguate (?<! and (?<= from (?<name> */
992    
993      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
994           *ptr != '\'')
995        continue;
996    
997      count++;
998    
999      if (name == NULL && count == lorn) return count;
1000      term = *ptr++;
1001      if (term == '<') term = '>';
1002      thisname = ptr;
1003      while (*ptr != term) ptr++;
1004      if (name != NULL && lorn == ptr - thisname &&
1005          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1006        return count;
1007    }    }
1008  return p;  
1009    return -1;
1010  }  }
1011    
1012    
# Line 778  for (;;) Line 1060  for (;;)
1060    
1061      case OP_CALLOUT:      case OP_CALLOUT:
1062      case OP_CREF:      case OP_CREF:
1063      case OP_BRANUMBER:      case OP_RREF:
1064        case OP_DEF:
1065      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1066      break;      break;
1067    
# Line 823  for (;;) Line 1106  for (;;)
1106    {    {
1107    int d;    int d;
1108    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1109    switch (op)    switch (op)
1110      {      {
1111        case OP_CBRA:
1112      case OP_BRA:      case OP_BRA:
1113      case OP_ONCE:      case OP_ONCE:
1114      case OP_COND:      case OP_COND:
1115      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1116      if (d < 0) return d;      if (d < 0) return d;
1117      branchlength += d;      branchlength += d;
1118      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1147  for (;;)
1147      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1148    
1149      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1150      case OP_CREF:      case OP_CREF:
1151        case OP_RREF:
1152        case OP_DEF:
1153      case OP_OPT:      case OP_OPT:
1154      case OP_CALLOUT:      case OP_CALLOUT:
1155      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1167  for (;;)
1167    
1168      case OP_CHAR:      case OP_CHAR:
1169      case OP_CHARNC:      case OP_CHARNC:
1170        case OP_NOT:
1171      branchlength++;      branchlength++;
1172      cc += 2;      cc += 2;
1173  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 910  for (;;) Line 1194  for (;;)
1194    
1195      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1196      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1197        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1198      cc += 4;      cc += 4;
1199      break;      break;
1200    
# Line 917  for (;;) Line 1202  for (;;)
1202    
1203      case OP_PROP:      case OP_PROP:
1204      case OP_NOTPROP:      case OP_NOTPROP:
1205      cc++;      cc += 2;
1206      /* Fall through */      /* Fall through */
1207    
1208      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 998  Returns: pointer to the opcode for Line 1283  Returns: pointer to the opcode for
1283  static const uschar *  static const uschar *
1284  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1285  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1286  for (;;)  for (;;)
1287    {    {
1288    register int c = *code;    register int c = *code;
1289    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1290    else if (c > OP_BRA)  
1291      /* XCLASS is used for classes that cannot be represented just by a bit
1292      map. This includes negated single high-valued characters. The length in
1293      the table is zero; the actual length is stored in the compiled code. */
1294    
1295      if (c == OP_XCLASS) code += GET(code, 1);
1296    
1297      /* Handle capturing bracket */
1298    
1299      else if (c == OP_CBRA)
1300      {      {
1301      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1302      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1303      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1304      }      }
1305    
1306      /* Otherwise, we can get the item's length from the table, except that for
1307      repeated character types, we have to test for \p and \P, which have an extra
1308      two bytes of parameters. */
1309    
1310    else    else
1311      {      {
1312      code += _pcre_OP_lengths[c];      switch(c)
1313          {
1314          case OP_TYPESTAR:
1315          case OP_TYPEMINSTAR:
1316          case OP_TYPEPLUS:
1317          case OP_TYPEMINPLUS:
1318          case OP_TYPEQUERY:
1319          case OP_TYPEMINQUERY:
1320          case OP_TYPEPOSSTAR:
1321          case OP_TYPEPOSPLUS:
1322          case OP_TYPEPOSQUERY:
1323          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1324          break;
1325    
1326  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1327          case OP_TYPEMINUPTO:
1328          case OP_TYPEEXACT:
1329          case OP_TYPEPOSUPTO:
1330          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1331          break;
1332          }
1333    
1334      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1335      by a multi-byte character. The length in the table is a minimum, so we have  
1336      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
     can use relatively efficient code. */  
1337    
1338      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1339      a multi-byte character. The length in the table is a minimum, so we have to
1340      arrange to skip the extra bytes. */
1341    
1342    #ifdef SUPPORT_UTF8
1343      if (utf8) switch(c)      if (utf8) switch(c)
1344        {        {
1345        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1347  for (;;)
1347        case OP_EXACT:        case OP_EXACT:
1348        case OP_UPTO:        case OP_UPTO:
1349        case OP_MINUPTO:        case OP_MINUPTO:
1350          case OP_POSUPTO:
1351        case OP_STAR:        case OP_STAR:
1352        case OP_MINSTAR:        case OP_MINSTAR:
1353          case OP_POSSTAR:
1354        case OP_PLUS:        case OP_PLUS:
1355        case OP_MINPLUS:        case OP_MINPLUS:
1356          case OP_POSPLUS:
1357        case OP_QUERY:        case OP_QUERY:
1358        case OP_MINQUERY:        case OP_MINQUERY:
1359        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1360        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1361        break;        break;
1362        }        }
1363  #endif  #endif
# Line 1072  Returns: pointer to the opcode for Line 1384  Returns: pointer to the opcode for
1384  static const uschar *  static const uschar *
1385  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1386  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1387  for (;;)  for (;;)
1388    {    {
1389    register int c = *code;    register int c = *code;
1390    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1391    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1392    else if (c > OP_BRA)  
1393      {    /* XCLASS is used for classes that cannot be represented just by a bit
1394      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1395      }    the table is zero; the actual length is stored in the compiled code. */
1396    
1397      if (c == OP_XCLASS) code += GET(code, 1);
1398    
1399      /* Otherwise, we can get the item's length from the table, except that for
1400      repeated character types, we have to test for \p and \P, which have an extra
1401      two bytes of parameters. */
1402    
1403    else    else
1404      {      {
1405      code += _pcre_OP_lengths[c];      switch(c)
1406          {
1407          case OP_TYPESTAR:
1408          case OP_TYPEMINSTAR:
1409          case OP_TYPEPLUS:
1410          case OP_TYPEMINPLUS:
1411          case OP_TYPEQUERY:
1412          case OP_TYPEMINQUERY:
1413          case OP_TYPEPOSSTAR:
1414          case OP_TYPEPOSPLUS:
1415          case OP_TYPEPOSQUERY:
1416          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1417          break;
1418    
1419  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1420          case OP_TYPEUPTO:
1421          case OP_TYPEMINUPTO:
1422          case OP_TYPEEXACT:
1423          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1424          break;
1425          }
1426    
1427        /* Add in the fixed length from the table */
1428    
1429        code += _pcre_OP_lengths[c];
1430    
1431      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1432      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1433      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1434    
1435    #ifdef SUPPORT_UTF8
1436      if (utf8) switch(c)      if (utf8) switch(c)
1437        {        {
1438        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1440  for (;;)
1440        case OP_EXACT:        case OP_EXACT:
1441        case OP_UPTO:        case OP_UPTO:
1442        case OP_MINUPTO:        case OP_MINUPTO:
1443          case OP_POSUPTO:
1444        case OP_STAR:        case OP_STAR:
1445        case OP_MINSTAR:        case OP_MINSTAR:
1446          case OP_POSSTAR:
1447        case OP_PLUS:        case OP_PLUS:
1448        case OP_MINPLUS:        case OP_MINPLUS:
1449          case OP_POSPLUS:
1450        case OP_QUERY:        case OP_QUERY:
1451        case OP_MINQUERY:        case OP_MINQUERY:
1452        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1453        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1454        break;        break;
1455        }        }
1456  #endif  #endif
# Line 1132  for (;;) Line 1465  for (;;)
1465  *************************************************/  *************************************************/
1466    
1467  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
1468  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1469  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
1470  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
1471  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1472    struck an inner bracket whose current branch will already have been scanned.
1473    
1474  Arguments:  Arguments:
1475    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1483  static BOOL
1483  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1484  {  {
1485  register int c;  register int c;
1486  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1487       code < endcode;       code < endcode;
1488       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1489    {    {
# Line 1157  for (code = first_significant_code(code Line 1491  for (code = first_significant_code(code
1491    
1492    c = *code;    c = *code;
1493    
1494    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1495    
1496      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1497        {
1498        code += _pcre_OP_lengths[c];
1499        do code += GET(code, 1); while (*code == OP_ALT);
1500        c = *code;
1501        continue;
1502        }
1503    
1504      /* For other groups, scan the branches. */
1505    
1506      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1507      {      {
1508      BOOL empty_branch;      BOOL empty_branch;
1509      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1173  for (code = first_significant_code(code Line 1519  for (code = first_significant_code(code
1519        }        }
1520      while (*code == OP_ALT);      while (*code == OP_ALT);
1521      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1522      c = *code;      c = *code;
1523        continue;
1524      }      }
1525    
1526    else switch (c)    /* Handle the other opcodes */
1527    
1528      switch (c)
1529      {      {
1530      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1531        cannot be represented just by a bit map. This includes negated single
1532        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1533        actual length is stored in the compiled code, so we must update "code"
1534        here. */
1535    
1536  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1537      case OP_XCLASS:      case OP_XCLASS:
1538      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1539      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1540  #endif  #endif
1541    
# Line 1233  for (code = first_significant_code(code Line 1585  for (code = first_significant_code(code
1585      case OP_NOT:      case OP_NOT:
1586      case OP_PLUS:      case OP_PLUS:
1587      case OP_MINPLUS:      case OP_MINPLUS:
1588        case OP_POSPLUS:
1589      case OP_EXACT:      case OP_EXACT:
1590      case OP_NOTPLUS:      case OP_NOTPLUS:
1591      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1592        case OP_NOTPOSPLUS:
1593      case OP_NOTEXACT:      case OP_NOTEXACT:
1594      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1595      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1596        case OP_TYPEPOSPLUS:
1597      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1598      return FALSE;      return FALSE;
1599    
1600        /* These are going to continue, as they may be empty, but we have to
1601        fudge the length for the \p and \P cases. */
1602    
1603        case OP_TYPESTAR:
1604        case OP_TYPEMINSTAR:
1605        case OP_TYPEPOSSTAR:
1606        case OP_TYPEQUERY:
1607        case OP_TYPEMINQUERY:
1608        case OP_TYPEPOSQUERY:
1609        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1610        break;
1611    
1612        /* Same for these */
1613    
1614        case OP_TYPEUPTO:
1615        case OP_TYPEMINUPTO:
1616        case OP_TYPEPOSUPTO:
1617        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1618        break;
1619    
1620      /* End of branch */      /* End of branch */
1621    
# Line 1250  for (code = first_significant_code(code Line 1625  for (code = first_significant_code(code
1625      case OP_ALT:      case OP_ALT:
1626      return TRUE;      return TRUE;
1627    
1628      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1629      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1630    
1631  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1632      case OP_STAR:      case OP_STAR:
1633      case OP_MINSTAR:      case OP_MINSTAR:
1634        case OP_POSSTAR:
1635      case OP_QUERY:      case OP_QUERY:
1636      case OP_MINQUERY:      case OP_MINQUERY:
1637        case OP_POSQUERY:
1638      case OP_UPTO:      case OP_UPTO:
1639      case OP_MINUPTO:      case OP_MINUPTO:
1640        case OP_POSUPTO:
1641      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1642      break;      break;
1643  #endif  #endif
# Line 1377  earlier groups that are outside the curr Line 1755  earlier groups that are outside the curr
1755  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before
1756  it, after it has been compiled. This means that any OP_RECURSE items within it  it, after it has been compiled. This means that any OP_RECURSE items within it
1757  that refer to the group itself or any contained groups have to have their  that refer to the group itself or any contained groups have to have their
1758  offsets adjusted. That is the job of this function. Before it is called, the  offsets adjusted. That one of the jobs of this function. Before it is called,
1759  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1760    
1761    This function has been extended with the possibility of forward references for
1762    recursions and subroutine calls. It must also check the list of such references
1763    for the group we are dealing with. If it finds that one of the recursions in
1764    the current group is on this list, it adjusts the offset in the list, not the
1765    value in the reference (which is a group number).
1766    
1767  Arguments:  Arguments:
1768    group      points to the start of the group    group      points to the start of the group
1769    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1770    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1771    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1772      save_hwm   the hwm forward reference pointer at the start of the group
1773    
1774  Returns:     nothing  Returns:     nothing
1775  */  */
1776    
1777  static void  static void
1778  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1779      uschar *save_hwm)
1780  {  {
1781  uschar *ptr = group;  uschar *ptr = group;
1782    
1783  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1784    {    {
1785    int offset = GET(ptr, 1);    int offset;
1786    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1787    
1788      /* See if this recursion is on the forward reference list. If so, adjust the
1789      reference. */
1790    
1791      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1792        {
1793        offset = GET(hc, 0);
1794        if (cd->start_code + offset == ptr + 1)
1795          {
1796          PUT(hc, 0, offset + adjust);
1797          break;
1798          }
1799        }
1800    
1801      /* Otherwise, adjust the recursion offset if it's after the start of this
1802      group. */
1803    
1804      if (hc >= cd->hwm)
1805        {
1806        offset = GET(ptr, 1);
1807        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1808        }
1809    
1810    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1811    }    }
1812  }  }
# Line 1475  Yield: TRUE when range returned; Line 1885  Yield: TRUE when range returned;
1885  */  */
1886    
1887  static BOOL  static BOOL
1888  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1889      unsigned int *odptr)
1890  {  {
1891  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1892    
1893  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1894    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1895    
1896  if (c > d) return FALSE;  if (c > d) return FALSE;
1897    
# Line 1492  next = othercase + 1; Line 1900  next = othercase + 1;
1900    
1901  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1902    {    {
1903    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1904    next++;    next++;
1905    }    }
1906    
# Line 1506  return TRUE; Line 1912  return TRUE;
1912  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1913    
1914    
1915    
1916  /*************************************************  /*************************************************
1917  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1918  *************************************************/  *************************************************/
1919    
1920  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called for unlimited repeats of certain items, to see
1921  changed during the branch, the pointer is used to change the external options  whether the next thing could possibly match the repeated item. If not, it makes
1922  bits.  sense to automatically possessify the repeated item.
1923    
1924  Arguments:  Arguments:
1925    optionsptr     pointer to the option bits    op_code       the repeated op code
1926    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1927    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1928    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1929    errorcodeptr   points to error code variable    ptr           next character in pattern
1930    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1931    reqbyteptr     set to the last literal character required, else < 0    cd            contains pointers to tables etc.
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
1932    
1933  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1934  */  */
1935    
1936  static BOOL  static BOOL
1937  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1938    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1939  {  {
1940  int repeat_type, op_type;  int next;
 int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
 int bravalue = 0;  
 int greedy_default, greedy_non_default;  
 int firstbyte, reqbyte;  
 int zeroreqbyte, zerofirstbyte;  
 int req_caseopt, reqvary, tempreqvary;  
 int condcount = 0;  
 int options = *optionsptr;  
 int after_manual_callout = 0;  
 register int c;  
 register uschar *code = *codeptr;  
 uschar *tempcode;  
 BOOL inescq = FALSE;  
 BOOL groupsetfirstbyte = FALSE;  
 const uschar *ptr = *ptrptr;  
 const uschar *tempptr;  
 uschar *previous = NULL;  
 uschar *previous_callout = NULL;  
 uschar classbits[32];  
   
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
 #endif  
1941    
1942  /* Set up the default and non-default settings for greediness */  /* Skip whitespace and comments in extended mode */
1943    
1944    if ((options & PCRE_EXTENDED) != 0)
1945      {
1946      for (;;)
1947        {
1948        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1949        if (*ptr == '#')
1950          {
1951          while (*(++ptr) != 0)
1952            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1953          }
1954        else break;
1955        }
1956      }
1957    
1958    /* If the next item is one that we can handle, get its value. A non-negative
1959    value is a character, a negative value is an escape value. */
1960    
1961    if (*ptr == '\\')
1962      {
1963      int temperrorcode = 0;
1964      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
1965      if (temperrorcode != 0) return FALSE;
1966      ptr++;    /* Point after the escape sequence */
1967      }
1968    
1969    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
1970      {
1971    #ifdef SUPPORT_UTF8
1972      if (utf8) { GETCHARINC(next, ptr); } else
1973    #endif
1974      next = *ptr++;
1975      }
1976    
1977    else return FALSE;
1978    
1979    /* Skip whitespace and comments in extended mode */
1980    
1981    if ((options & PCRE_EXTENDED) != 0)
1982      {
1983      for (;;)
1984        {
1985        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1986        if (*ptr == '#')
1987          {
1988          while (*(++ptr) != 0)
1989            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1990          }
1991        else break;
1992        }
1993      }
1994    
1995    /* If the next thing is itself optional, we have to give up. */
1996    
1997    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
1998      return FALSE;
1999    
2000    /* Now compare the next item with the previous opcode. If the previous is a
2001    positive single character match, "item" either contains the character or, if
2002    "item" is greater than 127 in utf8 mode, the character's bytes are in
2003    utf8_char. */
2004    
2005    
2006    /* Handle cases when the next item is a character. */
2007    
2008    if (next >= 0) switch(op_code)
2009      {
2010      case OP_CHAR:
2011    #ifdef SUPPORT_UTF8
2012      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2013    #endif
2014      return item != next;
2015    
2016      /* For CHARNC (caseless character) we must check the other case. If we have
2017      Unicode property support, we can use it to test the other case of
2018      high-valued characters. */
2019    
2020      case OP_CHARNC:
2021    #ifdef SUPPORT_UTF8
2022      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2023    #endif
2024      if (item == next) return FALSE;
2025    #ifdef SUPPORT_UTF8
2026      if (utf8)
2027        {
2028        unsigned int othercase;
2029        if (next < 128) othercase = cd->fcc[next]; else
2030    #ifdef SUPPORT_UCP
2031        othercase = _pcre_ucp_othercase((unsigned int)next);
2032    #else
2033        othercase = NOTACHAR;
2034    #endif
2035        return (unsigned int)item != othercase;
2036        }
2037      else
2038    #endif  /* SUPPORT_UTF8 */
2039      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2040    
2041      /* For OP_NOT, "item" must be a single-byte character. */
2042    
2043      case OP_NOT:
2044      if (next < 0) return FALSE;  /* Not a character */
2045      if (item == next) return TRUE;
2046      if ((options & PCRE_CASELESS) == 0) return FALSE;
2047    #ifdef SUPPORT_UTF8
2048      if (utf8)
2049        {
2050        unsigned int othercase;
2051        if (next < 128) othercase = cd->fcc[next]; else
2052    #ifdef SUPPORT_UCP
2053        othercase = _pcre_ucp_othercase(next);
2054    #else
2055        othercase = NOTACHAR;
2056    #endif
2057        return (unsigned int)item == othercase;
2058        }
2059      else
2060    #endif  /* SUPPORT_UTF8 */
2061      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2062    
2063      case OP_DIGIT:
2064      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2065    
2066      case OP_NOT_DIGIT:
2067      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2068    
2069      case OP_WHITESPACE:
2070      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2071    
2072      case OP_NOT_WHITESPACE:
2073      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2074    
2075      case OP_WORDCHAR:
2076      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2077    
2078      case OP_NOT_WORDCHAR:
2079      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2080    
2081      case OP_HSPACE:
2082      case OP_NOT_HSPACE:
2083      switch(next)
2084        {
2085        case 0x09:
2086        case 0x20:
2087        case 0xa0:
2088        case 0x1680:
2089        case 0x180e:
2090        case 0x2000:
2091        case 0x2001:
2092        case 0x2002:
2093        case 0x2003:
2094        case 0x2004:
2095        case 0x2005:
2096        case 0x2006:
2097        case 0x2007:
2098        case 0x2008:
2099        case 0x2009:
2100        case 0x200A:
2101        case 0x202f:
2102        case 0x205f:
2103        case 0x3000:
2104        return op_code != OP_HSPACE;
2105        default:
2106        return op_code == OP_HSPACE;
2107        }
2108    
2109      case OP_VSPACE:
2110      case OP_NOT_VSPACE:
2111      switch(next)
2112        {
2113        case 0x0a:
2114        case 0x0b:
2115        case 0x0c:
2116        case 0x0d:
2117        case 0x85:
2118        case 0x2028:
2119        case 0x2029:
2120        return op_code != OP_VSPACE;
2121        default:
2122        return op_code == OP_VSPACE;
2123        }
2124    
2125      default:
2126      return FALSE;
2127      }
2128    
2129    
2130    /* Handle the case when the next item is \d, \s, etc. */
2131    
2132    switch(op_code)
2133      {
2134      case OP_CHAR:
2135      case OP_CHARNC:
2136    #ifdef SUPPORT_UTF8
2137      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2138    #endif
2139      switch(-next)
2140        {
2141        case ESC_d:
2142        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2143    
2144        case ESC_D:
2145        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2146    
2147        case ESC_s:
2148        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2149    
2150        case ESC_S:
2151        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2152    
2153        case ESC_w:
2154        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2155    
2156        case ESC_W:
2157        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2158    
2159        case ESC_h:
2160        case ESC_H:
2161        switch(item)
2162          {
2163          case 0x09:
2164          case 0x20:
2165          case 0xa0:
2166          case 0x1680:
2167          case 0x180e:
2168          case 0x2000:
2169          case 0x2001:
2170          case 0x2002:
2171          case 0x2003:
2172          case 0x2004:
2173          case 0x2005:
2174          case 0x2006:
2175          case 0x2007:
2176          case 0x2008:
2177          case 0x2009:
2178          case 0x200A:
2179          case 0x202f:
2180          case 0x205f:
2181          case 0x3000:
2182          return -next != ESC_h;
2183          default:
2184          return -next == ESC_h;
2185          }
2186    
2187        case ESC_v:
2188        case ESC_V:
2189        switch(item)
2190          {
2191          case 0x0a:
2192          case 0x0b:
2193          case 0x0c:
2194          case 0x0d:
2195          case 0x85:
2196          case 0x2028:
2197          case 0x2029:
2198          return -next != ESC_v;
2199          default:
2200          return -next == ESC_v;
2201          }
2202    
2203        default:
2204        return FALSE;
2205        }
2206    
2207      case OP_DIGIT:
2208      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2209             next == -ESC_h || next == -ESC_v;
2210    
2211      case OP_NOT_DIGIT:
2212      return next == -ESC_d;
2213    
2214      case OP_WHITESPACE:
2215      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2216    
2217      case OP_NOT_WHITESPACE:
2218      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2219    
2220      case OP_HSPACE:
2221      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2222    
2223      case OP_NOT_HSPACE:
2224      return next == -ESC_h;
2225    
2226      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2227      case OP_VSPACE:
2228      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2229    
2230      case OP_NOT_VSPACE:
2231      return next == -ESC_v;
2232    
2233      case OP_WORDCHAR:
2234      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2235    
2236      case OP_NOT_WORDCHAR:
2237      return next == -ESC_w || next == -ESC_d;
2238    
2239      default:
2240      return FALSE;
2241      }
2242    
2243    /* Control does not reach here */
2244    }
2245    
2246    
2247    
2248    /*************************************************
2249    *           Compile one branch                   *
2250    *************************************************/
2251    
2252    /* Scan the pattern, compiling it into the a vector. If the options are
2253    changed during the branch, the pointer is used to change the external options
2254    bits. This function is used during the pre-compile phase when we are trying
2255    to find out the amount of memory needed, as well as during the real compile
2256    phase. The value of lengthptr distinguishes the two phases.
2257    
2258    Arguments:
2259      optionsptr     pointer to the option bits
2260      codeptr        points to the pointer to the current code point
2261      ptrptr         points to the current pattern pointer
2262      errorcodeptr   points to error code variable
2263      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2264      reqbyteptr     set to the last literal character required, else < 0
2265      bcptr          points to current branch chain
2266      cd             contains pointers to tables etc.
2267      lengthptr      NULL during the real compile phase
2268                     points to length accumulator during pre-compile phase
2269    
2270    Returns:         TRUE on success
2271                     FALSE, with *errorcodeptr set non-zero on error
2272    */
2273    
2274    static BOOL
2275    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2276      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2277      compile_data *cd, int *lengthptr)
2278    {
2279    int repeat_type, op_type;
2280    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2281    int bravalue = 0;
2282    int greedy_default, greedy_non_default;
2283    int firstbyte, reqbyte;
2284    int zeroreqbyte, zerofirstbyte;
2285    int req_caseopt, reqvary, tempreqvary;
2286    int options = *optionsptr;
2287    int after_manual_callout = 0;
2288    int length_prevgroup = 0;
2289    register int c;
2290    register uschar *code = *codeptr;
2291    uschar *last_code = code;
2292    uschar *orig_code = code;
2293    uschar *tempcode;
2294    BOOL inescq = FALSE;
2295    BOOL groupsetfirstbyte = FALSE;
2296    const uschar *ptr = *ptrptr;
2297    const uschar *tempptr;
2298    uschar *previous = NULL;
2299    uschar *previous_callout = NULL;
2300    uschar *save_hwm = NULL;
2301    uschar classbits[32];
2302    
2303    #ifdef SUPPORT_UTF8
2304    BOOL class_utf8;
2305    BOOL utf8 = (options & PCRE_UTF8) != 0;
2306    uschar *class_utf8data;
2307    uschar utf8_char[6];
2308    #else
2309    BOOL utf8 = FALSE;
2310    uschar *utf8_char = NULL;
2311    #endif
2312    
2313    #ifdef DEBUG
2314    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2315    #endif
2316    
2317    /* Set up the default and non-default settings for greediness */
2318    
2319  greedy_default = ((options & PCRE_UNGREEDY) != 0);  greedy_default = ((options & PCRE_UNGREEDY) != 0);
2320  greedy_non_default = greedy_default ^ 1;  greedy_non_default = greedy_default ^ 1;
# Line 1595  for (;; ptr++) Line 2345  for (;; ptr++)
2345    BOOL negate_class;    BOOL negate_class;
2346    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2347    BOOL is_quantifier;    BOOL is_quantifier;
2348      BOOL is_recurse;
2349      BOOL reset_bracount;
2350    int class_charcount;    int class_charcount;
2351    int class_lastchar;    int class_lastchar;
2352    int newoptions;    int newoptions;
2353    int recno;    int recno;
2354      int refsign;
2355    int skipbytes;    int skipbytes;
2356    int subreqbyte;    int subreqbyte;
2357    int subfirstbyte;    int subfirstbyte;
2358      int terminator;
2359    int mclength;    int mclength;
2360    uschar mcbuffer[8];    uschar mcbuffer[8];
2361    
2362    /* Next byte in the pattern */    /* Get next byte in the pattern */
2363    
2364    c = *ptr;    c = *ptr;
2365    
2366      /* If we are in the pre-compile phase, accumulate the length used for the
2367      previous cycle of this loop. */
2368    
2369      if (lengthptr != NULL)
2370        {
2371    #ifdef DEBUG
2372        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2373    #endif
2374        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2375          {
2376          *errorcodeptr = ERR52;
2377          goto FAILED;
2378          }
2379    
2380        /* There is at least one situation where code goes backwards: this is the
2381        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2382        the class is simply eliminated. However, it is created first, so we have to
2383        allow memory for it. Therefore, don't ever reduce the length at this point.
2384        */
2385    
2386        if (code < last_code) code = last_code;
2387    
2388        /* Paranoid check for integer overflow */
2389    
2390        if (OFLOW_MAX - *lengthptr < code - last_code)
2391          {
2392          *errorcodeptr = ERR20;
2393          goto FAILED;
2394          }
2395    
2396        *lengthptr += code - last_code;
2397        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2398    
2399        /* If "previous" is set and it is not at the start of the work space, move
2400        it back to there, in order to avoid filling up the work space. Otherwise,
2401        if "previous" is NULL, reset the current code pointer to the start. */
2402    
2403        if (previous != NULL)
2404          {
2405          if (previous > orig_code)
2406            {
2407            memmove(orig_code, previous, code - previous);
2408            code -= previous - orig_code;
2409            previous = orig_code;
2410            }
2411          }
2412        else code = orig_code;
2413    
2414        /* Remember where this code item starts so we can pick up the length
2415        next time round. */
2416    
2417        last_code = code;
2418        }
2419    
2420      /* In the real compile phase, just check the workspace used by the forward
2421      reference list. */
2422    
2423      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2424        {
2425        *errorcodeptr = ERR52;
2426        goto FAILED;
2427        }
2428    
2429    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
2430    
2431    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1623  for (;; ptr++) Line 2440  for (;; ptr++)
2440        {        {
2441        if (previous_callout != NULL)        if (previous_callout != NULL)
2442          {          {
2443          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2444              complete_callout(previous_callout, ptr, cd);
2445          previous_callout = NULL;          previous_callout = NULL;
2446          }          }
2447        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2462  for (;; ptr++)
2462    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2463         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2464      {      {
2465      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2466          complete_callout(previous_callout, ptr, cd);
2467      previous_callout = NULL;      previous_callout = NULL;
2468      }      }
2469    
# Line 1655  for (;; ptr++) Line 2474  for (;; ptr++)
2474      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2475      if (c == '#')      if (c == '#')
2476        {        {
2477        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2478        on the Macintosh. */          {
2479        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2480        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2481          if (*ptr != 0) continue;
2482    
2483          /* Else fall through to handle end of string */
2484          c = 0;
2485        }        }
2486      }      }
2487    
# Line 1672  for (;; ptr++) Line 2495  for (;; ptr++)
2495    
2496    switch(c)    switch(c)
2497      {      {
2498      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2499        case 0:                        /* The branch terminates at string end */
2500      case 0:      case '|':                      /* or | or ) */
     case '|':  
2501      case ')':      case ')':
2502      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2503      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2504      *codeptr = code;      *codeptr = code;
2505      *ptrptr = ptr;      *ptrptr = ptr;
2506        if (lengthptr != NULL)
2507          {
2508          if (OFLOW_MAX - *lengthptr < code - last_code)
2509            {
2510            *errorcodeptr = ERR20;
2511            goto FAILED;
2512            }
2513          *lengthptr += code - last_code;   /* To include callout length */
2514          DPRINTF((">> end branch\n"));
2515          }
2516      return TRUE;      return TRUE;
2517    
2518    
2519        /* ===================================================================*/
2520      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2521      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2522    
# Line 1711  for (;; ptr++) Line 2545  for (;; ptr++)
2545      *code++ = OP_ANY;      *code++ = OP_ANY;
2546      break;      break;
2547    
2548      /* Character classes. If the included characters are all < 255 in value, we  
2549      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2550      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
2551      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
2552      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2553        map as usual, then invert it at the end. However, we use a different opcode
2554        so that data characters > 255 can be handled correctly.
2555    
2556      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2557      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 1736  for (;; ptr++) Line 2572  for (;; ptr++)
2572        goto FAILED;        goto FAILED;
2573        }        }
2574    
2575      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2576        if the first few characters (either before or after ^) are \Q\E or \E we
2577        skip them too. This makes for compatibility with Perl. */
2578    
2579      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2580        for (;;)
2581        {        {
       negate_class = TRUE;  
2582        c = *(++ptr);        c = *(++ptr);
2583        }        if (c == '\\')
2584      else          {
2585        {          if (ptr[1] == 'E') ptr++;
2586        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2587                else break;
2588            }
2589          else if (!negate_class && c == '^')
2590            negate_class = TRUE;
2591          else break;
2592        }        }
2593    
2594      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
2595      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
2596      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2597    
2598      class_charcount = 0;      class_charcount = 0;
2599      class_lastchar = -1;      class_lastchar = -1;
2600    
2601        /* Initialize the 32-char bit map to all zeros. We build the map in a
2602        temporary bit of memory, in case the class contains only 1 character (less
2603        than 256), because in that case the compiled code doesn't use the bit map.
2604        */
2605    
2606        memset(classbits, 0, 32 * sizeof(uschar));
2607    
2608  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2609      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2610      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2611  #endif  #endif
2612    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
2613      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2614      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
2615      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
2616    
2617      do      if (c != 0) do
2618        {        {
2619          const uschar *oldptr;
2620    
2621  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2622        if (utf8 && c > 127)        if (utf8 && c > 127)
2623          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1786  for (;; ptr++) Line 2629  for (;; ptr++)
2629    
2630        if (inescq)        if (inescq)
2631          {          {
2632          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2633            {            {
2634            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2635            ptr++;            ptr++;                            /* Skip the 'E' */
2636            continue;            continue;                         /* Carry on with next */
2637            }            }
2638          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2639          }          }
2640    
2641        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1806  for (;; ptr++) Line 2649  for (;; ptr++)
2649            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2650          {          {
2651          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2652          int posix_class, i;          int posix_class, taboffset, tabopt;
2653          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2654            uschar pbits[32];
2655    
2656          if (ptr[1] != ':')          if (ptr[1] != ':')
2657            {            {
# Line 1836  for (;; ptr++) Line 2680  for (;; ptr++)
2680          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2681            posix_class = 0;            posix_class = 0;
2682    
2683          /* 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
2684          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
2685          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
2686          white space chars afterwards. */          result into the bit map that is being built. */
2687    
2688          posix_class *= 3;          posix_class *= 3;
2689          for (i = 0; i < 3; i++)  
2690            /* Copy in the first table (always present) */
2691    
2692            memcpy(pbits, cbits + posix_class_maps[posix_class],
2693              32 * sizeof(uschar));
2694    
2695            /* If there is a second table, add or remove it as required. */
2696    
2697            taboffset = posix_class_maps[posix_class + 1];
2698            tabopt = posix_class_maps[posix_class + 2];
2699    
2700            if (taboffset >= 0)
2701            {            {
2702            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2703            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;  
             }  
2704            else            else
2705              {              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;  
             }  
2706            }            }
2707    
2708            /* Not see if we need to remove any special characters. An option
2709            value of 1 removes vertical space and 2 removes underscore. */
2710    
2711            if (tabopt < 0) tabopt = -tabopt;
2712            if (tabopt == 1) pbits[1] &= ~0x3c;
2713              else if (tabopt == 2) pbits[11] &= 0x7f;
2714    
2715            /* Add the POSIX table or its complement into the main table that is
2716            being built and we are done. */
2717    
2718            if (local_negate)
2719              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2720            else
2721              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2722    
2723          ptr = tempptr + 1;          ptr = tempptr + 1;
2724          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2725          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2726          }          }
2727    
2728        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2729        of the specials, which just set a flag. Escaped items are checked for        of the specials, which just set a flag. The sequence \b is a special
2730        validity in the pre-compiling pass. The sequence \b is a special case.        case. Inside a class (and only there) it is treated as backspace.
2731        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2732        it marks a word boundary. Other escapes have preset maps ready to        to 'or' into the one we are building. We assume they have more than one
       or into the one we are building. We assume they have more than one  
2733        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2734    
2735        if (c == '\\')        if (c == '\\')
2736          {          {
2737          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2738            if (*errorcodeptr != 0) goto FAILED;
2739    
2740          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2741          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2742            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2743          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2744            {            {
2745            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1890  for (;; ptr++) Line 2749  for (;; ptr++)
2749            else inescq = TRUE;            else inescq = TRUE;
2750            continue;            continue;
2751            }            }
2752            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2753    
2754          if (c < 0)          if (c < 0)
2755            {            {
2756            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2757            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2758            switch (-c)  
2759              /* Save time by not doing this in the pre-compile phase. */
2760    
2761              if (lengthptr == NULL) switch (-c)
2762              {              {
2763              case ESC_d:              case ESC_d:
2764              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
# Line 1923  for (;; ptr++) Line 2786  for (;; ptr++)
2786              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2787              continue;              continue;
2788    
2789  #ifdef SUPPORT_UCP              case ESC_E: /* Perl ignores an orphan \E */
             case ESC_p:  
             case ESC_P:  
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
2790              continue;              continue;
 #endif  
   
             /* Unrecognized escapes are faulted if PCRE is running in its  
             strict mode. By default, for compatibility with Perl, they are  
             treated as literals. */  
2791    
2792              default:              default:    /* Not recognized; fall through */
2793              if ((options & PCRE_EXTRA) != 0)              break;      /* Need "default" setting to stop compiler warning. */
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
2794              }              }
           }  
   
         /* Fall through if we have a single character (c >= 0). This may be  
         > 256 in UTF-8 mode. */  
2795    
2796          }   /* End of backslash handling */            /* In the pre-compile phase, just do the recognition. */
2797    
2798        /* A single character may be followed by '-' to form a range. However,            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2799        Perl does not permit ']' to be the end of the range. A '-' character                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
       here is treated as a literal. */  
2800    
2801        if (ptr[1] == '-' && ptr[2] != ']')            /* We need to deal with \H, \h, \V, and \v in both phases because
2802          {            they use extra memory. */
         int d;  
         ptr += 2;  
2803    
2804              if (-c == ESC_h)
2805                {
2806                SETBIT(classbits, 0x09); /* VT */
2807                SETBIT(classbits, 0x20); /* SPACE */
2808                SETBIT(classbits, 0xa0); /* NSBP */
2809  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2810          if (utf8)              if (utf8)
2811                  {
2812                  class_utf8 = TRUE;
2813                  *class_utf8data++ = XCL_SINGLE;
2814                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2815                  *class_utf8data++ = XCL_SINGLE;
2816                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2817                  *class_utf8data++ = XCL_RANGE;
2818                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2819                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2820                  *class_utf8data++ = XCL_SINGLE;
2821                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2822                  *class_utf8data++ = XCL_SINGLE;
2823                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2824                  *class_utf8data++ = XCL_SINGLE;
2825                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2826                  }
2827    #endif
2828                continue;
2829                }
2830    
2831              if (-c == ESC_H)
2832                {
2833                for (c = 0; c < 32; c++)
2834                  {
2835                  int x = 0xff;
2836                  switch (c)
2837                    {
2838                    case 0x09/8: x ^= 1 << (0x09%8); break;
2839                    case 0x20/8: x ^= 1 << (0x20%8); break;
2840                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2841                    default: break;
2842                    }
2843                  classbits[c] |= x;
2844                  }
2845    
2846    #ifdef SUPPORT_UTF8
2847                if (utf8)
2848                  {
2849                  class_utf8 = TRUE;
2850                  *class_utf8data++ = XCL_RANGE;
2851                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2852                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2853                  *class_utf8data++ = XCL_RANGE;
2854                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2855                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2856                  *class_utf8data++ = XCL_RANGE;
2857                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2858                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2859                  *class_utf8data++ = XCL_RANGE;
2860                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2861                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2862                  *class_utf8data++ = XCL_RANGE;
2863                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2864                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2865                  *class_utf8data++ = XCL_RANGE;
2866                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2867                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2868                  *class_utf8data++ = XCL_RANGE;
2869                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2870                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2871                  }
2872    #endif
2873                continue;
2874                }
2875    
2876              if (-c == ESC_v)
2877                {
2878                SETBIT(classbits, 0x0a); /* LF */
2879                SETBIT(classbits, 0x0b); /* VT */
2880                SETBIT(classbits, 0x0c); /* FF */
2881                SETBIT(classbits, 0x0d); /* CR */
2882                SETBIT(classbits, 0x85); /* NEL */
2883    #ifdef SUPPORT_UTF8
2884                if (utf8)
2885                  {
2886                  class_utf8 = TRUE;
2887                  *class_utf8data++ = XCL_RANGE;
2888                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2889                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2890                  }
2891    #endif
2892                continue;
2893                }
2894    
2895              if (-c == ESC_V)
2896                {
2897                for (c = 0; c < 32; c++)
2898                  {
2899                  int x = 0xff;
2900                  switch (c)
2901                    {
2902                    case 0x0a/8: x ^= 1 << (0x0a%8);
2903                                 x ^= 1 << (0x0b%8);
2904                                 x ^= 1 << (0x0c%8);
2905                                 x ^= 1 << (0x0d%8);
2906                                 break;
2907                    case 0x85/8: x ^= 1 << (0x85%8); break;
2908                    default: break;
2909                    }
2910                  classbits[c] |= x;
2911                  }
2912    
2913    #ifdef SUPPORT_UTF8
2914                if (utf8)
2915                  {
2916                  class_utf8 = TRUE;
2917                  *class_utf8data++ = XCL_RANGE;
2918                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2919                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2920                  *class_utf8data++ = XCL_RANGE;
2921                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2922                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2923                  }
2924    #endif
2925                continue;
2926                }
2927    
2928              /* We need to deal with \P and \p in both phases. */
2929    
2930    #ifdef SUPPORT_UCP
2931              if (-c == ESC_p || -c == ESC_P)
2932                {
2933                BOOL negated;
2934                int pdata;
2935                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2936                if (ptype < 0) goto FAILED;
2937                class_utf8 = TRUE;
2938                *class_utf8data++ = ((-c == ESC_p) != negated)?
2939                  XCL_PROP : XCL_NOTPROP;
2940                *class_utf8data++ = ptype;
2941                *class_utf8data++ = pdata;
2942                class_charcount -= 2;   /* Not a < 256 character */
2943                continue;
2944                }
2945    #endif
2946              /* Unrecognized escapes are faulted if PCRE is running in its
2947              strict mode. By default, for compatibility with Perl, they are
2948              treated as literals. */
2949    
2950              if ((options & PCRE_EXTRA) != 0)
2951                {
2952                *errorcodeptr = ERR7;
2953                goto FAILED;
2954                }
2955    
2956              class_charcount -= 2;  /* Undo the default count from above */
2957              c = *ptr;              /* Get the final character and fall through */
2958              }
2959    
2960            /* Fall through if we have a single character (c >= 0). This may be
2961            greater than 256 in UTF-8 mode. */
2962    
2963            }   /* End of backslash handling */
2964    
2965          /* A single character may be followed by '-' to form a range. However,
2966          Perl does not permit ']' to be the end of the range. A '-' character
2967          at the end is treated as a literal. Perl ignores orphaned \E sequences
2968          entirely. The code for handling \Q and \E is messy. */
2969    
2970          CHECK_RANGE:
2971          while (ptr[1] == '\\' && ptr[2] == 'E')
2972            {
2973            inescq = FALSE;
2974            ptr += 2;
2975            }
2976    
2977          oldptr = ptr;
2978    
2979          if (!inescq && ptr[1] == '-')
2980            {
2981            int d;
2982            ptr += 2;
2983            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
2984    
2985            /* If we hit \Q (not followed by \E) at this point, go into escaped
2986            mode. */
2987    
2988            while (*ptr == '\\' && ptr[1] == 'Q')
2989              {
2990              ptr += 2;
2991              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
2992              inescq = TRUE;
2993              break;
2994              }
2995    
2996            if (*ptr == 0 || (!inescq && *ptr == ']'))
2997              {
2998              ptr = oldptr;
2999              goto LONE_SINGLE_CHARACTER;
3000              }
3001    
3002    #ifdef SUPPORT_UTF8
3003            if (utf8)
3004            {                           /* Braces are required because the */            {                           /* Braces are required because the */
3005            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
3006            }            }
# Line 1981  for (;; ptr++) Line 3012  for (;; ptr++)
3012          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
3013          in such circumstances. */          in such circumstances. */
3014    
3015          if (d == '\\')          if (!inescq && d == '\\')
3016            {            {
3017            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3018            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3019    
3020            /* \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
3021            was literal */            special means the '-' was literal */
3022    
3023            if (d < 0)            if (d < 0)
3024              {              {
3025              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3026              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3027                else if (d == -ESC_R) d = 'R'; else
3028                {                {
3029                ptr = oldptr - 2;                ptr = oldptr;
3030                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3031                }                }
3032              }              }
3033            }            }
3034    
3035          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3036          the pre-pass. Optimize one-character ranges */          one-character ranges */
3037    
3038            if (d < c)
3039              {
3040              *errorcodeptr = ERR8;
3041              goto FAILED;
3042              }
3043    
3044          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3045    
# Line 2022  for (;; ptr++) Line 3060  for (;; ptr++)
3060  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3061            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3062              {              {
3063              int occ, ocd;              unsigned int occ, ocd;
3064              int cc = c;              unsigned int cc = c;
3065              int origd = d;              unsigned int origd = d;
3066              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3067                {                {
3068                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3069                      ocd <= (unsigned int)d)
3070                    continue;                          /* Skip embedded ranges */
3071    
3072                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3073                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3074                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3075                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3076                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3077                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3078                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3079                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3080                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3081                  d = ocd;                  d = ocd;
3082                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3124  for (;; ptr++)
3124          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
3125          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3126    
3127          for (; c <= d; c++)          class_charcount += d - c + 1;
3128            class_lastchar = d;
3129    
3130            /* We can save a bit of time by skipping this in the pre-compile. */
3131    
3132            if (lengthptr == NULL) for (; c <= d; c++)
3133            {            {
3134            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3135            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3137  for (;; ptr++)
3137              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3138              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3139              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3140            }            }
3141    
3142          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3160  for (;; ptr++)
3160  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3161          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3162            {            {
3163            int chartype;            unsigned int othercase;
3164            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3165              {              {
3166              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3167              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3186  for (;; ptr++)
3186          }          }
3187        }        }
3188    
3189      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3190      loop. This "while" is the end of the "do" above. */  
3191        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3192    
3193      while ((c = *(++ptr)) != ']' || inescq);      if (c == 0)                          /* Missing terminating ']' */
3194          {
3195          *errorcodeptr = ERR6;
3196          goto FAILED;
3197          }
3198    
3199      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3200      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3201      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3202      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3203      single-bytes only. This is an historical hangover. Maybe one day we can  
3204      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3205        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3206        operate on single-bytes only. This is an historical hangover. Maybe one day
3207        we can tidy these opcodes to handle multi-byte characters.
3208    
3209      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3210      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2163  for (;; ptr++) Line 3214  for (;; ptr++)
3214      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3215    
3216  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3217      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3218            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3219  #else  #else
3220      if (class_charcount == 1)      if (class_charcount == 1)
3221  #endif  #endif
# Line 2210  for (;; ptr++) Line 3259  for (;; ptr++)
3259    
3260      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3261      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode. If there are no characters < 256,
3262      we can omit the bitmap. */      we can omit the bitmap in the actual compiled code. */
3263    
3264  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3265      if (class_utf8)      if (class_utf8)
# Line 2220  for (;; ptr++) Line 3269  for (;; ptr++)
3269        code += LINK_SIZE;        code += LINK_SIZE;
3270        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3271    
3272        /* 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;
3273        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3274    
3275        if (class_charcount > 0)        if (class_charcount > 0)
3276          {          {
3277          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3278            memmove(code + 32, code, class_utf8data - code);
3279          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3280          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;  
3281          }          }
3282          else code = class_utf8data;
3283    
3284        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3285    
# Line 2254  for (;; ptr++) Line 3296  for (;; ptr++)
3296      if (negate_class)      if (negate_class)
3297        {        {
3298        *code++ = OP_NCLASS;        *code++ = OP_NCLASS;
3299        for (c = 0; c < 32; c++) code[c] = ~classbits[c];        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3300            for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3301        }        }
3302      else      else
3303        {        {
# Line 2264  for (;; ptr++) Line 3307  for (;; ptr++)
3307      code += 32;      code += 32;
3308      break;      break;
3309    
3310    
3311        /* ===================================================================*/
3312      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3313      has been tested above. */      has been tested above. */
3314    
# Line 2331  for (;; ptr++) Line 3376  for (;; ptr++)
3376        }        }
3377      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3378    
     /* 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;  
       }  
   
3379      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3380      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
3381      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 3409  for (;; ptr++)
3409          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3410          }          }
3411    
3412          /* If the repetition is unlimited, it pays to see if the next thing on
3413          the line is something that cannot possibly match this character. If so,
3414          automatically possessifying this item gains some performance in the case
3415          where the match fails. */
3416    
3417          if (!possessive_quantifier &&
3418              repeat_max < 0 &&
3419              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3420                options, cd))
3421            {
3422            repeat_type = 0;    /* Force greedy */
3423            possessive_quantifier = TRUE;
3424            }
3425    
3426        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3427        }        }
3428    
3429      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3430      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-
3431      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3432      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3433        currently used only for single-byte chars. */
3434    
3435      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3436        {        {
3437        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3438        c = previous[1];        c = previous[1];
3439          if (!possessive_quantifier &&
3440              repeat_max < 0 &&
3441              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3442            {
3443            repeat_type = 0;    /* Force greedy */
3444            possessive_quantifier = TRUE;
3445            }
3446        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3447        }        }
3448    
# Line 2403  for (;; ptr++) Line 3456  for (;; ptr++)
3456      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3457        {        {
3458        uschar *oldcode;        uschar *oldcode;
3459        int prop_type;        int prop_type, prop_value;
3460        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3461        c = *previous;        c = *previous;
3462    
3463          if (!possessive_quantifier &&
3464              repeat_max < 0 &&
3465              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3466            {
3467            repeat_type = 0;    /* Force greedy */
3468            possessive_quantifier = TRUE;
3469            }
3470    
3471        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3472        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3473          previous[1] : -1;          {
3474            prop_type = previous[1];
3475            prop_value = previous[2];
3476            }
3477          else prop_type = prop_value = -1;
3478    
3479        oldcode = code;        oldcode = code;
3480        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2443  for (;; ptr++) Line 3508  for (;; ptr++)
3508          }          }
3509    
3510        /* 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
3511        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3512        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
3513        one less than the maximum. */        one less than the maximum. */
3514    
# Line 2470  for (;; ptr++) Line 3535  for (;; ptr++)
3535    
3536          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3537          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
3538          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3539          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3540          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3541    
# Line 2486  for (;; ptr++) Line 3551  for (;; ptr++)
3551  #endif  #endif
3552              {              {
3553              *code++ = c;              *code++ = c;
3554              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3555                  {
3556                  *code++ = prop_type;
3557                  *code++ = prop_value;
3558                  }
3559              }              }
3560            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3561            }            }
3562    
3563          /* 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
3564          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3565            UPTO is just for 1 instance, we can use QUERY instead. */
3566    
3567          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3568            {            {
# Line 2505  for (;; ptr++) Line 3575  for (;; ptr++)
3575            else            else
3576  #endif  #endif
3577            *code++ = c;            *code++ = c;
3578            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3579                {
3580                *code++ = prop_type;
3581                *code++ = prop_value;
3582                }
3583            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3584            *code++ = OP_UPTO + repeat_type;  
3585            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3586                {
3587                *code++ = OP_QUERY + repeat_type;
3588                }
3589              else
3590                {
3591                *code++ = OP_UPTO + repeat_type;
3592                PUT2INC(code, 0, repeat_max);
3593                }
3594            }            }
3595          }          }
3596    
# Line 2524  for (;; ptr++) Line 3606  for (;; ptr++)
3606  #endif  #endif
3607        *code++ = c;        *code++ = c;
3608    
3609        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3610        defines the required property. */        define the required property. */
3611    
3612  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3613        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3614            {
3615            *code++ = prop_type;
3616            *code++ = prop_value;
3617            }
3618  #endif  #endif
3619        }        }
3620    
# Line 2571  for (;; ptr++) Line 3657  for (;; ptr++)
3657      /* 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
3658      cases. */      cases. */
3659    
3660      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3661               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3662        {        {
3663        register int i;        register int i;
3664        int ketoffset = 0;        int ketoffset = 0;
3665        int len = code - previous;        int len = code - previous;
3666        uschar *bralink = NULL;        uschar *bralink = NULL;
3667    
3668          /* Repeating a DEFINE group is pointless */
3669    
3670          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3671            {
3672            *errorcodeptr = ERR55;
3673            goto FAILED;
3674            }
3675    
3676        /* 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
3677        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
3678        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 3707  for (;; ptr++)
3707          /* 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
3708          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
3709          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
3710          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3711          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3712            doing this. */
3713    
3714          if (repeat_max <= 1)          if (repeat_max <= 1)
3715            {            {
3716            *code = OP_END;            *code = OP_END;
3717            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3718            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3719            code++;            code++;
3720            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2637  for (;; ptr++) Line 3732  for (;; ptr++)
3732            {            {
3733            int offset;            int offset;
3734            *code = OP_END;            *code = OP_END;
3735            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3736            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3737            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3738            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3752  for (;; ptr++)
3752        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3753        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3754        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
3755        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3756          forward reference subroutine calls in the group, there will be entries on
3757          the workspace list; replicate these with an appropriate increment. */
3758    
3759        else        else
3760          {          {
3761          if (repeat_min > 1)          if (repeat_min > 1)
3762            {            {
3763            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3764            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3765              potential integer overflow. */
3766    
3767              if (lengthptr != NULL)
3768                {
3769                int delta = (repeat_min - 1)*length_prevgroup;
3770                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3771                                                                (double)INT_MAX ||
3772                    OFLOW_MAX - *lengthptr < delta)
3773                  {
3774                  *errorcodeptr = ERR20;
3775                  goto FAILED;
3776                  }
3777                *lengthptr += delta;
3778                }
3779    
3780              /* This is compiling for real */
3781    
3782              else
3783              {              {
3784              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3785              code += len;              for (i = 1; i < repeat_min; i++)
3786                  {
3787                  uschar *hc;
3788                  uschar *this_hwm = cd->hwm;
3789                  memcpy(code, previous, len);
3790                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3791                    {
3792                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3793                    cd->hwm += LINK_SIZE;
3794                    }
3795                  save_hwm = this_hwm;
3796                  code += len;
3797                  }
3798              }              }
3799            }            }
3800    
3801          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3802          }          }
3803    
# Line 2677  for (;; ptr++) Line 3805  for (;; ptr++)
3805        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3806        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,
3807        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
3808        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3809          replicate entries on the forward reference list. */
3810    
3811        if (repeat_max >= 0)        if (repeat_max >= 0)
3812          {          {
3813          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3814            just adjust the length as if we had. For each repetition we must add 1
3815            to the length for BRAZERO and for all but the last repetition we must
3816            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3817            paranoid checks to avoid integer overflow. */
3818    
3819            if (lengthptr != NULL && repeat_max > 0)
3820              {
3821              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3822                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3823              if ((double)repeat_max *
3824                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3825                      > (double)INT_MAX ||
3826                  OFLOW_MAX - *lengthptr < delta)
3827                {
3828                *errorcodeptr = ERR20;
3829                goto FAILED;
3830                }
3831              *lengthptr += delta;
3832              }
3833    
3834            /* This is compiling for real */
3835    
3836            else for (i = repeat_max - 1; i >= 0; i--)
3837            {            {
3838              uschar *hc;
3839              uschar *this_hwm = cd->hwm;
3840    
3841            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3842    
3843            /* 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 3853  for (;; ptr++)
3853              }              }
3854    
3855            memcpy(code, previous, len);            memcpy(code, previous, len);
3856              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3857                {
3858                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3859                cd->hwm += LINK_SIZE;
3860                }
3861              save_hwm = this_hwm;
3862            code += len;            code += len;
3863            }            }
3864    
# Line 2720  for (;; ptr++) Line 3881  for (;; ptr++)
3881        /* 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
3882        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3883        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
3884        correct offset was computed above. */        correct offset was computed above.
3885    
3886          Then, when we are doing the actual compile phase, check to see whether
3887          this group is a non-atomic one that could match an empty string. If so,
3888          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3889          that runtime checking can be done. [This check is also applied to
3890          atomic groups at runtime, but in a different way.] */
3891    
3892        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
3893            {
3894            uschar *ketcode = code - ketoffset;
3895            uschar *bracode = ketcode - GET(ketcode, 1);
3896            *ketcode = OP_KETRMAX + repeat_type;
3897            if (lengthptr == NULL && *bracode != OP_ONCE)
3898              {
3899              uschar *scode = bracode;
3900              do
3901                {
3902                if (could_be_empty_branch(scode, ketcode, utf8))
3903                  {
3904                  *bracode += OP_SBRA - OP_BRA;
3905                  break;
3906                  }
3907                scode += GET(scode, 1);
3908                }
3909              while (*scode == OP_ALT);
3910              }
3911            }
3912        }        }
3913    
3914      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 3919  for (;; ptr++)
3919        goto FAILED;        goto FAILED;
3920        }        }
3921    
3922      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
3923      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
3924      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
3925      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.
3926      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
3927        but the special opcodes can optimize it a bit. The repeated item starts at
3928        tempcode, not at previous, which might be the first part of a string whose
3929        (former) last char we repeated.
3930    
3931        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
3932        an 'upto' may follow. We skip over an 'exact' item, and then test the
3933        length of what remains before proceeding. */
3934    
3935      if (possessive_quantifier)      if (possessive_quantifier)
3936        {        {
3937        int len = code - tempcode;        int len;
3938        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
3939        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
3940        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
3941        tempcode[0] = OP_ONCE;        len = code - tempcode;
3942        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
3943        PUTINC(code, 0, len);          {
3944        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
3945            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
3946            case OP_QUERY: *tempcode = OP_POSQUERY; break;
3947            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
3948    
3949            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
3950            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
3951            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
3952            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
3953    
3954            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
3955            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
3956            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
3957            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
3958    
3959            default:
3960            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
3961            code += 1 + LINK_SIZE;
3962            len += 1 + LINK_SIZE;
3963            tempcode[0] = OP_ONCE;
3964            *code++ = OP_KET;
3965            PUTINC(code, 0, len);
3966            PUT(tempcode, 1, len);
3967            break;
3968            }
3969        }        }
3970    
3971      /* 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 3978  for (;; ptr++)
3978      break;      break;
3979    
3980    
3981      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
3982      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
3983      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
3984      of any of them means that this is not a referencing group. They were      parenthesis forms.  */
     checked for validity in the first pass over the string, so we don't have to  
     check for syntax errors here.  */  
3985    
3986      case '(':      case '(':
3987      newoptions = options;      newoptions = options;
3988      skipbytes = 0;      skipbytes = 0;
3989        bravalue = OP_CBRA;
3990        save_hwm = cd->hwm;
3991        reset_bracount = FALSE;
3992    
3993        /* First deal with various "verbs" that can be introduced by '*'. */
3994    
3995        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
3996          {
3997          int i, namelen;
3998          const uschar *name = ++ptr;
3999          previous = NULL;
4000          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4001          if (*ptr == ':')
4002            {
4003            *errorcodeptr = ERR59;   /* Not supported */
4004            goto FAILED;
4005            }
4006          if (*ptr != ')')
4007            {
4008            *errorcodeptr = ERR60;
4009            goto FAILED;
4010            }
4011          namelen = ptr - name;
4012          for (i = 0; i < verbcount; i++)
4013            {
4014            if (namelen == verbs[i].len &&
4015                strncmp((char *)name, verbs[i].name, namelen) == 0)
4016              {
4017              *code = verbs[i].op;
4018              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4019              break;
4020              }
4021            }
4022          if (i < verbcount) continue;
4023          *errorcodeptr = ERR60;
4024          goto FAILED;
4025          }
4026    
4027        /* Deal with the extended parentheses; all are introduced by '?', and the
4028        appearance of any of them means that this is not a capturing group. */
4029    
4030      if (*(++ptr) == '?')      else if (*ptr == '?')
4031        {        {
4032        int set, unset;        int i, set, unset, namelen;
4033        int *optset;        int *optset;
4034          const uschar *name;
4035          uschar *slot;
4036    
4037        switch (*(++ptr))        switch (*(++ptr))
4038          {          {
4039          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4040          ptr++;          ptr++;
4041          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4042            if (*ptr == 0)
4043              {
4044              *errorcodeptr = ERR18;
4045              goto FAILED;
4046              }
4047          continue;          continue;
4048    
4049          case ':':                 /* Non-extracting bracket */  
4050            /* ------------------------------------------------------------ */
4051            case '|':                 /* Reset capture count for each branch */
4052            reset_bracount = TRUE;
4053            /* Fall through */
4054    
4055            /* ------------------------------------------------------------ */
4056            case ':':                 /* Non-capturing bracket */
4057          bravalue = OP_BRA;          bravalue = OP_BRA;
4058          ptr++;          ptr++;
4059          break;          break;
4060    
4061    
4062            /* ------------------------------------------------------------ */
4063          case '(':          case '(':
4064          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4065    
4066          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4067            group), a name (referring to a named group), or 'R', referring to
4068            recursion. R<digits> and R&name are also permitted for recursion tests.
4069    
4070            There are several syntaxes for testing a named group: (?(name)) is used
4071            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4072    
4073            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4074            be the recursive thing or the name 'R' (and similarly for 'R' followed
4075            by digits), and (b) a number could be a name that consists of digits.
4076            In both cases, we look for a name first; if not found, we try the other
4077            cases. */
4078    
4079            /* For conditions that are assertions, check the syntax, and then exit
4080            the switch. This will take control down to where bracketed groups,
4081            including assertions, are processed. */
4082    
4083            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4084              break;
4085    
4086            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4087            below), and all need to skip 3 bytes at the start of the group. */
4088    
4089            code[1+LINK_SIZE] = OP_CREF;
4090            skipbytes = 3;
4091            refsign = -1;
4092    
4093          if (ptr[1] == 'R')          /* Check for a test for recursion in a named group. */
4094    
4095            if (ptr[1] == 'R' && ptr[2] == '&')
4096            {            {
4097            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4098            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4099            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4100            }            }
4101    
4102          /* Condition to test for a numbered subpattern match. We know that          /* Check for a test for a named group's having been set, using the Perl
4103          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4104    
4105          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4106            {            {
4107            int condref;                 /* Don't amalgamate; some compilers */            terminator = '>';
           condref = *(++ptr) - '0';    /* grumble at autoincrement in declaration */  
           while (*(++ptr) != ')') condref = condref*10 + *ptr - '0';  
           if (condref == 0)  
             {  
             *errorcodeptr = ERR35;  
             goto FAILED;  
             }  
4108            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4109            }            }
4110          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
4111          set bravalue above. */            {
4112          break;            terminator = '\'';
4113              ptr++;
4114          case '=':                 /* Positive lookahead */            }
4115          bravalue = OP_ASSERT;          else
4116          ptr++;            {
4117          break;            terminator = 0;
4118              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4119              }
4120    
4121          case '!':                 /* Negative lookahead */          /* We now expect to read a name; any thing else is an error */
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
4122    
4123          case '<':                 /* Lookbehinds */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
         switch (*(++ptr))  
4124            {            {
4125            case '=':               /* Positive lookbehind */            ptr += 1;  /* To get the right offset */
4126            bravalue = OP_ASSERTBACK;            *errorcodeptr = ERR28;
4127            ptr++;            goto FAILED;
4128            break;            }
4129    
4130            case '!':               /* Negative lookbehind */          /* Read the name, but also get it as a number if it's all digits */
4131            bravalue = OP_ASSERTBACK_NOT;  
4132            recno = 0;
4133            name = ++ptr;
4134            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4135              {
4136              if (recno >= 0)
4137                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4138                  recno * 10 + *ptr - '0' : -1;
4139            ptr++;            ptr++;
           break;  
4140            }            }
4141          break;          namelen = ptr - name;
4142    
4143          case '>':                 /* One-time brackets */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4144          bravalue = OP_ONCE;            {
4145          ptr++;            ptr--;      /* Error offset */
4146          break;            *errorcodeptr = ERR26;
4147              goto FAILED;
4148              }
4149    
4150          case 'C':                 /* Callout - may be followed by digits; */          /* Do no further checking in the pre-compile phase. */
4151          previous_callout = code;  /* Save for later completion */  
4152          after_manual_callout = 1; /* Skip one item before completing */          if (lengthptr != NULL) break;
4153          *code++ = OP_CALLOUT;     /* Already checked that the terminating */  
4154            {                       /* closing parenthesis is present. */          /* In the real compile we do the work of looking for the actual
4155            int n = 0;          reference. If the string started with "+" or "-" we require the rest to
4156            while ((digitab[*(++ptr)] & ctype_digit) != 0)          be digits, in which case recno will be set. */
4157              n = n * 10 + *ptr - '0';  
4158            if (refsign > 0)
4159              {
4160              if (recno <= 0)
4161                {
4162                *errorcodeptr = ERR58;
4163                goto FAILED;
4164                }
4165              if (refsign == '-')
4166                {
4167                recno = cd->bracount - recno + 1;
4168                if (recno <= 0)
4169                  {
4170                  *errorcodeptr = ERR15;
4171                  goto FAILED;
4172                  }
4173                }
4174              else recno += cd->bracount;
4175              PUT2(code, 2+LINK_SIZE, recno);
4176              break;
4177              }
4178    
4179            /* Otherwise (did not start with "+" or "-"), start by looking for the
4180            name. */
4181    
4182            slot = cd->name_table;
4183            for (i = 0; i < cd->names_found; i++)
4184              {
4185              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4186              slot += cd->name_entry_size;
4187              }
4188    
4189            /* Found a previous named subpattern */
4190    
4191            if (i < cd->names_found)
4192              {
4193              recno = GET2(slot, 0);
4194              PUT2(code, 2+LINK_SIZE, recno);
4195              }
4196    
4197            /* Search the pattern for a forward reference */
4198    
4199            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4200                            (options & PCRE_EXTENDED) != 0)) > 0)
4201              {
4202              PUT2(code, 2+LINK_SIZE, i);
4203              }
4204    
4205            /* If terminator == 0 it means that the name followed directly after
4206            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4207            some further alternatives to try. For the cases where terminator != 0
4208            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4209            now checked all the possibilities, so give an error. */
4210    
4211            else if (terminator != 0)
4212              {
4213              *errorcodeptr = ERR15;
4214              goto FAILED;
4215              }
4216    
4217            /* Check for (?(R) for recursion. Allow digits after R to specify a
4218            specific group number. */
4219    
4220            else if (*name == 'R')
4221              {
4222              recno = 0;
4223              for (i = 1; i < namelen; i++)
4224                {
4225                if ((digitab[name[i]] & ctype_digit) == 0)
4226                  {
4227                  *errorcodeptr = ERR15;
4228                  goto FAILED;
4229                  }
4230                recno = recno * 10 + name[i] - '0';
4231                }
4232              if (recno == 0) recno = RREF_ANY;
4233              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4234              PUT2(code, 2+LINK_SIZE, recno);
4235              }
4236    
4237            /* Similarly, check for the (?(DEFINE) "condition", which is always
4238            false. */
4239    
4240            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4241              {
4242              code[1+LINK_SIZE] = OP_DEF;
4243              skipbytes = 1;
4244              }
4245    
4246            /* Check for the "name" actually being a subpattern number. */
4247    
4248            else if (recno > 0)
4249              {
4250              PUT2(code, 2+LINK_SIZE, recno);
4251              }
4252    
4253            /* Either an unidentified subpattern, or a reference to (?(0) */
4254    
4255            else
4256              {
4257              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4258              goto FAILED;
4259              }
4260            break;
4261    
4262    
4263            /* ------------------------------------------------------------ */
4264            case '=':                 /* Positive lookahead */
4265            bravalue = OP_ASSERT;
4266            ptr++;
4267            break;
4268    
4269    
4270            /* ------------------------------------------------------------ */
4271            case '!':                 /* Negative lookahead */
4272            ptr++;
4273            if (*ptr == ')')          /* Optimize (?!) */
4274              {
4275              *code++ = OP_FAIL;
4276              previous = NULL;
4277              continue;
4278              }
4279            bravalue = OP_ASSERT_NOT;
4280            break;
4281    
4282    
4283            /* ------------------------------------------------------------ */
4284            case '<':                 /* Lookbehind or named define */
4285            switch (ptr[1])
4286              {
4287              case '=':               /* Positive lookbehind */
4288              bravalue = OP_ASSERTBACK;
4289              ptr += 2;
4290              break;
4291    
4292              case '!':               /* Negative lookbehind */
4293              bravalue = OP_ASSERTBACK_NOT;
4294              ptr += 2;
4295              break;
4296    
4297              default:                /* Could be name define, else bad */
4298              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4299              ptr++;                  /* Correct offset for error */
4300              *errorcodeptr = ERR24;
4301              goto FAILED;
4302              }
4303            break;
4304    
4305    
4306            /* ------------------------------------------------------------ */
4307            case '>':                 /* One-time brackets */
4308            bravalue = OP_ONCE;
4309            ptr++;
4310            break;
4311    
4312    
4313            /* ------------------------------------------------------------ */
4314            case 'C':                 /* Callout - may be followed by digits; */
4315            previous_callout = code;  /* Save for later completion */
4316            after_manual_callout = 1; /* Skip one item before completing */
4317            *code++ = OP_CALLOUT;
4318              {
4319              int n = 0;
4320              while ((digitab[*(++ptr)] & ctype_digit) != 0)
4321                n = n * 10 + *ptr - '0';
4322              if (*ptr != ')')
4323                {
4324                *errorcodeptr = ERR39;
4325                goto FAILED;
4326                }
4327            if (n > 255)            if (n > 255)
4328              {              {
4329              *errorcodeptr = ERR38;              *errorcodeptr = ERR38;
# Line 2876  for (;; ptr++) Line 4337  for (;; ptr++)
4337          previous = NULL;          previous = NULL;
4338          continue;          continue;
4339    
4340          case 'P':                 /* Named subpattern handling */  
4341          if (*(++ptr) == '<')      /* Definition */          /* ------------------------------------------------------------ */
4342            case 'P':                 /* Python-style named subpattern handling */
4343            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
4344              {
4345              is_recurse = *ptr == '>';
4346              terminator = ')';
4347              goto NAMED_REF_OR_RECURSE;
4348              }
4349            else if (*ptr != '<')    /* Test for Python-style definition */
4350            {            {
4351            int i, namelen;            *errorcodeptr = ERR41;
4352            uschar *slot = cd->name_table;            goto FAILED;
4353            const uschar *name;     /* Don't amalgamate; some compilers */            }
4354            name = ++ptr;           /* grumble at autoincrement in declaration */          /* Fall through to handle (?P< as (?< is handled */
4355    
           while (*ptr++ != '>');  
           namelen = ptr - name - 1;  
4356    
4357            for (i = 0; i < cd->names_found; i++)          /* ------------------------------------------------------------ */
4358            DEFINE_NAME:    /* Come here from (?< handling */
4359            case '\'':
4360              {
4361              terminator = (*ptr == '<')? '>' : '\'';
4362              name = ++ptr;
4363    
4364              while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
4365              namelen = ptr - name;
4366    
4367              /* In the pre-compile phase, just do a syntax check. */
4368    
4369              if (lengthptr != NULL)
4370              {              {
4371              int crc = memcmp(name, slot+2, namelen);              if (*ptr != terminator)
4372              if (crc == 0)                {
4373                  *errorcodeptr = ERR42;
4374                  goto FAILED;
4375                  }
4376                if (cd->names_found >= MAX_NAME_COUNT)
4377                  {
4378                  *errorcodeptr = ERR49;
4379                  goto FAILED;
4380                  }
4381                if (namelen + 3 > cd->name_entry_size)
4382                {                {
4383                if (slot[2+namelen] == 0)                cd->name_entry_size = namelen + 3;
4384                  if (namelen > MAX_NAME_SIZE)
4385                  {                  {
4386                  *errorcodeptr = ERR43;                  *errorcodeptr = ERR48;
4387                  goto FAILED;                  goto FAILED;
4388                  }                  }
               crc = -1;             /* Current name is substring */  
4389                }                }
4390              if (crc < 0)              }
4391    
4392              /* In the real compile, create the entry in the table */
4393    
4394              else
4395                {
4396                slot = cd->name_table;
4397                for (i = 0; i < cd->names_found; i++)
4398                {                {
4399                memmove(slot + cd->name_entry_size, slot,                int crc = memcmp(name, slot+2, namelen);
4400                  (cd->names_found - i) * cd->name_entry_size);                if (crc == 0)
4401                break;                  {
4402                    if (slot[2+namelen] == 0)
4403                      {
4404                      if ((options & PCRE_DUPNAMES) == 0)
4405                        {
4406                        *errorcodeptr = ERR43;
4407                        goto FAILED;
4408                        }
4409                      }
4410                    else crc = -1;      /* Current name is substring */
4411                    }
4412                  if (crc < 0)
4413                    {
4414                    memmove(slot + cd->name_entry_size, slot,
4415                      (cd->names_found - i) * cd->name_entry_size);
4416                 &nbs